transactions on environment and electrical engineering issn 2450-5730 vol 1, no 3 (2016) © petr krejci, pavel santarius, radomir gono & zdenek brunclik  abstract—in 1997 cooperation began between the faculty of electrical engineering and computer science and the power company cez on a joint project that involved monitoring selected voltage quality parameters. over the course of several years measurements were carried out at 59 points of the distribution system at a low voltage (lv) level and their power supply nodes at medium voltage (mv) and high voltage (hv) levels, which constituted more than 100 measured points. the measurements continued until 2012 (with a one-year break in 2009) and with a total scope of five cycles of three-year measurements. the last phase of the measurements was focused, inter alia, on the evaluation of 15th and 21st voltage harmonics. from 2013, the cooperation focused again on monitoring flicker in localities with the occurrence of elevated levels of this voltage quality parameter. at these sites, the monitoring of flicker was carried out for two weeks in spring and autumn for a period of three years. the basis for evaluating the quality parameters is the standard en 50160. in addition the results of annual continual monitoring for one selected locality on mv level will be presented in the paper. in addition the possibility of utilization of intelligent electrometers for monitoring of some quality parameters will be referred in the paper. index terms— flicker, unbalance, voltage harmonics, voltage quality. i. comprehensive long-term monitoring of some parameters of the voltage quality in the regional power engineering company as already mentioned, in 1997 cooperation began between the faculty of electrical engineering and computer science and the power company cez on a joint project that involved monitoring selected voltage quality parameters. among the main parameters to be monitored were selected harmonic voltages, flicker and voltage unbalance. given the large number of measurement points, the measurements were divided into six partial stages (the whole range that was measured is shown in fig. 1), in order to complete the entire measuring cycle within this research was partially supported by the sgs grant from vsb technical university of ostrava (no. sp2016/95 and no. sp2016/146). petr krejci, pavel santarius, radomir gono faculty of electrical engineering and computer science vsb – technical university of ostrava, ostrava, czech republic three years. after that period the measurements were repeated, and because they involved five three-year measurement cycles, the monitoring of these quality parameters lasted for 15 years. according to the requirements of the en 50160 standard [1], the measurement of the voltage quality parameters was carried out for one week, with a measurement interval of 10 minutes. a diagram of the connection of the measuring points is shown in fig. 2. fig. 1. locality of measurement at single feeder points the measured data was evaluated in all phases and on all voltage levels:  selected voltage harmonics (3rd, 5th, 7th, 9th, 11th)  flicker  unbalance (n) petr.krejci@vsb.cz, pavel.santarius@vsb.cz, radomir.gono@vsb,cz zdenek brunclik cez distribuce, a.s., czech power company (cez), ostrava, czech republic zdenek.brunclik@cez.cz long-term monitoring of flicker and some other parameters of voltage quality petr krejci, pavel santarius, radomir gono, zdenek brunclik fig. 2. ideal connection diagram of measuring points the results of this long-term measurement have been published continuously, e.g. [2]. this paper summarises the results and particularly the trends in the changes in the voltage quality parameters over the monitoring period of 15 years. ii. conclusions – the trends of changes continuous results of the measurements have already been published several times, e.g. [2]. in the tables below the changes in the selected voltage quality parameters (selected harmonics, flicker and unbalance) in the lv, mv and hv network during the five monitoring cycles are summarised. an example of an evaluation of selected voltage quality parameters (fifth harmonic, flicker and unbalance) in the lv networks is shown in fig. 3. table i. trends of changes in the selected voltage quality parameters in the lv networks lv ii.-i. lv iii.-ii. lv iv.-iii. lv v.-iv. u_03 (%) 0.056 -0.016 -0.004 0.00 u_05 (%) 0.157 0.204 -0.334 -0.11 u_07 (%) 0.095 0.204 -0.017 -0.04 u_09 (%) 0.006 0.026 0.010 0.02 u_11 (%) 0.015 0.075 0.024 0.05 pst (-) 0.157 -0.034 0.008 0.05 plt (-) 0.118 -0.005 -0.022 -0.05 n (%) 0.088 0.043 -0.066 -0.02 table ii. trends of changes in the selected voltage quality parameters in the mv networks mv ii.-i. mv iii-ii. mv iv-iii. mv v.-iv. u_03 (%) 0.056 -0.026 -0.019 -0.02 u_05 (%) 0.335 0.181 -0.303 -0.14 u_07 (%) 0.111 0.193 -0.022 -0.09 u_09 (%) -0.018 -0.009 -0.006 0.00 u_11 (%) 0.016 0.036 0.021 -0.01 pst (-) 0.135 -0.093 0.048 -0.06 plt (-) 0.077 -0.079 0.022 -0.07 n (%) 0.135 -0.052 -0.058 -0.01 table iii. trends of changes in the selected voltage quality parameters in the hv networks hv ii.-i. hv iii-ii. hv iv-iii. hv v-iv. u_03 (%) 0.183 0.089 -0.067 -0.045 u_05 (%) 0.065 0.106 -0.082 -0.057 u_07 (%) 0.040 0.073 -0.004 0.013 u_09 (%) -0.050 -0.012 -0.002 -0.012 u_11 (%) -0.001 0.001 0.022 -0.021 pst (-) 0.113 -0.024 0.050 -0.091 plt (-) 0.131 -0.080 0.050 -0.103 n (%) -0.017 0.169 -0.176 0.031 the most important conclusions: a) as regards harmonics, the results are relatively positive; the values of individual harmonic components are significantly below the values of compatible levels and the fifth harmonic was the most considerable at all voltage levels. the extreme values were only found at one point in the lv network, where it exceeded the third harmonic in 1999 (10.7%) and in 2002 (12.6%) and the ninth harmonic in 2002 (2.1%). the probable reason for this was resonance in the lv network; b) as for unbalance, the changes are also quite small. a small exceeding was only found at three points in the mv network; c) as for flicker, the situation was worse. exceeding the compatibility level was found at many measuring points and on all voltage levels; d) the trends of changes in the selected voltage quality parameters during the years 1997 to 2012 do not show extreme swings or a long-term increase in the selected voltage quality parameters. iii. measuring and evaluation of 15th and 21st harmonics in the last monitoring cycle (2010-2012) the 15th and 21st voltage harmonics in the lv distribution network were measured and evaluated. the measuring and evaluation were performed by analysers with a measuring uncertainty of 0.1% of the first harmonic. the compatibility level of 0.5% was not exceeded at any measurement point. the results of the evaluation of the 15th and 21st voltage harmonics in the lv distribution network are shown in fig. 4. hv (110 kv) 1 2 mv (22 kv) lv (0,4 kv) 3 i u hv (110 kv) 11 22 mv (22 kv) lv (0,4 kv) 3 i u 3 i u fig. 3. example of evaluation of selected voltage quality parameters (fifth harmonic, flicker and unbalance) in the lv networks fig. 4. the results of the measuring and evaluation of the 15th and 21st voltage harmonics in the lv distribution network iv. long-term measuring of flicker in 2014 we started the long-term monitoring of flicker [3] at the feeding points at which a problem with compatibility limit violation of flicker was detected. the measuring was done during a two-week period in spring and in autumn. in the figures below examples of the evaluation of pst and plt in the spring (fig. 5) and autumn of 2014 (fig. 6) are shown. from the present results of the flicker monitoring it is possible to state that at some feeding points the values of plt were more than 1. the greatest value was 1.7. fig. 5. the results of the measuring of flicker – spring 2014 fig. 6. the results of the measuring of flicker – autumn 2014 v. quality parameters monitoring at all times as of 2001, there are analysers qwave (manufactured by lem) fitted to distribution points of 110 kv so as it is possible to register as much information on individual parameters of voltage quality and events in the distribution system, as possible. qwave power measures, simultaneously, all voltage quality parameters and compares it with the limit values according to the csn en50160 standard, and furthermore, it also renders the current analysis. qwave light is a simplified version, evaluating only current for its all guaranteed and indicative parameters. the rules for operation of distribution networks (dn) contain annex 3 (quality of electrical power in the distribution system, manners of determination and evaluation). based on these rules, there must be quality analyser of the electrical power supply fitted at all times, as of january 1, 2006, for all hv supply terminals, and as of january 1, 2007, for all supplies from dn 110 kv. the data acquired by these analysers are being continuously processed and archived. ref. [4] on the fig. 7 you can see the illustration of the continuously monitoring power quality parameters at the selected place os8 mv distribution network. on the diagram there are also placed the results acquired by the cyclic monitoring at the same place as it was described in the previous part of the contribution. for example, you can see that the values of flicker acquired by the cyclic monitoring makes approximately one quarter of the maximal value acquired by the yearly monitoring. fig. 7. 5th harmonic and flicker (plt) in mv network os-8 vi. use of intelligent electrometers for monitoring of quality parameters characteristic parameters of voltage within low and high voltage networks are introduced in the standard čsn en 50160. the revised normative čsn en 50160 further defines parameters for very-high voltage networks. current intelligent electrometers usually provide data that is not in compliance with the normative mentioned above, but they afford relevant data for energy companies usable in operation. as for usage in operation of distribution, the most crucial issues are long-term monitoring of voltage deviations and their evaluation in compliance with standard čsn en 50160. further important values are overvoltage, falls and short-time blackouts typically with 1s sample period (thus quite not in accordance with čsn en 50160). yet this data can give a power company relevant information, because events longer than 1 second still report about the conditions of distribution network and during the changes (usually rising) they indicate the error states. unfortunately, these data are recorded as events, but the number or logged events are limited and set low. harmonics and thd, even evaluated until low frequencies only (till 10th or 25th harmonic multiple), can provide relevant information. for example, when the third harmonic element rises, it can indicate the problem of power transformer. a significant rise of any harmonic or thd indicates the problem with resonances in distribution network. vii. conclusions three pq problems are summarised in the paper: the results of long-term (15 years) monitoring of selected voltage quality parameters and evaluation of trends in changes of these parameters; the most severe changes were registered for flicker; measurement and evaluation of the 15th and 21st voltage harmonics in the lv distribution network. the compatibility level of 0.5% was not exceeded at any measurement point; from the present results of the flicker monitoring it is possible to state that at some feeding points the values of plt were more than 1. the greatest value was 1.7. 0,00 1,00 2,00 3,00 4,00 40 41 42 43 44 45 46 47 48 49 50 51 52 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 [% ] 5th harmonic, 95% values, mv, (compatibility level 6%) locality os-8 continual monitoring (qwave) cyclical monitoring (ena) 0,00 0,20 0,40 0,60 0,80 1,00 40 41 42 43 44 45 46 47 48 49 50 51 52 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 [ ] plt, 95% values, mv, (compatibility level 1) locality os-8 continual monitoring (qwave) cyclical monitoring (ena) the results from continuously monitoring of power quality parameters manifest that the measured values can differ during the year and therefore the continuously monitoring is wellfounded. current intelligent electrometers usually provide data that is not in compliance with the normative mentioned above, but they afford relevant data for energy companies usable in operation. acknowledgment this research was partially supported by the sgs grant from vsb – technical university of ostrava (no. sp2016/95 and no. sp2016/146). references [1] en 50160 ed.3:2010 “voltage characteristics of electricity supplied by public electricity networks”. [2] p. krejci, p. santarius, p. bilík, r. čumpelík “the periodical and continual pq monitoring of selected distribution networks in czech republic,” electrical power quality and utilisation 2011, lisabon, 2011, pp. 1-6. [3] m. tesarova, m. kaspirek “evaluation of long-term voltage dip monitoring in the distribution system,” international scientific symposium on electrical power engineering 2011, košice, 2013, pp.268271. [4] p. krejci, p. santarius, r. hájovský, r. velička, r. čumpelík “power quality monitoring in selected distribution networks in czech republic,” electromagnetic disturbances 2011, bialystok, 2011, p. 145148. transactions on environment and electrical engineering issn 2450-5730 vol 2, no 2 (2017) © turgay yalcin, muammer ozdemir, pawel kostyla, zbigniew leonowicz investigation of supra-harmonics through signal processing methods in smart grids turgay yalcin1, muammer ozdemir1 1ondokuz mayis university, electrical&electronic engineering faculty, samsun, turkey turgay.yalcin@omu.edu.tr , ozdemirm@omu.edu.tr pawel kostyla2, zbigniew leonowicz2 2wroclaw university of technology faculty of electrical engineering, wroclaw, poland pawel.kostyla@pwr.wroc.pl, leonowicz@ieee.org abstract— nowadays supra-harmonic distortion studies are gaining attention day by day in power quality research area. when handling communication systems especially power line carrier (plc) systems in frequency range 2150 khz, they are suitable for causing electromagnetic interference (emi) to other systems. this study shows results of analysis employing advanced method called ensemble empirical mode decomposition (eemd) to describe supra-harmonic distortion. unlike the traditional method (short time fourier transformstft), eemd gives extensive representation for supra-harmonic components index terms—ensemble empirical mode decomposition; power quality; short time fourier analysis; supra-harmonics; i. introduction a novel and significance increasing day by day hazardous risk to smart grid systems called supra ‐ harmonics or emissions in 2 khz ‐150 khz frequency band. this threat can affect capacitors, lose communication contacts with smart meters. the important single fault source operations from photovoltaic inverters (pvs). naturally fuel cells, battery chargers also wind turbines, can produce this serious threat [1, 2, 3]. supra‐ harmonics also disturb domestic appliances, semi-conductor manufacturing devices, medical equipments, security systems even transportation controls. plc produces low‐impedance for emissions in subharmonics frequency range. the highest levels are commonly by virtue of plc. as a result, power grids are worked out to transfer power at 50 hz however, line also carries 2 khz‐ 150 khz electromagnetic components [4, 5, 6, 7, 17]. in this study, national instruments pwr crio data recorder was used to acquire distortions in power systems. sampling frequency was selected 1 mhz for measurements. table i. gives the information of pv systems components properties. fig 1 shows pv system at the faculty of electrical engineering wroclaw university of technology. table i. pv system components properties monocrystalline c-si polycrystalline p-si the panel thin-film cigs efficiency 14,90% 15,50% 11,80% maximum power 190w 240w 90w maximum voltage 36,5v 30v 60,8v maximum current 5,2a 8a 1,48a dimensions 1580 × 808 × 35mm 1680 × 1040 × 35mm 1196 × 636 × 36mm weight 17,2 kg 20 kg 14,5 kg the angle of inclination to the horizontal  =400  = 400  =400 azimuth  =1350  =2250  =1350 fig 1. photovoltaic system at the faculty of electrical engineering wroclaw university of technology ii. signal processing methods a. short time fourier transform traditionally stft is applied to measured data from domestic appliances and sunny mini central pv inverter. the results from the stft are presented in a spectrogram. spectrograms used for signal processing owing to show supraharmonics. stft has drawbacks about representation magnitude and frequency bands [8]. in algorithm we used stft with hamming sliding window (5 ms) for decompositon. . respectively fig 2 and 5 shows stft spectrograms and 2 hz 120 khz frequency band for current signals of domestic appliances lcd tv and laser printer. mailto:turgay.yalcin@omu.edu.tr mailto:ozdemirm@omu.edu.tr mailto:pawel.kostyla@pwr.wroc.pl mailto:leonowicz@ieee.org fig 2. spectrogram of the current lcd tv in figure 2 is illustrated constant or continuous frequency emissions at 17. 6 khz, 53 khz and 88 khz bands. b. ensemble emprical mode decomposition (eemd) emd has been profitably performed for non-stationary signal processing. the emd could decompose the complicated signal function into a number of intrinsic mode functions (imfs)[9,10,11]. the algorithm has major drawbacks of mode mixing, end effects and etc[12,13,14,15,16]. therefore, in this work we performed eemd method for generating imfs in order to analyze supra-harmonics. we focused on pattern frequency band which is dominated in power grid. the mathematical background of eemd algorithm (fig 3): i. add noise, wn(t), to target signal s1(t). s2(t)=s1(t)+wn(t). ii. used emd [9, 10, 11, 12, 16] algorithm for decomposing the final signal s2(t). iii. continue steps (i) and (ii) till the trial numbers. when new imf combination cij(t) is succeeded, predict the ensemble mean of the last intrinsic mode function (imf). (selected nstd: 0 and number of ensemble: 1 we used eemd like emd.) the aimed output:    tn i ijj tctceemd 1 )()]([ (1) tn: trial numbers, i: iteration number and j: imf scale [13, 14, 15]. fig. 3. the representation of the eemd algorithm[16] fig 4. frequency imfs components for current of lcd tv figure 4 shows the frequency spectrum of imf 3imf 5. when we look for the spectrum it is clearly illustrated that imf 3 represents the frequency component (17.58 khz). moreover, imf 4 also shows frequency (52.73 khz) band. 0 1 2 3 4 5 6 7 8 9 x 10 4 -4 -2 0 2 4 current plot for lcd tv time samples c u r r e n t [a ] spectrogram of lcd tv time f r e q u e n c y ( h z ) 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0 2 4 6 8 10 x 10 4 -40 -20 0 20 40 60 17.67 khz 17.33 khz 53 khz 49 khz 88.04 khz 0 1 2 3 4 5 6 7 8 9 10 11 x 10 4 0 2 4 6 8 10 12 14 single-sided amplitude spectrum of imfs of current signal (lcd tv) frequency (hz) |y (f )| imf 3 imf 4 imf 5 17.58 khz 87.89 khz 52.73 khz fig 5. spectrogram of the current laser printer in figure 5 is illustrated continuous and fluctuant frequency emission at 15 khz and 44.67 khz bands. figure 6 shows the frequency spectrum of imf 3imf 5. when considering the spectrum it is comprehensively shown that imf 3 represents the frequency component (15.14 khz). furthermore, imf 4 also shows frequency (44.92 khz) band. fig 6. frequency imfs components for current of laser printer fig 7. spectrogram of the current power line with pv (sampling frequency: 1 mhz) in figure 7 is illustrated continuous and fluctuant frequency emissions at 16 khz output plc frequency and daylight emission at 32 khz pv inverter signature bands. 0 1 2 3 4 5 6 7 8 9 x 10 4 -5 0 5 current plot for laser printer time samples c u r r e n t [ a ] spectrogram of laser printer time f r e q u e n c y ( h z ) 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0 1 2 3 4 5 x 10 4 -40 -20 0 20 40 60 15 khz 44 .67 khz 39 khz 12.33 khz 0 2 4 6 8 10 12 x 10 4 0 1 2 3 4 5 6 7 8 9 single-sided amplitude spectrum of imfs of current signal (laser printer) frequency (hz) |y (f )| imf 3 imf 4 imf 5 15.14 khz 44.92 khz 74. 71 khz 0 1 2 3 4 5 6 7 8 9 x 10 4 -30 -20 -10 0 10 20 30 current of pv time samples c u r r e n t [ a ] spectrogram of pv time f r e q u e n c y ( h z ) 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0 1 2 3 4 5 x 10 4 -40 -20 0 20 40 60 80 48. 33 khz 32. 33 khz 16. 33 khz 16 khz transactions on environment and electrical engineering issn 2450-5730 vol 2, no 2 (2017) © turgay yalcin, muammer ozdemir, pawel kostyla, zbigniew leonowicz fig 8. imf components of pv current signal fig 9. frequency components of imfs figure 8 shows the imf components of pv signal. for measurements for pv, fig 9 shows frequency spectrum of imf 2imf 5. when we look for the spectrum it is clearly illustrated that imf 3 has plc frequency component (15.63 khz). moreover, imf 4 also shows pv converter frequency (32.23 khz) band. table ii. shows the relationship between signal and the imfs components. imfs are sorted from higher frequency to lower. algorithm routine stops till the end of getting monotonic function imf 9 called residual imf 7 (r=0.9998) strongly related with the signal. 0 1 2 3 4 5 6 7 8 9 10 x 10 4 0 20 40 60 80 100 120 140 single-sided amplitude spectrum of imfs current signal (pv) frequency (hz) | y ( f ) | imf2 imf3 imf4 imf5 32.23 khz 31.25 khz 63.48 khz 15.63 khz 47.85 khz table ii. correlation coefficient between signal and imfs r (correlation coefficient) signal,imf1 0,0006 signal,imf2 -0,0014 signal,imf3 -0,0005 signal,imf4 0,0087 signal,imf5 0,0049 signal,imf6 0,0079 signal,imf7 0,9998 signal,imf8 0,0511 signal,imf9 0,1524 iii. conclusions instead of traditional methods such as stft, eemd gives more accurate results, determining the pv inverter and plc frequency bands with high exactitude. it was shown that eemd method can be used for spectral analysis of supraharmonics and can be also applied for pattern recognition of supra-harmonics in smart grids with pv systems. with the help of the proposed method supraharmonic analysis and pattern detection of them easily inquire into. for future study, this results will discuss with another signal processing methods. filter design for measuring supra-harmonics in smart grids will be investigated in the light of the analysis. references [1] e.o.a. larsson, m.h.j. bollen, m.g. wahlberg, c.m. lundmark, and s.k. rönnberg, measurements of high-frequency (2–150 khz) distortion in low-voltage networks, ieee transactions on power delivery, vol.25, no.3 (july 2010), pp.1749-1757. [2] s.k. rönnberg, m.h.j. bollen, m. wahlberg, interaction between narrowband power-line communication and end-user equipment, ieee transactions on power delivery, vol.26, no.3 (july 2011), pp.2034-2039. [3] m. bollen, m. olofsson, a. larsson, s. rönnberg and m. lundmark, standards for supraharmonics (2 to 150 khz), ieee emc magazine, vol.3 (2014), quarter 1, pp.114-119. [4] s. rönnberg, m. wahlberg, m. bollen, conducted emission from a domestic customer in the frequency range 2 to 150 khz with different types of lighting, int. conf. electricity distribution (cired), frankfurt, june 2011. [5] a. gil-de-castro, s. k. rönnberg and m. h. j. bollen, "harmonic interaction between an electric vehicle and different domestic equipment," in international symposium on electromagnetic compatibility, gothenburg, 2014, pp. 991-996. [6] s. rönnberg, m. bollen, a. larsson, grid impact from pvinstallations in northern scandinavia, int. conf. electricity distribution (cired), stockholm, june 2013. [7] conducted emissions in the 2khz 150khz band supra-harmonics, pqube®applicationnotev3.0, http://www.powersensorsltd.com/download/appnotes/ [8] m. bollen, h. hooshyar, s. rönnberg, spread of high frequency current emission, int. conf. electricity distribution (cired), stockholm, june 2013. [9] k.-m. chang,“arrhythmia ecg noise reduction by ensemble empirical mode decomposition”, sensors, 10, 2010, pp. 6063 6080. [10] n.e. huang, z. shen., s.r. long, m.l. wu, h.h. shih, q. zheng, n.c. yen, c.c. tung, h.h. liu, “the empirical mode decomposition and hilbert spectrum for nonlinear and nonstationary time series analysis”, proc. roy. soc. london a, vol. 454, 1998, pp. 903–995. [11] z. wu, n.e. huang, “a study of the characteristics of white noise using the empirical mode decomposition method,” proc. roy. soc. london a, 2002. http://www.powersensorsltd.com/download/appnotes/ about the authors turgay yalcin received the b.sc. erciyes university (eru), kayseri, turkey in 2006 and m.sc. degrees in electrical engineering from ondokuz mayıs university (omu), samsun, turkey, in 2010 and he is currently pursuing the ph.d. degree in electrical engineering from ondokuz mayıs university (omu). his areas of interest are identification of power quality disturbances, signal processing methods and machine learning algorithms. muammer ozdemir received the b.sc. and m.sc. degrees in electrical engineering from black sea technical university (ktü), trabzon, turkey, in 1988 and 1991, respectively, and the ph.d. degree in electrical engineering from the university of texas at austin (ut), austin, tx, usa, in 2002.currently, he is an assistant professor with the department of electrical and electronics engineering, ondokuz mayıs university (omu), samsun, turkey. his areas of interest are power systems harmonics, power quality, and power system analysis. pawel kostyla in 1998 was awarded the title of doctor of science at the wroclaw university of science and technology. from 1999 until now he has been working as an assistant professor at the department of electrical engineering. he holds the position of laboratory manager of the theoretical electrical engineering. author and coauthor of publications from the author's interest areas such as artificial neural networks and methods of digital signal processing in automation and electrical engineering, algorithms of digital signal processing and electrical measurements, development and testing of new methods of measuring electrical parameters, quality of electricity. zbigniew leonowicz (ieee m’03– sm’12) became a member (m) of ieee in 2003 and a senior member (sm) in 2012. he received the m.sc., ph.d. and habilitate doctorate (dr sc.) degrees, all in electrical engineering, in 1997, 2001, and 2012, respectively. he has been with the department of electrical engineering, wroclaw university of technology, since 1997 where he currently holds position of associate professor. his current research interests are in the areas of power quality, control and protection of power systems, renewables, industrial ecology and applications of signal processing methods in power systems. [12] t.yalcin, o.ozgonenel, “feature vector extraction by using empirical mode decomposition from power quality disturbances”, ieee siu, fethiye, mugla, 2012. [13] o.ozgonenel, t. yalcin, i. guney, u. kurt, “a new classification for power quality events in distribution system”, electric power system research (epsr), 95, 2013, pp. 192-199. [14] z.wu, n.e. huang, “ensemble empirical mode decomposition: a noiseassisted data analysis method”, adv. adapt. data. anal., 1, 2009, pp.1–41. [15] z. wang, q. zhu, j. kiely, r. luxton, “hilbert huang transform impedance measurement data for cellular toxicity monitoring” international conference on networking, sensing and control, 2009, pp. 767-772. [16] ,t yalcin, m ozdemir, “an implementation of exploratory start for power quality disturbance pattern recognition”, transactions on environment and electrical engineering, vol 1, no 3, 2016, pp 86-93. [17] t. yalcin, m ozdemir, p. kostyla, z leonowicz, “analysis of supra‐ harmonics in smart grids”, eeeic 2017., 6-9 june 2017 , doi: 10.1109/eeeic.2017.7977812 http://ieeexplore.ieee.org/search/searchresult.jsp?searchwithin=%22authors%22:.qt.zhiyan%20wang.qt.&newsearch=true http://ieeexplore.ieee.org/search/searchresult.jsp?searchwithin=%22authors%22:.qt.quan%20zhu.qt.&newsearch=true http://ieeexplore.ieee.org/search/searchresult.jsp?searchwithin=%22authors%22:.qt.j.%20kiely.qt.&newsearch=true http://ieeexplore.ieee.org/search/searchresult.jsp?searchwithin=%22authors%22:.qt.r.%20luxton.qt.&newsearch=true https://doi.org/10.1109/eeeic.2017.7977812 i. introduction ii. sıgnal processıng methods a. short time fourier transform b. ensemble emprical mode decomposition (eemd) iii. conclusıons turgay yalcin received the b.sc. erciyes university (eru), kayseri, turkey in 2006 and m.sc. degrees in electrical engineering from ondokuz mayıs university (omu), samsun, turkey, in 2010 and he is currently pursuing the ph.d. degree in electrical eng... muammer ozdemir received the b.sc. and m.sc. degrees in electrical engineering from black sea technical university (ktü), trabzon, turkey, in 1988 and 1991, respectively, and the ph.d. degree in electrical engineering from the university of texas at a... paper title (use style: paper title) transactions on environment and electrical engineering issn 2450-5730 vol 1 , no 2 (2016) © mohammad tauquir iqbal & mohd tariq comparative analysis of llc resonant and quasi resonant converter for photo voltaic emulator mohammad tauquir iqbal*^, mohd tariq+^, * damodar valley corporation, kolkata, india, 700052 +school of electrical and electronic engineering, nanyang technological university, singapore, 639798 ^ formerly at, department of electrical engineering, indian institute of technology, kharagpur, india, 721302 tauquir.iqbal@gmail.com, tariq.iitkgp@gmail.com abstract— the paper presents a comparative analysis of the two resonant based converter topologies. the converter topologies selected are llc resonant converter and the quasi resonant converter. the modeling, analysis and control of both the converter is done and presented here for the case of the photovoltaic (pv) emulator. a pv emulator is basically a dc-dc converter having same electrical characteristics that of solar pv panel. the emulator helps to achieve real characteristics of pv system in a better way in an environment where using actual pv systems can produce inconsistent results due to variation in weather conditions. the complete system is modelled in matlab® simulink simpowersystem software package. the simulation results obtained from the matlab® simulink simpowersystem software package for both topologies under steady and dynamic conditions are analyzed and presented. an evaluation table is also presented at the end of the paper, presenting the effectiveness of each topology. keywords—buck converter, llc resonant converter, quasi resonant buck converter, photo voltaic emulator, simulink. i. introduction the pollution caused by the burning/ consumption of fossil fuels in power stations, automotive vehicles etc. has led the society and researchers to think on the environmental lines. energy obtained from naturally repetitive and persistent flows of energy occurring in the local environment is defined as renewable energy. solar energy is a good example of renewable energy as it repeats day after day [1, 2]. solar energy is being seen as one of the best renewable source of energy specialty in disconnected areas and it is becoming a popular solution to target energy problems in disconnected areas. moreover, pv panels also finds application in independent systems for the production of electricity, such as solar home systems (shs), street lighting, community facilities, etc. in isolated/ disconnected areas [3]. power obtained from solar array depends upon solar insolation, climate etc. [4]. hence, all the research and development activities required in the areas of solar energy requires a variable, stable and repeatable pv source for design and testing. hence a pv generator emulator is required and its main function is to reproduce the i–v curve of a practical pv panel. the development of the simulator was initiated/needed for the testing of pv applications such as three phase grid connected inverters or mppt charge controllers as shown in fig. 1. in fig. 1, a dc-dc converter is used as solar pv emulator. the input is taken from a dc supply/source. the output of the pv emulator is given to the three phase grid connected inverter under test. initially these tests were initially conducted on physical/real pv arrays, but many issues like changing weather etc. are associated with these types of tests [5]. there are many types of solar pv panels available in the market, and it is uneconomical to buy all of these for testing to find the right product in terms of efficiency. a pv emulator is handy here as it can give the characteristics of all panels at different temperature and varying weather conditions, thus helping in the correct selection of real pv panel suitable to the particular requirement/ weather conditions. simulation of a solar panel under various irradiances and temperatures is done using the mathematical model approach [6, 7]. in photovoltaic systems, switched power dc-dc converters are widely used to transform power between one voltages to other and also mainly used in maximum power point tracker (mppt) [8]. dc-dc converter has property of variable resistance which plays an important role to emulate solar i-v characteristics and its respective p-v curves of pv array [9-11]. a well designed solar pv emulator should have the following two features: 1) it should predict nearly same static i-v characteristics of real solar arrays and panel under various weather condition and load conditions. 2) it should be able to give satisfactory result under partial shading condition with more than one peak and step [12]. the dc/dc converter when designed properly can precisely describe the voltage-current and voltage-power characteristics of pv cell/array [13]. many authors have contributed to the small-signal modeling of lcc resonant topology. a small-signal model for lcc llc/ quasi converter as sol ar pv emulator dc dc suppl y 3 phase inverter vdc idc dc vout iout emulating r eal pv panel gri d a c vg(t) i a (t) i b (t) ic (t) fig. 1 pv emulator connection to a grid connected 3 phase inverter. mailto:tauquir.iqbal@gmail.com mailto:tariq.iitkgp@gmail.com resonant converter with lc filter has been well explored for high-frequency applications. dynamic modeling of lcc resonant topology with capacitive output filter for high-power applications has also been demonstrated by the authors in [1415]. a new topology for quasi-square-wave converters has been developed in 1988, and its detailed analysis and modeling is available in literature [16]. with all these research available, but still a comparative evaluation of the above topologies as a pv emulator is missing in the literature, and hence presented in this paper. ii. llc resonant converter based pv emulator pv emulator can be implemented using llc resonant converter (shown in fig. 2). this types of resonant converter can be operated under zero-voltage switching (zvs) for the high voltage side switch and zero-current switching (zcs) of the rectifier diodes for the low voltage side when designed properly. the output impedance of the resonant converter can be regulated from zero to infinite without serial or shunt resistor with the frequency modulation control. therefore, this type of inverter has significant higher than conventional pwm converter for this application. voltage gain of the llc resonant converter is given by equation (1). where, 𝐴 = 𝐿𝑟 𝐿𝑚 , 𝜔𝐿 = 2π𝑓𝐿 = 1 √(𝐿𝑟+𝐿𝑚)∗𝐶𝑟 and 𝑄𝐿 = 𝑅𝑖 ∙ 𝜔𝐿 ∙ 𝐶𝑟 𝜔𝐿 = 2π𝑓𝐿 = 1 √𝐿𝑟∗𝐶𝑟 iii. quasi resonant buck converter based pv emulator zero voltage switching can be obtained when capacitor is connected parallel across switch and zero current switching is obtained when an inductor is connected in series with the switch. circuit diagram of quasi resonant buck converter is shown in fig. 3. regulation of the output voltage is achieved by changing the effective duty cycle, performed by varying the switching frequency of the switch. thus, changing the effective on-time of the mosfet in a zvs design. the foundation of this conversion is simply the volt-second product equating of the input and output. it is somehow identical to that of pulse width power conversion, and mostly not like those of its electrical dual of energy transfer system, the zero current switched converters. regulation of the output voltage is accomplished by adjusting the effective duty cycle, performed by varying the conversion frequency. this changes the effective on-time in a zvs design. the foundation of this conversion is simply the volt-second product equating of the input and output. it is virtually identical to that of square wave power conversion, and vastly unlike the energy transfer system of its electrical dual, the zero current switched converters. during the zvs switch off-time, the l-c tank circuit resonates. output voltage can be regulated by varying its fig. 3 circuit diagram of quasi resonant buck converter llc resonant converter current-driven transformer with rectifier fig. 2 simulink block diagram of current mode control of llc resonant converter |𝑀𝑉 (𝜔)| = | 𝑉𝑜 𝑉𝑖𝑛 | = 1 2𝑛{√(1+𝐴)2[1−(𝜔𝐿 𝜔)⁄ 2 ]2+(1 𝑄𝐿)⁄ 2 ((𝜔 𝜔𝐿)(𝐴 1+𝐴⁄⁄ ))−(𝜔𝐿 𝜔)⁄ ) 2} (1) switching frequency. the voltage across the switch start resonating from zero to its peak, and back down again to zero. at this point the switch can again be switched on, and lossless zero voltage switching is achieved. because, the resonant tank discharges the output capacitance of the mosfet switch (coss), it do not contribute to power loss dissipated in the switch. therefore, the mosfet transition losses become zero regardless of circuit parameter i.e. operating frequency and input voltage. therefore, it helps in improving the efficiency of the resonant converter. it also helps in decreasing heat losses associated with the mosfet. this property of resonant converter makes zero voltage switching a suitable for high frequency, high voltage converter designs. moreover, the gate drive requirements also reduced significantly in a zvs design due to the lack of the gate to drain (miller) charge, which is deleted when v& i equals zero. 𝑉𝐼𝑁𝑚𝑎𝑥 = 𝑉𝐼𝑁𝑚𝑖𝑛 = 𝑉𝐼𝑁 = 30 𝑣𝑜𝑙𝑡 (2) 𝑉𝐷𝑆𝑚𝑎𝑥 = 𝑉𝐼𝑁𝑚𝑎𝑥 (1 + 𝐼𝑂𝑚𝑎𝑥 /𝐼𝑂𝑚𝑖𝑛 ) (3) from equation 2 and 3, choosing vdsmax to be 6 times more than vin & iomax=4 amp iomin will be .8 amp to achieve zero voltage switching. a resonant tank period frequency of fr=500 khz will be used then zr=vinmax/iomin = 30/.8=37.5 ohm l_r=zr/ωr=37.5/2π*50000 = 11.9 µh cr=1/zr ωr = 8.45 nf iv. simulation results and discussion simulation is done in matlab® simulink environment. the parameters used for llc resonant and quasi resonant buck converters are listed in table 1 and table 2 respectively.the output voltage is 0-21 volts, and output current is 0-4 amperes. the percentage ripple in output voltage and current is kept below 1%. switching frequency is kept to about 100 khz. resonant impedance, inductance and capacitance has been calculated for quasi resonant converter and reported in the table as 75 ohm, 23.8μh and 4.225 nf respectively. fig. 4 and 5 shows the frequency response of output voltage gain with different q factor and differnt inductor ratios respeectively. fig. 6 shows resonant current and magnetizing current for maximum power, whereas fig. 7 shows diode current for the same case. it can be seen that diode is turning off with zcs (zero current switching). hence, there will not be any spike in the secondary diode current which increases overall efficiency. fig. 8 shows output voltage at maximum power. fig. 9 shows i-v characteristics of i-v curve and their corresponding frequency is shown in fig. 10. it can be observed that points “e”, “f” and “g” are located in region 2 to obtain high efficiency in high output-power operations. it can be seen from above fig. that in order to achieve full i-v characteristics frequency have to vary from 80 khz to 175 khz. but for half insolation as shown in fig.9, in order to achieve open circuit voltage and short circuit frequency of llc converter have to vary more. it is very difficult to build high variable frequency pulse generator and it will also increase gate driver circuit loss to a significant value. therefore, efficiency of emulator will decrease. fig. 11 shows resonant current at maximum power. fig. 12 and 13 shows switch voltage v/s duty cycle of quasi buck table 1. simulation parameters of llc resonant converter simulation set up parameters rating input voltage 400 volts output voltage output current maximum power 0-21 volts 0 -4 amp 66 watts resonant frequency 100 khz percentage ripple in current percentage ripple in voltage less than 1% less than 1% table 2. simulation parameters of quasi resonant buck converter simulation set up parameters rating output voltage 0-21 volts output current switching frequency 0 -4 amp 80-120 khz percentage ripple in current percentage ripple in voltage output inductance of the converter (l) output capacitance of the converter (c) resonant impedance (zr) resonant inductance (lr) resonant capacitance (cr) output min current (iomin) less than 1% less than 1% 1 mh 1 mf 75 ohm 23.8μh 4.225 nf 0.4 a region 1 (zvs) region 2 (zvs) region 3 (zcs) rl increases fig. 4 frequency response of output voltage gain of the llc resonant converter fig. 5 frequency response of output voltage gain with different inductor ratios resonant converter. we can observe from fig. 12 that zvs of switch is achieved and maximum switch voltage is six times of input voltage as designed. if we want to achieve zero voltage for current less than 0.8 amps, we have to increase maximum switch voltage. suppose we want to achieve zero voltage for current up to 0.4 amps. then switch voltage will be 11 times of input voltage as shown in fig. 13. v. conclusions two types of dc-dc converter has been studied, analyzed and simulated in the simulation environment for photo voltaic emulator. resonant converter like llc is very efficient for photo voltaic emulator if designed for particular solar insolation, but when this converter is used for variable solar insolation then the variation in switching frequency increase. it is impractical to build a gate driver circuit of wide frequency generator and also gate drive loss become significant while operating. quasi buck also gives very good efficiency at maximum power point. but as discussed in section 4, in order to achieve full characteristics switch voltage have to increase. in order to achieve zvs at open circuit condition, peak voltage will go to infinity. there is also problem at different solar insolation because for low value of solar insolation zvs cannot be achieve on full i-v characteristics if short circuit current is less than iomin. references [1] f. robert, m. ghassemi, and a. cota. “solar energy: renewable energy and the environment”. crc press, 2009. [2] m. tariq, s. bhardwaj and m. rashid, “effective battery charging system by solar energy using c programming and microcontroller”, american journal of electrical power and energy systems, 2(2), 41-43. 2013. ilm ilr fig. 6. output current at maximum power operating in region 2 id3 id4 fig. 7. output current at maximum power operating in region 2 fig. 8. output voltage at maximum power fig. 9. the v–i curve of the llc resonant converter fig. 10. frequency response of output voltage gain fig. 11 resonant current at maximum power figure 12. switch voltage vs. duty cycle fig. 13. switch voltage vs. duty cycle [3] m. tariq and k. shamsi, “application of ret to develop educational infrastructure in uttar pradesh”. international journal of recent trend in engineering, aceee, vol. 4, pp 187-190, 2010. [4] a. tariq, m. asim, and m. tariq. "simulink based modeling, simulation and performance evaluation of an mppt for maximum power generation on resistive load." 2nd international conference on environmental science and technology ipcbee, singapore, vol.6, feb. 2011. [5] d.c. lu dylan and q. n. nguyen, “a photovoltaic panel emulator using a buck-boost dc/dc converter and a low cost micro-controller”, solar energy, elsevier, volume 86, issue 5, pp 1477-1484, may 2012. [6] s. yuliang, "a photo-voltaic array simulator", acta energiae solaris sinica, vol.18, pp.448-451, 1997. [7] l. hongliang, h. mingzhi and y. xiaojie, "investigation of photovoltaic array simulators based on different kinds of pwm rectifiers," communications, circuits and systems, icccas. international conference on , pp.737-741, 23-25 july 2009 [8] duran, e.; galán, j.; sidrach-de-cardona, m.; andujar, j.m., "a new application of the buck-boost-derived converters to obtain the i-v curve of photovoltaic modules," power electronics specialists conference, 2007. pesc 2007. ieee, vol., no., pp.413-417, 17-21 june 2007. [9] m. tariq and s. yuvarajan. modeling and analysis of self excited induction generator with electronic load controller supplying static loads. canadian journal on electrical and electronics engineering, 4(1), pp.9-13. 2013. [10] m. t. iqbal, m. tariq, and m.s.u. khan, m.s.u, “fuzzy logic control of buck converter for photo voltaic emulator”. in 2016 4th international conference on the development in the in renewable energy technology (icdret) (pp. 1-6). ieee, 2016. , january. [11] m. tariq and m. t. iqbal. power quality improvement by using multipulse ac-dc converters for dc drives: modeling, simulation and its digital implementation. journal of electrical systems and information technology, 1(3), pp.255-265, 2014. [12] k. ahmed, t. labella and j. s. lai. "high efficiency photovoltaic source simulator with fast response time for solar power conditioning systems evaluation." power electronics, ieee transactions on 29.(3) pp 12851297, 2014. [13] cirrincione, m.; di piazza, m.c.; marsala, g.; pucci, m.; vitale, g., "real time simulation of renewable sources by model-based control of dc/dc converters," industrial electronics, 2008. isie 2008. ieee international symposium on, pp.1548,1555, june 30 2008-july 2 2008. [14] e. x. yang, f. c. lee, and m. m. jovanovic, “small-signal modeling of lcc resonant converter,” in proc. ieee power electron. spec. conf., jun.1992, pp. 941–948. [15] j. f. lazar and r. martinelli, “steady-state analysis of the llc series resonant converter,” in proc. ieee appl. power electron.conf. expo.mar. 2001, vol. 2, pp. 728–735. [16] v. vorperian, "quasi-square-wave converters: topologies and analysis," power electronics, ieee transactions on , vol.3, no.2, pp.183,191, apr 1988.  transactions on environment and electrical engineering issn 2450-5730 vol 1, no 3 (2016) © ladislav novosad & zdenek hradilek  abstract—this paper focuses mainly on performance analysis for three biogas stations situated within the territory of the czech republic. this paper contains basic details of the individual biogas stations as well as description of their types. it also refers to the general description of the measurement gauge involved, with specifications of its potential use. the final part of this paper deals with the analysis of course data obtained, with special regard to voltage, current, active power and reactive power data.) index terms— biogas station; power analysis; measurement database; black out; co-generation units; distribution grid i. introduction he distribution networks (dn) have experienced the rise in volumes of electric power generated by renewable energy resources (res) over the recent years. vast majority of these resources work within the distribution network on decentralized basis and this practice brings a lot of new problems. such problems are due to the very principle employed by certain resources. the dependency on weather conditions among photovoltaic and wind power plants would be a typical example. the question is how to ensure safe and stable operation or even development of such system. that is why one needs to know characteristics of particular res, also including the biogas stations (bs). the reason above is therefore addressed by this paper with respect to operation of bs and provision of details referring to the variability of output produced over time. ii. description of measurement these requirements were accommodated by several longterm measurements conducted right at biogas stations. these bs are situated at three various locations within the territory of czech republic, serving as both resources of electric power as well as heat. the generated heat is extracted from combustion products and cooling of the co-generation units by means of heat exchangers. it is then used for heating applied in bs technologies, fermenters, adjacent farm and office buildings. nevertheless, most of the thermal energy is wasted, and conducted away to the surroundings through forced ventilation using electric fans and coolers. the utilization of thermal this research was partially supported by the sgs grant from vsb technical university of ostrava (no. sp2016/95). energy is not the primary aim of this article, and it is therefore not discussed further. the measurement at each of the three biogas stations was conducted using the automatic digital measurement devices – the network analyzer bk elcom ena 330 and monitor distribution grids mds-u(mds10). this is a complete systems for monitoring and analysis of electric power quality. the devices was employed to record effective values of phase voltage and current in one-minute increments. the voltage readings were taken right from the bus bars inside the main distribution board of bs; monitoring of current was performed using flexible coils – ampflex. the device then completed calculations of remaining values, as the active or reactive power, as well as the apparent power automatically and saved these details in its internal memory. fig. 1. demonstration of a connection measurement equipment in the bs 1 the electric power generated by all three bss is supplied to the main distribution board, which is linked with a transformer station rated for 0.4/22kv further connected to the 22kv distribution grid (dg). the measurement of electric values also involved assessment of electric power quality in compliance with csn en 50160. the characteristics subject to assessment included mainly the magnitude and drops of voltage, the total harmonic distortion of voltage, flicker and voltage unbalance. description of measurements on biogas stations ladislav novosad, zdenek hradilek t fig. 2. demonstration of a co-generation unit in the bs 1 biogas station the network analyzer bk ena 330 used in this case complies with the requirements focused on measurement equipment and the measurement procedure set forth by standards čsn en 50160, čsn en 61000-4-7, čsn en 61000-4-15 and čsn en 61000-4-30. [4]. this research was partially supported by the sgs grant from vsb technical university of ostrava (no. sp2016/95). a. biogas station at location 1 the first measurement was conducted at the bs in location 1 (bs 1). the biogas station itself is situated within the premises of a piggery farming enterprise. the input raw material for the wet fermentation process is then represented mainly by pig's slurry and maize silage. this bs provides the installed electrical capacity of 1090 kw. transformation of biogas into electric power is handled by four co-generation units. there are three identical compression ignition units delivering the output of 250 kwe and one spark ignition unit with the output of 340 kwe. (1 x kgj: agrogen type bga 222 340 kwe and 3 x cgu: schnell type es 2507 250k we).  the measurement was conducted in the following period: from 9.1.2014 to 24.2.2014 58 days in total b. biogas station at location 2 another measured bs was the one in location 2 (bs 2). this station forms a part of the collective farm nearby; this is also the point for supply of the input raw materials for fermentation. the main input raw materials for wet fermentation at this bs is grass silage, wheat silage, maize silage, pig's or cow's slurry. this bs provides the installed electrical capacity of 1186 kwe. conversion of biogas into electric power is handled by two cogeneration units ge jenbacher jgs 312 gs-b.lc with the electrical power of 549 kw and ge jenbacher jgs 312 gsb.lc with the electric power of 637 kw. [1]  the measurement was conducted in the following period: from 20.8.2014 to 19.9.2014 31 days in total c. biogas station at location 3 the last measurement was performed at location 3 (bs 3). this bs also forms an immediate part of the collective farm enterprise dealing both with livestock and crops production. the extraction of biogas is also based on wet fermentation with the raw material represented by maize silage and farmyard slurry. these are supplied from the said collective farm enterprise. electric power and heat are generated using three cogeneration units (cgus) made by schnell, fitted with scania combustion engines. the total installed electrical capacity of the bs is 750kw (2xkgj with the electrical power of 265kw and 1xkgj delivering the electrical power of 250kw).  the measurement was conducted in the following period: from 20.8.2015 to 19.9.2015 31 days in total iii. creating a database of measurement the method described above yielded large amounts of data. the initial processing of these data was done using software that comes with the measurement equipment. comprehensive waveforms of individual variables depending on time were created with this software. furthermore, the data were exported and processed using ms office excel 2007. statgraphics centurion xv.ii was subsequently used for the statistical evaluation. an example of a part of thus created database is shown in fig. 3. [3] fig. 3. sample of database of measured data iv. assessment of the data measured the next part of this paper deals with the analysis of course data obtained, with special regard to voltage, current, active power and reactive power data. figs 4 to 6 demonstrate the course of the entire active power (p) from the individual bss measured. the total three-phase active power (p) is then equal to the sum of power ratings of individual phases (p1+p2+p3). fig. 4. course of entire active power over the whole period observed at the location of bs 1 fig. 5. course of entire active power over the whole period observed at the location of bs 1 fig. 6. course of entire active power over the whole period observed at the location of bs 3 the graphs of active power and the chart depicted here also show that operation of all three bss encountered black outs or drops of power supplied to the grid. all these bss show basically the same nature of these drops of the power supplied or black outs, as the case may be. the total duration of these black outs or drops is almost negligible with respect to the total operation time, it ranges between approximately 3,5 and 73,7 hours, representing approx. 0,5 – 5,4 % of the total operation time of cgu in individual bss. as far as this period is concerned, individual stations would suffer approx. 2,7-69 min. of total power black outs (approx. 0.004 – 0.084 % of the total operation time of individual bss). the stability of cgus at individual biogas stations implies that the supply of electric power is stable and predictable. this is evident mainly during operation of bs 1 and bs 3. these two bss operated in a fairly stable manner throughout the majority of measurement period, their active power ranged within values approximating the nominal values. the only unit showing larger fluctuation of the output active power was bs 2. however, this fluctuation was mainly due to the incorrect control of bs, i.e. the failure to obey procedures to feed the fermented material. a. total black out of the active power (p) supply each total black out resulted in a step change of the active power from the regular operating level to the zero supply values. all three biogas stations were disconnected from the grid upon the black out, so no values could be measured. the resumed operation then brought gradual rise of the supplied power up to the regular operating level. the average time to restoration of regular operation after the black out of all cgus was below 10 minutes at each bs. however, total black outs of cgus are a very rare phenomenon. b. total black out of the active power (p) supply another condition occurring during the operation more frequently than total black outs was the drop of supplied power, as mentioned above. as far as the nature of their course was concerned, these drops showed identical characteristics at all three bss. these drops were characteristic for reduction of the active power of around 100-350kw. the apparent major cause of these drops was represented by shutdown of one or more cgus for either operating (insufficient volume of biogas generated) or technical (failure/outage) cause. since each of the bss mentioned above is provided with a different number and type of co-generation units, the output power available during such drops actually differs. the table no. 1 below shows the basic statistical data describing operation of all three co-generation units, i.e. the data includes basic details about the power supply to the grid. the entire measurement period showed significant fluctuations of grid voltage at all three bss. the voltage course characteristics of this fluctuation were identical for all three stations and not affected by the operation of cgus. for demonstration of the voltage course see fig. 7. the magnitude of voltage fluctuation was characterized by values from approx. 227v to 245v, i.e. with the total range of about 15.5v. the maximum and minimum deviations of voltage therefore did not exceed the limits defined for the relevant standard (± 10% un); meaning values within the range from 207v to 253v. table i basic information of cgus operation basic information of cgus operation basic information on power supply bps 1 bps 2 bps 3 total measurement period (h) 1361 693,1 1197,97 total period of decreased power supply (h) 73,7 3,5 46 total period of full power failures (h) 1,15 0,25 0.045 trouble-free operation (no black out or power failures) 1287,3 689,6 1151.97 full supply power failure 2 1 1 further findings also revealed a certain imbalance of voltage between individual phases. the difference between voltage levels at particular phases amounted up to approx. 2v. this imbalance was then projected at all three bss to a more or less extent. (fig.8) fig. 7. course of voltage at bs 3 (colors – individual phases) fig. 8. box plot of individual phase voltages v. evaluation of electric power quality as already mentioned in the introductory part, each measurement conducted on particular bs was accompanied by evaluation of electric power quality in compliance with csn en 50160. the outcomes of measurement then imply that the network parameters as the frequency, flicker and voltage unbalance complied with conditions given by the relevant standard and applicable to all three bss subject to measurement. the spectra of harmonics of higher degree and currents were almost identical for all the bss measured. figs 9 and 10 show the sample spectrum of harmonic voltage of higher levels with designated permitted limits in accordance with csn en 50160 applicable to bs 2. none of the limits were exceeded, the values approximating these limits are the fifth and seventh harmonics, which exceed a half of the limits permitted. fig. 9 shows the spectrum of harmonic current of higher levels from the first (basic) up to the fiftieth harmonic. the only features visible throughout the entire spectrum are harmonics of lower levels, i.e. the second, the third, the fourth, the fifth, the sevenths and the eleventh one. since the spectrum of higher degree harmonics show degrees above the spectrum of harmonic currents caused mainly by synchronous generators in co-generation units, these higher degrees of voltage are most likely dragged in form the superior 22 kv grid. fig. 9. spectrum of harmonic voltage of higher levels with designated permitted limits in accordance with csn en 50160 – bps2 fig. 10. spectrum of harmonic current of higher levels with designated permitted limits in accordance with csn en 50160 – bs 2 vi. evaluation of bs measurement results as far as the operation of all three bss is concerned, the measurement has proven these sources of electric power as strongly stable. as already mentioned above, these do not show any negative parameters with respect to the electric power quality either. the comparison relating to operation of individual bss has not disclosed any significant differences. these stations were almost identical with respect to the nature of both their operation and their black outs. when compared to the remaining two bss, the greatest differences were shown by bs 2. the operation of all three cgus shows drops or black outs of electric power. the apparently most frequent reason for full failures of all cgus at a particular bs was the very configuration of cgu protection features, which mostly respond to extraordinary phenomena in the grid (voltage/undervoltage). these short-term changes could not be recorded owing to the measurement sampling characteristic (1 min). however, there are more and more occasions with the drop of electric power supply to the grid during operation. the review of operating logs available for particular bss clearly shows three potential causes of this operating condition. the apparently most frequent cause is the insufficient generation of biogas during the fermentation process or even its low quality. the other one would be regular technical attitudes to maintenance and unexpected black outs due to a technical problem. the simplest method to reduce drops in electric power production would be to exclude falls caused by insufficient generation of biogas. this problem is associated with the technical procedures during fermentation but the option to store the biogas produced in particular. storage of biogas produced, i.e. keeping its stock for periods of insufficient production, relates to the volume of gas tank. the vast majority of existing bss within the territory of czech republic has been designed for uninterrupted operation, i.e. the biogas produced is consumed by cgus almost immediately. in case the volume of biogas generated is excessive or if one or more cgus have failed, the biogas cannot be stored and it must be combusted on the safety burner (flare), which seems highly uneconomical. this problem can be potentially resolved by construction of new gas tanks or even extension of storage space inside fermentation tanks. biogas stored in these facilities could be then used to compensate for production lacks or higher demand for electric power or heat. vii. conclusion the aim of this paper was to find reminiscence in the course of power supplied from individual bss or even to explain the causes of black outs with the attempt to suggest a way for their further elimination. the results obtained by measurement show the high reliability of electric power supply to the grid (the duration of black outs per bs is equal to approximately 0.0029 % of the operating hours). the supply of electric power from cogeneration units combusting biogas can be therefore generally described as very stable and also easily predictable as a renewable energy resource. the greatest impact on power supply is seen at quality, that is the lack of biogas supplied (percentage of methane ch4 in biogas), strongly dependent on the quality of primary (organic) product. this paper also serves as a source of information about stability of output from co-generation units for further research. the research will deal with cooperation among other renewable resources of electric power (solar or wind power plants) and present an opportunity for stabilization of the power supplied to 22kv grids. the last but not the least purpose of this paper is to contribute towards potential utilization of the bs for storage of electric power for its future re-use during emergency situations in the grid, for example black-outs. references [1] l. novosad, z. hradilek, “power analysis of co-generation units at biogas station”, proceedings of the 16th international scientific conference on electric power engineering (epe) 2015 dlouhé stráně, 2015, isbn 978-1-4673-6788-2 [2] l. novosad, z. hradilek, p. moldřík, “analysis of data measured in tosanovice biogas station”, elnet 2015: 12th workshop : ostrava, 24th november 2015, pp. 28-36 [3] m. dummer, m. litschmannova, “statistika i.”. statistika i. vsb – technical university of ostrava, 1997, 80-7078-496-2 [4] z. hradílek, “elektroenergetika průmyslových a distribučních zařízení,” 1.vyd. všb – tu ostrava. 2008. isbn 978-80-7225-291-6.,p.315, 2008. [5] l. novosad, z. hradilek, “analysis of energy balances three biogas stations in czech republic”, proceedings of the 16th ieee international conference on environment and electrical engineering, 2016 http://eeeic.eu/ http://eeeic.eu/ paper title (use style: paper title) transactions on environment and electrical engineering issn 2450-5730 vol 3, no 1 (2018) © jiansheng zhang, gang zhang, yaokui gao, yong hu stair-like multivariable generalized predictive control of pulverizing system in thermal power plants zhang jiansheng 1 electricity production department shenhua guoneng energy group corporation, limited beijing, china zjs3241@163.com zhang gang 2 school of control and computer engineering north china electric power university beijing, china 1562820108@qq.com gao yaokui 3 school of control and computer engineering north china electric power university beijing, china gaoyaokui05@126.com hu yong 4 school of control and computer engineering north china electric power university beijing, china ncepu_hu@yahoo.com abstract—pulverizing system is an important part in the clean and efficient utilization of coal in thermal power plant, the optimal control of the system is an important way to achieve this goal. this paper presents a stair-like multivariable generalized predictive control scheme for a pulverizing system. this scheme focuses on the problem of predictive control algorithm in practical application, especially when it incorporates feedforward control ideas. simulation results showed that the scheme are able to realize the decoupling control of the pulverizing system, avoid the problem of matrix inversion, reduce the amount of calculation, and has certain engineering application value. keywords—power plants; pulverizing system; predictive control; i. introduction “rich in coal but poor in oil and gas” is a distinctive feature of china’s present energy structure. the national potential assessment of coal resources shows that china’s total coal resources are 5.9 trillion tons, which accounts for 94% of the total primary energy resources; however, the oil and natural gas resources account for only 6%. the total energy consumption in 2016 is about 4.36 billion tons of standard coal, of which, 2.7 billion tons of coal were consumed, which accounts for 62% of the total energy consumption; in which, the coal consumed for power generation accounts for 53% [1,2] . further, coal-fired power generation capacity accounted for more than 60% of the total installed power generation capacity in china (about 14 billion kilowatts). therefore, the clean and efficient use of coal in china is crucial, especially in coal-fired power plants, which will be of great significance in alleviating the pressure on china's resources and the environment, and will ensure the sustainable development of the china’s energy system. in coal-fired power plants, the clean and efficient use of coal is affected by many factors, such as coal quality, type and dryness, distribution of primary and secondary air, burner structure, operating conditions of units, etc. these factors involve the pulverizing, air distribution, desulfurization, denitration, dust removal, and coordination system, which make it difficult to analyze them integrally. in this paper, we mainly study the optimization control of pulverizing system to improve the stability and economy of boiler combustion, thereby achieving the clean and efficient use of coal in coalfired power plants. the pulverizing system is a typical three-input, threeoutput, nonlinear, and time-varying system, and there is a serious coupling between each variable. the traditional control system generally consists of three independent single-loop, that is, the mill outlet temperature is controlled by the cold air valve, the primary air flow is controlled by the hot air valve, and the output of the pulverizing system is controlled by the coal feeder, this control method fails to achieve decoupling control of pulverizing system; the output of pulverizing system is generally controlled by the coal feeder indirectly, and its control accuracy is very poor. in addition, the mill outlet temperature is the main factor affecting the degree of dryness and ignition heat of pulverized coal, which is affected by both the raw coal moisture content, the coal feed flow, the primary air flow and the primary air temperature. among them, the raw coal moisture is an uncontrollable variable, the coal feed flow is controlled with the change of unit load, the primary air flow is controlled with the change of coal feed flow, none of the three can be used as control method for mill outlet temperature, thereby, the mill outlet temperature is essentially controlled by primary air temperature at the inlet of coal mill. the higher the primary air temperature at the inlet of the coal mill, the lower the pulverized coal moisture at the outlet of the coal mill, the this paper is supported by national natural science foundation of china (51776065); control technology on operating flexibility of the thermal power unit based on its energy management (2017) mailto:zjs3241@163.com 2 lower the latent heat of vaporization and the ignition heat required for the combustion, which is more conducive to the safe, stable and economical operation of the boiler; however, the excessive primary air temperature at the inlet of coal mill may cause spontaneous combustion of pulverized coal or even an explosion accident, which seriously affect the safety of milling equipment. therefore, it is of great importance to study the optimal control technology of pulverizing system to realize the safe, stable and economical operation of milling system, and improve the stability and economy of boiler combustion. recently, some advanced intelligent control algorithms has been applied to the design of pulverizing control system. combined the pid algorithm and the predictive control algorithm，sun [3] et al. proposes a pid-gpc predictive control algorithm, based on this algorithm, a control scheme for the pulverizing system is designed. the simulation show that the algorithm has better robustness than general feedforward decoupling pid control and gpc control, however, this algorithm fails to consider some practical engineering problems and is not conducive to engineering applications. on the basis of a t-s fuzzy model of a pulverizing system, zhang [4] et al. present a tracking control scheme for a pulverizing system, and some important performance indicators was considered to ensure the real-time performance of the control system, the simulation verified the effectiveness and real-time performance of the control system, however, this cannot achieve decoupling control of the pulverizing system. considering the effect of coal moisture on the energy balance of the coal mill, zeng [5] et al. established a dynamic model of a coal mill and designed an optimized control scheme for the coal mill, the simulation results also show the accuracy of the model and the effectiveness of the control scheme, however, the control scheme is only for the coal mill, the control variable is the inlet primary air temperature and primary air flow, which are not direct control variables of pulverizing system. through modeling and analysis, gao [6] et al. proposed an estimation signal for pulverized coal flow at the outlet of coal mill, and this signal was integrated into the design of an intelligent control scheme for the pulverizing system. finally, a control scheme for the pulverizing system based on state space prediction control was designed, the simulation indicated that the output control precision of the pulverizing system was improved and the ability to resist disturbances was enhanced, however, the control scheme does not consider matrix inversion and tracking switching issues from an engineering perspective, therefore, it is not conducive to engineering applications. in summary，the above research content are hard to apply to engineering practice, and only remain in the simulation stage. this article will directly address the engineering applications: 1) the multivariable generalized predictive control algorithm is adopted to realize the decoupling control of pulverizing system, thereby avoiding the coupling fluctuations of the controlled variables; 2) the stair-like solution idea is adopted to solve the control law of the predictive controller, thereby avoiding the inversion problem in the process of solving diophantine equation; and 3) the feedforward experience in traditional control schemes is integrated into the design of this control scheme. this paper is organized as follows. section 1 provides a brief introduction for pulverizing systems and simulation models. section 2 deduces the stair-like multivariable generalized predictive control algorithm. section 3 designs an optimized control scheme of the pulverizing system on the basis of the algorithm deduced in the former section. section 4 simulates and verifies the proposed control scheme. section 5 presents the conclusion of this paper. ii. brief introduction for pulverizing systems and simulation models a typical positive-pressure, direct-fired, pulverizing system is mainly composed of a coal feeder, a coal mill, a primary fan, a sealed fan, a separator, and a burner (see fig. 1). the raw coal is fed into the coal mill via a coal feeder, and then ground to pulverized coal; the primary air is boosted by a fan and divided into two parts, one part directly enters the cold air duct, while the other part is heated by an air preheater and then enters the hot air duct. these two parts of wind are mixed and then sent to the coal mill. the mixed primary air temperature and primary air flow are controlled by a cold air valve and a hot valve, these two valves cooperate to complete the drying and conveying tasks of the pulverized coal. in addition, the sealed fan is used to seal the coal mill to prevent the pulverized coal from leaking. 1.boiler furnace, 2.air preheater, 3.air blower, 4. coal feeder, 5. coal mill, 6. separator, 7. primary fan, 8. sealing fan, 9. burner fig.1 schematic of a medium speed mill, positivepressure, direct-fired pulverizing system the research work in this paper is based on a model of a mps positive-pressure, direct-fired, pulverizing system established in [5]. this model which is established based on the mass balance and energy balance of a coal mill, the specific form of the model is as follows: where, 𝑊𝑎𝑖𝑟 is primary air flow, kg/s; 𝜃𝑖𝑛 is primary air temperature, kg/s; 𝑀𝑐 is raw coal content in coal mill, kg; 𝑀𝑝𝑓 is coal powder content in coal mill, kg; 𝜃𝑜𝑢𝑡 is coal mill outlet temperature, °c; 𝑢𝐿 is valve opening of cold air, %; 𝑢𝐻 is valve opening of hot air, %; 𝑊𝑐 is coal feed flow, kg/s; 𝑀𝑎𝑟 is raw coal moisture, %. the input of the model are 𝑢𝐿, 𝑢𝐻, and 𝑊𝑐. the output of the model are 𝑊𝑎𝑖𝑟, 𝜃𝑜𝑢𝑡, and 𝑊𝑝𝑓. the states of the model are 𝜃𝑖𝑛, 𝑊𝑎𝑖𝑟, 𝑀𝑐, 𝑀𝑝𝑓, and 𝜃𝑜𝑢𝑡. the time-varying parameters is 𝑀𝑎𝑟. 1 2 3 7 4 5 6 8 9 3 { 𝑊𝑎𝑖𝑟̇ = −0.0971𝑊𝑎𝑖𝑟 + 0.183𝑢𝐿 +0.551𝑢𝐻 − 22.2 𝜃𝑖�̇� = −0.272𝜃𝑖𝑛 −198 + 7.98𝑢𝐿+192.0𝑢𝐻 (0.183𝑢𝐿+0.551𝑢𝐻)(8𝜃𝑖𝑛×10 −5+0.995) 𝑀𝑐̇ = −0.452𝑀𝑐 +𝑊𝑐 𝑀𝑝𝑓̇ = 0.452𝑀𝑐 − (0.00285𝜃𝑖𝑛 +0.778)𝑊𝑎𝑖𝑟 2 𝑀𝑝𝑓 × 7.95 × 10 −4 𝜃𝑜𝑢𝑡̇ = 1 4171.7 (1.1𝑀𝑐 + 0.233𝑀𝑝𝑓 + 9.42𝑊𝑎𝑖𝑟 − 2.414𝑊𝑎𝑖𝑟𝜃𝑜𝑢𝑡 + 2.151𝑊𝑎𝑖𝑟𝜃𝑖𝑛 + 𝑡𝑒𝑚𝑝𝐴+ 𝑡𝑒𝑚𝑝𝐵) , (1) 𝑡𝑒𝑚𝑝𝐴 = 2.17𝑢3(1.88𝜃𝑜𝑢𝑡+2499)(𝑀𝑎𝑟− 1.1𝑀𝑎𝑟 𝜃𝑜𝑢𝑡 0.45) ( 1.1𝑀𝑎𝑟 𝜃𝑜𝑢𝑡 0.45−100) , (2) 𝑡𝑒𝑚𝑝𝐵 = −2.17𝜃𝑜𝑢𝑡𝑊𝑐(0.01𝑀𝑎𝑟 −1.0)( 4.62𝑀𝑎𝑟 𝜃𝑜𝑢𝑡 0.45( 1.1𝑀𝑎𝑟 𝜃𝑜𝑢𝑡 0.45−100) −1.09), (3) iii. optimal control of the pulverizing system in consideration that predictive control algorithms generally perform well in strong coupling multivariable systems without being decoupled [8-10] , such an algorithm is adopted as the core of the control system design in this paper, and a stair-like solution idea was adopted to avoid matrix inversion problems. a. stair-like multivariable generalized predictive control algorithm assume that the system is based on the following discretetime carima model [8-10] : 𝑨(𝑧−1)𝒚(𝑘) = 𝑩(𝑧−1)𝒖(𝑘 − 1)+ 𝝃(𝑘)/δ (4) where 𝒚(𝑘) is the system's m-dimensional output; 𝒖(𝑘) is the system's p-dimensional input; 𝝃(𝑘) is the system's mdimensional noise vector; and: 𝑨(𝑧−1) = 1 +𝑨1𝑧 −1 +⋯+𝑨𝑛𝑎𝑧 −𝑛𝑎, 𝑩(𝑧−1) = 𝑩0 +𝑩1𝑧 −1 + ⋯+ 𝑩𝑛𝑏𝑧 −𝑛𝑏, where 𝑨𝑖 is a m × m dimension matrix, and 𝑩𝑖 is a m × m dimension matrix. assume that the objective function of the control system is as fellow: 𝑱 = ∑‖�̂�(𝑘 +𝑗|𝑘) −𝒚𝒅(𝑘 + 𝑗)‖𝑰𝑚 2 𝑁 𝑗=1 + ∑ ‖𝚫𝒖(𝑘 +𝑗 − 1)‖𝚲 2𝑁𝑢 𝑗=1 (5) where �̂�(𝑘 + 𝑗|𝑘) is a j-step prediction for 𝑦(𝑘); 𝚲 is a positive semi-definite matrix, generally take 𝚲 = 𝑑𝑖𝑎𝑔(𝜆1,⋯,𝜆𝑝) , and 𝜆𝑖 ≥ 0 ; 𝒚𝑑(𝑘 + 𝑗) is the softening sequence vector of the set value, which generated by: { 𝒚𝑑(𝑘) = 𝒚(𝑘) 𝒚𝑑(𝑘 +𝑗) = 𝜶𝒚𝑑(𝑘 + 𝑗 −1) +(𝑰𝑚 − 𝜶)𝒚𝑟(𝑘) (𝑗 = 1,⋯,𝑁) (6) where 𝜶 = 𝑑𝑖𝑎𝑔(𝛼1,⋯,𝛼𝑚) , and 0 ≤ 𝛼𝑖 < 1; 𝒚𝑟(𝑘) is an m-dimensional set value vector. introduce the following diophantine equations: 𝑰 = 𝑬𝑗∆𝑨+ 𝑧 −𝑗𝑭𝑗 𝑗 = 1,⋯,𝑁, 𝑬𝑗𝑩 = 𝑮𝑗 + 𝑧 −𝑗𝑯𝑗 𝑗 = 1,⋯,𝑁, where, 𝑬𝑗 = 𝑬 (0) +𝑬(1)𝑧−1 +⋯+ 𝑬(𝑗−1)𝑧−(𝑗−1), 𝑭𝑗 = 𝑭 (0) + 𝑭(1)𝑧−1 +⋯+ 𝑭(𝑛𝑎)𝑧−𝑛𝑎, 𝑮𝑗 = 𝑮 (0) + 𝑮(1)𝑧−1 +⋯+ 𝑮(𝑗−1)𝑧−(𝑗−1), 𝑯𝑗 = 𝑯 (0) +𝑯(1)𝑧−1 + ⋯+ 𝑯(𝑛𝑏−1)𝑧−(𝑛𝑏−1), and 𝐸(𝑖) , 𝐹(𝑖) are m-order square matrixes, 𝑮(𝑖) , 𝑯(𝑖) are p × m dimension matrixes. definition: �̂�(𝑘) = ( �̂�(𝑘 + 1|𝑘) ⋮ �̂�(𝑘 +𝑗|𝑘) ) 𝑚×𝑁 , ∆𝑼(𝑘) = ( ∆𝒖(𝑘) ⋮ ∆𝒖(𝑘 +𝑁𝑢 − 1) ) 𝑝×𝑁𝑢 resolving the diophantine equations, and then the predictive equations can be obtained as fellow: �̂�(𝑘) = 𝑮∆𝑼(𝑘)+ 𝒀0(𝑘), (7) 𝒀0(𝑘) = 𝑭𝑗(𝑧 −1)𝒚(𝑘) +𝑯𝑗(𝑧 −1)∆𝑼(𝑘 − 1), (8) where 𝑮 = [ 𝑮(0) ⋯ 𝑮(1) 𝑮(0) ⋯ ⋮ ⋮ ⋱ 𝑮(𝑁𝑢−1) 𝑮(𝑁𝑢−2) ⋯ 𝑮(0) ⋮ ⋮ ⋱ ⋮ 𝑮(𝑁−1) 𝑮(𝑁−2) ⋯ 𝑮(𝑁−𝑁𝑢)] let the increment of future control variables be : ∆𝒖(𝑘) = 𝛅, ∆𝒖(𝑘 +𝑖) = β∆𝒖(𝑘 +𝑖 −1) = β𝑖𝛅, 1 ≤ 𝑖 ≤ 𝑁𝑢 ∆𝑼(𝑘) = (∆𝒖(𝑘) ∆𝒖(𝑘 +1) ⋯ ∆𝒖(𝑘 +𝑁𝑢 −1)) 𝑇 = (𝛅 β𝛅 ⋯ β𝑁𝑢−1𝛅)𝑇 = (1 β ⋯ β𝑁𝑢−1)𝑇𝛅 4 𝑮∆𝑼(𝑘) = [ 𝑮(0) ⋯ 𝑮(1) 𝑮(0) ⋯ ⋮ ⋮ ⋱ 𝑮(𝑁𝑢−1) 𝑮(𝑁𝑢−2) ⋯ 𝑮(0) ⋮ ⋮ ⋱ ⋮ 𝑮(𝑁−1) 𝑮(𝑁−2) ⋯ 𝑮(𝑁−𝑁𝑢)] [ 1 β ⋮ β𝑁𝑢−1 ] 𝛅 = [ 𝑮(0) 𝑮(1) + β𝑮(0) ⋮ 𝑮(𝑁𝑢−1) + β𝑮(𝑁𝑢−2) +⋯+β𝑁𝑢−1𝑮(0) ⋮ 𝑮(𝑁−1) + β𝑮(𝑁−2) + ⋯+ β𝑁−𝑁𝑢𝑮(0) ] 𝛅 = �̃�𝛅 therefore, the predictive equations can be written as fellow: �̂�(𝑘) = �̃�𝛅 +𝒀0(𝑘), (9) min𝛅 𝑱 = (�̃�𝛅+𝒀0(𝑘)− 𝒀𝒅) 𝑇 (�̃�𝛅+𝒀0(𝑘)− 𝒀𝒅)+ 𝚲(1 + β2 + ⋯+β2(𝑁𝑢−1))𝛅2 , (10 ) minimize the objective function 𝜕𝑱 𝜕𝜹 = 0, and then obtain the control law as: 𝛅 = �̃�𝑻(𝒀𝒅 −𝒀0) �̃�𝑻�̃� +𝚲(1 +β2 +⋯+β2(𝑁𝑢−1))𝛅2 in the control process, only the current control amount ∆𝒖(𝑘) = ∆𝒖(𝑘 − 1)+ 𝛅 is implemented. b. overall control scheme considering that the pulverizing system is a multi-input, multi-output, and non-linear system, the inputs and outputs are strongly coupled, in order to fundamentally realize the decoupling control of the coupled system, a multivariable decoupling control scheme for milling system is designed based on multivariate predictive control algorithm; since the change of coal feed flow affects both the primary air temperature and primary air flow, the coal feed flow is used as feedforward to improve the accuracy of the prediction model. the details are as shown in fig.2. pulverizing system pid wairsp mgpc wcsp uc uh wairpv wcpv ul toutsp toutpv fig.2 overall control scheme for the pulverizing system iv. simulation and validation in order to verify the effectiveness and accuracy of the control scheme, a simulation experiment was conducted on the mill outlet temperature, primary air flow, and coal powder flow at the outlet of the mill respectively, and a 1% white noise was added to the coal supply to reproduce the internal coal disturbance. the specific verification process is as follows: (1) at 500 seconds, the set value of the pulverized coal flow rate at the mill outlet was increased from 9.67 kg/s to 11.36 kg/s, while keeping the other set values constant. as can be seen from figure 6, the opening of the cold air valve is reduced, and the opening of the hot air valve is increased, this is due to the increase in the amount of coal feed flow requires more energy to dry the raw coal (figs.3); the increased coal feed flow causes the action of cold and hot air valve, thereby resulting in a temporary deviation of primary air flow and mill outlet temperatures (figs.4); and it can be seen from fig. 5 that since the set value of mill outlet temperature is constant, the pulverized coal moisture quickly recovers after a temporary deviation. (2) at 1500 seconds, the set value of mill outlet temperature was increased from 71.98°c to 75.98°c, while keeping the other set values constant. as can be seen from figure 3-4, the opening of the cold air valve is reduced, and the opening of the hot air valve is increased, the mill outlet temperature rises and stabilizes to its set value; as the temperature of the mill outlet rises, the pulverized coal is sufficiently dried, resulting in a decrease in pulverized coal moisture and stabilizing to a new steady state value (fig. 5). (3) at 2500 seconds, the set value of primary air flow was increased from 24.6 kg/s to 28.91 kg/s, while keeping the other set values constant. as can be seen from figure 3-4, the hot and cold air flaps are opened at the same time, and the primary air flow rate increases and stabilizes to its new set value. fig.3 curve for control variables 5 fig.4 curve for controlled variables fig.5 moisture of raw coal and coal powder v. conclusion in this paper, a control scheme for the pulverizing system based on stair-like multivariable generalized predictive control algorithm is designed. this scheme focuses on the problem of predictive control algorithm in practical application, the pulverized coal at the outlet of coal mill is proposed as a new control target of the pulverizing system’s output. simulation results showed that the scheme can realize decoupling control of the pulverizing system, avoid the problem of matrix inversion, reduce the amount of calculation, and has certain engineering application value, which is of great significance for realizing the clean and efficient utilization of coal in thermal power plants. references [1] bugge j, kjær s, blum r. high-efficiency coal-fired power plants development and perspectives[j]. energy, 2006, 31(10): 1437-1445. [2] xu c, yuan y. analysis of present operating situation of large size coal fired utility boilers[j]. electric power, 2003, 1: 000. [3] sun l, sun j, miao y, et al. application research of pid-gpc algorithm in the ball mill system[j]. open automation & control systems journal, 2015, 7(1):157-166. [4] fei m, zhang j. robust fuzzy tracking control simulation of mediumspeed pulverizer[m]. systems modeling and simulation. springer japan, 2007: 78-82. [5] zeng d, hu y, gao s, et al. modelling and control of pulverizing system considering coal moisture[j]. energy, 2015, 80: 55-63. [6] gao y, zeng d, liu j, et al. optimization control of a pulverizing system on the basis of the estimation of the outlet coal powder flow of a coal mill[j]. control engineering practice, 2017, 63:69-80. [7] gao y, zeng d, liu j. modeling of a medium speed coal mill[j]. powder technology, 2017. [8] xiaotian l i, xin w, wang z, et al. a stair-like generalized predictive control algorithm based on multiple models switching[j]. ciesc journal, 2012, 63(1):193-197. [9] wu g, peng l x, sun d m. application of stair-like generalized predictive control to industrial boiler[c]// ieee international symposium on industrial electronics. ieee, 1992:218-221 vol.1. [10] qiu x, xue m, sun d, et al. the stair-like generalized predictive control for main-steam pressure of boiler in steam-power plant[c]// intelligent control and automation, 2000. proceedings of the, world congress on. ieee xplore, 2000:3165-3167 vol.5. i. introduction ii. brief introduction for pulverizing systems and simulation models iii. optimal control of the pulverizing system a. stair-like multivariable generalized predictive control algorithm b. overall control scheme iv. simulation and validation v. conclusion references paper template of wjms (use “title of paper” style) transactions on environment and electrical engineering issn 2450-5730 vol 1, no 2 (2016) © prasenjit d. wakode, mohd tariq & tapas kumar bhattacharya experimental analysis of linear induction motor under variable voltage variable frequency (vvvf) power supply prasenjit d. wakode1, mohd tariq2, tapas kumar bhattacharya3 1, 2, 3 department of electrical engineering, indian institute of technology (iit), kharagpur, india 1presently at, tool & gauge division (tgm), bharat heavy electricals limited, bhopal, india 2presently at, school of electrical and electronic engineering, nanyang technological university, singapore abstract. this paper presents the complete analysis of linear induction motor (lim) under vvvf. the complete variation of lim air gap flux under ‘blocked linor’ condition and starting force is analyzed and presented when lim is given vvvf supply. the analysis of this data is important in further understanding of the equivalent circuit parameters of lim and to study the magnetic circuit of lim. the variation of these parameters is important to know the lim response at different frequencies. the simulation and application of different control strategies such as vector control thus becomes quite easy to apply and understand motor’s response under such strategy of control. keywords: linear induction motor, variable voltage variable frequency, experimental, air gap flux density, blocked linor. 1. introduction a linear motor can be obtained by cutting a rotary motor along it’s radius from the center axis of the shaft to the external surface of the stator core and unrolling the cut motor to get a flat construction of (previously) annular stator and cylindrical rotor [1]. since the stator and rotor both have finite diameters, hence their lengths in linear version will also be finite and hence machine with mere such a construction will obviously not be of practical use. this is because as any of the part (stator or rotor) start moving it goes slowly out of the influence of the other part and stops moving after some time when it comes completely out of influence of the other part. hence it is necessary that either both the members be infinitely long-which is again not practical or at least one of the two members be very long and the other one is of some finite length [2-3]. the lims performance is affected by the conductor reactance, resistance and the construction of the secondary structure [4]. in the linear structure, the moving part which may or may not be the rotor of the induction motor is called as the ‘linor’. in general terms, the part which we excite by electrical supply to windings is called as the ‘primary’ and the other part is called as the ‘secondary’. the lim has many merits in comparison to the rotary induction motor (rim): higher ability to exert thrust on the secondary without mechanical contacts, greater acceleration or deceleration, less wear of the wheels etc. [5].the main advantage of lim is its open magnetic structure gives us access to its air gap magnetic field very easily and measurements of this field can be carried out with search coil. also the starting force measurements are easy to carry out with simple instruments such as spring balance placed horizontally and one end attached to a rigid support. recently the finite element modeling (fem) approach has gained greater importance in modeling of lim [6]. many research has been done considering attraction force and transverse edge effect of lim [7]. lim finds tremendous application in transportation. many countries has come up with transport network based on lim [8-10]. other application of induction motor beside transportation are in railway engines, self-excited induction generators for distributed generation in rural areas [11-13], etc. this paper is organized as follows: section 2 describes the experimental setup and procedure to obtain the vvvf supply and the methods used to measure the air gap flux density and starting thrust of lim under ‘blocked linor’ condition. the experimental values of flux density and starting thrust under vvvf supply are plotted in section 3. observations and conclusions from these plots are stated in section 4. 2. the experimental set-up to obtain vvvf supply: the schematic connection diagram of the supply arrangement which is used to obtain vvvf supply for lim testing is shown in figure 1. in the above figure the mechanically coupled arrangement of a dc machinesynchronous machine set up has been done. the dc machine used was of the separately excited type. the field terminals of dc machine are marked as f-ff and the armature terminal are marked as a-aa as shown. rf and ra are the internal winding resistances of the dc machine field winding and armature winding respectively. in the arrangement used for lim testing, rated constant voltage was given to field winding throughout the testing. the armature was given dc supply through a 3 phase diode bridge rectifier .the ac input to bridge rectifier was given by means of a variac (or autotransformer) so that we can have smooth control over the armature voltage. speed of the dc machine was varied by means of varying the bridge rectifier output through autotransformer. an ammeter was connected in series with the armature so that the armature current can be observed and it can be maintained within the safe limit set up by rated armature current limit of dc motor. the synchronous machine rotor dc field voltage can be varied by means of varying the variable resistance r1 connected in series with the dc rotor field. the rheostat for r1 is of appropriate current rating. the alternator’s 3 phase armature output was given to lim which is the regulated vvvf supply. the ammeter connected to one of the armature phases will give the lim phase current ‘iph’ and also the alternator armature current can be observed so as to maintain it within it’s safe limit specified by its rated value. the lim line voltage (√3 vph) is obtained by voltmeter connected in parallel between any of the two phases as shown in fig. (1). the real power input to lim is measured by means of ‘two wattmeter method’. for the power factor of cos (∅), fig. 1 schematic connection diagram to obtain vvvf supply to lim primary 3 phase power input to lim = w1+w2 = 3.vph.iph.cos (∅) (watts) (1) i. rating of different components: a) dc machine specifications: armature rating and specification: 3.75 bhp, 110v, 32.6a, 1200 rpm, ra =1 ω field ratings: 110v, 1.8a, rf = 61 ω b) synchronous machine specifications: stator armature ratings: the stator armature has two sets of 3 phase output terminals which can be used at two different frequencies viz. 50hz and 60hz .the voltage and current ratings are shown in table 1 for these two frequencies: table 1: the voltage and current ratings at two frequencies. frequency f (hz) output, (kva) line voltage (v) line current (a) speed (rpm) no of poles, p 50 3.0 130 13.3 1000 6 60 3.0 125 13.9 1200 6 the per phase armature resistance = 0.35 ω dc rotor field ratings: 110v, field dc winding resistance = 10 ω lim specifications: 3 phase, 50 hz, 1.5kw, 110v, pole pitch=9.65 cm ii. method used for obtaining vvvf supply principle: the rotating dc voltage supplied rotor’s field with speed nr rpm induces voltage in the stator armature of synchronous machine. the magnitude of this induced voltage depends on the speed of rotation nr rpm of synchronous machine and also the air gap flux produced by the dc supplied rotor i.e. the amount of current in the dc rotor winding. the frequency of the ac induced voltage depends on the rotor speed nr of alternator. control mechanism: to change the frequency of the output voltage , we increase the speed of the dc motor which acts as a prime mover for the alternator .the speed of dc motor is varied by ‘armature voltage control’ method by varying armature voltage using autotransformer. the main problem with this is that increasing the dc motor speed will increase the frequency as well as the magnitude of the alternator output voltage. so the alternator rotor dc field is needed to be reduced so as to obtain the same voltage at higher frequency. if, an alternator with p no of poles is rotating at a speed of nr rpm, then output frequency f (hz) of the armature voltage is given by, f = p.nr 120 hz (2) in the air gap, constant flux 𝜑𝜑 wb is set up by the dc field of rotor at standstill. if the rotor is rotating then, the flux 𝜑𝜑 linked with each phase is actually a time varying quantity 𝜑𝜑(t) as viewed from the stator, and is given by φ(t) =φmax .cos(ω.t) (3) where ω =2πf rad/sec. if the output armature voltage has frequency f hz , then the instantaneous induced voltage in each phase of the armature having per phase no. of turns tph and winding factor of kw is : eph.a(t) = ω.tph. φmax.cos (ω.t).kw (volts) (a phase induced voltage) (4) ephb(t) = ω.tph.φmax.cos �ω. t − 2π 3 �.kw (volts). (b phase induced voltage) (5) eph.c(t) = ω.tph.φmax.cos �ω. t + 2π 3 �.kw (volts) (c phase induced voltage) (6) as the armature windings are distributed at a spatial displacement of 120° in space w.r.t. each other, hence the induced voltages in each phase will be 120° phase shifted from one another as described by eq. (4) to (6). hence, we have to always adjust both 𝜑𝜑𝑚𝑚𝑚𝑚𝑚𝑚 (by adjusting the dc current of the alternator rotor) and frequency ‘f’ (by adjusting armature voltage of dc motor) at a time if we want to achieve same voltage at different frequencies or vice versa. iii. advantages and disadvantages of this supply arrangement: a) advantages: • this arrangement is the ideal and easiest method to obtain a constant v/f supply at different frequencies. from eq. (4) to (6), it can be observed that if the dc current of alternator dc field is kept constant i.e. if φmax is kept constant then if we neglect stator leakage reactance, stator winding resistance voltage drop of alternator and primary winding resistance and leakage reactance voltage drops in lim, then for lim, the induced voltage to frequency i.e. eph/ ω ratio is constant at different frequencies ω rad/sec. • this method is very rugged and robust method to obtain v/f as once the dc rotor current is adjusted at some value, then by just varying the autotransformer voltage, we achieve almost constant v/f across the load terminal at armature. • more robust and easy to use than the pwm inverter. b) disadvantages: • in running condition of alternator, as the load to armature increases, then the alternator speed reduces. to keep the alternator to run at same frequency irrespective of the armature terminals load variation, the alternator should be synchronized with the desired output frequency by some other alternator. this is somewhat tedious process because each time we want the different frequency then that time first of all that synchronization should be carried out and then the load is to be connected. • this method is mostly suitable for the constant loads connected to the armature of alternator. • the lim at ‘blocked linor’ condition acts as a constant load and hence this method for obtaining vvvf supply suits best for such condition. iv) testing of lim prototype by this setup: in case of testing of lim by using this setup, the measurement of starting thrust at ‘blocked rotor’ test can be easily carried out at different voltages and frequencies as load seen by armature of alternator is constant during ‘blocked linor’ condition. when lim starts moving, the load seen by the alternator armature gradually reduces, hence because of this the speed of alternator nr and hence armature supply frequency f increases and hence also the induced voltage as per eq.(3) to (6) . as the secondary length of the ss-lim prototype used is around 4 meters which is a short distance for lim, there is no appreciable increase in the supply frequency f at running condition of lim for this short distance of 4 metre. hence, for the speed measurement on available setup of lim prototype used, this method works sufficiently accurate. to obtain voltages at low frequencies from alternator, the dc motor should be run at low speeds. at such low speeds if the dc field excitation of alternator is increased, the load to dc motor increases and it can increase to such extent that the dc motor armature current becomes more than the rated value i.e. dc motor can get overloaded if we try to get more magnitude of alternator output voltage at low frequencies. so there is a limitation on the maximum voltage that we can obtain from alternator at a given frequency. for operating voltages and currents of lim at different frequencies, this setup is sufficient. 3. measurement i) measurement of air gap flux at vvvf supply under ‘blocked linor’ condition: the flux density magnitude at a particular point on lim primary is different than that of at other points at different locations because of edge effects, end effects, etc. the following fig. (3) shows the variation of maximum air gap flux bmax vs lim primary phase current at different supply frequencies. this air gap flux is measured by fixing the search coil at one specific particular point on the lim primary iron core surface in the air gap. as seen from fig (3) for the same primary phase current, the air gap flux density is reduced gradually. this is because, as we increase the supply frequency 𝜔𝜔 , the leakage reactance xl1= 𝜔𝜔l1 and hence the voltage drop across the primary leakage reactance becomes more and more prominent hence, the voltage across the magnetizing branch also goes on reducing and finally the magnetizing current is also reduced at increased frequencies as can be seen from the equivalent circuit of lim as shown in fig(2) . primary leakage reactance =iph.xl1= iph.𝜔𝜔l1 (7) we know that the magnetizing current is the main component of supply current which sets up the air gap magnetic flux. hence, at increased frequencies, for the same applied voltage magnitude, as we have lesser magnetizing current, the air gap flux reduces. ii) measurement of starting thrust by spring balance at vvvf supply: the fig. (5) shows the measurement of starting thrust (n) at ‘blocked linor’ condition by spring balance method vs primary phase current (a) at air gap = 0.7cm air gap at different supply frequency . from fig.(4), it can be observed that at same supply current ,as the supply frequency is increased, the stating thrust gets gradually reduced .this is because the reduction of air gap flux as observed from fig.(3) . fig. (2)per phase exact equivalent ckt of lim considering core losses at ‘blocked linor’ condition in phasor form, the force acted upon the lim primary is, �⃗�𝐹 = j1��⃗ × b��⃗ (n). (8) fig (3). maximum air gap flux density y component bmax (tesla) vs primary supply phase current (a) at different frequencies at ‘blocked rotor’ condition at 0.7 cm air gap fig (4). starting thrust vs primary supply phase current (a) at different frequencies at ‘blocked rotor’ condition at 0.7 cm air gap as observed from fig.(3), at different increasing frequencies, amplitude of b��⃗ is reduced at same primary current iph (i.e. at same primary linear current density amplitude j1��⃗ ) at different increasing frequencies. hence it is obvious that as per eq. (8), the starting thrust �⃗�𝐹 will decrease at constant j1��⃗ and reduced air gap flux density b��⃗ . 4. conclusion in this paper an experimental analysis of linear induction motor under variable voltage variable frequency (vvvf) power supply has been done. this analysis is required keeping in mind the tremendous application lim is having in transport nowadays. detail discussion of the experimental setup is given in the paper along with its merit and demerit. maximum air gap flux density y component bmax (tesla) vs primary supply phase current (a) at different frequencies as well as starting thrust vs primary supply phase current (a) at different frequencies at ‘blocked rotor’ condition is presented and discussed for better understanding of the subject. acknowledgements the work was done in the electrical machine laboratory of electrical engineering department, indian institute of technology (iit), kharagpur, india. the authors are thankful to the staff of the electrical machine laboratory. references 1) k.venkatratnam, “special electrical machines” universities press india pvt. ltd. (2009), 2) wei xu, jian guo zhu, yongchang zhang, zixin li, yaohua li, yi wang, youguang guo, yongjian li, “equivalent circuits for single sided linear induction motors” ,ieee transactions on industry applications, vol. 46, no. 6, ,pages.2410-2422 november/december 2010 3) wei xu; jianguo zhu; youguang guo; yi wang; yongchang zhang; longcheng tan, "equivalent circuits for single-sided linear induction motors," energy conversion congress and exposition, 2009. ecce 2009. ieee , vol., no., pp.1288,1295, 20-24 sept. 2009 4) b.-j. lee , d.-h. koo and y.-h. cho "investigation of linear induction motor according to secondary conductor structure", ieee trans. magn., vol. 45, no. 6, pp.2839 -2842 2009 5) wei xu, jianguo zhu, longcheng tan, youguang guo, shuhong wang, and yi wang, "optimal design of a linear induction motor applied in transportation," in proc. ieee int. conf. industry technology, , pp. 791-795. feb. 2009 6) s. c. ahn , j. h. lee and d. s. hyun "dynamic characteristic analysis of lim using coupled fem and control algorithm", ieee trans. magn., vol. 36, no. 4, pp.1876 -1880 2000 7) a. k. rathore and s. n. mahendra "simulation of secondary flux oriented control of linear induction motor considering attraction force and transverse edge effect", proc. int. conf. elect. eng., pp.158 -163 2003 8) r. thornton , m. t. thompson , b. m. perreault and j. fang "linear motor powered transportation", proc. ieee, vol. 97, no. 11, pp.1754 -1757 2009. 9) r. hellinger and p. mnich "linear motor-powered transportation: history, present status, and future outlook", proc. ieee, vol. 97, no. 11, pp.1892 -1900 2009. 10) wei xu, yaohua li, guangsheng sun, and jinqi ren, "performance study on high power linear induction motor in transportation," in proc. int. conf. electrical machines and systems, pp.1025-1027, oct. 2007. 11) m. tariq and s. yuvarajan. modeling and analysis of self excited induction generator with electronic load controller supplying static loads. canadian journal on electrical and electronics engineering, 4(1), pp.9-13, 2013. 12) m. tariq, s. yuvarajan, and p. wakode. digital simulation of electronic load controller with reduced thd for self-excited induction generator. iup journal of electrical and electronics engineering, 6(4), p.36, 2013. 13) m. tariq and s. yuvarajan, simulink based modeling, analysis and simulation of self excited induction generator for use in remote areas. iu-journal of electrical & electronics engineering, 13(1), pp.1623-1628. 2013. comprehensive_analysis_of_pre_charge_sequence_inautomotive_battery_packs_footer.pdf comprehensive analysis of pre-charge sequence in automotive battery systems murat kubilay ozguc, eymen ipek, kadir aras and koray erhan software&electronics, avl research&engineering, istanbul, turkey abstract—electric vehicles (ev) have brought promising technologies for future mobility solutions. as one of the key components of evs, battery systems have fundamental functions which disconnect the battery during parking and in case of failure. to provide a safe system, specialized high voltage (hv) electromechanical switches are used to perform these major functions such as switch on, switch off or pre-charging. due to these components can be easily damaged, expensive, heavy and bulky, a solution based on pure semiconductors may be desired to accomplish these operations. many studies were exhibited on ev battery systems regarding developing solid-state systems for hv switchgear. developing technology on semiconductor devices allows to make a safety concept based on only solid-state components. this study presents a comprehensive analysis of pre-charge sequences between conventional and semiconductor switchgear to be used in electric vehicle battery systems. spice simulations are presented to investigate advantages and drawbacks of these systems. index terms—electric vehicles, battery systems, li-ion, precharge, semiconductor i. introduction climate change concern has become a major driver for co2 reduction regulations, and this brings the necessity of zero-emission transportation which ensures a decrease in local air pollution and noise emissions. therefore, electric vehicles (ev) have received tremendous attention. by using evs, engineers and scientists not only provide a cleaner and quieter atmosphere but also drastically reduce operating costs compared to gas-powered vehicles. electric vehicles spend approximately 0.015 $/km, while ice vehicles spend 0.08 $/km [1]. the increasing need for electric vehicles brought high power and energy requirements. due to these requirements, battery packs of passenger vehicles rated at 400-600 v, sports and commercial vehicles have battery packs at the levels of 800-1000 v [2]. in addition to that, as an energy source, lithium-ion (li-ion) batteries are utilized in electric vehicles. usually, dc currents of approximately 300 a are needed when riding such vehicles over longer periods of time. electric parts big in geometric size that are not simple to package in a battery scheme are needed to manage heat losses when carrying such currents. this will also lead to elevated cost and weight. evs with voltages up to 800 v are currently targeted which will decrease the current levels in heavy load driving conditions. as a result, wiring and connectors are relatively small and can, therefore, overcome the issues described above. conventional battery disconnect unit (bdu) of ev batteries basically consist of contactors, fuses, sensors, battery management system and connectors. usage of semiconductors instead of contactors also brings the reduction of size and increase in efficiency [3]. automotive applications require reliable systems over the lifetime in terms of mechanical shocks, vibrations, electrical instability, temperature and humidity conditions. even tough mechanical and electro-mechanical components are proven against these, there is a chance to improve system performance. especially in automotive high voltage (hv) battery systems, electromechanical parts can present a bottleneck. for instance, mechanical relays can require certain mounting positions considering their contact working axis. over the past few years, the use of sic-based power semiconductor alternatives has shown enormous growth, relying on its revolution. the driving forces behind this growth of the industry are the following trends: saving energy, reducing the size, integrating the system and improving reliability [4]. despite semiconductors’ challenges, they can be used in automotive battery systems in lots of areas thanks to their compact design. either pre-charge relay or main relay can be replaced by semiconductor switches. this will lead to a more reliable system in terms of vibration and shock. also, this allows that the system is a thousand times faster for fault diagnosis and response. besides, it is possible to decrease weight up to 60% and the volume reduction is up to 80% [5]. studies on this subject generally present a solution which is having a semiconductor in parallel or in series to contactor [6]–[8]. on the other hand, some researches propose a method which consists a dc-dc converter in series with main relay is used for pre-charging action [9]–[11]. however, it is aimed to remove the main relay and control the power flow by using stand alone semiconductor. moreover, the rapid progress of semiconductor technologies will enable this approach to become widespread and used in practice. at the inverter input, filter capacitors exist, that generate a severe inrush current when the circuit is closed. if this current is not limited, it may damage the cells, contactors or other battery system components. the functional requirement of the high voltage pre-charge circuit is to minimize the peak current out from the power source by slowing down the dv/dt of the input voltage. in this paper, it is compared the characteristics of the conventional pre-charging with pre-charging with semiconductor switches by spice simulation results. transactions on environment and electrical engineering issn 2450-5730 vol 4, no 1 (2020) © murat kubilay ozguc, eymen ipek, kadir aras, koray erhan ii. system parameters tesla model s is chosen because it is a popular vehicle and easy to reach the system specifications. the battery pack specifications of the tesla model s 100p is selected to determine the simulation parameters in fig. 1. as shown in table i, the maximum voltage of the package is 403.2 vdc within the package configuration is 16 modules in series that every module has 6 series 86 parallel cell configuration. panasonic ncr18650b cells are used in the selected battery pack with nominal cell resistances of 18 mω and considering the contact resistances and internal resistances of the bdu components, the total equivalent resistance of the pack is calculated as 22.10 mω by using eq.(1). equivalent inductance values of the elements in the package and the cables going from the package to the inverter are accepted as 30 μh in total. the values in the table are official values besides the cell inner resistance, stray inductance and dc link capacity. these values are determined according to avl know-how. rbs = ((rdcir + 2 ∗ rc) ∗ ns np ) ∗ nm + rmb + rbdu (1) where; • rbs: battery system resistance • rdcir: dc internal resistance • rc: contact resistance of the cells • rmb: module busbar resistance • rbdu: bdu components’ resistance • nm: number of modules • ns: number of cells in series • np: number of cells in parallel by using eq.(1), battery system resistance is calculated as 22.10 mω. the busbar plates that make the connection of modules are accepted as they have resistance of 1 mω and bdu components accepted as they have 1 mω internal resistance. table i system parameters [12] parameter value vehicle model tesla model s battery capacity 100 kwh pack configuration 16 ms (6s86p) nominal cell voltage 3.6 maximum cell voltage 4.2 cell inner resistance (@3.6 v) 18 mω nominal pack voltage 345.6 v maximum pack voltage 403.2 v stray inductance 30 μh battery system resistance 22.10 mω dc link capacity (tolerance) 550 μf (m%10) voltage difference 5 v iii. conventional pre-charge when dc power source is applied to a capacitive load, the step response of the voltage input will cause the input capacitor to charge. the capacitor charging starts with inrush current fig. 1. simulation circuit with contactor. and ends with an exponential decay down to the steady state condition. the current drawn by a capacitor can be calculated by using eq.(2). ic = c ∗ dv dt (2) where; • ic: current passing through capacitor • c: dc capacitance of battery system in farads • dv : voltage change in volts • dt: time change in seconds as can be seen in eq.(2), the peak inrush current depends upon the capacitance c and the rate of change of the voltage (dv/dt). the functional requirement of the high voltage precharge circuit is to minimize the peak current out from the power source by slowing down the dv/dt of the input power voltage. upon completion of the pre-charging sequence, the pre-charge resistor is switched out of the power supply circuit and returns to a low impedance power source for normal mode. in order to find inrush current, pre-charge resistor value can be calculated with eq.(3). in this simulation, pre-charge time is decided as 120 ms and voltage difference between dc-link capacitor and battery pack is determined as 5 v. with precharge time of 120 ms, pre-charge resistance is calculated as 49.79 ω using eq.(3). however, in the simulation, it is rounded up to 50 ω. rpre = − t c ∗ ln( δu ubat.max ) (3) where; • rpre: resistance value of pre-charge resistor in ω • δu: voltage change in volts • ubat.max : maximum battery system voltage in volts in the simulation circuit, the capacity value is taken as 550 μf and start of the sequence have 10 ms delay to get a coherent output on graphs. the result of the simulation circuit shown in figure 1 can be seen in fig. 2, the capacity voltage has reached 398 v within 120 ms. at the end of 180 ms, there is no voltage difference between the capacity and the battery pack. meanwhile, the pre-charge current reaches the peak value of 8 a at the start time (i.e. 10 ms), then exponentially reduces and cuts off around 180 ms. fig. 2. simulated pre-charge waveforms. as shown in fig. 3, the maximum power dissipated on precharge resistor is 3.25 kw. consequently, in 120 ms sequence total energy loss is 44.71 ws. fig. 3. power dissipation on pre-charge resistor. iv. pre-charge with semiconductor on the other hand, the pre-charge sequence can be achieved by using semiconductor switches. a silicon-based switch allows not only removing pre-charge contactor but also resistor. at low to medium power levels which require few hundred volts of blocking capability, mosfets are ideal semiconductors to use in this application because they are capable of fast switching time against majority carrier devices, lower switching loss due to fast rise and fall times, uncomplicated gate drive and low rds(on) to increase the efficiency by decreasing the voltage drop during steady state operation. since this is an only a new and innovative method can be applied, it is assumed that only a single mosfet to use which’s parameters can be seen in table ii. however, to cover power ratings of general battery packs, semiconductors in the market need to be used in parallel. table ii selected mosfet parameters [13] parameter value unit vds 650 v rds(on), max 0.048 ω continuous drain current (tc=100c) 40 a pulsed drain current (tc=25c) 228 a power dissipation 500 w because of being automotive compatible, having low onresistance and covering voltage spikes with its high voltage rating of 650 v, infineon ipw65r048cfda mosfet model is used in these spice simulations which can be seen in fig. 4. fig. 4. simulation circuit with semiconductor. a. pre-charging with pwm method the capacitor charge can be done via pwm switching of semiconductors since a switching action will block inrush currents from the battery until the difference is acceptable. after the capacitors reached a certain voltage level, semiconductors can be turned-on fully. as seen in fig. 5, dc link voltage rises while pre-charge current stays at same level with approximately 10 a in mosfet’s on state. fig. 5. pwm method dc-link capacitance waveforms. this method can adjust the peak value of inrush current and settling time by pwm control sequence of mosfet as shown in fig. 6. however, it has a limitation due to switching losses of pwm operation and needs proper heatsink design. fig. 6. mosfet voltage waveforms. as shown in fig. 7, the maximum power dissipated on mosfet during pwm operation reaches to 5.275 kw peak. consequently, in 40 ms sequence, total energy loss is 44.60 ws. fig. 7. power dissipation on mosfet. b. pre-charging with controlling the turn–on switching of mosfet limitation of inrush current comes from the gate charge characteristic of mosfet when the device is turned on. the gate charge characteristic is originated from the equivalent capacitance of the mosfet [14]. the quicker the capacitance is charged and discharged will dictate the easier the system turns on or off. the most efficient way to obtain and control the mosfet switching mechanism is by using the gate-charge transfer curves provided on datasheet [15]. fig. 8 shows the turn–on gate–charge transfer curve from spice simulation. this curve shows how the voltage curve pattern of vgs and vds changes for limiting the drain current which is also same current that flows through dc-link capacitor. pre–threshold region and the constant current are used for rising of vgs to device threshold voltage vth at a linear slope. when vgs has reached the vth, the drain current rises to its steady state region. after that, the drain–source voltage starts its transition and the gate–charge transfer curve starts to fig. 8. mosfet gate charge transfer curve. level off. as the voltage across the drain–gate reduces even more. dc link capacity voltage as can be seen in fig. 9 reaches its expected value. gate voltage is increased and mosfet starts conduct as full on mode. since changes in drain–source voltage affects drain current, the pre-charging time collaborates with gate voltage directly. the drain voltage control capability will allow the dvds/dt to be fully controlled independently of the load condition. the ability to control dvds/dt will allow the capacitive load or resistive load to be controlled by the inrush current [7]. fig. 9. linear mode method dc-link capacitance waveforms. as shown in fig. 10, the maximum power dissipated on mosfet during turn-on switching operation reaches to 2400 kw peak. consequently, in 40 ms sequence total energy loss is 44.65 ws. oscillations which are seen in fig. 5, 6, 7, 8, 9 will occur in real system. therefore, it is necessary to use snubber circuit for absorbing oscillations during switching in experimental studies. fig. 10. power dissipation on mosfet. v. conclusion as shown in table iii, pre-charge sequence can be performed 80 ms faster by using semiconductor, while maximum power loss varies depending on which active pre-charge method is used. in contrast to the 2.34 kw power dissipation value of the pre-charge sequence with the conventional resistance, peak values reach the value of 5.28 kw in the pwm method and 2.38 kw in turn-on control method. since the energy stored in the capacitor depends on voltage applied and the capacity of capacitor, in all 3 simulations energy losses are similar. however, in addition to these results, it should be taken into consideration that semiconductor solutions will be much lighter and take up less space than mechanical relay solutions. on the other hand, semiconductor switchgears need cooling unlike mechanical relays. according to all findings in this paper, it can be clearly seen that the usage of stand-alone semiconductor switch gear empowers pre-charging action in automotive battery systems with its notable advantages compared to a conventional mechanical relay system. moreover, solid-state technology is more reliable against mechanical shocks, electrical abnormalities, and environmental conditions. besides, there is no strict requirement of mounting of solid-state circuits. semiconductors can react in microseconds during electrical failures. thanks to superiorities of semiconductors comparing electro-mechanical parts, one can say that they will surpass their current usage areas. in the near future, they will be used in many additional areas as well as automotive battery systems. references [1] siang fui tie, chee wei tan “a review of energy sources and energy management system in electric vehicles”, renewable and sustainable energy reviews, vol. 20, pp 82-102, april 2013. [2] zvei, ”voltage cases for electric mobility” p. 11-12. available: zvei.org/fileadmin/user upload/presse und medien/publikationen/ 2014/april/voltage classes for electric mobility/voltage classes for electric mobility.pdf [accessed november 20, 2019]. [3] avl trimerics gmbh, high performance 800v e-motor for automotive application, available: https://bit.ly/2k9q4ys [accessed november 20, 2019]. [4] friedrichs, p., buschkuehle, m. (2016). the future of power semiconductors : rugged and high performing silicon carbide transistors. table iii comparison between switchgears mechanical semiconductor relay switchgear parameter pre-charge pwm method turn-on resistor control method time [ms] 120 40 40 maximum power 3.24 5.28 2.38 dissipation [kw] energy loss 44.71 44.60 44.65 [ws] volume [l] >1 <1 weight [kg] >1 <1 cooling not necessary required [5] rößler, werner. ”when do we get the electronic battery switch?.” in advanced microsystems for automotive applications 2014, pp. 165177. springer, cham, 2014. [6] parrish, r., elankumaran, k., gandhi, m., nance, b., meehan, p., milburn, d., ... & brenz, a. (2011). voltec battery design and manufacturing (no. 2011-01-1360). sae technical paper. [7] mensah-brown, a. k., hashim, h. r., blakemore, b. c., & gale, a. r. (2017). u.s. patent no. 9,573,474. washington, dc: u.s. patent and trademark office. [8] burkman, w. e., & sturza, j. (2018). u.s. patent application no. 15/469,012. [9] soldati, a., imamovic, e., & concari, c. (2019). bidirectional bootstrapped gate driver for high-density sic-based automotive dc/dc converters. ieee journal of emerging and selected topics in power electronics. [10] blakemore, b. c., gale, a. r., degner, m. w., mensah-brown, a. k., & wang, c. l. (2015). u.s. patent application no. 14/250,231. [11] mensah-brown, a. k., gale, a. r., blakemore, b. c.,& wang, c. l. (2017). u.s. patent no. 9,796,288. washington, dc: u.s. patent and trademark office. [12] tesla motors. ”tesla model s. palo alto” available: http://my. teslamotors.com/de\ de/models/design, [accessed 2015]. [13] infineon, “650v coolmos cfda power transistor” ipw65r048cfda datasheet, mar. 2012. [14] lee, eun-ju, jung-hoon ahn, seung-min shin, and byoung-kuk lee. ”comparative analysis of active inrush current limiter for high-voltage dc power supply system.” in ieee vehicle power and propulsion conference 2012, pp. 1256-1260. ieee 2012. [15] motorola inc, applitaction note an1542. kubilay ozguc was born in 1996 in istanbul/turkey. he received his bsc in electrical engineering from yıldız technical university (ytu), turkey in 2018. then he started to his msc in electrical engineering at istanbul technical university (itu) in 2019 and he is a msc student in itu. he had worked on power electronics and renewable energy during his bsc. he is currently working as battery development engineer in the field of battery systems at avl r&d company, turkey. eymen ipek was born in 1995 in istanbul/turkey. he received his bsc in electrical engineering from istanbul technical university (itu), turkey in 2017. then he started to his msc in electrical engineering at istanbul technical university in 2017 and he is a msc student in itu. he had worked on power electronics, electric & hybrid electric vehicles, liion batteries and battery systems topics during his bsc. he is currently holding position of battery development engineer in the field of battery systems at avl r&d company, turkey. he has published 2 papers about battery systems for electric & hybrid electric vehicles. kadir aras was born in 1994 in antalya/turkey. he received his bsc in electrical engineering from istanbul technical university (itu), turkey in 2018. he had worked on renewable energy sources, electric & hybrid electric vehicles, li-ion batteries and battery systems topics during his bsc. he is currently holding position of battery development engineer in the field of battery systems at avl r&d company, turkey. he has published 1 paper about battery systems for electric & hybrid electric vehicles. koray erhan was born in 1987 in turkey. he received his bsc in electrical engineering from yildiz technical university, turkey in 2010. then he completed his msc in electrical engineering at istanbul technical university in 2013. finally, he got his phd at department of energy systems engineering, kocaeli university in 2018. he became a research assistant in 2010 at istanbul technical university and in 2013 at kocaeli university. he has published more than 30 papers in different subjects including electric & hybrid electric vehicles photovoltaic power generation systems, renewable energy sources, energy storage technologies, and smart grid integration and automation systems. he has been a referee in sci and other indexed journals. he is currently holding position of battery development engineer at avl r&d company, turkey. transactions on environment and electrical engineering issn 2450-5730 vol 4, no 1 (2020) © turgay gucukoglu, haluk sari, koray erhan feasibility check of electrification in istanbul metrobus line turgay gucukoglu, haluk sari and koray erhan software & electronics, avl research & engineering, istanbul, turkey abstract—today, climate change is a significant effect of population growth, especially in big cities. transport in metropolitan cities is the most important reason for climate change with the contribution of co2 pollution that threatens human health and the environment. electrified transport systems can therefore provide a suitable solution to air pollution and health problems. this study investigates feasibility of applying electric buses to metrobus line in istanbul. initially, metrobus line data regarding to number of vehicles, number of stops, route length is gathered from official istanbul transportation system to define how much energy is needed for the routes. then, it is analyzed how to re-charge proposed batteries for allocation of chargers on the stations where is applicable. keywords—electric bus, metrobus, electrification analysis i. introduction with the urbanization in the world, the population of the city increases day by day. the increase in population causes serious damage to the city's natural environment and air quality. public transportation plays a major role in cities to ensure urban sustainability in order to mitigate the effects of population growth. the decision which was initiated by major metropolises to gradually discontinue the purchase of fossil fuel buses has played an important role in zeroemission public transport systems. moreover, improvements in battery capacities and fast charging systems have enabled many bus companies to expand their production and r&d (research and development) activities in the field of electric buses [1]. the buses used for public transportation can be divided into six types. these are diesel, compressed natural gas (cng), biofuels, hybrid diesel, hydrogen, and electric bus (e-bus) [2,3]. although diesel buses are the most widely used bus type, hybrid and e-buses are alternative to conventional diesel buses [4]. furthermore, many of diesel bus manufacturers such as man, volvo, iveco, vdl, and irizar also manufacture e-buses. e-buses have been used in cities for short routes early on [5,6]. generally, the ranges of e-buses are between 30-300 km and battery capacities change between 76-340 kwh. the range of e-bus depends on some features such as battery capacity, charging methodology, cooling, driving cycle, etc. [1,6]. in this study, the case of replacement conventional buses with the e-buses is analyzed in accordance with the obtained data from a metrobus line in istanbul, turkey. in conclusion, it is determined that which line is effective to transform the buses to electrified buses instead of conventional buses. in consideration of the data obtained, it is analyzed on which lines buses can be turned into electric buses. metrobus line in istanbul is the largest compared to similar applications around the world. accordingly, electrifying such a big transportation line is a huge challenge. in addition, many studies are being carried out on the electrification of this line in turkey. the results obtained here will serve as an example for similar applications in the world. in this context, relevant additions have been made to the study. ii. e-bus there has been a significant increase in e-bus production worldwide in recent years. in the european region, e-bus production accounts for 9% of global production. the size of e-bus fleet in european region and the usa is given in fig. 1. fig. 1. the size of e-bus fleet in european region and usa [7]. currently, the e-bus production leader in the world is china. there are currently 400,000 e-buses used in china. china accounts for 90% of the electric bus production in the world and 75% of the batteries of these electric buses in the world. lfp (lithium iron phosphate (lifepo4)) batteries are used more than %97 in e-buses [8]. distribution of estimated bus market share in 2020 is given in fig. 2. in turkey, the first electric bus was used in konya in 2016 for public transportation. at the beginning of 2017, it was announced that 200 electric buses will be bought by iett (istanbul electric tram and tunnel operations) in istanbul. the first e-bus domestic production was done by bozankaya. with e-karat, turkey has gone into producing electric buses [9]. the other e-bus manufacturer except bozankaya are otokar, karsan, bmc, and temsa in turkey [10]. e-bus manufacturers in turkey is given in fig. 3. e-buses differ in themselves in many ways, such as energy source type, charging strategy, charging refueling interface, on-board energy source, drive motor, drive topology, cooling and heating. fig. 2. distribution of estimated bus market share in 2020 [10]. fig. 3. e-bus manufacturers in turkey [11]. lfp, nmc (lithium nickel manganese cobalt oxide (linimncoo2)), and lto (lithium titanite (li2tio3)) are generally preferred as battery chemistry in electric buses. in addition to battery technology, it is used as a power source in addition to the battery in fuel cells and supercapacitors. rails, low, medium and high voltage are used in addition to the built-in battery and h2 tank to provide the necessary power to the electric buses. opportunity, in motion, and depot are charging strategies used in e-bus. manual, pantograph, induction, trolleybus current collector, and battery swapping methods are used to charge the battery. the most common types of electric motors used in electric buses are permanent magnet synchronous, electrically excited synchronous, asynchronous, and switched reluctance and central motor, and wheel hub motor are used as drive topologies. body type of e-buses can be divided into 4 as 12 m single-deck, 18 m single-deck, 24 m bi-articulated and double-deck. electric air-conditioning is used as a cooling system and electric resistance heating, electric heat pump, and fuel heating are preferred heating systems in e-buses [12]. according to unece (united nations economic commission for europe) regulations, there are vehicle categories which classify a land vehicle for regulatory purposes. in that categorization, m class is defined as ‘vehicles having at least four wheels and used for the carriage of passengers (e.g., standard car with 2, 3, 4 doors).’ m class vehicles are also spread into 3 categories as m1, m2 and m3 which is directly referring to buses with the definition as ‘vehicles used for the carriage of passengers, comprising more than eight seats in addition to the driver's seat, and having a maximum mass exceeding 5 tones’ [13]. the buses may also be categorized according to different specifications. single decker, articulated, coach, midi-bus, mini-bus and double decker buses can be a categorization according to different use-cases while low floor, low entry and high floor buses can be another category according to design criteria. in this study, further specifications are evaluated regarding articulated class bus. electrification feasibility of metrobus line in istanbul/turkey is studied in the next sections. iii. proposed e-bus concept for metrobus the metrobus line where the study is carried out is in istanbul. this metrobus line operates intercontinental and main target is to reduce traffic jam between crowded centers. with the capacity to carry 950,000 passengers per day, it is the largest among its kind in the world [14]. there are 8 lines on the metrobus route with 44 stations. the first and last stops of each of these lines are different. in addition, the number of buses operating on each line varies daily. table i shows the numerical data of the metrobus line. there is a total of 495 buses moving along the line. these buses are grouped for each line. the distance covered for the shortest line is 11 km and the distance covered for the longest line is 52 km. line and route information for each line is given in table ii. route map is also seen in figure 4. when calculating the required battery capacity values, it is considered that a bus has the capacity to cover the longest line twice. the energy consumption value is 250 kwh / 100 km for the 19.725-meter bus and 239 kwh / 100 km for the 18.125-meter bus respectively. these consumption values are calculated by considering vehicle dynamics such as air drag force, rolling resistance, gear box loses, axle loses, etc. table i metrobus line general data route number number of vehicles turnaround time (one cycle) track length (km) number of stops average vehicle speed (km/h) average turnaround time (s) total range per vehicle (km) 34 92 125 30 27 34 82 423 34a 22 94 22 20 30 256 88 34as 100 162 41.5 35 31 97 493 34bz 128 154 40 39 31 72 562 34c 70 115 29 26 35 99 361 34g 15 180 52 44 35 720 260 34u 38 40 11 6 38 63 32 34z 30 52 11.5 8 31 104 700 total 495 n/a n/a n/a n/a n/a n/a table ii route for each metrobus line station numbers 34 34a 34as 34bz 34c 34g 34u 34z 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34-44 fig. 4. metrobus line route on the map (blue line) in the analyses, energy consumption value which is greater than the other is taken into consideration. one of the advantages of this is the production-based price advantage of using one type of battery. figure 5 gives percentage of energy losses due to vehicle dynamics and auxiliary loads. detailed consumption data for buses is shown in table iii. in the calculations, the average speed value of the buses is taken as 40 km/h [15]. the passenger carrying capacity of buses is a factor affecting on total weight. both buses can carry a total of 160 passengers. in the calculations, the average occupancy rate is taken as 60% using with official data (iett). fig. 5. energy losses rates the energy and power densities of the battery packs used in electric buses play important role which directly effects on vehicle range. battery design should be made by considering the weight of the buses and the constant carrying of passengers throughout the day as it requires full time operation during a day. in this type of application, battery system is requested to deliver high number of cycles in demand. that demand is provided by lfp type cell which is suitable for high cycle life-time. accordingly, the average number of cycles is assumed as 6000. the energy consumption values of buses carrying passengers on each line is given in table iv. the total desired energy value plays important role in determining the battery capacity of buses. auxiliaries 4.90% rolling resistance 36.87% air drag 32.96% other 9.19% gbx losses 2.90% braking energy 9.99% axle losses 3.20% 19.725 m bus auxiliaries 4.90% rolling resistance 35.33% air drag 34.50% other 9.19% gbx losses 2.90% braking energy 9.99% axle losses 3.20% 18.125 m bus required energy for buses to make two cycles on the longest line (52 km) is approximately 520 kwh. however, this value is the usable energy required by bus. batteries should not charge and discharge between 0% and 100% soc (state of charge) values. at that point, the concept of soc window is defined. the energy value is given in here represents usable energy. soc window value was determined as 5-92% in order to extend the life of the batteries. in this case, installed battery capacity value is calculated as 600 kwh. according to the energy values obtained in table iv, daily charge per vehicle is calculated in table v. the purpose of this calculation is to determine how many times a day the bus with 600 kwh battery capacity needs to be charged. calculations are made separately for all lines. according to obtained data, buses operating on line 34 should be charged 187 times a day. considering buses that start to operate in this line, each bus needs to be charged about 2 times a day. these values are given in table v for each line code. total desired bus/charge number for all metrobus line is 1026. this value indicates that total chargers installed along the line should charge 1026 vehicles per day. a bus with a 600-kwh battery capacity will charge between 3 and 6 hours as average. when a vehicle's charging time is assumed to be approximately 4 hours, a charger can charge 6 battery packs in 24 hours. that means 1026 vehicle batteries can be recharged for approximately 24 hours of continuous operation by approximately 171 chargers. on the metrobus line, there are five stations where buses can be parked for required charging time. referred station numbers are 1, 8, 19, 33 and 44. therefore, chargers shall be in those 5 stops. however, it is not possible to install those number of chargers due to limited space in the stations located in city centre. when the places that are suitable for charging stations are examined, it is not possible to place more than 50 chargers in total considering available bus parking lot. based on these results, it is possible to charge a total of 300-350 buses in 24 hours. this is equal to approximately 30-35% of the total number of buses. with this assumption, the lines which are suitable for the operation of electric buses are shown in green in the column of daily vehicle / charge number in table v. lines that are not applicable for electrification are marked with red. on the other hand, battery lifetime is also assessed. maximum charge-discharge cycle of applied cells is assumed as 6000. based on these values, battery lifetime of buses operating in each line is given in table v. battery lifetime which is less than 7 8 years is not effective compared to internal combustion engine in terms of tco (total cost ownership). table iii required energy values for e-buses required energy content for 2 cycle on the longest route 34g model average energy consumption required energy content unit selected energy content citaro 239kwh/100km 496.69 kwh 600 capacity 250kwh/100km 519.77 kwh 600 table iv energy consumption values for one cycle route -> 34 34a 34 as 34 bz 34 c 34 g 34 u 34 z distance (km) -> 30 22 41.5 40 29 52 11 11.5 energy consumption (kwh) citaro 71.6 52.5 99.1 95.5 69.2 124.2 26.3 27.5 energy consumption (kwh) capacity 75.0 55.0 103.7 100.0 72.5 129.9 27.5 28.7 table v number of daily charges per line line route number of daily bus trips trips per bus number of daily charges total bus/charge number battery eol (years) feasibility for e-bus required charger 34 1296 14 2.03 187 8 + 187 34a 88 4 0.42 9 39 + 9 34 as 1188 12 2.37 237 7 34 bz 1798 14 2.70 346 6 34 c 872 12 1.74 122 9 + 122 34 g 75 5 1.25 19 13 + 19 34 u 110 3 0.15 6 110 + 6 34 z 1827 61 3.37 101 5 total n/a n/a n/a 1026 n/a n/a 343 according to that assumption, green lines in the battery eol (end of life) column in table v are compatible with electrification and the red marked lines are not feasible. when both columns (total bus/charge number and battery eol) are examined together, it is seen that 5 lines (34, 34a, 34c, 34g, 34u) are suitable for electrical conversion of transportation. iv. conclusion today, demand for electric vehicles is limited by the introduction of some parameters. in this context, the cost of batteries is fundamental, especially in commercial vehicles that are constantly in use. with the decrease in battery costs in the following years, demand for electric vehicles is expected to increase. when the results obtained in this study are reviewed, operating time (distance) is very important especially in electric bus applications. electric bus is not advantageous in terms of tco value, since operating distance is above a certain value, especially as the battery life ends very quickly. as a result, obtained values in the study are considered and it is not possible to completely electrify the metrobus line. however, in order to reduce emission values, it is reasonable to convert some routes to electric buses. for future studies, the feasibility study shall be performed in order to re-charge the batteries on all buses. initially, the whole transportation system should be designed so that the operation of the electric buses can be carried out. as an example, the charging station network infrastructure should be installed and the parking spaces where buses can wait while charging should be considered. taking everything into account, the problem could be focus on optimization techniques and linear programming to determine the electric buses recharging scheduling for a transportation network in istanbul/turkey. references [1] yilmaz, c. (2018). economic evaluation of urban electric bus charge stations: case of eindhoven, the netherlands. msc thesis, istanbul technical university, graduate school of natural and applied sciences, istanbul. [2] alves, b. b., sethi, k., dodero, a. l., guerrero, a. h., puga, d., yeghyaian, e., bose, r. (2019). green your bus ride: clean buses in latin america (no. 133929). [3] “maria xylia, towards electrified public bus transport: the case of stockholm doctoral thesis, kth royal institute of technology industrial engineering and management department of energy technology energy and climate studies, stockholm, sweden, 2018.” [4] göhlich, d., kunith, a., & ly, t. (2014). technology assessment of an electric urban bus system for berlin. wit trans. built environ, 138, 137-149. [5] lajunen, a., & lipman, t. (2016). lifecycle cost assessment and carbon dioxide emissions of diesel, natural gas, hybrid electric, fuel cell hybrid and electric transit buses. energy, 106, 329-342. [6] teoh, l.e., khoo, h.l., goh, s.y., chong, l.m., (2018). scenariobased electric bus operation: a case study of putrajaya, malaysia. int. j. transp. sci. technol, 7, pp. 10-25. [7] houbbadi, a., pelissier, s., trigui, r., redondo-iglesias, e., & bouton, t. (2019, may). overview of electric buses deployment and its challenges related to the charging-the case study of transdev. [8] de-leon, s., e-bus battery market 2019. https://www.emove360.com/wp-content/uploads/2019/10/e-busbattery-market-2019.pdf, date of access: 24.12.2019 [9] mahmoud, m., garnett, r., ferguson, m., & kanaroglou, p. (2016). electric buses: a review of alternative powertrains. renewable and sustainable energy reviews, 62, 673-684. [10] https://zeeus.eu/uploads/publications/documents/zeeus-ebus-report2.pdf, date of access: 11.11.2019 [11] evtrader, https://evtrader.com/c/electric-bus-manufacturers/, date of access: 24.12.2019 [12] göhlich, d., fay, t. a., jefferies, d., lauth, e., kunith, a., & zhang, x. (2018). design of urban electric bus systems. design science, 4. [13] unece trans/wp.29/1045 special resolution no. 1 [14] https://www.iett.istanbul/tr/main/pages/tarihce/2, date of access: 22.12.2019 [15] https://www.iett.istanbul/tr/main/hatlar/, date of access: 22.12.2019 turgay gucukoglu was born in 1988 in kahramanmaras/turkey. he received his bsc in electrical engineering from yildiz technical university (ytu), turkey in 2011. then, he completed his msc in control and automation engineering at yildiz technical university in 2019.he is currently holding position of senior battery development engineer in the field of battery systems at avl r&d company, turkey. he has published 1 paper about battery systems for electric & hybrid electric vehicles. haluk sarı was born in 1988 in kırklareli/turkey. he graduated as an automotive engineer from ruse “angel kanchev” university, bulgaria in 2012. he has 6 year-experience in field of xev powertrain and battery systems. he is currently holding position of battery development engineer at avl r&d company, turkey. koray erhan was born in 1987 in turkey. he received his b.sc. in electrical engineering from yildiz technical university, turkey in 2010. then he completed his msc in electrical engineering at istanbul technical university in 2013. finally, he got his phd at department of energy systems engineering, kocaeli university in 2018. he became a research assistant in 2010 at istanbul technical university and in 2013 at kocaeli university. he has published more than 30 papers in different subjects including electric & hybrid electric vehicles photovoltaic power generation systems, renewable energy sources, energy storage technologies, and smart grid integration and automation systems. he has been a referee in sci and other indexed journals. he is currently holding position of battery development engineer at avl r&d company, turkey. https://www.emove360.com/wp-content/uploads/2019/10/e-bus-battery-market-2019.pdf https://www.emove360.com/wp-content/uploads/2019/10/e-bus-battery-market-2019.pdf https://www.emove360.com/wp-content/uploads/2019/10/e-bus-battery-market-2019.pdf https://www.emove360.com/wp-content/uploads/2019/10/e-bus-battery-market-2019.pdf https://zeeus.eu/uploads/publications/documents/zeeus-ebus-report-2.pdf https://zeeus.eu/uploads/publications/documents/zeeus-ebus-report-2.pdf https://zeeus.eu/uploads/publications/documents/zeeus-ebus-report-2.pdf https://zeeus.eu/uploads/publications/documents/zeeus-ebus-report-2.pdf https://evtrader.com/c/electric-bus-manufacturers/ https://evtrader.com/c/electric-bus-manufacturers/ https://www.iett.istanbul/tr/main/pages/tarihce/2 https://www.iett.istanbul/tr/main/pages/tarihce/2 https://www.iett.istanbul/tr/main/hatlar/ https://www.iett.istanbul/tr/main/hatlar/  transactions on environment and electrical engineering issn 2450-5730 vol 1, no 3 (2016) © l. zemite, j. gerhards, m. gorobetz, a. levchenkov  abstract— reliability analysis of distribution systems has been attracting increasing attention. a special concern pertains to the distribution networks on which most failures occurs. the optimization of distribution system of breakers and power switches is a possible strategy to improve reliability. the paper describes development procedure for modelling restoring after a fault and calculating associated reliability indices and customers’ outage costs. the developed model of the network and reliability and outage costs calculating algorithm is suitable for multi-criteria analysis of the network. proposed reliability and outage costs calculation algorithm is based on monte carlo simulation and genetic algorithm. index terms— distribution network, simulation, power supply reliability. i. introduction ith the development of economy and mankind the electric distribution networks and technical and technological solutions of the equipment connected to them are also changing resulting in the changes of its application opportunities and requirements to quality of the supplied energy. taking into account the conditions of the free market and increasing demands of the customers in an uninterrupted electric power supply the effectiveness of capital investments are expected to be determined as well as losses resulted from the supply interruptions and electric supply reliability should be calculated. the basic task of an electric supply network operator is to provide a customer with energy supply of a necessary level of reliability and quality with as low financial expenses as possible. the evaluation of interruptions risks of an electric supply system requires to know structure of the network, its load and customers data [1]. ii. methods for calculation of the losses resulted from the electric supply interruptions the losses resulted from the electric supply interruptions that have economic and social influence on the society can be divided into direct and indirect losses. direct losses are connected with undelivered electric energy. indirect losses are not connected with the interruptions themselves but with their consequences. different types of losses calculations depend on different durations of interruptions, distribution of the customers’ groups, methods of results calculations, methods of data obtaining, etc. [2, 3]. according to the calculation types the methods of losses calculation can be divided into three subgroups – analytical, simulation and methods of customers’ interview [2, 3]. after choice of methods – analytical, simulation or interview, the direct and indirect losses resulted from interruptions should be analyzed and calculated. the factors influencing the reliability can be divided into subgroups according to the customers of electric energy, undelivered energy or power, duration of interruptions, frequency as well as combining these subgroups in different ways. the customers groups are divided taking into account equal electric energy consumption and equal interruption losses (fig.1.) [4]. methodology of calculation of electric supply losses from interruptions the values of the electric supply reliability and losses resulted from the interruptions are calculated with the accounting of the following factors: structure of electric supply network, undelivered electric energy, distribution functions of the interruptions durations probabilities, the duration of electric supply system interruptions elimination for reserved, unreserved and auxiliary elements as well as the losses of the network, society and customers resulted from the interruption [5, 6, 7]. the purpose of the calculations is to consider different scenarios as well as to calculate the losses form the interruptions. for the calculation of the losses the following tasks are optimization of distribution system reliability l. zemite, j. gerhards, m. gorobetz, a. levchenkov w observance of the reliability influencing factors in methods for calculation of the losses resulted from the electric supply interruptions seperate customerssociety private customers group of customers commercial customers agricultural customers public customers industrial customers unsupplied energy unsupplied power duration of interruption intensity of interruption frequency of interruption different combinations fig. 1. factors influencing the losses from electric supply interruptions. defined:  to model the network and select the criteria for reliability and from interruptions resulted losses;  to provide an opportunity to calculate the losses resulted from the electric supply interruptions;  to develop the methods for calculations of losses resulted from the interruptions for different periods of time and models of network taking into account consumption of electric energy, loading factor, length of the line, structure of the network, number of the customers, expenses for the interruptions elimination and capital investments, etc. and analyze the obtained results. in the calculations of total losses for different scenarios for the analysis of capital investments scenario the several factors considered in the multi-criteria analysis are taken into account (fig.2.) [4, 8]. 20kv distribution network of latvia is analyzed in details and calculated. it resulted in a developed model of network the data determined below (fig.3.). the reservation of electric supply is possible along the connecting line for supply sources a2, a3 and a4. in normal regime the power switches qf1, qf5, qf3 are in on condition. in normal regime power switches qf4 and qf2 are in off condition. points 16, qf4 and qf5 are distribution places. in general case for a particular group of customers the losses resulted from the interruptions depend on the number of customers (n), on the month fm(t), on the day of a week fn(t), and time of a day fh(t). thus the average incomes of the correspondent customers c(d) (€/year) and td – duration of interruptions, h (1) [3] dhnm t,n,d tc(d)(t)f(t) f(t)fn=)ecost(  (1) for the determining of the efficiency of the planned capital investments in the network with several possible solutions that can change the model of the network, the economic calculations are required. while calculating the level of the losses from interruptions and capital investments the capital investments expenses, ageing of the elements and the expenses for the elimination of the supply system interruptions. the expenses are formed from (2), where c – losses from interruptions during the calculation period, €/year, cki – capital investments expenses, €/year, cani – interruptions elimination expenses, €/year, ceui – direct and indirect losses of the customers, €/year, n – number of new elements, m – number of customers [8] mincc=c m 1i eui n 1i ki   anc . (2) iii. descriptions of calculation of electric supply reliability and losses from interruptions a – star algorithm is a heuristic method for the way search in given graph. the algorithm detects whether there way from the starting point to the end point. there are developed algorithm modifications, which are intended for checking whether the customer is connected to a power source. in the algorithm modifications there are taken into account in additional restrictions – the reserve source searching in the case of network element interruption in accordance with network node positions [9]. if to take into account the features of the network structure according to a particular customer it is possible to obtain a more accurate duration of the interruptions and frequency for each customer. a4 t1 t2 t3 t7 t4 t5 t6 t8 t9 t10 t11 t12 t13 t14 t15 t16 t18 t17 t21 t19 t20 a1 a2 a3 l1 l3 l4 l 7 l26 l 1 6 l17 l18 l 1 9 l 2 0 l21 l22 l 2 3 l 2 4 l 2 5 1 2 44 qf1 l5 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 qf2 qf3 qf4 qf5 49 50 l2 51 l6 l 8 l 9 l 1 0 l 1 1 l 1 2 l 1 3 l 1 4 l 1 5 45 46 47 48 52 53 54 55 main line, connecting lines, branch line, a – supply source, a1, a4 – supply source of main line, a2, a3 – supply source of connecting lines, l1 – l26 – lines, qf1 – qf5 – power switch, t1–t21 – transformer substations (ts), t16 – transformer substation connected to main line, t2, t10 – transformer substation connected to connecting line, 1, 2, 3,…, 55 – breakers, qf4, qf5, 16 – places of distributing fig. 3. model of distribution system. start choose of network distribution model averagerural area urban area elimination of interruption calculations calculation losses resulting from interruptions users’ disposable incomes month factor weekday factor daytime factor number of users ts load factor ts power location of failure interruption duration hypthesis electrical energy market price costs of power switches depreciation of elements lending interest rate inflation operational costs loan period capital costs life time of elements interruption prevention costs results decision end interruption frequency costs of breakers group of users lenght of lines interruption duration distribution network level society level users level labor costs distance to failure direct losses resulting of interruptions indirect losses resulting from interruptions fig. 2. multi–criteria analysis of calculation of electric supply reliability and losses resulted from interruptions. reservation the elements of electric supply can be divided into 3 basic groups – unreserved, reserved and auxiliary. reservation the elements of electric supply can be divided into 3 basic groups – unreserved, reserved and auxiliary. unreserved elements are those which in case of interruptions customers cannot be provided with electric supply along the lines or from other sides. reserved elements are those which in the case of interruptions customers can be supplied from other sides. auxiliary elements are the elements that in case of interruptions for customers can restore electric supply for the element under consideration, as soon as interrupted element is interrupted [9]. the frequency of the elements interruptions regarding to the considered transformer substations (ts) can be found as (3), where ][xn – frequency of failures of unreserved elements, ][x ta  – frequency of failures of unreserved ts, ][xnlin – frequency of failures of unreserved lines, ][xn at  – frequency of failures of unreserved breakers, ][xn js  – frequency of failures of unreserved power switches, ][xr – frequency of failures of reserved elements, ][xr lin  – frequency of failures of reserved lines, ][xr at  – frequency of failures of reserved breakers, ][xr js  – frequency of failures of reserved power switches, ][xp – frequency of failures of auxiliary elements, – frequency of failures of auxiliary lines, – frequency of failures of auxiliary breakers, – frequency of failures of auxiliary power switches:         ].[][][][ ],[][][][ ],[][][][][ xpxpxpxp xrxrxrxr xnxnxnxxn jsatlin jsatlin jsatlinta    (3) the total frequency of failures of the elements interruptions is calculated with (4) ].[][][][ xpxrxnx   (4) the total frequency of the elements interruptions can be calculated as (5), where – number unreserved elements, – number reserved elements, – number of auxiliary elements ][][][][ xapxarxanxa  . (5) the total duration of the elements interruptions is (6), where – duration of interruption of unreserved elements, – duration of interruption of reserved elements, – duration of interruption of auxiliary elements tpxaptrxartnxanx  ][][][][ . (6) a. calculations of interrupted energy assessment rate the statistical data of several countries were compared, the calculated interrupted energy assessment rate (iear) was determined for all groups of customers; the values of maximum and minimum undelivered electric energy were calculated [10]. b. calculations of losses from interruptions by monte carlo method the electric supply reliability and the losses resulted from the electric supply interruption for the models with additional manually indicated localizations of power switches are calculated on the basis of monte carlo modelling method. the calculations take into account the analysis of the factors influencing the reliability for the urban and rural areas electric supply network models as well as for the average model with the purpose to calculate the expected duration of the interruptions without the analysis of previous power switches localizations [8, 11]. c. calculations of losses from interruptions by genetic algorithm with optimized numbers and localizations of power switches the optimization of the number and localization of the power switches is realized with the help of genetic algorithm (ga), taking into account capital expenses and the expenses of elimination of supply system interruption [8, 10, 12]. the calculation of losses from interruptions according to the number and localization of additionally placed power switches in main lines and connection lines is realized with the aim to define an optimal number and localization of power switches taking into account the capital investments for power switches [10]. d. calculations of losses from interruptions by genetic algorithm with optimized numbers and localizations of breakers the criteria of the optimization of the power switches resulted in the decision to optimize the breakers in the main lines and connecting lines, taking into account the expenses of the capital investments and interruptions elimination. the optimization of the number and localization of breakers in main and connecting lines is realized with the help of genetic algorithm for the electric supply network models of rural area, urban area and average type. capital investments and expenses for the interruptions elimination without loan, for the loan period of 10 years and for the loan period of 25 years. iv. results of calculations of electric supply losses resulted from interruptions a. a. results of interrupted energy assessment rate iear calculated for an average electric supply network model is demonstrated in fig. 4. ][xp lin  ][xp at  ][xp js  ][xa ][xan ][xar ][xap ][x tn tr tp iear calculated for an electric supply network model in rural areas is demonstrated in fig. 5. iear calculated for an urban electric supply network model is demonstrated in fig. 6. the summarized data show that the losses from the interruptions in 20 kv electric supply network in latvia are equivalent to those in finland, norway, the netherlands, usa and sweden [4, 11, 13]. b. results of calculations of losses from interruptions by monte carlo method the purpose is to calculate the parameters of reliability and losses resulted from the interruptions. application of monte carlo method gives an opportunity to obtained more accurate results using distribution of probabilities and reducing the number of assumptions. for an average electric supply model with one additional manually indicated power switch the total losses resulted from the interruptions are reduced for 14 %, for the models of rural areas – for 1.6 %, urban areas – for 19 %. for an average electric supply model with two additional manually indicated power switches the total losses resulted from the interruptions are reduced for 19 %, for the models of rural areas – for 3.2 %, urban areas – for 19 %. for an average electric supply model with three additional manually indicated power switches the total losses resulted from the interruptions are reduced for 22 %, for the models of rural areas – for 21 %, urban areas – for 22.8 % (fig. 7). the optimal localization of the power switches in the network is a significant factor for improving of electric supply reliability and, therefore, decreasing of the losses from interruptions. the absence of the optimal localization of sectioning elements in the network can cause an increasing or insignificant decreasing of the losses resulted from the interruptions as well as the manually indicated localizations of the sectioning elements do not give an opportunity to calculate the efficiency of capital investments for the reliability level increasing [8]. the reducing of the losses from interruptions are directly proportional to ts power, number of customers, consumption of electric energy and incomes of the customers. c. results of calculations of losses from interruptions by genetic algorithm with optimized numbers and localizations of power switches the optimization of the number and localization of power switches in main and connecting lines is realized with the help of genetic algorithm for the electric supply network models of rural area, urban area and average type. the total losses of the distribution network, society and customers resulted from the interruptions after the optimization of the number and localization of power switches and for different loan periods are given in fig.8. after the optimization of the number and localization of power switches for the case of no loan the losses resulted from the interruptions for the average network model are reduced for 14 %, for the model of rural area electric supply network – for 12.6 %, and for the urban area model – for 33.6 %. total losses from interruptions for the loan period of 10 years in the model of average network are reduced for 8 %, in the model of rural area network – for 7 %, and in the model of urban area – for 30 %. fig. 4. average iear. fig. 5. iear in a rural area. fig. 6. iear in a urban area. fig. 7. the total losses resulting from interruptions. total losses from interruptions for the loan period of 25 years in the model of average network are reduced for 5.8 %, in the model of rural area network – for 11 %, and in the model of urban area – for 20 %. summarizing the results we can conclude that for the models of average and rural areas electric supply networks the capital investments for the purchasing of power switches are proportionally decreased. the total losses resulted from the interruptions are significantly decreased as a result of sectioning for the model of urban area electric supply network. this is connected with higher consumption of electric energy in the urban regions and higher incomes of the customers [8]. taking into account the total losses from the interruptions the additional connection of power switches can be considered from the smart grids development and maintaining easy point of view. d. results of calculations of losses from interruptions by genetic algorithm with optimized numbers and localizations of breakers note that the optimization of the number and localization of the breakers does not give an opportunity to achieve a minimum level of the losses resulting from the interruptions. besides, the correspondence of the optimal number of the breakers to that of the power switches, the breakers cannot provide the operation of the network without the interruptions that is why the minimized expenses include the customers’ losses from the interruptions. fig.9. represents the total losses of the distribution network, customers and society for an optimal number of breakers for the loan period of 10years, 25 years and without the loan. after the optimization of the number and localization of breakers for the case of no loan the losses resulted from the interruptions for the average network model are reduced for 2 %, for the model of rural area electric supply network – for 1%, and for the urban area model – for 9%. total losses from interruptions for the loan period of 10 years in the model of average network are reduced for 6 %, in the model of rural area network – for 3 %, and in the model of urban area – for 9 %. total losses from interruptions for the loan period of 25 years in the model of average network are reduced for 8 %, in the model of rural area network – for 3 %, and in the model of urban area – for 7 % (fig. 9). taking into account the total losses from the interruptions, the additional connection of breakers can be considered from the perspective of smart grid development and easy maintenance. v. conclusions the proposed methods can be applied for the calculations of the interruption risks of the customers and compensation of the losses resulted from the interruptions. the possible future modifications of the proposed methods are applicable in the solutions of different problems related to the analysis of electric supply interruptions. the overview, analysis and systematization of the methods of distribution networks reliability calculation, optimization methods of reliability improvement, the alternative of the reliability improvement and the losses resulted from the electric supply interruptions give an opportunity to search for the solutions of capital investments on the level of distribution networks, customers and society taking into account the tendencies of development and perspective technologies. optimization of the number of power switches and localization completed with the help of genetic algorithm, in accordance with the loan term and properties of the network model, allows reducing the losses from interruptions from 5.8 % to 33.6 %. optimization of the number of breakers and localization completed with the help of genetic algorithm, in accordance with the loan term and properties of the network model, allows reducing the losses from interruptions from 1 % to 9 %. the risks, losses, planning and advantages of the electric supply network should be analyzed in accordance with economic, environment, electric supply quality, probability of interruptions, the risks of the changes in legislation, etc. estimating the solutions of the electric supply reliability improvement most of the attention should be turned to the selection of optimal number and localization of the sectioning elements that provides an immediate improvement of the electric supply reliability. taking into account the total losses from the interruptions the additional connection of power switches and /or breakers can be considered from the smart grids development and easy maintenance point of view, but not from the losses decreasing fig. 8. total losses resulted from the interruption for the case of power switches optimization. fig. 9. total losses resulted from the interruption for the case of breakers optimization. point of view. the volume of the possible losses resulted from interruptions can be variable with the increasing of electric energy consumption and/or changing of the customer properties. acknowledgement this research work has been supported by latvian council of science (project no. 673/2014). l. zemite, j. gerhards, m. gorobetz, a. levchenkov, riga technical university faculty of electrical and power engineering, riga, latvia. references [1] brown r. e. electric power distribution reliability. – marcel dekker, new york, 2002. – p. 365. [2] alevehag k. impact of dependencies in risk assessments of power distribution systems. – licentiate thesis, royal institute of technology, school of electrical engineering, electric power systems, stockholm: sweden, 2008. – p. 155. [3] helseth a. modeling reliability of supply and infrastructural dependency in energy distribution systems. – thesis for the degree of philosophy doctor, trondheim, norwegian university of science and technology, 2008. – p. 132. [4] chayakulkheeree k. economy of power system reliability. – a training workshop on power system economics and planning, asian institute of technology, 2006. – p. 68. [5] kjølle g. h., samdal k., singh b., kvitastein o. a. customer costs related to interruptions and voltage problems: methodology and results. – power systems, ieee transactions on, vol.23, no.3, aug. 2008. – pp. 1030. – 1038. [6] billinton r. reliability of power supplies. in electronics and power , vol.24, no.4, , april 1978. – pp. 307. – 310. [7] theil a., theil m. medium voltage network reliability: efficiency oriented supply restoration strategies. – 15th pscc, session 8, paper 5, liege, 2005. – pp. 1–6. [8] distribuzione c. n. optimal placement of automation devices in enel distribution network. – in electricity distribution – part 1, 2009. cired 2009. 20th international conference and exhibition, 8–11 june 2009. – pp. 1. – 4. [9] haghifam m. r. optimal allocation of tie points in radial distribution systems using a genetic algorithm.– eur. trans. elect. power, vol. 14, no. 2, 2004. – pp. 85. – 96. [10] zemīte l., gerhards j., gorobecs m., ļevčenkovs a. optimization of switch allocation in power distribution systems – international workshop on deregulated electricity market issues (demsee 2015), ungārija, budapešta, 2014.. – pp.166. – 173. [11] sljivac d., nikolovski s., kovac z. distribution network restoration using sequential monte carlo approch. – the 9th international conference on probabilistic methods applied to power systems, june 11–15, 2006. – p. 6. laila zemite is graduated phd in 2016, assistant professor of riga technical university institute of power and electrical engineering. results of research activity are published in various international scientific proceedings and journals in fields of distribution system reliability, power system development, planning and control. she is leader of various national projects and international projects. janis gerhards is phd professor, riga technical university, institute of institute of power and electrical engineering. his fields of interests are optimization theory, reliability research in free electricity market conditions, distribution system reliability, automatization, power system development, planning and control. he was leading various national and international projects, author of many patents, books and publications. mikhail gorobetz is graduated phd. in 2008, assistant professor and leading researcher of riga technical university institute of industrial electronics and electrical engineering. results of research activity are published in various international scientific proceedings and journals in fields of adaptive control, neural networks, genetic algorithms, modelling and simulation of dynamic processes. he is leader of various national projects and international projects. he is author of many study books and patented inventions. аnatoly levchenkov, phd professor, riga technical university, institute of industrial electronics and electrical engineering, institute of railway transport. he got engineer diploma in electrical engineering in 1969, phd degree in 1978. his fields of interests are optimization theory, group decision support systems, negotiation support systems, scheduling, logistics, intelligent transport systems, evolutionary algorithms for embedded systems. he was leading various national and international projects, author of many patents, books and publications. paper title (use style: paper title) transactions on environment and electrical engineering issn 2450-5730 vol 2, no 2 (2017) © maneesh kumar raminder kaur an analytical approach for transmission expansion planning with generation variations maneesh kumar raminder kaur indian institute of technology, roorkee, india p.e.c university of technology, chandigarh, india abstract—today, the transmission expansion planning for an existing power system under different constraints, is one of the major challenge for power engineers. there are several reasons; one of the reasons is rapid growth in load and inadequate capacity addition. hence, it is important and essential to implement an algorithm for the transmission expansion which works well and have a good feasibility under certain assumptions and available constraints. transmission expansion planning used in proposed paper works on an analytical algorithm which has been implemented for specific load condition with variations in generation. for that purpose, an economical factor has been measured that considers economical aspects of the line to be added for expansion. the prescribed analytical approach is also implemented for its feasibility check on a practical case study system. index terms— central electricity authority(cea), distributed generation(dg), economic index (ei), power system operator (pso), power world simulator (pws), renewable energy sources (res), transmission expansion planning (tep). i. introduction any power system can only work reliably, when there exists some favorable system condition and constraints for which it is designed to work upon. with increase in complexities, planning for a suitable and reliable transmission system becomes very difficult for a power engineer. hence, there must always be a coordination between three entities i.e. generation, transmission and distribution for enhancement of overall system stability. while working upon transmission expansion planning the main consideration emphasis on the system stability and reliability. in many developing countries the majority of power system planning comes from inefficient capacity utilization, very high line losses, unpredictable growth in electricity demand that exceeds the available capacity addition [1]. hence, optimal solutions are needed to overcome from these problems. this will include a suitable scheduling and operation of existing capacities and detecting the bottlenecks in generation, transmission, etc. in this paper the emphasis is on an optimal solution for the transmission expansion under existing load conditions to reduce expenditures for new transmission corridors [2]. the methodology used here is based on mwmile method under a deregulated power system environment which considered line length as well as cost associated with the line being added. the methodology is simple to implement also, convergence rate for the applied approach is fast compare to other similar approaches for transmission expansion planning. in tep, we try to implement cost effective approaches and other measures like series compensation, reconductoring, addition of one or more lines in the existing system and also alteration of the transmission voltage level. the responsibility of power system operators (pso) such as utilities, is to provide suitable expansion to transmission corridor so that power transfer with in the corridor become more reliable and system stability should increase [4,11]. there is a unique effect of restructuring and deregulation of the power industry on the objectives of tep. therefore, there have been an increased the uncertainties in the system and also the idea behind tep greatly influenced [5]. due to these changes, new approaches and criteria are desirable for tep. the expansion approach used in this paper has uniqueness in terms of its easy applicability and fast convergence unlike some other methods; also the solutions coming after iteration are more practical as seen from the case study. ii. transmission expansion planning with respect to generation change a. assuming a 60%t load increment in overall network fig1. ieee 14 bus system with 60% uniform load increment on pws here, the red arrows show the overloading in the system and hence, to overcome from this we have to add new alternate lines in the network. the system data after simulating the system on pws with a 60% load increment is given in c. below. a60% load increment is based on 5-7 years planning criteria of cea [3,9,10]. complete simulation has been done using pws and matlab. fig.1 shows the ieee 14 bus system which is overloaded under given load conditions. ei is obtained for each case using matlab through a program. b. assuming a solar plant of 75mw capacity at bus no. 10 we have considered a solar plant with a capacity of 75mw connected at bus 10. while working on pws we assumed constant atmospheric conditions i.e. the output of the sources is constant to 75mw irrespective to the changes in weather conditions. fig 2. distributed source added at bus no. 10 with already 60% load increment now we add a distributed source (solar farm with capacity of 75 mw) at bus 10 and check the loading of the network.fig.2 shows a capacity addition at bus 10 as per availability. below is the system data in case of 60% load increment and new generation at bus 10. here red color digits show the overloading in the network. from the system data it is observed that there are many lines in system under consideration are overloaded and hence, there is a need to remove overloading within the system. the algorithm used here are being applied on the transmission lines only that means calculation of total number of alternate lines added or removed are only transmission lines of the system and not the transformers. the simulation results show the different columns like bus numbers between which the lines are connected, number of circuits, status of the circuit i.e. whether they are either closed or open, connected transformers between bus no. 4-7, 4-9 and 5-6, the mw flow, mvar flow, mva flow also mva limit between different line are also obtained from the simulation. losses occurred from power flow between different lines can also be observed from the simulation. the alternate lines needed for expansion are calculated in table 1 and will be described in later paragraphs. c. simulation results with 60% load increment and distributed source at bus 10 table 1. additional lines needed with 60% load increment and distribution generation at bus 10. from bus to bus % loading total lines alternate lines 1 2 186.1 3 3-1=2 1 5 164.6 3 3-1=2 2 3 246.6 4 4-1=3 2 4 110.5 2 2-1=1 2 5 114.3 2 2-1=1 4 5 142 2 2-1=1 5 6 159.8 2 3-1=1 (transformer) 6 13 122.3 2 2-1=1 9 10 127.4 2 2-1=1 10 11 199.6 3 3-1=2 table 1 shows the number of alternate lines added in the system based on the criteria mentioned below. here, 80% loading is considered while adding alternate line into the system. 20% margin is taken for line safety and simplicity [3]. therefore, each line is having a maximum loading of 80%, if it is more than that existing percentage loading divided by 80% loading gives the new alternate lines. fig.3 is obtained after adding alternating lines based on 80% loading condition as mentioned above. approaches used for tep should also consider different uncertainties present in the system either from load side or from generation side i.e. future generation location, intermittency of res production, costs of fuel, growth in load etc.) [6-7] fig3. network with new alternate lines iii. optimization of the above network using optimal algorithm fig. 3.1 flowchart for analytic algorithm for tep fig.3.1 shows the algorithm used to obtain the optimization model of the network under consideration using economical factor known as economic index (ei) described in coming section a, below. there are certain steps used under applied analytical algorithm and are described as follows: step 1: run the power world simulator to check that no route is overloaded in the network system data. step 2: add another alternate line on this route in case any route is overloaded keeping the line loading limit to 80% of its capacity. step 3: number of alternate lines to be added = % loading of overloaded line between the two buses divided by 80% loading. step 4: build the suppositional/ redundant network by adding all the alternate lines. step 5: calculate the economic index of all the transmission lines. step 6: tabulate the alternate lines in the increasing order of their economic index. step 7: remove the alternate line with minimum economic index from the redundant network. step 8: reconnect the line which was removed in step 5 if there is overloading in any line in the network. step 9: remove the alternate line with next higher minimum economic index in the order. step 10: continue the removal of alternate lines with minimum economic index fulfilling our criteria for no overloading of any line met in the process. step 11: repeat the algorithm until removal of all the added alternate lines has been tested. at last we will be left with the alternate lines which reduce the percentage overloading of overloaded lines and maximizes the objective function i.e. the overall increase in the economic efficiency of system expressed as a ratio of sigma of economic index of all transmission lines and the total number of lines in the system at each step of removal of alternate line in the transmission network expansion planning algorithm. the analytical algorithm used in paper for finding out the optimal solution, is basically a set of instruction or a set of some specific steps which are used to find out overall task of optimization. a. economic index (ei) economic efficient lines can be obtained from a novel introduced factor which is acknowledged as economic index (ei) and is defined as [2]: ei = ii / ci = [e(pi)* di* ρ] / ci where, ii: profit prospect of alternative line i ci: cost of alternative line i pi: actual power flow di: length of line i ρ: pre-determined unit cost factor of each transmission line if ei>ej, this implies that the line i is more economically efficient than line j. higher is the economic index more efficient will be the line. iteration 1. without removal of any alternating line. fig.4 below shows the chart for ei of network when there is no exclusion of any alternate line. hence, summation of all ei divided by existing number of line gives the average ei of the network. fig4. economic index with no exclusion of any alternate line average (ei). for the existing number of lines is 1389.6/31=43.42 above figure shows the lines with minimum and maximum (ei) in ascending order i.e. minimum and maximum efficient lines. hence, the choice of removal of lines is based on this order. the least efficient line is preferred to remove first while maintaining system stability considerations i.e. 80%loading condition should also be maintained. table 2. route that can be removed from system while maintaing 80% loading condition route ei status 10-11 14 not removed 9-10 31 not removed 6-13 41.6 not removed 2-5 43.2 removed 4-5 44.8 removed 2-4 53.6 not removed 2-3 54 removed 1-5 70.4 not removed 1-2 135.8 not removed table 2 shows the routes that can be excluded based on the 80% loading condition. it can be seen that route number 2-5, 4-5 and 2-3 can be removed from the network for the stability and loading conditions along with obtained ei. iteration 2. exclusion of route 2-5 from the network the average ei for existing number of lines from below figure after the exclusion of less efficient line 2-5 will be= 1425.4/30=47.51. fig.5 shows the chart for ei of network with exclusion of route 2-5. fig5. economic index after exclusion of less efficient line 2-5 iteration 3: exclusion of route 4-5 the average ei of the system for existing number of lines from below figure after the exclusion of less efficient lines 2-5, 4-5 will be = 1419.4/29= 48.94. fig.6 shows the chart for ei of network with exclusion of route 4-5 from the network. fig6. economic index after exclusion of inefficient line 4-5 iteration 4: exclusion of route 2-3 the average ei of the system for existing number of lines from below figure after the exclusion of less efficient lines 2-5, 4-5 and 2-3 will be=1425.8/28= 50.92= 49.1. fig.7 shows the chart for ei of network with exclusion of route 2-3 from the network. fig7.economic index after exclusion of less efficient line 2-3 b. nework data corresponding to optimal system  total number of transmission lines (without any exclusion) = 31  base network transmission lines=17  total transformers in the network=4  total alternate lines added to network= 14 (excluding a transformer between bus 5-6) so that (17+14=31=total transmission lines)  total alternate lines left in network=11(14-3 removed lines) 0 20 40 60 80 100 120 140 160 4 -7 4 -9 5 -6 5 -6 7 -8 3 -4 6 -1 1 1 2 -1 3 1 0 -1 1 1 0 -1 1 1 0 -1 1 1 3 -1 4 9 -1 0 9 -1 0 6 -1 2 6 -1 3 6 -1 3 2 -5 2 -5 7 -9 9 -1 4 4 -5 4 -5 2 -4 2 -4 2 -3 2 -3 2 -3 2 -3 1 -5 1 -5 1 -5 1 -2 1 -2 1 -2 lines no. e  total transmission lines left in network= 28 (31-3 removed lines) therefore, the optimization algorithm provides a new network that needs only 28 lines out of 31 which corresponds to the most economical and efficient network. fig8. new network after optimization fig.8 shows an optimized network that is obtained after applying analytical algorithm in the network. table 3 below shows overall ei pattern for all four iteration cases also table 4 shows overall economic index with all available routes. table 3. pattern for economic index (ei) s.n. line to be removed ei(overall) lines in network ei per line 1 no removal 1389.6 31 43.42 2 2-5 1425.4 30 47.51 3 4-5 1419.4 29 48.94 4 2-3 1425.8 28 50.92 table 4. overall economic index (ei) with all available routes s.n. route ei (%) 1 10-11 14 2 9-10 31 3 6-13 41.6 4 2-5 43.2 5 4-5 44.8 6 2-4 53.6 7 2-3 54 8 1-5 70.4 9 1-2 135.8 fig 9. graph between route and economic index fig10. graph for average economic index with removed line corresponding to ei order. case study: a practical power system of dehradun (in uttarakhand, an indian state) area has been considered here to implement analytical algorithm. the overall system data for the power system under consideration is given in table 5 in 2007. there are three load buses situated at three locations known as majra, purkul and bindal as shown in fig.11. there are four generator buses placed at four different locations known as khodari, dhakrani, dhalipur, and kulhal one bus at rishikes considered as slack bus. similarly, table 6 shows the system data of 2015. table 5. bus data of dehradun system in 2007 s. no. bus name location 1. generator bus khodari (60mw) dhakrani(12mw) dhalipur(16mw) kulhal(12mw) 2. slack bus rishikesh(70mw) 3. load bus majra (80mw) purkul (35mw) bindal(50mw) 43,42 47,51 48,94 50,92 38 40 42 44 46 48 50 52 no removal 2-5 4-5 2-3routes ei table 6. bus data of dehradun system in 2015 s. no. bus name location 1. generator bus khodari (60mw) dhakrani(12mw) dhalipur(16mw) kulhal(12mw) 2. slack bus rishikesh(70mw) 3. load bus majra (96mw) purkul (69mw) bindal(81mw) iteration 1a: the power system network data of 2007 and 2015 are taken with corresponding increase in load in the duration from 2007-2015 as shown in fig.11. the below network is obtained after first iteration with an increment in load between 2007 to 2015. the line length data and type of conductor used are obtained from the manual and power map available at power transmission corporation of uttarakhand state website [12]. it is observed from table 6 that in 2015, the load has increased on all the three load buses in the areas of majra from 80 mw to 96 mw, purukul from 35 mw to 69 mw and bindal from 50 mw to 81 mw. fig.11 dehradun area power system on pws (with 2007-2015 load change) fig.12 overloaded system after first iteration from fig. 12 is observed that there is overloading in the bus between majra and rishikesh. the algorithm when applied to the network along with the loading condition of the lines shows one more line is required in the areas between majra and rishikesh. table.7 additional lines required after load increment from bus to bus % loading total lines alternate lines rishikesh majra 159 2 2-1=1 from table 7 it is observed that there is only one additional line is needed between rishikesh and majra for removing the system overloading and hence to obtain optimized network as there is no scope to remove any additional line from the system from the view point to system loading corresponding to 80% as considered throughout the system. hence, system in fig. 13 below, is the optimized system which is also the actual or practical system present in dehradun. fig.13 shows the optimal network. fig.13 optimized power system of dehradun (2015) after addition of only one line table. 8 ei of the system with no res from bus to bus ei 5.0 6.0 50.8 4.0 6.0 52.4 4.0 6.0 55.4 6.0 8.0 74.6 6.0 7.0 81.2 6.0 7.0 81.2 3.0 4.0 85.4 3.0 5.0 89.2 7.0 8.0 97.6 1.0 2.0 120.0 2.0 3.0 143.4 table.8 shows the ei of the dehradun system when there is no res in system. bus numbering is obtained as per table 9. table also gives require data like mw flow etc. from the above table average ei = 931.2305/11 = 84.685. table. 9 system data obtained from pws after first iteration. iteration 1b: considering one res of 25mw capacity at heavily loaded bus of majara. the consideration behind capacity addition is as per “24x7 power for all uttarakhand” a joint initiative by government of india and government of uttarakhand, manual. there is a proposal of a capacity addition of renewable energy sources into existing system. the addition of solar power (grid connected as well as off grid) for the year 2016-17, has an average of approximately 24 mw (32.1 mw for 2016; 15.4 mw for 2017) [9-10]. so, a solar capacity addition of 25 mw is taken in our case study. fig.14 dehradun system with one res of 25mw at majra bus table. 10 ei of the system with one res from bus to bus e.i 4.0 6.0 51.0 4.0 6.0 54.0 5.0 6.0 59.8 3.0 4.0 82.2 6.0 7.0 82.4 6.0 7.0 82.4 6.0 8.0 82.4 7.0 8.0 87.6 3.0 5.0 92.4 1.0 2.0 120.0 2.0 3.0 143.4 from table 10 above, the average ei of the system is 937.5835/11 = 85.23. hence, from the above two iterative cases it is concluded that with an additional res at bus 6 the overall ei of the system improved and the overall system loading is enhanced. fig.15 graph for economic index with and without res figure 15 above shows the increase in average ei of the system after applying analytical algorithm. vi. conclusion and future scope there are several approaches available for power system expansion planning with different available constraints. the approach given in this paper is based upon mw-mile method comprises for deregulated power systems. the approach is simple and easy to implement for different configurations of power system e.g. load variations, generation variations, any available contingency in the system or other. proposed methodology is applied to a practical power system which shows its ease of implementation and importance. based on different scenarios, the given approach can be implemented in future on the following described areas. 1. the proposed methodologies in future can also be applied on garver’s 6 bus, ieee-9 and ieee-30 etc. bus systems and also on practical systems to see the responses against various transmission system constraints. 2. the prescribed system can also be used as stand-alone system with storage system. 3. the cost analysis of the system under consideration can also be done. 4. transmission losses and their reduction could be part of the planning process. 925 930 935 940 iteration 1a iteration 1b ei 5. the load end can be used as input (predefined) in order to identify transmission plans. the consideration of transmission constraints in the generation planning process would be also a possible subject for future work. 6. in addition to this other software like etap, matpower, etc. can also be used. references [1] junhua zhao and john foster, “flexible transmission network planning considering the impacts of distributed generation”, energy economics and managenent group, school of economics, queensland university. availableat: http://eemg.uq.edu.au/filething/get/194/01.pdf [2] xinsong zhang, yue yuan, boweng wu, qiang li, “a novel algorithm for power system planning associated with large-scale wind farms in deregulated environment” ieee 4th international conference on electric utility deregulation and restructuring and power technologies (drpt), 2011, china [3] kaur raminder, kaur tarlochan, kumar maneesh, verma shilpa, “optimal transmission expansion planning under deregulated environment: an analytical approach” ieee 1st international conf. on power electronics, intelligent control and energysystems (icpeices) 2016 india. [4] rajeev kumar gajbhiye,, devang naik, sanjay dambhare, and s. a. soman, “an expert system approach for multi-yearshort-term transmission system expansionplanning: an indian experience” ieee trans. on power systems, vol. 23, no. 1, february 2008 [5] m. oloomi, h. m. shanechi, g. balzer and m. shahidehpour, “transmission planning approaches in restructured power systems,” in proc. ieee/power eng. soc. power tech. conf.,bologna, italy, vol.2, 2003. [6] i. i. skoteinos, g. a. orfanos, p. s. georgilakis and n. d.hatziargyriou “methodology for assessing transmission investments in deregulated electricity markets”, ieee trondheim powertech,2011 [7] risheng fang and david j. hill, “a new strategy for transmission expansion in competitive electricity markets”, ieee transactions on power systems, vol. 18, no. 1, february 2003 [8] v. s. k. murthy balijepalli, and s. a. khaparde, “novel approaches for transmission system expansion planning including coordination issues” ieee power and energy society general meeting, july 2010 [9] cea general guidelines for substation, new delhi, india, june 2012 [10] manual on transmission planning criteria, new delhi, india january 2013 [11] t. tachikawa,hiroyuki kita, hideharu sugihara, ken-ichi nishiya and jun hasegawa “a study of transmission planning under a deregulated environment in in power system,” in proc. ieee int. conf. electric utility deregulation and restructuring, pp. 649–654,2000 [12] power transmission corporation of uttarakhand ltd. website available: www.ptcul.org maneesh kumar (m’15–s’16) was born in haridwar district of uttarakhand state in indian on feb. 1987. he received his bachelor degree in electrical engineering from college of technology, g.b. pant university of agriculture and technology pantnagar, uttarakhand in 2009. he received his master’s degree in electrical engineering from p.e.c university of technology, chandigarh, india in 2015 currently he is pursuing his phd. degree from indian institute of technology, roorkee, india, since dec. 2015. he also worked as a lecturer for more than 3 years (from 2009 to 2013) and also assistant professor for 6 months (from july 2015 to dec. 2015). his major research area includes power system. renewable energy systems, microgrids etc. raminder kaur has received her bachelor degree from p.e.c chandigarh, india in 1984 and also her master’s degree from the same institute in 1991. she has submitted her phd. thesis in 2017 and currently she is working as an assistant professor in p.e.c university of technology chandigarh, india. she has more than 30 years of teaching experience. her area of interest is power system planning, distributed generation, renewable energy systems etc. http://ieeexplore.ieee.org/xpl/mostrecentissue.jsp?punumber=5577387 http://www.ptcul.org/ i. introduction ii. transmission expansion planning with respect to generation change a. assuming a 60%t load increment in overall network b. assuming a solar plant of 75mw capacity at bus no. 10 c. simulation results with 60% load increment and distributed source at bus 10 iii. optimization of the above network using optimal algorithm a. economic index (ei) b. nework data corresponding to optimal system vi. conclusion and future scope references  transactions on environment and electrical engineering issn 2450-5730 vol 3, no 1 (2019) © vishnu sidaarth suresh  abstract—load flow studies are carried out in order to find a steady state solution of a power system network. it is done to continuously monitor the system and decide upon future expansion of the system. the parameters of the system monitored are voltage magnitude, voltage angle, active and reactive power. this paper presents techniques used in order to obtain such parameters for a standard ieee – 30 bus and ieee-57 bus network and makes a comparison into the differences with regard to computational time and performance of each solver. the objective being to first understand the working of each solver and then come to conclusions regarding the best one keeping in mind the network size and complexity so that it can extended to bigger networks for analysis. the methods are evaluated in this study using matpower which is a tool meant for academical purposes and not intended for on-line use. index terms—load flow, ieee 30 bus, ieee 57 bus numerical methods. i. introduction he load flow problem is an important tool for the operation and control of power systems. it gives the system operator information regarding active power, reactive power demand and consumption, voltage magnitude and voltage angle at every bus within the system which enables the operator to execute an appropriate schedule for dispatch of power. this information is also useful while planning expansion of power systems and helps maintain power system stability [1]. there are many techniques in-order to address the load flow problem [2-4], the techniques are numerical methods that are used to solve non-linear equations in order to obtain the steady state parameters of the system. in [5] network design and load flow analysis were carried out using etap and the resulting conclusions were taken as considerations for future expansion of power systems. in [6] load flow studies are performed using newton-raphson and decoupled load flow methods and a comparison is made amongst systems with and without unified power system controllers. [7] used ‘distflow’ for comparison of different numerical methods based solvers for the load flow problem. [8] uses a power system analysis toolbox called ‘mipower’ to study the performance of gaussseidel method on an ieee-3 bus system. [9] presents a unique vishnu suresh, phd candidate – wroclaw university of science and technology, faculty of electrical engineering, wybrzeze wyspianskiego 27,50-370 wroclaw, e-mail: vishnu.suresh@pwr.edu.pl. . power flow iterative algorithm and it is applied to a modified ieee – 30 consisting of two wind farms in order to validate the model. [10] provides a novel method called nonsy load flow in which the study has conducted load flow analysis using data that is unsynchronized and is obtained from diesel generators and the main substation in their network. once this data is obtained other parameters of the network are solved using backward/forward sweep methods. this study makes a comparison of the performance of the methods using matpower applied to two standard ieee test bus cases. matpower is a useful toolbox in matlab to solve the load flow problem, it is developed by the power system engineering research center at cornell university [4]. it is intended for academical use and understanding the different methods for solving load flow problems. in this paper we compare solving of the load flow problem for a standard ieee-30 and 57 bus test cases using gaussseidel, newton-raphson and fast decoupled load flow (fdlf) techniques in matpower and come to conclusions regarding the characteristics of each method. the reason for taking two test cases is to understand how the performance of the solvers varies with increased network size and complexity. moreover, such a comparison would enable the choosing an appropriate solver for analysis of city sized networks. the version of matpower used is 7.0b1, installed in matlab 2018b in a windows 10 64-bit system with an i5 core processor. the computational time in this study indicates the overall time take to obtain the solution whereas performance of each solver indicates the time taken per iteration and computational burden refers to the memory that is needed to run each solver. convergence is defined as a property of a solver to reach the solution vector, it represents the ability of a function to approach a limit as terms in the series increases. the ieee-30 bus test case system has a total of 6 generators, 24 loads, transmission lines at 1kv, 11kv, 33kv and 132kv along with capacitor banks at certain buses for reactive power compensation. the ieee-57 bus test case system has a total of 7 generators, 50 loads, transmission lines along with capacitor banks at certain buses for reactive power compensation. the test systems serve as a representative model to carry out power system studies and load flow analysis. the load flow problem involves solving for 4 parameters at every bus: active power(pi), reactive power (qi), voltage magnitude (vi) and voltage angle (i ) where i = 1,2,…..,n denotes the number of buses and if there are n buses then the total number of variables to be ascertained are 4n, but power flow studies usually assume bus types which usually comparison of solvers performance for load flow analysis vishnu suresh t mailto:vishnu.suresh@pwr.edu.pl keeps 2 out of 4 variables as constants, thereby reducing the number of variables to be solved to 2n. the bus types are summarized below [2, 3]:  pq bus/load bus: in this type of bus the total active power (pi) and reactive power (qi) at the bus are known, and is calculated as a difference between the active and reactive power injected and consumed in a bus. hence, the variables to be determined include voltage magnitude (vi) and voltage angle (i).  pv bus/voltage controlled bus/generator bus: this type of bus is usually preferred for power generating sources. here, the total active power injected and consumed is known (pi) and the voltage magnitude is maintained at a particular value by means of reactive power injection. hence, the unknown variables are total reactive power at the bus (qi) and voltage angle (i).  swing bus/slack bus/ reference bus: in this type of bus the voltage magnitude (vi) and the voltage angle (i) are known and the active power (pi) and reactive power (qi) are unknown. the slack bus is in-fact a fictitious concept that is created by a power system analyst in order to study the system [2]. in any load flow study, the total active and reactive power (complex power) at every bus is not known since the net complex power flow within the system is unknown including the total loses along transmission lines. therefore, it is a convention to choose the largest generator in a system to be the slack bus as it is understood that it is capable of producing active and reactive power according to the needs of the system. there is usually only one such bus chosen in a system as a reference. in the ieee test bus cases, the largest generator is chosen as the slack bus and the other sources are chosen as pv buses whereas the loads are modeled as load buses. once the buses are decided the equations to solve are (1). 𝑃𝑖 = |𝑉𝑖|∑ |𝑉𝑘||𝑌𝑖𝑘|cos⁡(𝜃𝑖𝑘 + 𝑘 − 𝑛 𝑘=1 ⁡𝑖) (1) 𝑄𝑖 = −|𝑉𝑖|∑ |𝑉𝑘||𝑌𝑖𝑘|sin⁡(𝜃𝑖𝑘 + 𝑘 − 𝑛 𝑘=1 ⁡𝑖) (2) where, i = 1,2,…..,n. yik – represents self and mutual admittances, between buses i and k and forms the bus admittance matrix ybus that is crucial to obtain the load flow solution. in order for static load flow equations to match reality as close as possible it is important to incorporate limits pertaining to all components in the network. the constraints are described as follows:  voltage magnitude constraints |𝑉𝑖|𝑚𝑖𝑛 ≤⁡|𝑉𝑖⁡| ≤⁡|𝑉𝑖|𝑚𝑎𝑥 (3)  voltage angle constraints |𝑖⁡⁡— ⁡𝑘| ≤ ⁡|𝑖⁡⁡— ⁡𝑘| 𝑚𝑎𝑥 ⁡⁡⁡⁡⁡ (4) this difference with regard to difference of angle during transfer of power between buses i and k is important for system stability.  constraints of sources to generate active and reactive power (𝑃𝑔𝑖)𝑚𝑖𝑛 ⁡≤⁡𝑃𝑖⁡ ≤⁡(𝑃𝑔𝑖)𝑚𝑎𝑥 (5) (𝑄𝑔𝑖)𝑚𝑖𝑛 ⁡≤ 𝑄𝑖⁡ ≤⁡(𝑄𝑔𝑖)𝑚𝑎𝑥 (6) pgi and qgi are the active and reactive power generated at bus i ii. numerical solvers a. gauss-seidel method this method is used to solve a set of non-linear algebraic equations. it is an iterative method and begins with an assumption of a solution vector. the assumption is made with regard to practical considerations. revised value of a variable is obtained by substituting in one of the equations in (1) the remaining present variables of the solution vector. then the solution vector is immediately updated with this new revised variable. this process is done for all variables in the solution vector in one iteration. the iterations continue until a certain degree of accuracy of the solution vector is obtained. the gauss-seidel method is very simple in terms of its usage to solve non-linear equations, also it is not necessary to store data from previous iterations to go to the next iteration. on the other hand, this method is very sensitive to the initial assumption of the solution vector, hence the speed of convergence depends on the closeness of the solution vector to the actual solution. in certain cases when the assumption is highly inaccurate the method might fail to converge [2, 3]. the application of this method to the power system is as follows: 1. first, the load demand (pdi and qdi) are obtained at all buses, then keeping relevant constraints in mind the active and reactive power generations (pgi and qgi) are allocated at all generating stations and since the largest generating station is kept as a reference bus the active and reactive power generation at this bus is allowed to change during the iterations. 2. the bus admittance matrix ybus is assembled with the available line and shunt admittance data 3. to begin the iterative process a flat voltage start is assumed and all buses are set to a voltage magnitude and angle of 10 o . then the voltages at every bus is recalculate by a rearranged version of equation (1) and the iterations continue until an acceptable accuracy is obtained. 𝑣𝑖 𝑝+1 − 𝑣𝑖 𝑝 < ⁡ℇ 4. once the voltage values of all buses are known then active and reactive power at the slack bus is obtained. 5. the last step of the process involves calculating the losses of the system using the line and shunt admittance data along with the known voltage values. these steps describe the method to obtain all parameters for pq buses since it begins with an assumption of active and reactive power demand and consumption at every bus. for pv buses the iterative method is different with regard to the assumptions made at the beginning of the iterative process, the detailed procedure is described in [3]. b. newton-raphson method this is a powerful tool for solving a set of non-linear equations, the advantages of this method are that it is not sensitive to the assumption of the solution vector made. the solution in this case converges in most cases as compared to the gauss-seidel method and it is done in a fewer number of iterations. the drawback of this method is increased computational burden and the need of additional storage space since it involves calculation of jacobian matrices and storage of values of previous iterations. at any iteration, the function is approximated by a tangent hyperplane and the problem is linearized into a jacobianmatrix equation [3]. the jacobian matrix consists of slopes of the tangent hyperplanes. ⁡f(x)⁡=⁡—⁡j. δx (7) the problem is solved for the correction δx the correction solved is then added to the previous value of x, so the new updated value is closer to the solution and this iteration process continues until an acceptable accuracy is obtained and the correction values in subsequent iterations are very small. the application of this method to the power system will be as follows: 1. for a pq bus for which the values of active and reactive power are known (pi and qi), an initial assumption of the solution for vi and i is made. substituting these values in equation (1) the calculated values for pi and qi are obtained then the corrected values are calculated. 𝛥𝑃𝑖 =⁡𝑃𝑠𝑝𝑒𝑐𝑖𝑓𝑖𝑒𝑑 ⁡−⁡𝑃𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑒𝑑 (8) ⁡⁡⁡𝛥𝑄𝑖 ⁡=⁡𝑄𝑠𝑝𝑒𝑐𝑖𝑓𝑖𝑒𝑑 ⁡− ⁡𝑄𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑒𝑑 (9) these values δpi and δqi correspond to f(x) in (6) and the corrected values for vi and i can be obtained by solving (6). 2. at the slack bus the values of voltage magnitude and angle vi and i is fixed. hence, there would be no equations pertaining to the slack bus in the jacobian. 3. once the corrected values δvi and δi are obtained, the next iteration is carried out by adding these corrected values to the previous values of vi and i and step 1 is repeated. this process continues until the corrected values are very small pertaining to an accuracy that is acceptable. 4. pv buses have constant voltage magnitude and active power at every bus and the iterative process is carried out to obtain values for reactive power and voltage angle. c. fast decoupled load flow methods. in transmission systems there is always an interdependence between the voltage angle and active power (-p) and between voltage magnitude and reactive power (v-q). hence, the coupling amongst -p and v-q is weak and this can be exploited in order to make the load flow problem simpler and reduce the computational burden on processing software. this can be done by solving the -p and v-q problems separately, that is to have two small submatrices for the variables and it is the basis for the decoupled load flow methods. the jacobian that is formulated in the newton-raphson method as mentioned above is simplified by eliminating the elements with weak coupling and it is usually about half of the elements that are eliminated in this manner. this could affect the true convergence of the solution but there is a trade-off between solution accuracy and reduced computational burden which is acceptable [2, 3]. iii. load flow analysis load flow analysis will be carried out for all three methods mentioned above on ieee-30 bus and ieee-57 bus test cases and the results of each will be discussed in-order to understand the advantages and disadvantages of both. a. ieee-30 bus the summary of load flow analysis of the ieee-30 bus test case, including the total amount of active and reactive power generated, consumed and line losses using gauss-seidel, newton-raphson and fdlf methods are presented in figures 1-3 respectively. the gauss-seidel method was able to arrive at the solution in 492 iterations and 0.45 seconds. the newton-raphson method was able to arrive at the solution in 2 iterations and 0.14 seconds. the fdlf method was able to arrive at the solution in 7 piterations and 6 q-iterations with a total of 13 iterations and in 0.16 seconds. this method takes advantage of the weak coupling between -p and v-q, hence the equations for both set of variables are solved separately and the number of iterations for the solution also differ. figure 1. system summary and load flow analysis using gauss – seidel numerical method. figure 2. system summary and load flow analysis using newton-raphson numerical method figure 3. system summary and load flow analysis using fast decoupled load flow method b. ieee-57 bus to study the effects of increasing the network size on the performance of numerical solvers, the ieee-57 bus test case is used and the results are compared with those obtained from the ieee-30 bus test case. figures 4-6 represent load flow summary, including the total amount of active and reactive power generated, consumed and line losses using the all 3 methods respectively. the gauss-seidel method was able to arrive at the solution in 518 iterations and 0.59 seconds figure 4. system summary and load flow analysis using gauss – seidel numerical method figure 5. system summary and load flow analysis using newton-raphson method figure 6. system summary and load flow analysis using fast decoupled load flow method. the newton-raphson method was able to arrive at the solution in 3 iterations and 0.15 seconds. the fdlf method was able to arrive at the solution in 7 piterations and 7 q-iterations with a total of 14 iterations and in 0.17 seconds. this method takes advantage of the weak coupling between -p and v-q, hence the equations for both set of variables are solved separately and the number of iterations for the solution also differ. c. convergence of methods figure 7. convergence of gauss-seidel method (ieee-30 bus) figure 7 describes the convergence of the gauss-seidel method (ieee-30 bus) and in-comparison with figure 8 it can be inferred that the slope of convergence is quite gradual in this method and the number of iterations are much higher when compared to the newton-raphson and fdlf methods (ieee – 30 bus) as seen in figure 8. figure 8 which describes the convergence of newton-raphson and fdlf methods, it can be seen that the newton-raphson method takes lesser number of iterations, and from figure 3 the time taken by the fdlf method is 0.16 seconds compared to 0.14 seconds for the newton-raphson method from figure 2, hence it can be concluded that the per iteration is much higher in the newtonraphson method when compared to the fdlf method. this is because the assumptions taken in the fdlf method reduce the computational burden hence accelerating the iterative process. it should also be noted that there is no significant improvement in the overall time taken for the load flow analysis between newton -raphson and fdlf methods. figure 8. convergence of newton-raphson and fdlf method (ieee-30 bus) figure 9 represents the convergence of the gauss-seidel method (ieee-57 bus) to the solution. it can be noticed that the convergence is gradual and it takes a total of 518 iterations for the method to finish. figure 10 represents the convergence of the newton-raphson and fdlf (ieee – 57 bus) methods. it can be noticed that the convergence in figure 10 is much steeper and the solution is obtained in 3 iterations for the newton-raphson method and 14 iterations (7 – p iterations and 7 – q iterations) for the fdlf method. figure 9. convergence of gauss-seidel method (ieee-57 bus) figure 10. convergence of newton-raphson and fdlf method (ieee-57 bus) iv. conclusions a. results table i load flow results for ieee – 30 bus characteristics gauss-seidel newtonraphson fdlf iterations 492 2 7 p-iterations 6 q-iterations total – 13 time 0.45 0.14 0.16 time/iteration 0.0009 0.07 0.0114 convergence gradual steep steep computational burden low high higher than gauss-seidel, lower than newtonraphson table ii load flow results for ieee – 57 bus characteristics gaussseidel newtonraphson fdlf iterations 518 3 7 p-iterations 7 q-iterations total – 14 time 0.59 0.15 0.17 time/iteration 0.0011 0.05 0.0121 convergence gradual steep steep computational burden low high higher than gauss-seidel, lower than newton raphson b. discussions the comparison of different methods to solve the load flow problem yields the following results, the gauss-seidel method takes less time to perform one iteration when compared to the newton-raphson method, this is because of the fewer number of arithmetic operations involved in completing an iteration, as the calculation of the jacobian which is an inherent part of the calculations for the newton raphson method. the newton-raphson method has a faster rate of convergence because of its quadratic convergence characteristics. the technique is said to ‘home-in’ to the solution. for the gauss-seidel method the number of iterations increase with the network size i.e. higher the number of buses in the network, the longer it takes for the method to find a solution, this evident from the fact that it takes 518 iterations and 0.59 seconds for the ieee-57 bus test case to find a solution compared to 492 iterations and 0.45 seconds for the ieee-30 bus test case. the relationship is not as proportional in the newton-raphson method as the time taken for the ieee-57 test bus case with this method is 0.15 seconds and 3 iterations whereas for the ieee-30 bus case it is 0.14 seconds and 2 iterations representing only a marginal increase in the computational time. this conclusion holds also for bigger networks with a much higher number of buses [2,3]. the gauss-seidel method is relatively easier to implement and does not require a lot of memory, whereas the newtonraphson method is complex to implement and does require higher memory and processing capacity. the gauss-seidel method is very sensitive to the selection of the slack bus, in some cases the method is also known to not converge to a solution hence, making the first step of choosing a solution vector very crucial. inversely, the newton-raphson method is not so sensitive to the selection of the slack bus and almost always converges to a solution. the fdlf method in both cases (ieee – 30 and 57 bus test cases) takes more iterations and more time to arrive at the solution. it is important to remember that the fdlf method takes into account certain assumptions while searching for the solution making it as fast as the newton-raphson method with the advantage of reduced computational needs such as memory and processing capability. it can be hence concluded that both newton-raphson and fdlf methods are efficient and can be extended to bigger and more complex networks but the computational advantage that the fdlf method provides can lead to cost savings. therefore, the selection of a methods depends on the overall finances involved in solving load flow issues along with speed and accuracy. therefore, this paper has described and compared the application of 3 methods in solving the load flow problem for 2 standard ieee bus test cases and conclusions arrived at commensurate with the objective. future study in this regard is to extend the analysis to networks containing renewable energy sources that are unpredictable in their output which makes the load flow problem more complicated and to include time series analysis. the methods can also be executed on other tools and make a comparison as to which tools are most efficient for performing load flow analysis references [1] afolabi, o. a., ali, w. h., cofie, p., fuller, j., obiomon, p., & kolawole, e. s. (2015). analysis of the load flow problem in power system planning studies. energy and power engineering, (7), 509–523. https://doi.org/10.4236/epe.2015.710048 [2] stott, b. (1974). review of load-flow calculation methods. proceedings of the ieee, 62(7), 916–929. https://doi.org/10.1109/proc.1974.9544g. o. young, “synthetic structure of industrial plastics,” in plastics, 2nd ed., vol. 3, j. peters, ed. new york: mcgraw-hill, 1964, pp. 15–64. [3] d. p. kothari and i. j. nagrath, “power system engineering,” 2nd edition, tata mcgraw hill, new delhi, 2007 . [4] r. d. zimmerman, c. e. murillo-sanchez, and r. j. thomas, \matpower: steadystate operations, planning and analysis tools for power systems research and education," power systems, ieee transactions on, vol. 26, no. 1, pp. 12{19, feb. 2011. doi: 10.1109/tpwrs.2010.2051168 [5] prabhu, j. a. x., sharma, s., nataraj, m., & tripathi, d. p. (2016). design of electrical system based on load flow analysis using etap for iec projects. 2016 ieee 6th international conference on power systems, icps 2016, 1–6. https://doi.org/10.1109/icpes.2016.7584103. [6] model, u. m. (2000). power flow analysis of power system with upfc, 00(c), 2–5. dept. of electric power & automation engineering tianjin university, tianjin (300072) [7] g.m. gilbert, d.e. bouchard, a. y. c. (2010). comparison of load. international journal of engineering and advanced technology, 850– 853. department of electrical and computer engineering royal military college of canada kingston, ontario [8] kaur, s., singh, a., & khela, r. s. (2015). load flow analysis of ieee-3 bus system by using mipower software. international journal of engineering research & technology (ijert), 4(03), 9–16. [9] phan, t. t., nguyen, v. l., hossain, m. j., to, a. n., & tran, h. t. (2016). an unified iterative algorithm for load flow analysis of power system including wind farms. in l-s. lê, t. k. dang, j. küng, n. thoai, & r. wagner (eds.), 2016 international conference on advanced computing and applications acomp 2016: proceedings (pp. 105-112). piscataway, nj: institute of electrical and electronics engineers (ieee). https://doi.org/10.1109/acomp.2016.024 [10] bahmanyar, a., estebsari, a., bahmanyar, a., & bompard, e. (2020). nonsy load flow : smart grid load flow using non-synchronized measurements. 2017 ieee international conference on environment and electrical engineering and 2017 ieee industrial and commercial power systems europe (eeeic / i&cps europe), (646568), 1–5. https://doi.org/10.1109/eeeic.2017.7977509 https://doi.org/10.1109/acomp.2016.024 i. introduction ii. numerical solvers a. gauss-seidel method b. newton-raphson method c. fast decoupled load flow methods. iii. load flow analysis a. ieee-30 bus b. ieee-57 bus c. convergence of methods iv. conclusions a. results b. discussions references paper title (use style: paper title) transactions on environment and electrical engineering issn 2450-5730 vol 1, no 3 (2016) © micaela caserza magro, paolo pinceti, luca rocca can we use iec 61850 for safety related functions? micaela caserza magro, paolo pinceti, luca rocca department diten university of genova genova, italy paolo.pinceti@unige.it abstract—safety is an essential issue for processes that present high risk for human beings and environment. an acceptable level of risk is obtained both with actions on the process itself (risk reduction) and with the use of special safety systems that switch the process into safe mode when a fault or an abnormal operation mode happens. these safety systems are today based on digital devices that communicate through digital networks. the iec 61508 series specifies the safety requirements of all the devices that are involved in a safety function, including the communication network. also electrical generation and distribution systems are processes that may have a significant level of risk, so the criteria stated by the iec 61508 applies. starting from this consideration, the paper analyzes the safety requirement for the communication network and compare them with the services of the communication protocol iec 61850 that represents the most used protocol for automation of electrical plants. the goal of this job is to demonstrate that, from the technical point of view, iec 61850 can be used for implementing safety-related functions, even if a formal safety certification is still missing. keywords—communication protocols, fieldbus, functional safety, iec 61508, iec 61850 i. introduction the automation system of an electrical generation or distribution system is today based on intelligent electronic devices (ied) connected through a digital network. each intelligent device controls, protects, supervises, measures a section of the plant. in medium voltage switchgears we normally have an intelligent device per cubicle, in high voltage substations one or more per bay. all the status signals, interlocks, closing or tripping commands are transmitted through the communication network. some general purpose protocols can be used for communicating with intelligent devices (e.g. modbus rtu, ethernet tcp/ip and others), but also specific protocols for electrical plants exist, like the series iec 60870 and dnp3. all these protocols can be effectively used for the configuration and the interrogation of the ieds, and support all the functions that are typical of a scada (supervisory, control, and data acquisition). on the other hand, none of these protocols supports real-time functions, so they cannot be used for commands or interlocks. the only protocol that supports all the functions required for the substation automation is the more recent iec 61850. it has services both for scada and for real time functions. with the use of iec 61850, protection logic, safety interlocks, commands from real time functions (e.g. load-shedding) are transmitted digitally. fig. 1 shows a typical architecture of the automation system in a substation, with different protocols used for different functions:  iec 61850 for real-time control  iec 61870 for remote control  tcp/ip for remote maintenance  usb for local set-up and commissioning fig. 1. typical architecture for substation automation with a full-digital architecture, also the commands related to safety functions are transmitted through the communication protocol. in electrical plants most safety commands lead to the de-energization of a circuit or a machinery (e.g. the emergency stop of a motor, fire alarm, etc.). in some cases, the safety electrical command is related to a risk of the process and requires the energization of a circuit (e.g. starting a ventilation fan in a tunnel, starting an emergency generator, switching from one source to another, etc.). fig. 2 shows an example of interaction between an emergency stop push button and the trip of one or more circuit breakers. when the button is pushed, the safety plc receives the status through the safety fieldbus and sends a trip command to the station control unit on the iec 61850 network. then the trip command is sent on the iec 61850 bus, normally via goose message, to the ieds that command the relevant circuit breakers. the paper analyses the requirements of communication protocols when used for carrying out safety functions, with reference to the international standards series iec 61508. the transmission mechanisms of iec 61850 are then considered to check if it can be used for safety related functions. fig. 2. example of safety system interaction with iec 61850 ii. functional safety concepts safety is a concept dealing with the reduction of the risk of physical injury or of the damage to the people/environment health. the overall risk must be below the so-called acceptable risk. there are several concepts of safety, and this paper focuses on the functional safety, which involves the operation of the industrial processes or machineries. functional safety [1] is a part of the overall safety, and it is based on the principle that a system or equipment operates correctly in response to its inputs. functional safety cannot be determined without considering the system as a whole and the environment with which it interacts. a process or a machinery has an intrinsic level of risk. the risk is the probability of happening of a fault multiplied by the consequences of the fault, in terms of damages to people or to the environment. if the risk of a certain fault or malfunction is above the acceptable risk, it is necessary to introduce a safety related function to reduce it. such a function has the role to intervene for avoiding the fault or for reducing the consequences of the fault; in other words, for reducing the risk below the level of acceptable risk. the safety related systems can be implemented using any technology, but there is the constraint to respect some requirement of integrity. this means that it is mandatory that any safety related system has an adequate integrity level for assuring the proper operation. integrity is here intended as the probability that a specific function/system is properly working when requested. the required safety integrity level (sil) of a safety function must be adequate to the risk of the process/machine malfunctioning. the higher the risk of a fault the higher the required sil. today, most safety related functions use electrical and/or electronic and/or programmable electronic (e/e/pe) technologies. e/e/pe safety functions are regulated by the iec 61508 series [2]. one of the concept within the iec 61508 is the definition of safety function as the coordinated operation of some basic elements, like in fig. 3:  sensors, that are responsible for measuring or detecting the abnormal operation,  analogic/digital conversion, if needed,  logic solver, within one programmable logic controller plc, that may be programmable or not. it is responsible for implementing the safety logic according to the inputs coming from the sensors,  analogic/digital conversion, if needed,  actuators, that are responsible for acting on the process to drive the overall system in a safe condition. the concept and performance of sil applies to the overall safety function, thus the probability of failure on demand (pfd) of each component should be compliant to the requirements identified to achieve the requested pfd of the complete safety function. the iec 61508 defines four level of sil: from 1 to 4. the higher the sil the lower the pfd for the safety function and the higher the reduction of the risk. note that pfd is a value that indicates the probability of a system failing to respond to a demand. the average probability of a system failing to respond to a demand in a specified time interval is referred as pfdavg. fig. 3. example of a safety function table i. values of pfd according to the sil value safety concepts are applied both to the hardware and to the software related to safety functions. this means that also the software should comply with the required pfd. the software is both the firmware and the specific application software. this is an important aspect also for the operation and the connections between the components of the safety function. traditionally, the connection between the field devices and the logic solver uses copper cables for transmitting on/off or 4/20 ma signals. today, with the wide use of digital communication networks there is the need of using digital communication protocols even for safety functions. a communication system contains a hardware part (cables and communication chips) and a firmware part, responsible for the definition of the telegram, the technique for accessing the medium, the mechanism for transmitting, and so on. for safety applications, it is necessary that also the communication protocol comply with integrity requirements. iec 61508 allows the use of digital communication protocol for safety functions, but it requires that methods are implemented to detect transmission errors. in a quantitative way, iec 61508 requires that the communication system use no more than the 1% of the budget pfd for the safety function (see błąd! nie można odnaleźć źródła odwołania.). fig. 4. safety communication as a part of safety function iii. iec 61850 basics iec 61850 is an ethernet based protocol typical of electrical automation systems. its main goal is to exploit the ability of ieds from different manufacturers to exchange information used for protection, control and monitoring of an electrical substation. the standard defines this feature as interoperability [5]. the iec 61850 series consists of ten parts that specify all the various aspects of communication into details. section 7 is the core of the standard’s innovative concepts. the standard defines all the devices and functions that exist in a substation, and it organizes them into a set of logical nodes. a ln has a four letters standardized name in which the first letter defines the class the ln belongs, i.e. protection, control or monitoring. lns are the interface between the automation functions and the real world: the ied’s manufacturer, following its own engineering practices, implements the function while the structure of data and data exchange are standardized. section 7.2 defines the information models and the information exchange service models (acsi abstract communication service interface). the information model consists on the definition of four elementary classes that can model any type of logic device [6]:  server – it represents the visible behavior of the device from the outside. a server's role is to manage communication with the client and send information to the other servers,  logical device (ld) – it contains information managed and shared between different applications hosted in the same device. homogeneous info are grouped into logical nodes,  logical node (ln) – it contains the elementary data necessary for implementing the function the logical node refers to (i.e. overcurrent protection, measuring, breaker command, etc.),  data – it represents the value of interest with all the attributes that are used to describe it. the acsi comprises also the information exchange models needed to operate on data, which are:  data sets – used to group data,  substitution – supports replacement of a process value by another value,  setting group control block permits to switch from one set of setting values to another one,  report control block describes the model used to exchange information between a ln and a client. report generation can be triggered by a change of a process data value,  control blocks for generic substation event (gse) describes the exchange of hard real-time information. it is used for information changing sporadically and it provides simultaneous delivery of the same message to multiple devices using multicast/broadcast frames. the principal information exchange model for time critical information like tripping function or interlocking is called generic object oriented substation event (goose),  control blocks for transmission of sampled values used by measuring devices to send fast and cyclically analog sampled values,  control describes the services that a client use to control an ied,  time and time synchronization provides the time base to the substation automation system,  file transfer provides the exchange of large data. iv. remedial measures to detect errors and failures of a communication system if used in a safety function, the communication system must implement specific measures to detect communication errors and failures [3]. this is mandatory in order to avoid that any error or fault compromises the proper operation of the safety function. while iec 61508 does not restrict the use of communication technologies, iec 61784-3 [3] focuses on the use of fieldbus based functional safety communication systems. when using iec 61158 [4] based fieldbus structures without modifications in the definition of each communication layer, all the measures necessary to implement transmission of safety data in accordance with the requirements of iec 61508 shall be supported by an additional “safety communication layer”. fig. 5 describes the so-called “black channel” approach. the communication protocol of the selected fieldbus is not affected by any additional safety service. all the safety functions are in an additional layer, the safety layer, which is above the 7th layer (application) of the protocol. the safety messages are embedded within the standard protocol. fig. 5. implementation of a safety layer over a standard fieldbus communication system the role of the safety profile or layer is to detect all the possible errors that may lead to the loss or the corruption of the packets. the safety data must satisfy the following requirements:  trusted data, the safety data must be correct,  correct receiver, the receiver of the data must be correct,  just in time, the data must receive the destination at a proper time. the first step is to consider what are the possible errors in a digital communication system, as in iec 61784:  corruption: messages may be corrupted due to errors within a bus participant, due to errors on the transmission medium, or due to message interference,  unintended repetition: due to an error, fault or interference, old not updated messages are repeated at an incorrect point in time,  incorrect sequence: due to an error, fault or interference, the predefined sequence (for example natural numbers, time references) associated with messages from a particular source is incorrect,  loss: due to an error, fault or interference, a message is not received or not acknowledged,  unacceptable delay: messages may be delayed beyond their permitted arrival time window, for example due to errors in the transmission medium, congested transmission lines, interference, or due to bus participants sending messages in such a manner that services are delayed or denied,  insertion: due to a fault or interference, a message is inserted that relates to an unexpected or unknown source entity,  masquerade: due to a fault or interference, a message is inserted that relates to an apparently valid source entity, so a safety relevant participant, which then treats it as safety relevant, may receive a non-safety relevant message,  addressing: due to a fault or interference, a safety relevant message is sent to the wrong safety relevant participant, which then treats reception as correct. starting from the list of the possible errors in the communication system, the iec 61784 defines the measures that can be implemented in the communication stack in order to detect the errors and to set the system into a safe condition. proposed remedial measures are:  sequence number: a sequence number is integrated into messages exchanged between message source and message sink,  time stamp: in most cases the content of a message is only valid at a particular point in time. the time stamp may be a time, or time and date, included in a message by the sender,  time expectation: during the transmission of a message, the message sink checks whether the delay between two consecutively received messages exceeds a predetermined value. in this case, an error has to be assumed,  connection authentication: messages may have a unique source and/or destination identifier that describes the logical address of the safety relevant participant,  feedback message: the message sink returns a feedback message to the source to confirm reception of the original message. this feedback message has to be processed by the safety communication layers,  data integrity assurance: the safety-related application process shall not trust the data integrity assurance methods if they are not designed from the point of view of functional safety. therefore, redundant data is included in a message to permit data corruptions to be detected by redundancy checks,  redundancy with cross checking: in safety-related fieldbus applications, the safety data may be sent twice, within one or two separate messages, using identical or different integrity measures, independent from the underlying fieldbus. in addition to this, the transmitted safety data is cross-checked for validity over the fieldbus or over a separate connection source/sink unit. if a difference is detected, an error shall have taken place during the transmission, in the processing unit of the source or the processing unit of the sink. in a safety communication profile errors must be detected and the methods for doing so are those described above. it is not mandatory to implement all the protection measures, since it is necessary to implement only the measures that can avoid all the possible errors. iec 61784-3 defines the correlation between the errors and the possible detection methods (fig. 6). fig. 6. overview of the effectiveness of the various measures on the possible errors even when the messages are arriving in a correct (deterministic) manner the safety data still may be corrupted. thus data integrity assurance is a fundamental component of the safety communication layer to reach a required safety integrity level. suitable hash functions like parity bits, cyclic redundancy check (crc), message repetition, and similar forms of message redundancy shall be applied. generally, the communication channel shall not use the same hash function the superimposed safety communication layer uses. the safety code shall be functionally independent from the transmission code. the residual error rate is calculated from the residual error probability of the superimposed (safety) data integrity assurance mechanism and the transmission rate of safety messages. the number of destination sinks has to be considered for this calculation. the value of admissible residual error is related to the sil level of the safety function, like table ii. shows. table ii. relationship of residual error rate to sil level v. safety analysis of iec 61850’s communication services protection and interlocking functions are the main applications that may require a functional safe communication system. these functions use the goose model to exchange information. goose is based on a publisher/subscriber mechanism, and it uses a multicast transmission of data. if the value of one or more data configured in the goose application changes, one ied (the publisher) sends a message to a group of ieds (the subscribers) simultaneously within a single goose message. in addition to the data, a goose frame contains a set of parameters that describe the message itself (see fig. 7). fig. 7. goose service parameter mapping in particular, goose messages implement a time stamp parameter that contains the time at which one value configured in the data set has changed. at the same time, the parameter “state number” is incremented. this parameter represents a counter that increments each time a value change is detected within the data set, and a goose message has been sent. the frame contains also a sequence number that increment each time a goose message is sent. time stamp, state number and sequence number are remedial measures against unintended repetition, incorrect sequence, loss, unacceptable delay, insertion errors as defined in section iv. goose data exchange allows the use of vlans and priority tagging as defined in ieee 802.1 q. the use of vlans permits defining the set of ieds that shall receive the goose message. this feature prevents a safety relevant information to be delivered to the wrong device and to cause unintended events or errors: it is a remedial measure against masquerade and addressing errors defined in [3]. the goose message frame is terminated with a frame check sequence. fcs field contains a number calculated from the data in the frame. the receiver calculates autonomously this number, and compares it with the number in the fcs field. if the numbers are different, the receiver knows that an error in the communication is occurred and it discards the corrupted frame. furthermore, goose messages use a specific scheme of re-transmission to achieve the appropriate level of reliability (fig. 8). when an event occurs, the goose server generates a sendgoosemessage request and the current data set values are encoded in a goose message. the goose message is transmitted immediately and then retransmitted with a variable time interval (t1, t2, t3) not defined by the standard and gradually increasing the parameter “sequence number”. an effective scheme of retransmission is based on an exponential increment of the time between the frames. the time interval increments until it reaches the retransmission stable time t0 defined in the configuration of the goose application as tmax. t0 can be shortened by the event ((t0) in fig. 8). each message in the retransmission sequence contains a timeallowedtolive parameter that represents the time the receiver waits before the next retransmission. if the timeallowedtolive expires, the receiver reports a communication problem, and the system should be switched in a safe mode. the retransmission mechanism explained can be used as a time expectation measure defined in section iv. fig. 8. transmission time for events vi. conclusion all the safety buses existing today use the “black channel” approach to detect communication errors. iec 61850 does not implement any form of safety communication channel (black channel) as required by iec 61784-3. however, the analysis made in section v reveals that the standard natively implements a bunch of remedial measures to detect all the communication errors defined by iec 61784 [3]. the problem is that the standard does not define what must be done when a communication error is detected. if the communication between two ieds fails, the system has to switch in a safe condition. it is reasonable to think that, for electrical plants, the safe state is considered a circuit de-energized. a possible solution is to implement in the 61850 stack the functions energize-to-trip and de-energize-to-trip. another possible solution is to configure the ied to switch the system in safe mode when a communication error is detected using the ied’s configuration tool. the two solutions present a significant difference considering a hypothetical safe certification process. implementing functions in the 61850 stack means that the stack should be certified compliant to the iec 61508 standard; on the other hand, if a configuration software is used to program one or more safety function in a device, the software itself should be certified. while this work focuses on iec 61850 communication protocol, it is also important to remember that sil’s concepts applies to all the hardware and the software that compose the safety system. this means that not only the communication protocol should be certified, but also all the ieds and circuit breakers that compose the safety chain. further analysis will be made on a typical safety function that uses iec 61850 as communication protocol. this will be useful to understand if iec 61850 can satisfy the requirement of iec 61508 about the maximum use of 1% of the budget pfd for a typical safety function and to estimate what is the maximum possible safety integrity level of a safety system that uses a communication network based on iec 61850. references [1] iec 61508-0 “functional safety of electrical/electronic/programmable electronic safety-related systems – part 0: functional safety and iec 61508”, 2010 [2] iec 61508-1 “functional safety of electrical/electronic/programmable electronic safety-related systems – part 1: general requirements”, 2010 [3] iec 61784 “industrial communication networks – profiles – part 3: functional safety fieldbuses – general rules and profile definitions” [4] iec 61158 series “industrial communication networks – fieldbus specifications”, 2014 [5] iec 61850-1 “communication networks and systems for power utility automation – part 1 introduction and overview”, 2013 [6] iec 61850-7-2 “communication networks and systems for power utility automation – part 7-2: basic communication structure – abstract communication service interface (acsi)”, 2010 [7] iec 61850-8-1 “communication networks and systems for power utility automation – part 8-1: specific communication service mapping (scsm) – mappings to mms (iso 9506-1 and iso 9506-2) and to iso/iec 8802-3” [8] k-p brand, m.ostertag, “safety related, distributed functions in substations and the standard iec 61850”, 2003 bologna power tech conference, june 23th-26th, bologna, italy [9] j. hoyos, m. dehus and t. x. brown, "exploiting the goose protocol: a practical attack on cyber-infrastructure", proc. 2012 ieee globecom workshops, pp. 1508-1513 paper title (use style: paper title) transactions on environment and electrical engineering issn 2450-5730 vol 3, no 1 (2018) © dmitry ivanovich panfilov, ahmed elgebaly, alexander nikolaevich rozhkov, michael astashev s2 l4l2 s4 s7 l3 s1 l1 s3 s5 s6 v s is fig. 1. tsr with 25 power steps control strategy of thyristors switched svcs with high power quality d. i. panfilov ahmed e. elgebaly m. g. astashev alexander n. rozhkov department of industrial electronics moscow power engineering institute moscow, russia electrical power and machines department faculty of engineeringtanta university tanta, egypt g. m. krzhizhanovsky power engineering institute (jsc enin) moscow, russia department of industrial electronics moscow power engineering institute moscow, russia dmitry.panfilov@inbox.ru ahmed.elgebaly@feng.tanta.edu.eg astashev@eninnet.ru rozhkovan@eninnet.ru abstract—in this paper, new static var compensators svcs schemes for inductive and capacitive reactive power are developed. the provided schemes improve the flexibility and power system quality of svcs by developing new circuit topologies with new control strategy of the reactive power. new circuit schemes are introduced for thyristors switched reactors tsr and thyristors switched capacitors tsc to design harmonicfree svc with higher discrete number of reactive power levels. this paper provides the control algorithm and block diagram of the new svcs schemes. the switching strategies of tsr and tsc are described and implemented. the new scheme of tsc requires special modifications to decrease transient effects and implementation of specific switching strategies to acquire svc with high power quality indexes. keywords—static var compensators, thyristor switched reactors, thyristor switched capacitors i. introduction static var compensators svc play an important role in electrical power systems because of their capability to dynamically compensate the reactive power. the svc fundamentally is made up of inductive reactive power source, such as thyristor-controlled reactor tcr or thyristors switched reactors tsr, and capacitive reactive power source, for instance fixed capacitor fc and thyristors switched capacitors tsc [1]. the advance of svc has various research trends such as optimization of the device rating and the place of installation in the power network [2], improvement of power electronics structure in the schemes [3,4,5] and development of svc control system structure and operation [6]. tsr is considered as a solution to eliminate high order harmonics which are commonly produced in svc based on tcr scheme. the new tsr topologies have flexible performance than the simple equal parallel branches or binary tsr schemes [7]. the existed tsc has one of the following topologies: equal rating parallel tsc branches and binary tsc [8]. in the paper, new developed circuit topologies for both of tsr and tsc are developed with essential idea of increasing the number of steps of reactive power compensation with the same number of passive elements (reactors and capacitors). this paper demonstrates the principle of the new schemes operation. the characteristics of the new schemes that relates between the required and developed reactive power is provided in this paper. furthermore, the control strategy and construction of the new developed schemes are clarified. the modification of tsc operation during transients is implemented with special algorithm and circuitry. ii. circuits topologies of improved tsc and tsr tsr and tsc new schemes are developed to control the reactive power with zero harmonic content but with discrete manner. the earlier schemes of tsr control the developed reactive power by varying the equivalent inductance of the parallel connected reactors. the developed schemes control the reactive power not only by changing the parallel connected branches but also by series connection of passive elements [9]. fig. 1 illustrates a new circuit scheme for tsr which consists the study was performed at stock company g. m. krzhizhanovsky power engineering institute within the framework of the project "development of a controlled source of reactive power with the absence of the higher-order harmonics currents during the regulation of the electrical energy and with the improved technical and economicl indicators on the basis of the domestic component base of power electronics for automatic control of the voltage and the power flows in the electric power distribution networks of 6 110 kv (rfmefi57917x0140)" with the financial support of the ministry of education and science of the russian federation mailto:dmitry.panfilov@inbox.ru mailto:astashev@eninnet.ru mailto:rozhkovan@eninnet.ru four reactors and seven back-to-back thyristor switches. this scheme allows to obtain 25 different equivalent inductances; thus, it produces 25 reactive power steps. it can become evident that with additional three switches, the number of steps are increased from 16 steps in binary tsr to 25 steps in the new scheme [10]. table i shows the different equivalent inductances and the states of operation of the switches where + means series connection and // means parallel connection of reactors. the total apparent power of reactors in the new scheme equals the total apparent power of the reactor used in conventional tcr and binary tsr [10]. with the same manner, the number of reactors may be more than four, but the number of switches will be more than the number of reactor by 3. these three switches are s5, s6 and s7 in fig. 1. table i. the equivalent inductance obtained by new tsr with four reactors and seven switches no. equivalent inductance switches state s1 s2 s3 s4 s5 s6 s7 1 0 off off off off off off off 2 l1+l2 on on off off off off on 3 l2+l3 off on on off off off on 4 (l1//l3) + l2 on on on off off off on 5 l2 off on off off off on off 6 l1+l4 on off off on off off on 7 (l2//l4) + l1 on on off on off off on 8 l3+l4 off off on on off off on 9 (l2//l4) + l3 off on on on off off on 10 l1 on off off off on off off 11 (l1//l3) + l4 on off on on off off on 12 (l1//l3)+ (l2//l4) on on on on off off on 13 l1//l2 on on off off on on off 14 l3 off off on off on off off 15 l4 off off off on off on off 16 l2//l3 off on on off on on off 17 l2//l4 off on off on off on off 18 l1//l3 on off on off on off off 19 l1//l4 on off off on on on off 20 l1//l2//l3 on on on off on on off 21 l1//l4//l2 on on off on on on off 22 l3//l4 off off on on on on off 23 l3//l4//l2 off on on on on on off 24 l1//l4//l3 on off on on on off off 25 l1//l2//l3//l4 on on on on on on off fig. 2 demonstrates the new scheme of tsc which contains four capacitors and seven switches. this topology can produce 25 steps of operation such as that of the new tsr in fig. 1. this scheme is better than binary tsc that contains four branches because the new topology provides 25 steps instead of 16 steps which can produced by binary tsc. although the topologies of tsr and tsc have the same principle of operation by varying the developed reactive power by changing the equivalent reactance, the method of control of each scheme is different [11,12] as will explained in section iv. iii. characteristics optimization of the developed tsc and tsr topologies the characteristic of the svc system is the relation between the required reactive power as a reference from the control system and the actual produced reactive power from svc. the characteristics of the new tsr or tsc schemes have discrete form. consequently, it is very vital to adjust the svc characteristics to be most near to the continuous characteristic of svc with tcr. optimization technique can determine the parameters of each of the new developed schemes to obtain the smoothest variation in the produced reactive power [3]. in this section, the optimization technique is applied for tsr and with the same way it can be applied for tsc [3]. by assuming that, each equivalent inductance lx(n) produces the inductive reactive power qd:     2 d x v q n l n  (1) in the case of the capacitive reactive power, each equivalent capacitance cx(n) produces capacitive reactive power according the following equation    2d xq n v c n , where n is the number of the equivalent circuits which depend on the switches states in fig. 2. an objective function should be applied to optimize the distribution of the produced reactive power. this objective function o.f.d designates the difference between two consecutive developed reactive power (qd(n), qd(n+1)), while this difference shouldn’t exceed certain determined value δq.      1 1 . . 1 m d dn of d q n q n q       (2) where m is the total number of steps obtained by the developed scheme. this objective function is used for both of new tsr and tsc schemes. optimization technique that depends on genetic algorithm ga is used to solve this objective function. the constraint for the variables is the ratio between the maximum and the minimum variables (maximum and minimum inductance or capacitance of passive elements) doesn’t exceed 8. so, the parameters of tsr and tsc can be determined as a function of rated inductance leq and rated capacitance ceq correspondingly. the values of inductances to obtain the finest solution are as the following: ll = 5.1 leq, l2 = 10.8 leq, l3 = 3 leq, l4 =2.6 leq. while the value of capacitance for tsc optimal design c1 = 0.2 ceq, c2 = 0.0998 ceq, c3 = 0.3248 ceq, s2 c4c2 s4 s7 c3 s1 c1 s3 s5 s6 v s is l t1 t2 t1 t2 t1 t2 t1 t2 t1 t2 t1 t2 t1 t2 + vc0 + vc0 + vc0 + vc0 fig. 2. tsc with 25 power steps svc with tsr and tsc binductive tsr step number distribution unit for inductive and capacitive susceptance svc thyristors control block bsvc binductive s te p n u m b er btotal error-ve error+ve input output error * error voltage regulator gr (s) = ki/s dead zone block _ vm × vsvc isvc measuring and filter circuits h = 1/(tms+1) slope ksl measuring and filter circuits h = 1/(tms+1) _ power system bcapacitive bcapacitive s te p n u m b er tsc step number capacitors discharging strategy vref(new) + fig. 4. the block diagram of control system for thyristors switched svc c4 = 0.3754 ceq. the predetermined parameters offer the characteristic for each of tsr or tsc as demonstrated in fig. 3. this characteristic links between the required reactive power from the scheme and the discrete produced reactive power. it can be noted that this characteristic may be inductive or capacitive according to the type of the scheme. to get certain required reactive power, certain step of 25 steps should be realized consistent with its percentage. the grouping between the two new tsc and tsr schemes provides more smooth variation in the group characteristic. iv. algorithms of thyristors controlling in new svc schemes fig. 4 demonstrates the general block diagram of svc with discrete tsr and tsc. the required susceptance b (the required reactive power) is converted to required inductive susceptance and required capacitive susceptance [4]. the required susceptance b is converted by the control system to specified step according to the characteristics of tsr or tsc. this control system is essential for the discrete control svc schemes and has the required blocks to diminish the effect of the discrete characteristics [4]. the inductive reactive power control is occurred because of the changing of the inductance value consistent with the certain step as in fig. 5. the varying of inductance value is simple because it should be occurred at zero crossing of current. for inductive current, the current zero crossing occurs at voltage phase angles of 90° and 270° i.e. at the moment of maximum instantaneous voltage. therefore, the maximum time delay in this case equals half cycle of the fundamental frequency. fig. 6 illustrates the block diagram of the tsr control system which contains three main components; the first is the block responsible for the synchronization between the tsr and the power grid; the second is the block responsible for the determination of the required susceptance and therefore the required operated switches; the third block is the firing circuit of thyristors. fig. 7 shows the waveforms of tsr current and voltage throughout the changing from one step to another. if new step is required by the control system, the implementation of the new step happens only at the zero-crossing of the current or the maximum amplitude of instantaneous voltage. conversely, the capacitive reactive power control of tsr is more complicated. there is no opportunity to switch on a switch in series with sinusoidal power supply and capacitor because there is a need to avoid the sudden change of voltage on the capacitor which leads to excessive level of current which may lead to the damage of the switch and the capacitor. in some cases, when the firing angle of the switch occurs at zero crossing of voltage, the high value of current change di/dt is very hazard for the switch. consequently, connection of small damping inductors in series with capacitor will reduce the high values of current or its rate of change [11]. for the new scheme of tsc, it is preferred to install a reactor l in series with power supply as in fig. 2. this reactor plays vital role not only for switching on the capacitors with lower transients but also for preparing the capacitors for the next step. the developed topology depends on the series connection and parallel connection. nonetheless, the connection of capacitors in series needs the discharging of all capacitors to prevent residual dc charge on capacitors during producing of capacitive current. consequently, the control required q% 0 10 20 30 40 50 60 70 80 90 100 d e v e lo p e d q % 0 10 20 30 40 50 60 70 80 90 100 ideal new topology fig. 3. the optimal characteristics of the new tsr and tsc characteristics algorithm of tsc should contain the discharging process earlier than the starting of the new step of operation. after the capacitors discharging, the new connection of the capacitors starts at the next voltage zero crossing. fig. 8 illustrates the flow chart of the control algorithm of tsc to start new step of operation. consistent with this algorithm, if there is requirement of new value of susceptance b, the discharging process should be activated before the implementation of new value of b. fig. 9 demonstrates the detailed block diagram of the tsc control system which contains the blocks required for the changing of the required susceptance and the discharging of the capacitors. fig. 10 (a) demonstrates the waveforms of the supply voltage and the current of the tsc in the transition between two steps where the tsc current equals zero during the discharging period except the impulse current due to capacitor discharge. fig. 10 (b) shows the voltage of certain capacitor in the transition process the voltage is kept constant till the moment to discharge it to zero voltage. the special discharging process for capacitors is applied within only one quarter of the fundamental period to keep the fast response of the overall system. if the capacitors are charged with positive or negative charge from the previous state of operation, this energy could be regenerated to the power supply depending on the reactor l. therefore, the inductance l of installed reactor and the equivalent capacitance ceq should provide resonance frequency higher than the fundamental frequency ω0 to end the discharging process in very short time. the resonance frequency recommended in this paper equals 24 times the fundamental frequency. so, the value inductance l can be determined with maximum capacitance ceq and resonance frequency of 24 ω0.   2 0 1 24 eq l c  (3) for delta connected tsc 200 kvar, 400 v-line voltage, the value of ceq equals 1315 μf and the value of inductance equals 13.2 μh. the control algorithm for the discharging can be explained according to the following example: assume all capacitors have remaining voltages vc0 illustrated in fig. 2 which equals 220√2 v. the goal is the regeneration of this tsr characteristics tsr power circuit required b* developed b certain step of operation fig. 5. flow chart for controlling thyristors switched reactors thyristor switched reactors reactors group switches group network voltage block of measurement of sinusoidal supply voltage firing circuits for thyristors synchronization block determination of required thyristors determination of required inductance fig. 6. block diagram of tsr control system time (s) 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 -400 -300 -200 -100 0 100 200 300 400 voltage tsr current requiring of new step implementing of the new step fig. 7. tsr voltage and current waveforms during transition from one step to another tsc characteristics tsc power circuit developed bc if new steps of operation activate the discharging process yes certain step of operation certain step of operation no required bc* fig. 8. flow chart of algorithms for controlling thyristors switched capacitors stored energy to the supply to reduce the capacitor voltage to zero level. the starting of discharging happens when the waveform of voltage passes through zero to the positive values as in fig. 11 (a) and the thyristors t2 for all switches are fired in this moment. fig. 12 illustrates the flow of current during discharging of capacitors where all switches except s7 operate in the direction enables to discharge the capacitors. in this moment, the voltage of capacitors is more than the instant value of supply voltage; so, all capacitors produce currents passes through reactor l and thyristors t2 in all switches except s7. this current has negative sign with respect to the power supply as in fig. 11 (b). at the moment when the capacitors voltage reaches zero fig. 11 (c) (0.35 μs) the thyristors t2 in the switches s1, s2, s3 and s4 are changed to off state (open circuit). but, the inductor current continuous passing in the same direction through thyristors t2 in switches s5, s6 and s7. the positive voltage of the supply forces the current to decay to zero as in fig. 11 (b) in the period 0.35 to 1.1 μs. then the capacitors are ready to the next connection through the power supply at the next voltage zero crossing. with the same strategy the discharging of capacitors can be applied if the capacitors are charged with negative sign. v. conclusion this paper has provided new topologies of thyristors switched svc which control the reactive power with zero d e te r m in a ti o n o f r e q u ir e d c a p a c it a n c e synchronization block firing circuits for thyristors determination of required thyristors determination of the required operated switches to obtain the required capacitance control of regenerative discharge of charged capacitors determination of voltage supply polarity block of measurement of sinusoidal supply voltage thyristors switched capacitors capacitors group network voltage switches group current limiting reactor fig. 9. block diagram of new topology of tsc control system time(s) 1.12 1.14 1.16 1.18 1.2 1.22 1.24 -500 0 500 tsc current (a) supply terminal voltage (v) time(s) 1.12 1.14 1.16 1.18 1.2 1.22 1.24 -500 0 500 tsc current (a) capacitor voltage (v) requiring of new step implementing of the new step starting of capacitors discharging (a) (b) fig. 10. the waveform of the applied voltage on tsc and its current and a voltage on one capacitor fig. 11. discharging process for capacitor of tsc with regeneration of energy (a) the supply voltage, (b) the supply current and (c) the capacitors voltage harmonic content. the power circuitries for inductive and capacitive schemes of the new svc are introduced. the principle of operation of new svc has been explained to illustrate the basics of the control system algorithm. the block diagram of the control system of the new svc system has been introduced with some modifications than the conventional svc. the control algorithm for tsr of the new svc schemes has been explained. the control algorithm for tsc is developed including the switching strategy required for the discharging of capacitors. acknowledgment the study was performed at stock company g. m. krzhizhanovsky power engineering institute within the framework of the project " development of a controlled source of reactive power with the absence of the higher-order harmonics currents during the regulation of the electrical energy and with the improved technical and economical indicators on the basis of the domestic component base of power electronics for automatic control of the voltage and the power flows in the electric power distribution networks of 6110 kv (rfmefi57917x0140)" with the financial support of the ministry of education and science of the russian federation. references [1] l. gyugyi, “power electronics in electric utilities: static var compensators”, proceedings of the ieee, vol: 76, issue: 4, april 1988 [2] sharma, p. r. kumar, a. kumar, n. optimal location for shunt connected facts devices in a series compensated long transmission line. // turk j elec engin. 15, 3(2007), pp. 321-328. [3] d. i. panfilov, a. e. elgebaly, m. g. astashev and alexander n. rozhkov, “new approach for thyristors switched capacitors design for static var compensator systems” 19th international conference of young specialists on micro/nanotechnologies and electron devices june 29 july 3, 2018 [4] d. i. panfilov, a. e. elgebaly and m. g. astashev, “design and optimization of new thyristors controlled reactors with zero harmonic content”, 18th international conference of young specialists on micro/nanotechnologies and electron devices june 29 july 3, 2017 [5] d. i. panfilov, a. e. elgebaly and m. g. astashev, “design and evaluation of control system for static var compensators with thyristors switched reactors” ieee 58th international scientific conference on power and electrical engineering of riga technical university (rtucon), riga, latvia, 12-13 october 2017 [6] d.i. panfilov, a. e. elgebaly, “modified thyristor controlled reactors for static var compensators” 2016 ieee 6 th international conference on power and energy (pecon 2016) , melaka, malaysia, november 2016 [7] d. i. panfilov, a. e. elgebaly and m. g. astashev, “topologies of thyristor controlled reactor with reduced current harmonic content for static var compensators” 17th eeeic concerence, milan, italy, 6-9 june 2017 [8] a. e. elgebaly, d. i. panfilov and m. g. astashev “comparative evaluation of binary and conventional static var compensators” mechanics, materials science & engineering journal (mmse), vol. 17, 2018 [9] d. i. panfilov, a. e. elgebaly and m. g. astashev, “design and assessment of static var compensator on railways power grid operation under normal and contingencies conditions”, 16th eeeic conference, florence, italy, 7-10 june 2016 [10] d. i. panfilov, a. e. elgebaly and m. g. astashev, “thyristors controlled reactors for reactive power control with zero harmonics content”, 17th ieee international conference on smart technologies ieee eurocon 2017, ohrid, macedonia, 6 8 july 2017 [11] n. garcia ; a. medina “fast periodic steady state solution of systems containing thyristor switched capacitors”, 2000 power engineering society summer meeting seattle, washington usa, 16 20 july 2000 [12] a.n. vasconcelos, et.al, “detailed modeling of an actual static var compensator for electromagnetic transient studies”, ieee transactions on power systems, vol. 7, no. 1, february 1992. s2 c4c2 s4 s7 c3 s1 c1 s3 s5 s6 v s is l t1 t2 t1 t2 t1 t2 t1 t2 t1 t2 t1 t2 t1 t2 + vc0 + vc 0 + vc 0 + vc0 fig. 12. discharging path for positively charged capacitors of tsc http://www.mmse.xyz/ http://ieeexplore.ieee.org/search/searchresult.jsp?searchwithin=%22authors%22:.qt.n.%20garcia.qt.&newsearch=true http://ieeexplore.ieee.org/search/searchresult.jsp?searchwithin=%22authors%22:.qt.a.%20medina.qt.&newsearch=true i. introduction ii. circuits topologies of improved tsc and tsr iii. characteristics optimization of the developed tsc and tsr topologies iv. algorithms of thyristors controlling in new svc schemes v. conclusion acknowledgment references edited_trevino.pdf 1 theoretical analysis of the efficiency of a v2g wireless charger for electric vehicles alicia triviño, jose m. gonzalez-gonzalez, and jose a. aguado abstract—v2g (vehicle-to-grid) technology will report important benefits for the operation and safety of the grid. in order to facilitate the expansion of the v2g technology in a future, it is recommended to offer the drivers with easy to use methods to charge and discharge their ev batteries. in this sense, wireless chargers are expected to play a relevant role in the future electrical networks as it reduces the users’ intervention. the development of this kind of system is still open to improve them in terms of their operation, their compliance and their control. an important issue for the evaluation of these systems is the efficiency, which measures the power losses occurring in the system. this paper addresses a deep study about the losses in a bidirectional wireless charger. then, it provides with a mathematical model to characterize them. this model is validated by means of experimental results conducted in a 3.7kw prototype. index terms—v2g, wireless charge, wireless discharge, losses, efficiency, electric vehicle, inductively-coupled power system, icpt, bidirectional. i. introduction e lectric vehicles (ev) represent a clear eco-friendly mobility solution. firstly, it is able reduce co2 emissions [1–3]. on the other hand, it also helps for the integration of renewable energy sources [3, 4]. despite these potential benefits, the proliferation of electric vehicles must be carefully controlled as a big number of them stands out for a considerable aggregated load to the grid. market-driven solutions [5] aims at prompting the charge of the vehicles in those times when it is more convenient for the grid, avoiding the periods with a high demand too [6]. in a v2g context, the vehicles operate in an active discharge mode that needs to be also controlled [6]. specifically, when recommended, the vehicles could decide to deliver energy to the grid from their batteries. if correctly coordinated, this operation leads to important advantages for the grid. in order to promote the controlled charge and discharge modes and to obtain important advantages, it is highly recommended to reduce the user’s intervention. this can be achieved with wireless chargers for electric vehicles (ev) [7]. wireless power transfer technology can be realized by different techniques. by now, the most popular one is the one based on a pair of loosely coupled coils operating under resonant conditions. this is known as inductively coupled power transfer or icpt [7]. in this technique, one coil is placed in the pavement (named the primary side) and the other in the chassis (known as the pickup or secondary side). a. triviño, j. m. gonzalez-gonzalez and j. a. aguado are with the electrical engineering department, universidad de málaga, málaga, spain (e-mail: atc@uma.es). the current through the primary coil induces a voltage in the secondary coil, which is used to charge the battery. the magnetic field involved in this charger is ranged in the 20 khz 100 khz interval [8]. to generate this high-frequency sinusoidal current, power converters need to be included in the charger. the non-idealities of the switches of the power converters and the parasitic resistances of the reactive components are responsible for the losses in the whole system, which degrades the charger efficiency. this paper addresses the proposal and verification of a model to characterize the efficiency of a bidirectional wireless charger. the main contribution of our work are the following ones: • in contrast to [9, 10], our work is focused on a bidirectional wireless charger operating at 85 khz and 3.7 kw. the change in the frequency and on the power requires the use of specific semiconductors. in particular, sic mosfets are used [11, 12]. the particularities of these devices need to be considered for the model as they clearly affect in the losses of the system and in the efficiency. • it extends the work in [13] by focusing on the efficiency and adding a comprehensive analysis of the experimental results, which is the basis for the computation of this parameter in a prototype. the rest of the paper is structured as follows. section ii reviews some related works about the study of losses in wireless chargers basing on icpt technology. section iii presents the icpt wireless charger topology used in this work. the theoretical model to compute the efficiency and the losses is presented in section iv. section v evaluates the methods basing on the real measurements performed in a 3.7-kw prototype. finally, section vi describes the main conclusions of this work. ii. related work the losses of an ev wireless charger are mainly due to the semiconductors employed in the power converters and the parasitic resistance of the real reactive elements. ideal semiconductors in the power converters are operated in such a way that no losses result. however, the real behavior leads to conductive and switching losses for these elements. conductive losses are produced because of the difference of the element from an ideal switch. to understand the losses by this event, we can rely on semiconductors models, which represent the device as an equivalent circuit with parasitic components such as resistances or capacitors. conductive transactions on environment and electrical engineering issn 2450-5730 vol 3, no 1 (2018) © alicia trivino, jose m. gonzalez-gonzalez, and jose a. aguado 2 losses occur because of the resistances. when the current flows through the resistances, losses occur. on the other hand, the capacitors prevents the semiconductors from having instantaneous commutations, which implies switching losses. as for the reactive components, the conductive losses are due to the resistance offered by the cables on which they are built. an equivalent model for these elements adds a series resistance to the reactive component. it is known as the esr (equivalent series resistance) and its value greatly depends on the material of the reactive components. for instance, litz wire minimizes the skin effect and, in turn, the esr [14]. the work in [9] makes a deep theoretical study about the switching losses of a wireless charger. the impact of them depends on the fact whether they are hard or soft. a similar study but in a low-power application is proposed in [10]. the prototype under evaluation uses a different power transfer technique working at 6.78 mhz and transferring 2 w. the work in [15] presents the model for the computation of the switching losses for mosfet in a unidirectional wireless charger. however, they do not rely on the equivalent model and they employ the parameters observed in the experimental results to derive this type of loss. the use of the parasitic capacitors, as we do in the present work, eases the estimation of the losses before the prototype is built. our present work uses the model presented in [9] to derive the efficiency of an ev bidirectional wireless charger. it represents an extension to the study presented in [13] as it makes a deep analysis of the experimental results. iii. icpt wireless charger for ev in this work, we will focus on icpt technology for charging and discharging the battery of an ev. the generic scheme of a unidirectional wireless charger is represented in figure 1. fig. 1. structure of a unidirectional wireless charger. the charger is connected to the grid, whose alternating current has a frequency of 50 hz or 60 hz depending on the region. nevertheless, these frequency values are not enough to transfer power inductively. consequently, it is necessary to convert this current to high frequency current, using power electronics for that. firstly, the alternating current is converted to direct current using a rectifier. both single-phase and threephase rectifiers can be used. despite its higher cost, the three phase rectifier has some advantages as for example lower losses, lower ripple factor and higher transformer utilization factor. the direct current is necessary to obtain high frequency current using an inverter. the generated high frequency current flows through the primary coil, creating a magnetic field which mainly depends on the coil structures. the magnetic field created by the primary coil induces an alternating voltage in the secondary coil, which is rectified again to provide direct current, which is appropriate for the battery. additionally, a compensation system composed of capacitors is introduced on both sides to enable the operation under resonant conditions. when implementing a bidirectional wireless charger, the power flow must be also allowed from the battery to the grid. as a consequence, the power converters should have a dual behavior to cope with the two senses of the power flows. this modification can be observed in figure 2. fig. 2. structure of a bidirectional wireless charger. in order to allow the power flow in both senses, the primary inverter has to be capable to work also as a rectifier, while the rectifiers have to perform the functions of an inverter. this issue is solved using bidirectional ac/dc converters. as in the unidirectional wireless charger, the connection to the grid can be done through both a single-phase and three-phase converter. the three-phase converter has benefits to the singlephase converter because using three phases allows to convert the same power with less current per phase, thereby reducing the losses on the components of the converter. iv. theoretical model for the efficiency the efficiency (η) of a wireless charger measures the relationship between the real power delivered to the load (pload) and the real power generated by the source (psource). so that: η = pload psource (1) at resonant operation and with ideal components, all the power generated by the source is delivered to the battery, that is η equals 1. however, due to conduction and switching losses (lcond and lsw respectively), this parameter decreases as follows: η = psource − lcond − lsw psource (2) where the calculation of lcond and lsw is presented in the following subsection. despite having similar structures in both sides, wireless chargers can use different components in the primary and the secondary sides. one important component of the charger 3 i1 c1 rl1 l1 c2 i2 l2 rl2 m coilsprimary side secondary side dc-link battery fig. 3. topology of the analyzed bidirectional wireless charger. which can differ are the coils because having a bigger coil on the primary side can help with misalignments. different coils also require different compensation systems. due to this asymmetrical structure of the bidirectional wireless charger, it is necessary to define two efficiencies, one for each sense of the power flow. thus, we have ηch to characterize the power flow from the grid to the battery and ηdis to define the charger performance in the opposite sense. we can reformulate eq. 2 as: ηch = pgrid − l ch cond − l ch sw pgrid (3) and ηdis = pgrid − l dis cond − l dis switch pgrid (4) where pgrid represents the real power generated by the grid and pbat is the real power delivered by the ev battery. in the next subsections, we formulate the equations to estimate both efficiencies for the bidirectional wireless charger topology presented in fig. 3. a. charge mode in this operation mode, the dc/ac power converter in the primary side acts an inverter whereas the ac/dc power converter in the pickup is a rectifier. the conduction losses are: lchcond = l ch con,inv + l ch con,rec + l ch coils + l ch match (5) where lchcon,inv are the conductions losses of the inverter, lchcon,rec are the conductions losses of the rectifier, l ch coils are the losses in the coils and lchmatch are the losses in the compensation system. the superscript ch indicates that these values correspond to the charging mode. in order to simplify the control, a full-bridge topology with a phase-shift control of a duty cycle equal to 50% is employed. in this scheme, there is only one transistor in each leg in conduction. thus, the conduction losses of the inverter are: lchcon,inv = 2 · rds · î 2 1 (6) being î1 the rms (root-mean squared) current in the primary side and rds the internal resistance drain-source of the mosfet in the inverter. for the conduction losses of the rectifier, we also rely on the equivalent model of the diodes. in a full-bridge rectifier, two diodes are simultaneously conducting, both of them provoking some losses as: lchcon,rec = 2 · rd · î 2 rec + 2 · vth · îrec (7) where rd represents the internal resistance of the diodes, vth their forward voltage and îrec the rms value of the current traversing these elements. in a series-series compensation topology as we are using in this work, irec equals i2, that is, the secondary current. the coils and the matching networks also incur in conduction losses to the system due to their parasitic resistances. specifically, the losses of the coils are estimated as: lcoils = rl1 · î 2 1 + rl2 · î 2 2 (8) where rl1 and rl2 are the resistances associated to the primary and secondary coil respectively. the currents injected to the coils are the same as those in the capacitors as a series-series compensation network is used. thus, the conduction losses in the matching networks are: lmatch = rc1 · î 2 1 + rc2 · î 2 2 (9) where rc1 and rc2 are the resistances associated to the primary and secondary capacitors respectively 4 concerning the switching losses, they are due to the inverter. theoretically, these losses can be estimated by the addition of multiple terms, being the one due to the output capacitance (coss) the most relevant (lcoss). thus, we simplify as: lchsw ∼= lcoss = 1 2 fscossv 2 ds (10) being fs the switching frequency and vds the voltage between the drain and the source. b. discharge mode in the discharge mode, the power flow from the battery to the grid. making use of the dual behavior of the power converters, in this operation mode, the power converter attached to the battery acts as an inverter while the power converter in the primary side works as a rectifier. in this way, the conduction losses can be expressed as: ldiscond = l dis con,inv + l dis con,rec + l dis coils + l dis match (11) where the superscript ch indicates these values correspond to the discharging mode. although the conduction losses of the coils (lcoils) and the compensation system (lmatch) can be calculated following the same equations of the charging mode (eq. 8 and 9), the conduction losses of the inverter and the rectifier differ in the current values because in this mode they are working on the secondary and the primary side respectively. theses losses can be computed as: ldiscon,inv = 2 · rds · î 2 2 (12) ldiscon,rec = 2 · rd · î 2 1 + 2 · vth · î1 (13) finally, the switching losses are estimated following the same equation of the charging mode (eq. 10). v. validation of the model: experimental results the developed theoretical model is verified with a 3.7-kw bidirectional prototype based on icpt technology. the coils are both square but with different sizes. the compensation networks are uniresonant and in series with the coils. as for the magnetic field, it is generated at 85 khz as recommended by sae tir j2954 [16]. cree c2m0080120d sic mosfets are the components of the power converters as they support the power demanded and the switching frequency. the main properties of the charger, including the values related to the non-idealities of the components, are summarized in table i. firstly, we derive the theoretical results for the losses and the efficiencies considering the model presented in the previous section. these results are exposed in table ii. the resistance of the coils has the highest impact in the efficiency of the system. the rest of the components of the system have similar losses with the exception of the switching losses of the inverter which, thanks to the use of sic mosfets, are negligible. due to security reasons with the operation of the battery, the power transferred in the discharge mode has been reduced. table i parameters of the icpt system. charger specifications tx-rx parameters (prototype values) output 3.7 kw 300v l1[mh] 240.5 fs [khz] 85 l2[mh] 230.6 coils geometry c1[nf ] 14.3 primary coil [m2] 0.75 x 0.75 c2[nf ] 15.6 secondary coil [m2] 0.5 x 0.5 rl1 [mω] 196 c2m0080120d mosfet rl2 [mω] 143 rd[mω] 40 rc1 [mω] 67 vth[v ] 0.98 rc2 [mω] 52 coss[pf ] 80 m[mh] 54.5 rds[mω] 80 k = m(l1l2) 1/2 0.231 table ii computed losses. charge mode discharge mode lcon,inv[w ] 25 lcon,inv[w ] 3.3 lsw,inv[w ] 1 lsw,inv[w ] 1 lcon,rec[w ] 34 lcon,rec[w ] 15 lcoils[w ] 64 lcoils[w ] 11 lmatch[w ] 23 lmatch[w ] 4 to validate the computed results, the losses are also carried out by obtaining the real power which flows for each component of the system. to get these values is necessary to analyze the oscilloscope captures both at the input and at the output of the components, with which the current and voltage are taken into account. figure 4 shows an oscilloscope capture of the current and the voltage measure at the output of the primary dc/ac converter in charge mode. as can be observed, voltage measurement consists on a square signal which can be used to computed the real power assuming a fundamental harmonic approximation. the difference between input and output real power of each component corresponds to the losses. the measurements of the electrical signals in the prototype leads to the values exposed in table iii. the following subsections presents the calculation methods to obtain the results from the experimental measurements. using these equations and the measures of table iii, the prototype losses has been computed to validate the model. these losses are shown in table iv. as can be observed, the use of the model based on the nonidealities leads to higher losses estimation in comparison with those derived from the analysis of the waveforms. in particular, for the charge mode, the inverter is assumed to loss 26 w whereas the waveform analysis states that these losses 5 fig. 4. voltage (channel 1) and current (channel 2) measurements at the output of the primary dc/ac converter. table iii electrical signals measured in the prototype. electrical signals charge mode discharge mode vinvinput [v ] 288 298 vinvoutput [v ] 290 293 iinvinput [a] 12.56 4.56 vinvinput [a] 13.78 5.14 vrecinput [v ] 285 247 vrecoutput [v ] 288 250 irecinput [a] 13.74 6.02 irecoutput [a] 12.16 5.30 are 20 w. concerning the rectifier operation, this difference is 11 w. for the discharge mode, there are also some deviations. specifically, the difference in the inverter is 1.3 w whereas the rectifier is associated to a deviation equals to 1 w. taking into account the total values, the differences are significant. this is due to the errors in the measurements, which impact on both the waveform analysis and on the model based on the nonidealities. a. charge mode concerning the output of the primary inverter, the voltage is a square-wave of vinvoutput amplitude while the current is inphase and sinusoidal (with a peak value equals to iinvoutput ). the shape and phase of the output current is the consequence of forcing the system to operate under resonant conditions. for the active power, we must extract the peak value of the fundamental harmonic of the voltage signal (v 1invoutput ) and operate as follows: p chinvoutput = v 1invoutput · iinvoutput 2 = 4 · vinvoutput · iinvoutput π · √ 2 (14) thus, the losses in the primary inverter (lchinv) are: lchinv = p ch invinput − p chinvoutput (15) table iv losses computed from the electrical signals measured in the prototype. charge mode discharge mode lchinv[w ] 20 l dis inv[w ] 3 lcoils + lmatch[w ] 73 lcoils + lmatch[w ] 17 lchrec[w ] 23 l dis rec[w ] 14 the rectifier input consists of a square-wave voltage (with a peak value of vrecinput ) and a sinusoidal current wave with a peak value equal to irecinput . applying the decomposition of harmonics, we can state that: p chrecinput = 4 · vrecinput · irecinput π · √ 2 (16) the output of the rectifier in conjunction of the low-pass filter results in two constant signals for voltage and current. the voltage equals to vrecoutput while the current is irecoutput . this lead to an output power computed as follows: p chrecoutput = vrecoutput · irecoutput (17) consequently, the losses in the controlled rectifier (lchrec) are: lchrec = p ch recinput − p chrecoutput (18) b. discharge mode in the discharge mode, the battery of the ev provides energy to the grid so that the power flow is reverse to the previous mode. this means that the power converter attached to the battery acts as an inverter while the power converter in the primary side works as a rectifier. the waveforms of the converters change according to their new functionality but it follows a form equivalent to the previous case. thus, these are the losses that are defined in a different way in comparison with the afore mentioned definitions. alternatively, lcon,inv, lsw,inv, lcon,rec, lcoils and lmatch can be computed as previously. in the discharge mode, for the inverter attached to the battery: p disinvoutput = v 1invoutput · iinvoutput 2 = 4 · vinvoutput · iinvoutput π · √ 2 (19) p disinvinput = vinvinput · iinvinput (20) on the other hand, the losses in the primary inverter (ldisinv) are: ldisinv = p dis invinput − p disinvoutput (21) the rectifier input, which is now in the primary side, corresponds with a square-wave voltage (with a peak value of vrecinput ) and a sinusoidal current wave with a peak value equal to irecinput . applying the decomposition of harmonics, we can assure that: 6 p disrecinput = 4 · vrecinput · irecinput π · √ 2 (22) the output of the rectifier is now connected to the grid. the voltage equals to vrecoutput while the current is irecoutput . both signals are constant as a low-pass filter is used. this implies that the output power is computed as follows: p disrecoutput = vrecoutput · irecoutput (23) as a consequence, the losses in the controlled rectifier for the discharge mode (ldisrec) are: ldisrec = p dis recinput − p disrecoutput (24) vi. conclusion this paper presents a model for the estimation of the charge and discharge efficiency in a bidirectional icpt wireless charger for ev. specifically, the model relies on the equivalent circuit of the semiconductors and of the reactive components to derive the conduction and the switching losses of the system. the model is contrasted with the experimental results obtained in a 3.7 kw wireless charger at 85 khz. there exist some differences among them, which are assumed to be due to measurement errors. as future work, we intend to model the switching losses when the resonant conditions do not hold because of misalignments between the two coils. references [1] l. canals casals, e. martinez-laserna, b. amante garcı́a, and n. nieto, “sustainability analysis of the electric vehicle use in europe for co2 emissions reduction,” journal of cleaner production, vol. 127, pp. 425– 437, 2016. [2] e. a. moreira falcão, a. c. rodrigues teixeira, and j. r. sodré, “analysis of co2emissions and techno-economic feasibility of an electric commercial vehicle,” applied energy, vol. 193, pp. 297–307, 2017. [3] a. triviño, m. longo, and f. foiadelli, “reduction of co2 emissions using res to recharge evs : the spanish case,” international conference on renewable energies and power quality (icrepq’17), vol. 1, no. 15, 2017. [4] k. e. forrest, b. tarroja, l. zhang, b. shaffer, and s. samuelsen, “charging a renewable future: the impact of electric vehicle charging intelligence on energy storage requirements to meet renewable portfolio standards,” journal of power sources, vol. 336, pp. 63–74, 2016. [5] m. a. lopez, s. de la torre, s. martin, and j. a. aguado, “demandside management in smart grid operation considering electric vehicles load shifting and vehicle-to-grid support,” international journal of electrical power and energy systems, vol. 64, pp. 689–698, 2015. [6] k. m. tan, v. k. ramachandaramurthy, and j. y. yong, “integration of electric vehicles in smart grid: a review on vehicle to grid technologies and optimization techniques,” renewable and sustainable energy reviews, vol. 53, pp. 720–732, 2016. [7] k. a. kalwar, m. aamir, and s. mekhilef, “inductively coupled power transfer (icpt) for electric vehicle charging a review,” renewable and sustainable energy reviews, vol. 47, pp. 462–475, 2015. [8] a. triviño-cabrera, j. aguado, and j. gonzález, “analytical characterisation of magnetic field generated by icpt wireless charger,” electronics letters, vol. 53, no. 13, pp. 871–873, jun 2017. [9] g. di capua, j. a. a. sanchez, a. t. cabrera, d. f. cabrera, n. femia, g. petrone, and g. spagnuolo, “a losses-based analysis for electric vehicle wireless chargers,” in 2015 international conference on synthesis, modeling, analysis and simulation methods and applications to circuit design (smacd). ieee, sep 2015, pp. 1–4. [10] g. di capua, n. femia, g. petrone, g. lisi, d. du, and r. subramonian, “power and efficiency analysis of high-frequency wireless power transfer systems,” international journal of electrical power & energy systems, vol. 84, pp. 124–134, jan 2017. [11] n. oswald, p. anthony, n. mcneill, and b. h. stark, “an experimental investigation of the tradeoff between switching losses and emi generation with hard-switched all-si, si-sic, and all-sic device combinations,” ieee transactions on power electronics, vol. 29, no. 5, pp. 2393–2407, may 2014. [12] r. a. wood and t. e. salem, “evaluation of a 1200-v, 800-a all-sic dual module,” ieee transactions on power electronics, vol. 26, no. 9, pp. 2504–2511, sep 2011. [13] a. trivino, j. m. gonzalez-gonzalez, and j. a. aguado, “evaluation of losses in a bidirectional wireless power transfer system for electric vehicles,” in 2017 ieee international conference on environment and electrical engineering and 2017 ieee industrial and commercial power systems europe (eeeic / i&cps europe). ieee, jun 2017, pp. 1–5. [14] q. deng, j. liu, d. czarkowski, m. k. kazimierczuk, m. bojarski, h. zhou, and w. hu, “frequency-dependent resistance of litzwire square solenoid coils and quality factor optimization for wireless power transfer,” ieee transactions on industrial electronics, vol. 63, no. 5, pp. 2825–2837, may 2016. [online]. available: http://ieeexplore.ieee.org/document/7383291/ [15] a. moradi, f. tahami, and m. a. ghazimoghadam, “wireless power transfer using selected harmonic resonance mode,” ieee transactions on transportation electrification, vol. 3, no. 2, pp. 508–519, 2017. [online]. available: http://ieeexplore.ieee.org/document/7903716/ [16] sae international, “wireless power transfer for light-duty plugin/electric vehicles and alignment methodology (sae tir j2954).” [online]. available: http://standards.sae.org/wip/j2954/  transactions on environment and electrical engineering issn 2450-5730 vol 1, no 3 (2016) © lorenzo capineri  abstract— the aim of this research project is the architecture and the design of an electronic system for controlling domestic tactile switches to be integrated into a home automation system based on the knx standard. all the steps that led to the fulfillment of the finished prototype are reported, from the study and design of the capacitive tactile sensors and the electronic control board according to the specifications imposed by knx standard. the touch event detection is reached as a trade-off with the footprint requirements of the switch. experimental results of the fabricated prototype are presented to demonstrate the feasibility of this device. index terms— industrial electronics, capacitive switch, home automation, knx standard i. introduction he home automation industry is going through a phase of strong growth overall in europe, mainly thanks to the knowledge gained by users on potential of this technological approach in terms of greater safety, comfort, energy savings and ease of use. even the building designers show a greater interest in home automation, as an added value of dwelling and contributing factor to the increase in competitiveness of the offer. in parallel with the advancement of home automation solutions in recent years the offer of advanced functionalities and increasing levels of integration is enhanced. to this is added the modern conception of logic in living styles requiring housing with greater flexibility to match the changing needs of users in connection with new technologies. the home automation market manifests a growing interest especially for use in new or renovated houses with particular interest to the issue of security and the energy saving. from the users’ side the main requirements for new home automation devices are: ease of use: you must ensure easily accessible information preferring intuitive interfaces and the use of touch-screen. solutions based on open standards: that there must be the possibility for the user to replace the entire system or change a service without requiring a redesign of the facilities. the reliability of continued operation: can be obtained by providing flexible and easily programmable systems to suit the needs of customization of each user and an efficient maintenance service. the purpose of this work is to describe the research and development phase of an electronic system for controlling domestic tactile switches to be integrated into a home automation system based on the knx standard. the device designed tries to respond to the major needs of end-users of a home automation system, preferring a tactile technology to the control elements and using knx it is at present the open standard for home automation widespread in europe. as for the use of capacitive touch acquisition of control, it should be noted that in recent years, this technology has matured enough to provide greater robustness, reliability and greater user satisfaction compared to traditional switches mechanical. the main advantages of the proposed solution are: greater durability and reliability due to the absence of moving parts, which is due to the mechanical switches of the device wear. moreover, the absence of openings they make it possible to use in harsh environments where dust and moisture could penetrate the device. a sleek design that allows switches to be integrated perfectly into the home. the multi-functionality: hardly feasible with mechanical switches, it becomes possible with tactile technology that can implement different functions starting from a single control element. feedback to the user: the apparent lack of tactile switches do not provide the user with clear evidence of the activation, is instead resolved by providing the ability to have different types of feedback as the lighting of a led, a sound or one of the key vibration and differentiate them according to the function performed. the work presented in this paper is organized into the following sections: 1. the study of the state of the tactile capacitive acquisition techniques and solutions used in home automation systems; 2. the design of the sensors, the electronic control board and design and realization of a tactile switches module with capacitive sensing method implemented with a microcontroller lorenzo capineri department of information engineering university of florence t the software according to the specifications; 4. description of the realization and implementation of the electronics and the firmware; 5. experimental results obtained with a full operational prototype. ii. study of tactile capacitive acquisition techniques a simplified model to understand the principle on which is based the capacitive acquisition considers the electrode as one of the faces of a capacitor, while the second is represented by the environment in which it is located (forming the parasitic capacitance c0) or, in addition, by another conductive object such as a finger (forming a ct capacitance). the overall capacitance is connected to a measuring circuit that periodically acquires the value of the electrode capacity. this model shown in figure 1 neglects the detrimental effects due to the interaction of the finger with the scattered fields generated by the coupling between the electrode and the underlying ground plane. to analyze how occurs the change in capacity from the physical point of view, we make reference to figure 2. the coupling between the copper electrode on the upper surface of the printed circuit board (pcb) and the underlying ground plane generates a parasitic capacitance, usually of the order of 10 pf to 300pf. the charge distributions on the two capacitor plates determine an electric field: most of the energy will be concentrated directly between the two conducting surfaces but a part protrudes in the external area and gives rise to the electric field lines denominated "fringing fields". according to the general characteristics explained above, it is important to indicate the main parameters sizing of the electrodes and the choices done in this project: • presence of a double interface: the sensors consist of a double interface, the first made by the copper layer on the surface of the touch board and the plastic cover, the second by the plastic cover and the outer cover made of steel. at these will go to addition, in case of touch, the contact surface between the metal cover external and finger. • material and thickness of casing: polycarbonate, 1 mm • geometry: rectangular with rounded corners • dimensions: ratio of coverage and thickness of the electrode side at least 1: 4 • distance between the electrodes: ratio between the distance and longer side of electrode 1: 8 and use of adjacent key suppression (aks) method available for microcontroller implementation. • inserting a ground plane: reduces sensitivity. a. techniques of capacitive touch sensing after understanding the physical principle underlying the change in capacity determined by the contact of the electrode fingers, it must explain how this change can be detected and converted into a useful signal. there are numerous methods to detect the capacity increase caused by the proximity or contact of a conductor from: variation of the resonance frequency, frequency modulation or amplitude, the measurement of the charge/discharge, delay measurement time in reaching a threshold or duty cycle [1][2][3]. most of these methods require analog circuits inheriting the related problems as crosstalk and sensitivity to noise [4], this may be preferable for a digital approach that guarantees smaller dimensions and lower power consumption than analog solutions. some of the most commonly used acquisition techniques are described in the following paragraphs, divided on the basis of the expected degree of integration. in fact we can distinguish two main categories of implementation: software-based solutions, using microcontrollers for general use, and specialized controllers for capacitive acquisition. of course are also possible other solutions such as custom microcontrollers for interfacing capacitive sensors whose architecture include modules dedicated to scanning arrays of capacitive sensors and are easy to implement. in tactile sensing the recent trend is to prefer the use of custom microcontrollers or otherwise use microcontrollers already tailored for capacitive acquisition. each class of solutions, however, has its strengths and weaknesses and then depending on the application requirements, the designer must select the best approach. in this section are compared the two approaches based on dedicated microcontrollers and the other based on the use of a fig. 1. simplified model of change of capacitance from the value c0 to c0+ct when a finger is touching the dielectric layer top surface. fig. 2. effects of electrical field fringes. general purpose microprocessor requiring the programming of software for touch detection. b. solutions based on the use of dedicated controllers several off the shelf solutions have been analyzed and compared for the design of the final prototype: 1) fujitsu fma1127 the fma1127 is a controller for capacitive sensors, which converts the capacitance generated by the interaction between the human body and the electrode only in digital data without any analog signal processing. moreover, this solution increases the flexibility of design, provides good performance in terms of power consumption and stability. in figure 3 is shown the acquisition scheme adopted for this case. this scheme has two r-c delay lines: one connected to a reference capacity cr and the other one connected to the "sensing" capacity cm, that is, the electrode, subjected to the same environmental conditions. both are controlled by the same clock (clk) signal, which frequency can be programmed up to 20 khz to ensure fast response times and no negative impact on the system's susceptibility to the emi. the leading front of the clock increases more slowly because the larger charging time due to increased capacitance in the case of the touch event. in figure 3 can be observed that clk2 rises more slowly than clk1, and then reaches a threshold value considered "high", with a delay. therefore the two logic signals clkr and clkm generated by clk1 and clk2 respectively, have two rising edges with different delays with respect to clk signal. these two rising edges, however, arrive in phase to the flip-flop thanks to the parasitic compensation logic obtained with dr digital values dr[0: n], the digital delay dm [0: n] values are adjusted by a feedback control with an up/down counter until clkr_d and clkm_d are in phase. the cm of the measured capacitance value can then be derived from the difference between the digital values dr [0: n] and dm [0: n]. the cm capacity, as well as the difference between the delays of the two clocks on the lines, will be proportional to the difference dr [0: n] dm [0: n]. in a capacitive sensor, the value of the cm capacity changes depending on whether or not a finger is present on the conductive plate, while the cr capacity has small variations due to environmental conditions (e.g. temperature changes). during a touch event, cm varies rapidly from a minimum value (cm1 = c0) to greater (cm2 = c0 + ct), and this change must cause a signal variation to reach a threshold. in the reference delay line the value of dr[0: n] is configured during the initialization phase to a value slightly lower than that of dm1[0: n], which indicates the absence of a touch. so the difference dr[0: n] dm1[0: n] will give a negative delay value. the presence of the finger is detected when the difference (dr[0: n] dm2[0: n]) becomes positive, in fact dm2[0: n], related to the case cm2 = c0 + ct, will be less than the digital value dr [0: n]. 2) qtouch® in at42qt1070 controller qtouch® is a technology for the tactile acquisition developed and patented by atmel that employs both controllers dedicated to this type of detection is the use, for these purposes, the processors of the at-mega family (see figure 4). in this work the acquisition technique was initially estimated by the evaluation-board qtouch xplained. this card supports four capacitive touch interfaces acquisition: slider, wheel and two touch-buttons and provide the connection to a dedicated card containing the atxmega128a1 processor. in figure 4 the two blue circles represent two planar electrodes, each of which is connected to the gpio pin of the processor by means of a number of capacities and resistances visible in the lower part of the pcb which allow the detection of the capacity variation electrode. in this study the technique was deepened with the at42qt1070 controller tests and reproducing the operating diagram with discrete components using nxp lpc 1227 processor, as will be detailed below. the qtouch® technique is based on charge transfer from the electrode capacity of acquisition to a higher value. in the first phase the cx electrode, having unknown capacitance because its value will be different depending on whether the touch occurred or not, is charged to a known potential. the charge is then transferred to a measurement capacitor cs (see figure 5), the latter has a fixed value and greater than about one order of magnitude compared to cx. fig. 3. diagram of the acquisition technique blocks used for fujitsu fma1127 and the timing signals for the touch event detection. fig. 4. tactile capacity sensor card qtouch xplained (atmel). we can assume that the sensing electrode capacity is of the order of pf and the design for cs value is fixed in the range 1222nf. the charge transfer cycle is repeated until the voltage at the terminals of cs reaches a threshold voltage of vih. the level of the signal of interest is given by the number of charge transfer cycles from the capacity cx to cs taken to reach this voltage. placing a finger on the touch surface introduces an external capacity that increases the amount of charge transferred in each cycle (because it increases cx), thus reducing the total number of cycles necessary for cs to reach the threshold voltage. when the number of cycles falls below the aforementioned threshold, then the sensor is again placed in the touch detection phase. cx and cs are currently connected with a certain number of resistors rs in series. some of these series resistors are inserted in this design to enhance the performance of the system to emi and esd, the number and the value of these resistors must be chosen on the basis of the cx value, a total value of resistance is usually considered appropriate 10kω; the presence of additional resistance is due to the resistivity of the tracks that connect the capacitive sensor and to that of the electrode constituent material itself. the rc time constants that are determined tend to slow down the acquisition process, therefore it is important to measure the settling time relative to the individual pulses of the charge transfer in order to have fronts of the more rectangular as possible while ensuring reliable detection. a simple test may be performed with the oscilloscope by placing the tip of the probe on a coin resting on the surface of the sensor cover. c. experimental evaluation of two solutions for capacitive tactile switches after the general analysis reported in the previous sections, two possible solutions have been evaluated with experimental board and switches prototypes and their descriptions as follow. 1) microprocessor and custom software programming this solution was carried out in microprocessor and software configuration, employing at first the electrodes of the evaluation-board qtouch xplained atmel and subsequently two copper electrodes of the same type as the final touch card, realized on the surface of the pcb and having a central hole for the signaling led, with area1.6 cm x 1.5 cm. the processor is mounted on the nxp lpc1227 evaluation board iar kickstart kit. the software was developed in the environment iar embedded workbench. with this setup have been measured the duration of the charge transfer cycle for the touch and no-touch cases, that are 0,881ms and 1,784 ms respectively. an illustration of the waveforms recorded for the touch case is reported in figure 6. 2) atmel custom microcontroller at42qt1070 the second solution was based on the at42qt1070 controller atmel, configured comms mode and employing six channels of detection. the unused channels can be disabled by setting to zero the averaging factor parameter. the schematic of the electronic circuit is shown in figure 7. fig. 6. charge transfer cycle of 0,881 ms for the touch case, where is assumed for cx=5-10 pf and cs=12nf. fig. 7. tactile interface solution with dedicated microcontroller and two tactile switch keys (k0 and k1). fig. 5. illustration of atmel solution for capacitive acquisition based on the charge transfer. electrode capacitance is cx.. the cpu of this lpc1227 iar kickstart kit, communicates with the microcontroller via i2c line. the event of variation of the state of one of the buttons involves the transition of the change line from logic high to a low level that is signaled to the microprocessor via an interrupt allowing the savings of otherwise necessary resources for the periodic reading of the registers of the controller. with reference to figure 8; we can observe that the reading of the registers occurs only to the interruption, after which the change line returns high again. the change line (red color) goes from low level to high following the reading of first register, blue and yellow lines are reported respectively for the i2c clock and data lines and show the i2c standard reading protocol. the outcome of the comparison of these solutions, led to prefer the use of a custom microcontroller for the following reasons: • simplified management of multiple electrodes, • touch event capture rate • increased reliability of detection • saving microprocessor resources • fewer gpio pins and links • greater flexibility of use iii. electronic hardware design and implementation the system design consists of two main logical sections that correspond to implementation with two pcbs: the touch board realizes the acquisition of the touch (see figure 9); the control card implements the control section and the interface with the knx bus and the external expansion. the control board is composed of three main sections: the processing consisting of the microprocessor which communicates with the acquisition section via the i2c line and the mezzanine type connector (see figure 10). the system communicates with the knx bus through ncn5120 transceiver. besides allowing interfacing with the knx bus system on twisted pair networks, the device integrates two programmable dc/dc converters that allow to change the input voltage available from the knx bus line, to the voltage levels needed to power the integrated circuit. the extension section shown in figure 10 has been introduced to interface the cpu with other external devices, for example temperature sensors. the i2c line buffer is useful to perform a level shift for the communication with devices powered at different voltage levels. iv. software design and implementation in this section is briefly reported the software design. system management software covers both the acquisition of the user commands by the touch controller is communicating with the knx bus. to minimize the software development time has fig. 10. block scheme of the designed control board fig. 9. block scheme of the designed touch detection board for up to 6 capacitive tactile sensors. on the bottom right the prototype with 4 electrodes with the front metal panel and the pcb for reference electrodes fig. 11. prototype of the control board fig. 8. signals on lines sda, scl and change in case touch on key 1 been decided to use communication stack sw already certified for knx bus. the communication stack assigns different functions to the tactile switches, for example typical configurable functions required by domotic houses are: • remote relays • switch on / off • lighting control • window with sliding curtains v. prototype realization and firmware testing the prototypes of the control board and the integrated touch board with capacitive touch sensors have been realized also considering the possible problems arising from the possible certifications emi. after installing the firmware on the control card (see figure 11) through the uart port of the microprocessor and the pc on which is installed the boards of the programming tool, a series of electrical tests on the prototype were carried out. the complete device was assembled by using a 3d printing for the plastic case of the electronics and sealed connectors and the 4 tactile metal switches (see figure 12). in figure 13 is shown the final set-up with the power supply, the four tactile switches board with led signaling, the knx interface and electromechanical actuator (relay). a. power consumption analysis. a preliminary estimation of the power consumption of the prototype was done during the electronic design phase. the following list reports the estimate of power consumption of the main electronic components of the system: • microprocessor: 13mw (operating freq. 4,9152mhz, configuration active mode, peripherals all active) • touch controller : 3mw (parameter lp mode = 0) • transceiver: 150 mw (normal operating mode, no load, both dc/dc activated) • 6 leds: 79mw adding up all these value the total is ≈ 245 mw. the measurements on the final prototype results in: • current = 3.8ma @30v for knx bus: 114mw • 6 leds on: 79mw the total power consumption for these two components is 193 mw that is in good agreement with the preliminary estimate and it is acceptable for a domotic plant. vi. preliminary certification tests in order to submit a product on the european market it is necessary to verify compliance with the applicable eu directives, in our case the rohs (restriction of hazardous substances) and emc (electro magnetic compatibility). evidence of radiated and conducted emission has been measured at a test laboratory. the product has passed all tests except emc immunity to conducted interference, therefore, is necessary to improve the design of the prototype. for carrying out immunity tests it was necessary to study a mechanical equipment that can reproduce the touch: it is proposed a piston controlled via pc actuated by a magnetic coil that was moved at regular intervals for pressing one of the buttons, the plunger tip is aluminum covered and connected to a battery 1.5v to simulate the touch (see figure 14). the knx certification covers both the hardware part of the product software, checking the requirements imposed by the standard; choosing the hardware interface (ncn5120) and the communication stack already knx certified simplifies the procedure by reducing time and costs for the obtaining of the brand. fig. 12. final prototype assembly with the plastic cases developed with a 3d printer. on the right the four tactile metal switches. fig. 13. final demonstration system. fig. 14. electromagnetic actuation system for mimicking the touch on the metal switch with an aluminum rounded tip connected to a 1.5 v battery for the electrostatic stimulus. the assembly was installed in the anechoic chamber during emc tests. vii. conclusions the study of the architecture of an integrated system with tactile capacitive switches and a knx interface is presented with explanation of the design criteria. after the fabrication of a prototype with 4 tactile capacitive switches, a series of tests have been carried out to verify the correct operation of the of capacitive tactile signal acquisition which shown an excellent sensitivity and a good knx communication reliability. the total power consumption of the device is about 245 mw that is rather low for a domotic plan device. foreseen the industrialization of the prototype, preliminary tests have been developed according to emi and esd standards. this new device is useful for domotic plants to save energy consumption thanks to the versatile use of tactile switches function with limited additional cost for installation due to simple cabled connection of knx. acknowledgment the author wishes to acknowledge the contribution of atmel university program with industrial electronics laboratory at pin (prato, italy). references [1] a. ożadowicz, “communication reliability in the intelligent building systems, “ white paper, http://www.knx.org/fileadmin/downloads [2] atmel – at42qt1070 seven channel qtouch® touch sensor ic datasheet. http://www.atmel.com [3] m. lee, “the art of capacitive touch sensing” by cypress , http://www.cypress.com/file/72881/download [4] s. kim, w. choi, w. rim, y. chun “a highly sensitive capacitive touch sensor integrated on a thin-film-encapsulated active-matrix oled for ultrathin displays,” ieee transactions on electron devices, vol. 58 , pp. 3609 3615, aug. 2011 [5] directive 2011/65/eu of the european parliament and of the council of 8 june 2011 on the restriction of the use of certain hazardous substances in electrical and electronic equipment text with eea relevance [6] directive 2004/108/ec of the european parliament and of the council relating to electromagnetic compatibility. [7] cei en 50491-5-2 standard. [8] cei en 61000-6-3 standard. lorenzo capineri (m’83–sm’07) was born in florence, italy, in 1962. he received the m.s. degree in electronic engineering, in 1988, the doctorate degree in nondestructive testing, in 1993, and post-doctorate in advanced processing method for ground penetrating radar systems from the university of florence, in 1994.in 1995, he became an associate researcher and an associate professor of electronics with the department of information engineering (formerly department of electronics and telecommunications) with the university of florence, in 2004. he has worked on several research projects in collaboration with national and international companies, the italian research council (cnr), the italian space agency (asi), and the european space agency (esa), aea technology and ukaea, england, international science and technology centre (istc), moscow, russia, thales alenia space italia (tasi), texas instruments, usa, general electric (uk) , joint research centre (european commission), and nato. he has coauthored six italian patents, three book chapters, and around 200 scientific and technical papers. his research interests include the design of ultrasonic guided waves devices, buried objects detection with seismo-acoustic methods, and holographic radar. http://www.atmel.com/ http://www.cypress.com/file/72881/download proposal of heuristic regression method applied in descriptive data analysis: case studies proposal of heuristic regression method applied in descriptive data analysis: case studies flávio a. gomes, alfredo de o. assis, márcio r. da c. reis, viviane m. gomes, sóstenes g. m. oliveira, wanderson r. h. de araujo, wesley p. calixto abstract—the purpose of this paper is to use the hybridized optimization method in order to find mathematical structures for analysis of experimental data. the heuristic optimization method will be hybridized with deterministic optimization method in order to that structures found require not knowledge about data generated experimentally. five case studies are proposed and discussed to validate the results. the proposed method has viable solution for the analysis of experimental data and extrapolation, with mathematical expression reduced. index terms regression, heuristic, modeling, optimization. i. introduction this paper is an extended version of our paper published in 2016 ieee 16th international conference on environment and electrical engineering [1]. traditionally, researches show the need to express the variable behavior through functions that represent experimental data. in several areas, regression methods are used to establish the relationship between variables, such as in the image processing [2], analysis of concrete structures [3] [4], extraction of tone of voice [5], health area [6] and waste flow forecasting [7]. to [8], the regression analysis consists in the study of the dependence between variables, verifying the relationship of the explanatory variables towards the dependent variable to perform forecasts and previews. this study is necessary due the existent lack of knowledge of the algebraic expression that rules the system being analyzed. the absence of the function that describes the behavior of the system implies in simulations or experiments performing in order to define the outputs, every time the inputs are changed. several times, this requires time and effort, which can make the process of study the system unpractical. the experiments (real or simulated) provide, as output, discrete data, however, in most cases, there is needed a function that describes the data in a continuous way [9]. once the function that defines the system is found, many analyses can be performed, such as data prediction, which tries to obtain an output for a certain correspondent input beyond the predefined interval [10]. in case of forecasting of natural resources demand, the efficient use can be obtained based on the performed predictions. in many situations, even simulations take a considerable amount of time, making the system analysis process difficult. in order to solve this problem, we use regression to replace part of the system by an expression that represents it, decreasing the simulation time. in [1], a regression of collected data on a test bench of controlled rectifiers was performed. regression methods use techniques that seek flexibility and predictive capacity. many studies base themselves on polynomials and trigonometric functions for approximating data. however, regressions by hybrid functions, polynomial and trigonometric, present themselves more representative that each of them apart, overcoming limitations as the periodicity for polynomial or prediction for trigonometric series [11]. other methods are used for prediction and curve fitting, such as artificial neural networks in [3] and [4], which got better results that quadratic regressions and additives models in [7], which are compared to cubic smoothing splines. researches about regression seek effective parametrization methods, in order to improve the curve fitting. in [12] is used darwinian particle swarm optimization, p-spline method in [13], regularized algorithm of levinson-galerkin in [6], the least squares method to parametrize trigonometric series in [5] and in [14] has the solving of compound optimization criterion through weighted polynomial regression models. the purpose of this work is to present a methodology to determine mathematical expressions that represent the systems with the least number of possible terms. the main contribution is to reduce the edge effect due to the reduced number of terms. besides that, it contributes to the recognition of systems from the experimental data and also in assertive extrapolation at considerable intervals. the proposed methodology is based on the generalization of the power and trigonometric series and the application of optimization methods. section ii presents the theoretical background, section iii brings the proposed methodology and the results achieved are presented in section iv. ii. background according to [15], a bounded-input, bounded-output system (bibo) is stable when it is limited in respect of the space’s norm in which it is defined (l2, l∞). using the space: l2(ω) = {f(t) | ||f(t)||2 < ∞}, (1) the norm of (1) is defined by ||f(t)||2 = √∫ ω |f(t)|2dt, where ω is a subinterval in the real numbers and f(t) is a square-integrable function in ω. by analyzing the experimental data fex(x) of a bibo system, we have according to [16] that the collected data are represented by: transactions on environment and electrical engineering issn 2450-5730 vol 2, no 2 (2017) c⃝ flávio a. gomes, alfredo de o. assis, márcio r. da c. reis, viviane m. gomes, sóstenes g. m. oliveira, wanderson r. h. de araujo, wesley p. calixto fex(x) = fop(x) + ϵ, (2) since fop(x) represents the regression and ϵ is the random additive error of the process that does not depend on "x" and satisfies the homoscedasticity criterion, which is, that the variance of ϵ is constant. in this sense it is said that fop(x) is the regression that represents the system if the mean square error (mse) is as minimal as possible. therefore, the following optimization problem is generated: fop(x) = min x∈ω {||ϵ||22}, (3) where fop(x) depends on the used base for data interpolation. for the representation of these events, there is a wide collection of interpolation and extrapolation theories, being the polynomial approximation of weierstrass the main interpolation theorem. in this, it is shown that in the space of the continuous functions c[a,b] ⊂ l2[a, b] any function f ∈ c[a,b], where a, b ∈ r, can be approximated by a polynomial function [17]. extending its definition to the space of the analytic functions f ∈ c(−∞,∞), any function can be expressed as a power series. the standard methods vary from polynomial to trigonometric representations, using the base β1 for the power series or polynomial, given by (4), and the base β2 for the trigonometric series, given by (5). β1 = {1, x, · · · , xn, · · · } (4) and β2 = {1, sin( πx p ), sin( 2 · πx p ), · · · sin( nπx p ), · · · cos( πx p ), cos( 2 · πx p ), · · · cos( nπx p ) · · · } (5) the obtained approximations verify trends and represent data by means of functions [18]. thus, the regression methods are chosen depending on the characteristics of the problem. the bases β1 and β2 have properties of representation in the space of continuous functions in the interval [a, b]. when there is some kind of frequent oscillation, the base β1 is insufficient to extrapolate the polynomial regression interval, since to represent the trigonometric frequencies, there is the need to transform the polynomial regression into a series. however, the extrapolation problem is also present in the base β2, since it has limitations for data prediction for non-periodic functions [11] [5]. iii. metodology the proposed methodology will use hybridized optimization method (heuristic and deterministic) to determine parameters of predefined structures. based on experimental data, the optimization process will return the mathematical expression that will represent the dynamics of the system, as fig. 1. figure 1. flow of optimization process. these structures, based on polynomial, trigonometric, and exponential functions, enable to represent a significant amount of curves. regression will be performed by comparing the curve defined by the experimental data fex with the curve generated by structures, called optimized curve fop. structures that generalize the power and trigonometric series given by fop1 , fop2 and fop3 will be proposed in order to meet the different curve profiles. these structures are presented in expressions (6), (7) and (8), respectively. fop1 = a0 + n∑ i=1 ai · xbi (6) fop2 = a0 + n∑ i=1 ai · xbi · cos(ci · x + di) (7) fop3 = a0 + n∑ i=1 ai · xbi · cos(ci · x + di) · exp(ei · x) (8) where: a0, ai, bi, ci, di, ei ∈ r. unlike other methods [11] [14], the parameters of fop will assume values belonging to the set of real numbers. therefore, polynomials of the β1 base from the power series will be generalized to rational functions, well as trigonometric functions with fixed frequencies of the β2 base will be generalized to any real frequency. thus, it will be possible to express experimental data with smaller structures, compared to other regression methods, maintaining the power of prediction. based on the characteristics of experimental curve fex, the proposed methodology will select the structure that have greater proximity between the optimized curve fop and the experimental fex. thus, the optimization process will be applied following the expression (3), but due to the fact of working with discrete signals of finite duration in the optimization process, the calculus of approximation error or evaluation function faval will be given by: faval = √ n∑ i=1 (fexi − fopi)2 (9) where: n will be the number of fex points. before performing the regression, data set will be processed in order to select the characteristic intervals ik to assist in the optimization process, that will express the orderly domain j of the fex curve in j = i1 ∪ i2 ∪ · · · ∪ ik (10) where: k will be the number of intervals. the first regression interval will be the one that contains the initial point of fex curve. the method will be applied successively by the union of subsequent intervals given by expression (10). in order to define the intervals, experimental curve will be divided into parts, based on inflection points and variation at the ordinates axis. the inflection point is the main factor for choosing the structure and also the optimization method. this occurs because this concept is related to the change in the function’s variation rate, being characterized by the point at which the derivative of the function changes from increasing to decreasing and vice versa. this feature influences both at the choice of structure and the improvement of the optimization process. due to the fact that structures with several inflection points tend to be more oscillatory, this parameter directly influences the choice of structure that best fits the data. if we analyze the optimization aspect, by dividing the interval based on inflection points reduces the possibility of stopping the process in some local optimum point. in this sense, the way of choosing the structures from the simple characteristic of the experimental data is defined. the amount of inflection points will be the base parameter to define those intervals. if there are until 2 inflection points, it will mean that data set have no oscillatory characteristic. therefore, data set will be divided into 10 equal parts and the intervals will be chosen based on variation at ordinate axis on these parts. the highest variation will be chosen as reference and the set of intervals (j) of (10) will be compound by only those that will achieve variation higher than 30% in relation to the chosen reference. if there are 3 or more inflection points, it will mean that data set presents sinuosity and its analysis will be based on these oscillations. thus, the highest variation at ordinates axis for all set will be chosen as reference. the subinterval between first 3 inflection points will be chosen to check the higher variation at ordinates axis present in this subset. if this variation exceeds 5% of reference variation, then this subinterval will be selected as the set of interval (j) of (10) for analysis in the optimization process. if this variation does not overcome that percentage, the subinterval will be grouped with other more relevant. the following inflection points will continue being analyzed in search of variations that meet this restriction. these intervals will be passed for the optimization routine that hybridizes the heuristic methods, genetic algorithm, and deterministic, nelder-mead, in order to find the optimized parameters [19]. at the end, the result will be the values of structures parameters proposed and their respective evaluation functions faval of data set. the best result will selected and the parameter values will be replaced at the corresponding structure with the view to mount the function that describes the set of experimental data. iv. results in order to generate the set of experimental data, known and used functions have been used to evaluate regression processes in mathematics and statistics. these functions do not represent physical systems and still present problems of mapping by both interpolating polynomials and extrapolations. these functions were used as case studies as well as data collected from a test bench of controlled rectifiers. this choice was done due to: i) the possibility to perform extrapolation of original set, ii) the approximation error with the results obtained at the initial simulation can be measured, and iii) the success of optimization process can be verified. a. case study 1 the generating function of experimental data chosen for this first case study was given by: fex = 1 1 + x2 (11) this function was chosen because of presenting oscillation problem near the edges of interval analyzed using polynomial interpolation with polynomials of high order. this problem is known as runge phenomenon like cited in [20]. in the expression (11), x assumes 1000 values in the range 1 ≤ x ≤ 100. the smallest error was got by the structure that contains only polynomials derived from (6) and the eleven terms of final expression was given by: fop = −6.08 · 10−5 + 1.30 · x−2.07 + 5.62 · 10−5 · x−1.52 − 0.81 · x−2.89 − 9.54 · 10−4 · x−0.77 + 6.21 · 10−4 · x−0.40. (12) fig. 2 illustrates experimental and optimized curves obtained with faval = 1.25 · 10−2. in the same figure there is a cut at the point 75 showing the difference between both curves with instantaneous error of about 10−4. figure 2. case study 1. b. case study 2 for the second case study, the generating function of the chosen experimental data was given by: fex = sin(2 · x + 3) · exp(−0.5 · x) (13) this function was chosen because of presenting a difficult behavior to be mapped by the structures (6) and (7). it presents also different oscillations throughout data set analyzed. in the expression (13), x assumes 1000 values in the range 0 ≤ x ≤ 40. the smallest error was got by the most complete structure that contains polynomials, cosine, and natural exponential derived from (8). the eleven terms found of final expression was given by: fop = 2.18 · 10−9 − 1.00 · x4.27·10 −7 · cos(1.99 · x − 1.71) · exp(−0.50 · x) − 4.61 · 10−12 · x1.12 · cos(−0.39 · x + 1.48) · exp(0.14 · x). (14) fig. 3 illustrates experimental and optimized curves obtained with faval = 0.14. in the same figure there is a cut at the point 30 showing the difference between both curves with instantaneous error of about 10−8. figure 3. case study 2. c. case study 3 the chosen generating function of the experimental data for this third case study was given by: fex = x + 1 tan(x) (15) this function was chosen because it presents output data with negative values, increasing oscillation and also in order to compare with polynomial interpolation methods. in (15), x assumes 20 values in the interval 1 ≤ x ≤ 20. the smallest error was obtained by the structure that has polynomials and cosines (7) and the 25 terms of the final expression was given by (16). fop = 19.16 + 44.95 · x−0.11 · cos(6.31 · x + 10.42) + 11.32 · x−0.79 · cos(2.15 · x + 3.10) + 2.96 · 10−6 · x4.73 · cos(2.17 · x + 8.90) − 5.85 · 10−6 · x3.72 · cos(4.31 · 10−3 · x − 3.51 · 10−3) + 6.47 · 10−3 · x1.90 · cos(−0.22 · x − 1.22) − 5.19 · 10−5 · x2.97 · cos(0.71 · x + 19.87). (16) fig. 4 illustrates the experimental and optimized curves obtained with faval = 1.97 · 10−1. within the same figure, there is a cut at the point x = 3, which illustrates the difference between both curves, with the order of the distance between them of approximately 10−2. figure 4. case study 3. polynomial interpolations were also performed to the same generating function in (15) in order to compare the proposed method and this technique of curve fitting. two polynomials were found, one being 20 degree in (17) and the other nine degree in (18). fpol20 = 3.01 · 10−13 · x19 − 6.29 · 10−11 · x18 + 6.10 · 10−9 · x17 − 3.65 · 10−7 · x16 + 1.51 · 10−5 · x15 − 4.57 · 10−4 · x14 + 1.05 · 10−2 · x13 − 1.87 · 10−1 · x12 + 2.61 · x11 − 28.80 · x10 + 2.52 · 102 · x9 − 1.74 · 103 · x8 + 9.43 · 103 · x7 − 3.96 · 104 · x6 + 1.27 · 105 · x5 − 3.01 · 105 · x4 + 5.07 · 105 · x3 − 5.67 · 105 · x2 + 3.73 · 105 · x − 1.06 · 105. (17) fpol9 = −2.09 · 10−7 · x9 + 1.84 · 10−5 · x8 − 6.80 · 10−4 · x7 + 1.38 · 10−2 · x6 − 1.69 · 10−1 · x5 + 1.30 · x4 − 6.25 · x3 + 18.52 · x2 − 29.4 · x + 18.01. (18) fig. 5 illustrates the experimental and optimized curves by the proposed method and by the polynomials in (17) and (18). the approximation error of the proposed method was faval = 1.97 · 10−1, whereas using the polynomial of 20 degree the error was faval = 2.03·101 and the polynomial of nine degree with error of faval = 4.67 · 102. figure 5. proposed method and polynomial interpolation comparison. d. case study 4 in this case study, the errors of extrapolations made for the previous case studies were calculated in order to verify the efficiency of the proposed method. in addition to reduction of terms of the expressions found, the extrapolations showed that the curve fitting captured the essence of the systems studied. the case study of section iv-a was extrapolated until point 300 in order to show the curve fitting after the original interval. fig. 6 illustrates the experimental and optimized curves. the measured error for the new interval was faval = 1.13 · 10−2 and within the same fig. 6 there is a cut at the point x = 280, which illustrates the difference between both curves with the order of the distance between them being approximately 10−5. figure 6. extrapolation of the case study 1. for the case study of section iv-b, the extrapolation was performed both before and after the initial interval. in fig. 7, the explanatory variable x takes on values in the new interval −15 ≤ x ≤ 60 and again, it can be noticed that (14) follows the behavior of the experimental data curve. the measured error for the new interval was faval = 2.36 · 10−2 and within the same fig. 7, there is a cut close to the point x = −11.84, which illustrates the difference between the two curves, being the order of distance between them approximately 10−3. figure 7. extrapolation of the case study 2. for the case study of section iv-c, the extrapolation was performed a little after the initial interval, since the approximation error of the curves by the methods becomes difficult to be perceived graphically. the nine degree polynomial in (18) was unable to adjust the curve in the original interval, remaining in the extrapolation process. the 20 degree polynomial in (17) obtained a suitable approximation in the analyzed interval and diverged abruptly when the extrapolation occurred shortly after the original interval due to the edge effect or runge’s phenomenon [20] which is noticed in polynomial interpolations. in fig. 8, there are presented the experimental and optimized curves by the proposed method and by the interpolating polynomials. the explanatory variable x assumes values in the new range 5 ≤ x ≤ 21 and again, it can be noted that (16) follows the behavior of the experimental data, whereas the interpolating polynomials lose their ability of approaching. for the new interval, the measured errors by using the proposed method and (17) and (18) were faval = 7.27 ·10−1, faval = 6.68 · 102 and faval = 6.41 · 102, respectively. figure 8. extrapolation of the case study 3. e. case study 5 at the fifth case study were analyzed data collected at a test bench for studies of controlled rectifiers. these rectifiers provide dc voltage of variable output as from a fixed ac voltage. due to its ability to provide dc voltage continuously variable, the controlled rectifiers revolutionized the modern industrial control equipments. this converter was shown in fig. 9. figure 9. power converter circuit with rl load. in order to obtain the instantaneous value of voltage controlled output vo, the literature has the solutions given by (19) according to [21]. vo = ⎧⎪⎨ ⎪⎩ β √ 2 vab if ωt ≤ π6 β sin ( ωt + π 6 ) if π 6 + α ≤ ωt ≤ π 2 + α, β sin ωt′ if π 3 + α ≤ ωt′ ≤ 2π 3 + α, (19) where: ωt′ = ωt + π 6 and vab is the voltage (effective) of input line and β is the extinction angle of electric current described in [22]. a test bench has been developed for obtaining experimental data of the converter output voltage and the firing angles of keys. the collected data set was interpolated in order to also contain 1000 values, and then was applied the proposed method to obtain analytical expression that represent the voltage as a function just of the firing angle α. the smallest error was obtained by the structure of polynomials and cosines derived from (7) and the 21 terms of found expression was given by (20): fop = 263 + 40.9 · x0.98 · cos(3.78 · 10−4 · x + 1.57) + 3.67 · x0.16 · cos(8.20 · 10−2 · x + 3.01) − 2.84 · 10−4 · x2.62 · cos(−3.96 · 10−2 · x + 3.67) − 0.15 · x0.93 · cos(0.10 · x − 61.3) − 0.46 · x3.57·10 −5 · cos(0.22 · x + 3.92 · 10−2). (20) fig. 10 presents the characteristic experimental curve of voltage of converter controlled three phase operating with load rl (resistor-inductor) and the optimized curve obtained. the approximation error found was faval = 37.6. the set of terms was analysed to identify the importance of each of them in the composition of encountered error. it was noticed that removing the last term in expression (20) the new value was faval = 42.7, that is, with 17 terms it still maintain an acceptable approximation error. v. conclusion this work presented the hybrid optimization method to be applied in the development of descriptive analysis data figure 10. case study 5. structure. the study results indicate that the proposed method is able to formulate mathematical expressions, in the form of regression, allowing to explore the relationship between the dependent and independent or explanatory variables. the proposal finds values in the set of real numbers for the coefficients, exponents and frequency of structures that generalize the power and trigonometric series, in an attempt to minimize errors. this proposed method is able to find a continuous function expression that represents a set of experimental data described by a discrete function expression. another advantage is the extrapolation performed in an assertive form at first and second case studies without observe problems like runge phenomenon at the edges of analyzed sets. researches are still being developed in order to compare the proposed method with the traditional methods of regression. acknowledgment the authors would like to thank coordination for the improvement of higher education personnel (capes), the national counsel of technological and scientific development (cnpq) and research support foundation of goiás state (fapeg) for financial support research and scholarships. references [1] f. a. gomes, v. m. gomes, a. d. o. assis, m. r. d. c. reis, g. da cruz, and w. p. calixto, “heuristic regression method for descriptive data analysis,” 2016. [2] e. garcia, r. arora, and m. r. gupta, “optimized regression for efficient function evaluation,” 2012. [3] w. h. chien, l. chen, c. c. wei, h. h. hsu, and t. s. wang, “modeling slump flow of high-performance concrete using a back-propagation network,” 2010. [4] i.-c. yeh, “modeling slump flow of concrete using second-order regressions and artificial neural networks,” 2007. [5] k. steiglitz, g. winham, and j. petzinger, “pitch extraction by trigonometric curve fitting,” 1975. [6] t. strohmer, “a levinson–galerkin algorithm for regularized trigonometric approximation,” 2000. [7] a. antoniadis, i. gijbels, and a. verhasselt, “variable selection in additive models using p-splines,” 2012. [8] d. gujarati and d. porter, “econometria básica 5.ed.:,” 2011. [9] l. a. aguirre, “introdução à identificação de sistemas–técnicas lineares e não-lineares aplicadas a sistemas reais,” 2004. [10] m. k. goyal, “monthly rainfall prediction using wavelet regression and neural network: an analysis of 1901–2002 data, assam, india,” 2014. [11] r. l. eubank and p. speckman, “curve fitting by polynomialtrigonometric regression,” 1990. [12] m. s. couceiro, d. portugal, n. gonçalves, r. rocha, j. m. a. luz, c. m. figueiredo, and g. dias, “a methodology for detection and estimation in the analysis of golf putting,” 2013. [13] b. u. park, e. mammen, y. k. lee, and e. r. lee, “varying coefficient regression models: a review and new developments,” 2015. [14] h. dette, g. haller, et al., “optimal designs for the identification of the order of a fourier regression,” 1998. [15] p. t. chen g, “introduction to fuzzy sets, fuzzy logic, and fuzzy control systems,” 2000. [16] g. broniatowski, michel; celant, “interpolation and extrapolation optimal designs. 1, polynomial regression and approximation theory,” 2016. [17] r. glowinski, k. atkinson, and w. han, “theoretical numerical analysis: a functional analysis framework,” 2003. [18] h. a. schilling and s. l. harris, “applied numerical methods for engineers using matlab,” 1999. [19] w. p. calixto, a. paulo coimbra, j. c. d. mota, m. wu, w. g. silva, b. alvarenga, l. d. c. brito, a. j. alves, e. g. domingues, and d. p. neto, “troubleshooting in geoelectrical prospecting using real-coded genetic algorithm with chromosomal extrapolation,” 2015. [20] l. trefethen, approximation theory and approximation practice. society for industrial and applied mathematics, 2013. [21] m. rashid, “eletrônica de potência: circuitos, dispositivos e aplicações,” 1999. [22] m. r. c. reis, “comparative analysis of optimization methods applied of tuning pi controller,” 2014.  transactions on environment and electrical engineering issn 2450-5730 vol 1, no 3 (2016) © oluwafemi e. oni, kamati i. mbangula, and innocent e. davidson  abstract—power system stability is an essential study in the planning and operation of an efficient, economic, reliable and secure electric power system because it encompasses all the facet of power systems operations, from planning, to conceptual design stages of the project as well as during the systems operating life span. this paper presents different scenario of power system stability studies on a modified ieee 30-bus system which is subjected to different faults conditions. a scenario whereby the longest high voltage alternating current (hvac) line is replaced with a high voltage direct current (hvdc) line was implemented. the results obtained show that the hvdc line enhances system stability more compared to the contemporary hvac line. dynamic analysis using rms simulation tool was used on digsilent powerfactory. index terms—cct, commutation failure, dynamic voltage stability, hvdc, steady state analysis. i. introduction estriction in transmission network expansion due to reduced right of way (row) brings about long distance bulk power transfer. a minor fault on a heavily loaded line may result in cascading problem which can eventually lead to systems collapse if proper preventive measures are not taken. increase in heavy system load in major urban centre is another major concern. this now goes to the fact that proper systems planning and predictions goes a long way for stable and quality power transfer to major areas that are prone to load increase. ac lines have been the most commonly means of power transmission from one area to another especially in african countries. but with the inherent problems associated with ac lines such as stability problem, cascading effect, corona loss, synchronism problem, and the situation whereby the generating stations are located far away from load centres. ac lines are not suitable for such transmission being that it requires different compensating devices for a specific distance interval. solving these problems brings about usage of hvdc lines to transfer bulk power over long distance which also tends to improve the stability margin of the systems. this paper was submitted for review on august 1, 2016. this work was supported by eskom power plant engineering institute, eskom centre of excellence in university of kwazulu-natal. westville campus, south africa. o. e oni is with electrical engineering department, university of kwazulu-natal. durban 4041, south africa (e-mail: maxiphem@yahoo.com). k. n. i. mbangula was with department electrical engineering, university of kwazulu-natal. durban 4041, south africa. he is now with the department power systems fault such as loss of synchronization of a large power plant, tripping of a load and/or sudden disturbance on a transmission line, most times result in interconnecting systems to enter voltage instability state by not meeting active/reactive power demanded and acceptable voltage at each systems bus. this state can further lead to voltage collapse when all the voltage profile after disturbance is below acceptable limits in an important part of the power systems such that the different part of the systems controllers are stressed beyond their operational limit. thus, ability of the systems to remain practically intact and regain a state of operating equilibrium makes the system voltage stable. much research has been done on voltage stability analysis [1] , [2]. improvement through the use of fact devices was discussed in [3]-[5]. the effect of the automatic voltage regulator (avr), on-load tap changers and power systems stabilizer (pss) on voltage stability was discussed in [6] and [7], while [8] and [12] extensively discusses the impact of hvdc links on power system stability analysis and improvement. hvdc transmission became favorable when hvdc converter problems were reduced by the introduction of thyristor based switches. still a lot of improvement has been made in this area of interest, from hvdc cables, converter transformer, converter technology and topology, etc. and there is still a lot of ongoing research being conducted in this area. this paper presents an investigation into the impact of hvdc links on power systems operation by considering the line loadings and voltage profile. also, a dynamic approach with the use of real time simulation rms tools was also considered in analyzing the critical clearing time with and without hvdc line. the obtained results were compared to determine the extent to which the hvdc system helps in improving voltage stability of the network ii. voltage stability analysis voltage stability is the ability of a power system to be able to maintain an acceptable voltage profile at all buses in the power network when operated under healthy conditions or when of electrical engineering, university of namibia. ongwediva 3624, namibia (e-mail: imbangula@unam.na). i. e. davidson was with department of electrical engineering, university of kwazulu-natal. durban 4041, south africa. he is now with the department of electrical power engineering, durban university of technology. durban 4001, south africa (e-mail: innocentd@dut.ac.za). dynamic voltage stability studies using a modified ieee 30-bus system oluwafemi e. oni, kamati i. mbangula, and innocent e. davidson r subjected to systems disturbance. fig. 1 shows effect of load increase on voltage profile. more increase in load demand beyond the critical point can result in system collapse [1], [10], [13]. fig. 1. receiving end voltage, current and power as a function of load with hvdc line. a. static analysis static analysis of voltage stability involves the use of pv curves to investigate the maximum power that can be transmitted through a transmission line to a load considering the voltage profile of the load bus and the, reactive power needed for the load at specific load power. it can also made use of reduced jacobian matrix in (1)-(4) to analyse the voltage sensitivity of a particular bus to change in reactive power in that bus while the active power is kept constant [1], [10].                       vjj jj q p qvq pvp    (1) let δp=0, then,   vjjjjq pvpqqv  1  (2) vjq r  (3) qjv r  1 (4) where; δp incremental change in bus real power δq incremental change in bus reactive power δθ incremental change in bus voltage phase angle δv incremental change in bus voltage magnitude jr -1 is called the v-q sensitivity as it value determines how stable the system is. a positive value indicates a stable system, while a negative value indicates an unstable system. for a positive value, the smaller the sensitivity value, the more stable the system becomes, meaning as load increases, the value tends towards infinity signifying an unstable system condition. for the modal analysis that explains different snapshot of voltage sensitivity to reactive power, this is given by (5)-(9).   r j (5)  11   r j (6) qv   . 1 (7) since ξ-1=η, i.e. an identity matrix. therefore, qv   . 1  (8) qv 1  (9) (v=η∙∆v) is the vector for the modal voltage variations and (q=η∙∆q) is the modal of reactive power variations. b. dynamic analysis critical clearing time (cct) is one of the method used in analyzing the transient rotor angle stability of a power system. this involves the use of real time dynamic analysis to calculate, for a given defined fault, the maximum allowable clearing time in which the system remains transiently stable. this time frame gives the allowance to which the fault must be cleared or isolated from the rest of the system for the power system to remain in a stable state of operation. if the fault clearing time is longer than the cct, the power system will definitely become unstable [14], [15]. dynamic voltage stability model comprises of first order differential equations as shown in (10)-(12). ‘x’ connotes the state vector of the system and ‘y’ represent the network variable like bus voltage. eq. 10 becomes linear during the case of small disturbance around a steady state equilibrium point (xo, yo) and further eliminated to give (12). static bifurcation, capable of causing voltage collapse will occur when d is zero. thus an assumption of d≠0 is always made for dynamic bifurcation studies. ),( yxfx   (10) ),(0 yxg (11) xacbda dt xd    ')( 1 (12) dynamic voltage stability can then be performed by analyzing the eigenvalues of a΄. (a, b, c and d are appropriate dimensioned matrices) [1], [16], [17]. iii. methodology a modified ieee 30 bus test network was used for this study. modified in the sense that the ieee 30 bus test network is a representation of a portion of the american electric power system (in the midwestern us) as of december, 1961. fig. 2 shows the modified ieee 30 bus test network as setup on digsilent powerfactory. the modified network consist of 30 buses, 8 generators, 20 loads, 40 lines, 11 transformers, 1 shunt capacitor, and 1 shunt reactor. the main voltage level of the network is 400kv (nominal voltage) with nominal transmitting frequency of 50hz. the 132kv of bus 19, bus 20 and bus 21, and the 11kv of bus 15 and bus 18 was assumed on power factory for this study. a. model parameters the model parameters were based on calculations and use of standard ieee models, due to the fact that most data unit were not specified by different online sources for the ieee 30 bus systems. 1) load model the loads are modelled to be voltage dependent with constant active and reactive power demand for load flow calculation and for stability analysis according to (13). the value of kp is set to be 1 for active power (constant current behaviour) and 2 for the reactive power (constant impedance) [18]. kpkp u u q u u pjqp                   0 0 0 0 (13) 2) generator model all the generators are connected via a transformer. sm (synchronous machine) was used on powerfactory to name all the generators. table i shows the generator parameter as used on digsilent. sm 1.1 with 00 voltage angle and 1.00pu voltage set-point was used as the reference machine because it is connected to the bus with the highest fault level. all generator were set to be in voltage control mode. two generator type were used as shown in table ii, first is the 16.5kv, 500mva used by sm 3, 5 and 6 and the other synchronous machine uses the 20kv, 800mva type. transformer parameter used is shown in table iii. table i generator parameter sm mw mvar sm mw mvar 1.1 slack 5 4.1 720 5 1.2 700 0 4.2 720 0 2 320 0 5 390 0 3 407 0 6 385 5 table ii synchronous machine type data sm mva kv xd (pu) xl (pu) xq (pu) td΄ (s) tq΄ (s) typ1 800 20 2.2 0.3 2.2 2.1 0.28 typ2 500 16.5 1.3 0.15 2.0 1.0 1.0 table iii transformer type data sn mva kv uk & uk0 vector. group ultc trf 1 400 11/400 12.6 ynd1 -5 to 5 trf 2 600 16.5/400 14.88 ynd1 -4 to 4 trf 3 850 20/400 14.44 ynd1 -5 to 5 trf 4 400 132/400 11.82 ynyn -3 to 3 all generators are rated for realistic inertial time constant and modelled using the ieee controller model on powerfactory for the automatic voltage regulator, governor control as well as the power systems stabilizer (but the pss for all the generator are disable due to set point error). 3) hvdc model the hvdc network is modelled as depicted in fig. 3 with the parameter on table 1v. the 1350mw monopolar with 600kv operating voltage transmits power over a distance of 700km. the reliability of the hvdc network was put to test using a three phase fault at the inverter terminal (fault reactance of 10ω for 200ms. result shows a commutation failure at the inverter side of the converter which the voltage dependent current order limiter (vdcol) was activated as a result of reduction in dc voltage. the commutation failure was resolved when the rectifier controller reduces the dc current to allow minimum power across the link during fault conditions. with this, most of the dc faults are self-clearing with the help of a well-equipped controller. equations (14) and (15) show the fundamental law governing the hvdc systems. hvdc equivalent circuits’ equation is given by (16).   dcitii di ixv v 3cos23   (14)   dcrtrr dr ixv v 3cos23   (15) cilcr doidor d rrr vv i     coscoscos (16) hvdc control as set up on digsilent powerfactory software can be divided into two hierarchical levels: inverter control and rectifier control. a block definition that defines the transfer function in the form of graphical block diagrams and equation is first created for each of the controllers, and then a composite block frame is created for the overall control. this consists of the entire overview diagram showing all the slots interconnections and which object should be assign to a slot. after a common model is created from the block definition, they are then added into the composite model. fig. 4 and 5 from [19] give a simple overview diagram of the controller as connected to the converter. table iv hvdc data rectifier inverter ac voltage (kv) 400 400 firing angle control current control voltage control commutation reactance 13.445ω 13.445ω tap changer control α-control γ-control actual winding ratio 0.97 0.97 fig. 2. single line diagram of the modified ieee 30 bus system. fig. 3. monopolar hvdc model. fig. 4. rectifier controller. fig. 5. inverter controller. iv. simulation result the results consist of both the steady state power flow results which are depicted in bar graphs and transient stability analysis results which are configured using the real time rms simulation. a. steady-state load flow the steady state load flow results follows the newton raphson iterative equation for power flow calculation. fig. 4 to 8 show the steady load flow presented in a bar diagram format for the bus voltage magnitudes, the line loading, and load active and reactive power respectively. from the line loading diagram, the transmission line connecting bus 3 and 1, and the line connecting bus 8 and 3 is already overloaded beyond their ratings, even transmission line connecting bus 9 to bus 13. however, the hvdc line is out of service, so as to know the systems condition when only ac lines is in operation (meaning hvdc system is out of service). fig. 8 present active load demand in the ieee 30-bus system. b. time-domain stability simulation two operational scenarios were considered in this study. the first scenario is when the power is being transmitted using ac lines only, and the second scenario is when an existing longest ac line is replaced with an hvdc line and the results are depicted on a graph. bus voltage and generator excitation current are the main object of focus for this study. for the bus voltage magnitude graph, specific busbar that are prone to voltage instability was selected rather than use all the 30 buses of the network. fig. 6. busbar voltage magnitudes. fig. 7. line loading (ac lines only). fig. 8. load active power. 1) first scenario (without hvdc) different study cases was carried out while using the time domain simulation to investigate the weakest area of the network. this involve the use of critical clearing time to observe the maximum time a fault can stay on each of the element on the network (busbar and lines). three phase short circuit fault was placed on each of the busbar once at a time and their cct was estimated. it was observed that bus 8 has the least critical time. the same process was also carried out on the line (placing the fault at the beginning of the line) and cleared by switching off the line. fig. 9-11 show the bus voltage magnitude, generator rotor angle and its excitation current when a three phase fault placed at the beginning of ‘line 1_3’ was simulated and the fault cleared by switching the line off after 100ms. this is the maximum allowable time the fault can stay on the line for the system to be stable when using only ac lines to transmit power from one region to another. during the fault, the bus voltage dips down to about 0.11pu, but the generator exciter helps to restore the system back to stability by increasing the field current winding of the synchronous generator. thereby automatically adjust the field current to maintain the required terminal voltage. fig. 12-14 shows a situation whereby the fault stay more than expected time before the circuit breaker of the line was opened (120ms). this caused the generator to have yielded all its excitation limit and the systems enter instability when the voltage profile of all the buses cannot be met again. this causes the generator angle to swing 3600 off from the reference machine, a situation caused when the generator pole slipped. fig. 9. voltage plot during fault on ‘line 3_1’, cleared by switching off the line after 100ms. (without hvdc line). fig. 10. generators rotor angle (without hvdc line). fig. 11. generator excitation current (without hvdc line). fig. 12. voltage plot during fault on ‘line 1_3’, cleared by switching off the line after 120ms. (without hvdc line) fig. 13. generator rotor angle fault (without hvdc line) fig. 14. generator excitation current (without hvdc line) 2) second scenario (with hvdc line). when existing ac ‘line 8_27’ was replaced with a monopolar hvdc system as shown in fig. 3, the load flow result for the lines loading are shown below in fig. 15, with all lines loading within acceptable range. the same three phase fault scenario was carried out on ‘line 3_1’ and the fault cleared by isolating the line after 120ms while using hvdc line to interconnect ‘bus 1’ to ‘bus 28’, it was found out that the system was stable even until it reaches a maximum time of 150ms. further increase of fault clearing duration beyond this limit resorted in convergence error. fig. 16-18 shows the bus voltage magnitude, generator rotor angle and excitation current respectively. a case of commutation failure occur at the inverter side of the hvdc link during fault period, but with the help of the converter selection mode by blocking the fault current from transferring into the hvdc link, the systems was able to maintain stability. hvdc systems thus help to increase the time duration a fault can stay on the line before being isolated from the systems. all different study cases carried out on the systems prove the efficacy of hvdc link in power systems stability and control during system disturbance. fig. 15. line loading (with hvdc line). fig. 16. voltage plot during fault on ‘line 3_1’, cleared by switching off the line after 150ms. (with hvdc line). fig. 17. generators rotor angle (with hvdc line). fig. 18. excitation current. (with hvdc line). v. conclusion impact of hvdc scheme on ac systems short term voltage stability study was investigated in this study, and based on this study; it was found out that hvdc systems helps in enhancing voltage stability than the ac line, in that it helps to improve the critical clearing/ isolating time for disturbances on the systems. the effect of vdcol in hvdc link during systems disturbance was also analyzed. the strategies for improving voltage stability are thus proposed; that hvdc lines improve dynamic voltage stability of power systems. although, different facts devises and a well modelled generator controller helps in enhancing voltage stability of a system. however, this cannot be compared to the benefit which hvdc systems offers, namely; little line losses, long distance bulk power transfer, immunity to cascading effect, bi-direction power transfer, small right of way, asynchronous interconnection etc. initial cost of constructing converter station can be a little expensive, but the cost saved by transmission line construction with associated losses in dc systems outweigh the latter. and with the emergence of new power electronic converter and well rugged controller, hvdc system will be the best mode to transmit bulk power due to high efficiency and economics of transmission that it offers. vi. references [1] g. morison, b. gao, and p. kundur, "voltage stability analysis using static and dynamic approaches," ieee transactions on power systems, , vol. 8, pp. 1159-1171, 1993. [2] t. van cutsem and c. vournas, voltage stability of electric power systems vol. 441: springer science & business media, 1998. [3] m. noroozian, l. ängquist, m. ghandhari, and g. andersson, "improving power system dynamics by series-connected facts devices," ieee transactions on power delivery, , vol. 12, pp. 1635-1641, 1997. [4] y.-h. song and a. johns, flexible ac transmission systems (facts): iet, 1999. [5] s. gasperic and r. mihalic, "the impact of serial controllable facts devices on voltage stability," international journal of electrical power & energy systems, vol. 64, pp. 1040-1048, 2015. [6] c. vournas and m. karystianos, "load tap changers in emergency and preventive voltage stability control," transactions on power systems, ieee, vol. 19, pp. 492-498, 2004. [7] y. wang, d. j. hill, r. h. middleton, and l. gao, "transient stability enhancement and voltage regulation of power systems," ieee transactions on power systems, , vol. 8, pp. 620-627, 1993. [8] d. l. h. aik and g. andersson, "power stability analysis of multi-infeed hvdc systems," ieee transactions on power delivery, , vol. 13, pp. 923-931, 1998. [9] hammad, "stability and control of hvdc and ac transmissions in parallel," ieee transactions on power delivery, vol. 14, pp. 1545-1554, 1999. [10] o. e. oni, k.n.i. mbangula and i.e. davidson, “voltage stability improvement of a multi-machine system using hvdc," proceedings of the clemson university power systems conference (psc), march 8-11, 2016, clemson university, clemson, sc, usa. [11] k. n. i. mbangula, i. e. davidson and r. tiako, “improving power system stability of south africa’s hvac network using strategic placement of hvdc links”, cigre science & engineering journal (cse), vol. 5, june 2016, pp. 71-78. [12] k.n.i. mbangula, o.e. oni and i.e. davidson, “the impact of hvdc schemes on network transient rotor angle stability”. in proceedings of the 24th south african universities power engineering conference, 2628 january 2016, vereeniging, south africa, pp. 461 – 466, isbn 978-177012-386. [13] digsilent powerfactory: power system stability seminar digsilent buyisa (pty) ltd. [14] w. a. oyekanmi, g. radman, a. a. babalola, and t. o. ajewole, "effects of statcom on the critical clearing time of faults in multimachine power systems during transient stability analysis studies," in 2014 ieee 6th international conference on adaptive science & technology (icast), 2014, pp. 1-6. [15] r. kamdar, m. kumar, and g. agnihotri, "transient stability analysis and enhancement of ieee-9 bus system. electrical & computer engineering: an international journal (ecij) volume 3, number 2, june 2014" [16] j. chow and a. gebreselassie, "dynamic voltage stability analysis of a single machine constant power load system," in proceedings of the 29th ieee conference on decision and control, 1990, pp. 3057-3062. [17] b. h. lee and k. y. lee, "dynamic and static voltage stability enhancement of power systems," ieee transactions on power systems, , vol. 8, pp. 231-238, 1993. [18] digsilent powerfactory: technical reference documentation general load, gomaringen, germany, 2013. [19] d. kong, "advanced hvdc systems for renewable energy integration and power transmission: modelling and control for power system transient stability," doctor of philosophy, school of electronic, electrical and computer engineering, university of birmingham, birmingham, 2013. oluwafemi e. oni was born in nigeria on 17 september 1988. he received his bsc (honours) degree in electrical and electronic engineering from ekiti state university, ado ekiti, nigeria, in 2013. he then proceed to university of kwazulunatal, durban, south africa in 2015 for his msc degree in electrical engineering (currently handed in his thesis for examination). he was a system and maintenance engineer at egbin power thermal plant, lagos, nigeria, in 2012 and omotosho power plant, ore, nigeria, in 2013/2014. he is currently a research assistance with department of electrical engineering, university of kwazulu-natal. his research includes power systems stability analysis using high voltage direct current transmission scheme, integration of renewable energy into the grid using multi-terminal hvdc scheme, and smart grid systems using facts. mr. oni’s award and honors include mtn foundation scholarships, etisalat scholarship and ekiti state scholarship. kamati n.i. mbangula was born in namibia on 20 august 1989. he graduated with a bsc. (honours) degree in electrical engineering from the university of namibia (unam). he pursued his postgraduate studies in south africa at the university of kwazulu-natal (rsa), and carried out his research at the eskom centre of excellence in high voltage direct current (hvdc). his work experience includes working as a staff development fellow at unam, and working as a research assistant and lab technician at the eskom centre of excellence in hvdc. he is currently employed as a lecturer at unam. his fields of interests include power systems stability analysis, and low voltage reticulation systems design and analysis. innocent e. davidson (m’92–sm’02) received the bsc (hons) and msc degrees in electrical engineering from university of ilorin in 1984, and 1987 respectively. phd in electrical engineering from the university of cape town, rondebosch, south africa1998; and postgraduate diploma in business management from the university of kwazulu-natal, south africa, 2004; associate certificate, sustainable energy management, british columbia institute of technology, burnaby, canada, 2011. from 1994-1995, he was engineering inspector, rainbow energy project at easigas (pty) ltd, cape town; senior lecturer, university of pretoria (1999-2001); senior lecturer, department of electrical engineering, university of kwazulunatal (ukzn), 2001–2006; part-time instructor, graduate engineering program (power & energy), ukzn high voltage dc centre (2000-2008) a program co-offered by ukzn and eskom south africa’s electric utility. from 2005–2006, he was a visiting professor, powertech labs inc., surrey, bc, a world leading consortium in clean energy technologies, independent testing services, power system solutions and smart utility services. from 2007-2011 he was energy consultant in surrey, bc, implementing energy efficiency (electricity/gas) measures, british columbia provincial government’s mandate on climate change. he has been an invited guest writer for the ieee power and energy technical magazine as an expert on africa: “energizing africa’s emerging economy”, ieee power and energy, vol. 3, no 4, july/august 2005. he was associate professor of electrical engineering and research coordinator, university of namibia (2012-2014); director, eskom centre of excellence in hvdc engineering, ukzn (2014-2016). currently, he is a full professor of electrical engineering, durban university of technology, south africa. he is the author/co-author of over 150-refereed journal and conference papers. his research focus is on grid integration of renewable energy using smart technologies and innovation for smart cities. prof davidson is a member, western canada group of chartered engineers (wcgce); the institute of engineering and technology (iet canada) british columbia chapter; a chartered engineer, c.eng. united kingdom. he is a fellow of the south african institute of electrical engineers and a registered professional engineer, p. eng. (ecsa), south africa. transactions on environment and electrical engineering issn 2450-5730 vol 4, no 1 (2020) © vinícius henrique farias brito, josé carlos de oliveira, fabricio parra santilio  abstract— although there currently exists a wide range of voltage regulators that are commercially available, the search for devices with a simpler physical design remains the focus of research studies. following this line, an electromagnetic voltage regulator (evr) arrangement has been proposed. the evr is constituted of an autotransformer that supplies, via discrete taps, a series transformer that injects voltage for regulating the feeder voltage. even though its operating principle is shown as being similar to that of other devices on the market, the physical arrangement and operating strategy of evr show novelties which result in properties such as: economic attractiveness, constructive simplicity, and operational reliability. moreover, when installing voltage regulators, efficacy studies must be carried out to optimize equipment design. in this context, this paper aims at evaluating the factors that influence the effectiveness of the evr in restoring voltage variations according to the determinations imposed by regulatory agencies. the ultimate goal of this study is to determine the voltage deviation range that the evr is able to restore. to achieve this goal, a mathematical modeling of the evr is given and study cases are computationally carried out to investigate its performance when connected to a typical distribution feeder. index terms—computational modelling, distribution system, electromagnetic voltage regulator, performance evaluation, power quality, voltage regulation. i. introduction mong the electrical power supply requirements imposed on power utilities, power quality indices are included, such as long-duration and short-duration voltage variations, harmonic distortions, unbalances, etc. in this scenario, the issues related to the voltage magnitude variations at power frequency are particularly highlighted, since distinguished normative documents establish limits for these phenomena in terms of long and short duration [1]. in brazil, the definitions of voltage variation severity and this study was financed by the coordenação de aperfeiçoamento de pessoal de nível superior – brazil (capes) – finance code 001. vinícius henrique farias brito and josé carlos de oliveira are with the faculty of electrical engineering, federal university of uberlandia, uberlandia, brazil (e-mail: vinicius.brito@ufu.br; jcoliveira@ufu.br). fabricio parra santilio is with the electrical engineering department, federal university of mato grosso, cuiaba, brazil. (e-mail: fabricio.qee@gmail.com.br). duration are set by the brazilian electricity regulatory agency (aneel) in the technical standard titled electricity distribution procedures in the national electric system (prodist) module 8 [2]. this document classifies longduration voltage variations as changes in the rms value of the voltage over a period longer than 3 minutes. on the other hand, short-duration phenomena include those voltage variations manifested for periods shorter than the 3-minute limit. moreover, the directive also establishes the magnitude limits for long and short duration voltage variation events. when the voltage magnitude infringes the established limits, regulation or compensation processes are carried out in order to regulate the voltages to the acceptable levels defined by legislation. a wide range of equipment is currently employed to perform this task. in general, it is recognized that compensation devices are based on two basic strategies. the first comprises of voltage compensation by indirect methods, such as voltage control via static or dynamic devices associated with the control of reactive power flow. the second performs its function by acting directly on the voltage magnitudes using devices that change the voltage values via tap changers or direct injections of compensating voltage [3]. in the context of devices based on the control of reactive power, the simplest devices are capacitor and reactor banks – fixed or automatic [4]. another possibility, widely used in the past in large power systems, is the synchronous compensator [5]. further still, with the evolution of electronic switching technologies, commercial products that make use of the wellknown facts technology have arisen [6]. this group includes the static var compensators (svcs) and synchronous static compensators (statcoms). regarding compensation technology which acts directly on voltage magnitudes, traditional transformers with on-load tap changer (oltc) and no-load tap changer (nltc) [7] stand out, as well as other electromagnetic regulators based on tap changes to adjust the electrical quantities, such as the step voltage regulator (svr) [8]. additionally, other devices, based on electronic switching, are available on the market. this is, for example, the case of the dynamic voltage regulator (dvr) [9]. in light of the above, one recognizes, therefore, that there is a diversity of devices available on the market with operating properties capable of regulating the voltage at the load in modeling and performance evaluation of an electromagnetic voltage regulator via series compensation vinícius henrique farias brito, josé carlos de oliveira, fabricio parra santilio a mailto:vinicius.brito@ufu.br accordance with the required standards. however, the most widely used strategy to mitigate long-duration voltage variations, in distribution systems, is the direct compensation of the voltage magnitude [10]. furthermore, in terms of the voltage regulators based on the direct compensation of the voltage magnitudes mentioned previously, there exists in this same category a device proposed in [11], which is called the electromagnetic voltage regulator (evr). this device consists of a shunt autotransformer with taps supplying a transformer connected in series with the electrical feeder focused on the regulation process. the series transformer is responsible for the injection of a controlled reinforcement voltage, being that additive or subtractive, which aims at compensating the voltage at the load terminals. the fundaments that govern this proposal are found in [12], which shows that, via discrete switching, different taps of the autotransformer can be used to make the reinforcement voltage compatible with the required compensation level. the possibility of disconnecting the shunt autotransformer, as well as the switches, provides greater operational reliability for the network to which the evr is connected. that is, in the event of a failure or maintenance of the regulator, despite the loss of the regulation process, the power flow between the source and the load is not interrupted. in line with the aforementioned topological proposition, more recent works, such as [10] and [13], covering similar physical structures and commercial products have shown the feasibility of the arrangement focused upon in this paper. in fact, with a topology similar to the evr, [14] describes a new conception for a voltage compensator, which was installed on a rural medium voltage feeder in germany with high presence of distributed generation. the commercial equipment was called the line voltage regulator (lvr) and the results of its performance indicated great improvements in the voltage profiles of the rural medium voltage feeder. in addition to the operational effectiveness, the work highlights that the lvr presented a cost-benefit ratio higher than other options for voltage compensation, which significantly increased the network distributed generation capacity. within the same constructive and operational strategy, [15] showed that the solution proved to be efficient for different voltage levels. even though the efficacy of the evr injection has been verified for a constant load consumption [11] [12], its effectiveness for a load with dynamic behavior must be further investigated. in fact, variables such as the long-duration voltage variation magnitudes, the feeder and load parameters, and the available taps of the evr are features that will strongly affect the performance of the device. in this context, this paper aims at evaluating the relationship among these influencing factors since they define the range of voltage variations that the evr can restore to the adequate voltage level set in the standards. for a better understanding, initially, the physical arrangement and operating principle of evr are presented, as well as the development of its mathematical modeling in the frequency domain. in the following, a detailed performance evaluation study and computational case studies are carried out to achieve the goal proposed in this paper. ii. electromagnetic voltage regulator: physical arrangement and mathematical modeling before deriving the mathematical model of the evr, its physical arrangement is shown in the schematic diagram given in fig. 1. it can be noted that the device enables the control of the load voltage (bus 2) by injecting a series compensation voltage (additive or subtractive reinforcement), which, when summed to the supply voltage, leads to a controlled voltage at the load terminals. in fig. 1, the regulator is operating to restore the load voltage (bus 2) when an undervoltage occurs on bus 1. once this phenomenon happens, the control on the regulator detects it and selects the tap of the autotransformer that offers the most adequate level of compensating voltage for injection into the system. this compensation voltage when summed to the supply voltage restores the voltage at bus 2 to its required value or close to it. subtractive voltages are also feasible for injection by the series transformer, in the case of overvoltages. this can be carried out by changing the contacts using the swpp and swpn switches. fig. 1. schematic diagram of the electromagnetic voltage regulator. adapted from [11]. the equivalent electrical circuit of the evr and overall system arrangement, related to a specific steady state operational condition, is given in fig. 2. noteworthy here is that the equivalent circuit shown is applicable to a given selected tap when operating to restore undervoltages. the corresponding equation to establish the relationship among the system (evr, load, and feeder) parameters and the load voltage is given by (1). noted also is that the magnetizing branches of the transformers are disregarded. the variables in (1) are identified in (2) to (5). fig. 2. equivalent electrical circuit of the evr, feeder and load. 1 21 s l st at v v k k           (1) 2 1 . * atst at at st l z k z               (2)   2 * st s at at st l z z k z                (3) 1 2 2 at at at at e e e    (4) 1 2 st st st e e   (5) where: • sv  is the supply voltage. • lv  is the voltage at the load terminals. • at  is the transformation ratio of the autotransformer. • st  is the transformation ratio of the series transformer. • sz  is the supply network short-circuit impedance. • stz  is the total impedance of the series transformer referred to the primary side. • atz  is the total impedance of the autotransformer referred to the secondary side. • 1ste  and 2ste  are the voltages on the primary and secondary sides of the series transformer, respectively. • 1ate  and 2ate  are the voltages on the primary and secondary sides of the autotransformer, respectively. iii. a case study of the evr performance in order to carry out the performance investigation of the overall system arrangement and the evr effectiveness during the occurrence of voltages deviations from the standard values, a typical electrical system has been utilized. it consists of a radial feeder, whose parameters are found in table i. it is noteworthy that, without the action of the regulator, the connection of the rated load to the feeder leads to a voltage drop of about 3.5% at the series impedance, assuming the supply voltage is 1 pu. table i feeder and load characteristics parameter value rated voltage 13.8 kv network short circuit power 200 mva r/x ratio 0.5 load power factor 0.94 (lag) rated load power 10 mva as for the evr, the compensator is connected to the feeder, in accordance with the electrical circuit of fig. 2. its regulation range (reg. range) goes up to ± 20% in steps of 2.5%. the autotransformer has 8 taps, therefore, the evr, via its tap 8, is capable of providing a maximum line voltage of:  1 1 re . * 0.2*13.8 2.76 st rated st v g range pu v v kv kv    (6) in order to reduce the current level of the switches of the taps, the transformation ratio of the series transformer was chosen as αts = 0.5. one is reminded that under this condition, the current on the primary side of the series transformer will be 50% of the feeder (load) current. therefore, due to this transformation ratio of the series transformer, the voltage on the secondary side of the autotransformer must be twice the voltage value intended for injection. taking tap 8 as an example, to achieve a compensating voltage of 20%, it is necessary that the voltage on the secondary side of the autotransformer be 40% of the feeder rated voltage. similar reasoning applies to the other taps. regarding the rated power of the transformers, both autotransformer and series transformer have the same rated power, as determined by (7).     re 1 re 2.5 st at load st at g range pu s s s g range pu s s mva           (7) where: • st s is the rated power of the series transformer. • at s is the rated power of the autotransformer. • load s is the rated load power. table ii provides the characteristics of the two electromagnetic units that make up the evr. the impedances and resistances are in line with typical designs. table ii series transformer and autotransformer characteristics data power (mva) primary/secondary winding voltages zcc (%) rcc (%) series transformer 2.5 2.76/5.52 kv 5 1 autotransformer 2.5 13.8 kv/taps 5 1 autotransformer taps – secondary winding voltage 5.52 kv (tap 8) – 4.83 kv (tap 7) 4.14 kv (tap 6) – 3.45 kv (tap 5) 2.76 kv (tap 4) – 2.07 kv (tap 3) 1.38 kv (tap 2) – 0.69 kv (tap 1) once the feeder, the load, and the design characteristics of the regulator are defined, the premises for the studies are listed as follow:  the studies performed are associated with phenomena classified as long-duration voltage variations.  according to the criteria defined in [2], the voltage values between 0.93 pu and 1.05 pu are considered as adequate. this voltage range is delimited by the green region of fig. 3.  when the limits of this range are violated, the action of the voltage regulator is represented, for each tap of the autotransformer, by lines that are changed according to the tap used. fig. 3 shows, initially, that the supply network suffers a 0.15 pu voltage drop. then, with the load disconnected, the load voltage reaches the value of 0.85 pu. once such a voltage variation is detected by the equipment control, the evr starts to operate using tap 4 of the autotransformer – the corresponding reinforcement voltage is 2.76 kv, therefore, at-tap4 is equal to 5, as shown in (8). thus, still in the condition of the disconnected load, (9) shows that the load voltage increases to 0.94 pu. this value is obtained by (1) when k1 and k2 are equal to zero, as such; it corresponds to the maximum compensation achieved by tap 4. 4 13.8 5 2.76 at tap kv kv     (8) 4 0.85 0.94 0.5 1 1 5 s l st at tap v puv           (9) where: 4at tap   is the transformation ratio of the autotransformer when tap 4 is selected. therefore, under the above-mentioned conditions, tap 4 was sufficient for restoring the voltage to the appropriate range. however, as the load is connected and its power increases the complex coefficients k1 and k2 change, and despite the injection of the voltage from tap 4 being maintained, as expected, the load voltage gradually decreases; the decay rate is defined by the feeder and series transformer impedances. fig. 3 shows that, starting from a 3 mva load, tap 4 is no longer sufficient to adjust the load voltage to the adequate range. then, the control changes to tap 5, and a new line starts in fig. 3, with a similar decay behavior, however, starting with an adequate voltage value. when the load becomes equal to 7.9 mva, the situation repeats once more. fig. 3. correlation between the load voltage and the load power for an undervoltage of 0.85 pu. the results of the performance studies carried out so far reveal that the voltage compensation efficacy decreases with the increase of the load power, for each tap, as expected. therefore, in order to maintain the load voltage within the adequate range, it becomes necessary to change the tap to compensate for the loss of regulation efficiency caused by the increase in the k1 and k2 coefficients, according to (1). moreover, the dynamics of the control and switching system will define the regulator response time for the voltage regulation. from another evaluative aspect, performance studies are now carried out for a constant 10 mva load under different undervoltages that occurred on the network. in doing so, the behavior of the set source-feeder-evr-load is considered for various undervoltage magnitudes, and the effectiveness of the voltage regulation is evaluated for the 8 taps, as shown in fig. 4. fig. 4 shows that under different undervoltages, as the taps of the autotransformer change from 1 to 8 (increasing the reinforcement voltage), the load voltage also rises as desired. the graph shows that each tap determines an undervoltage limit for which the evr is able to ensure that the load voltage is within the adequate range. consequently, the most critical undervoltage magnitude that the evr is effective in restoring is set by tap 8. the figure indicates that for a 10 mva load the regulator is capable of compensating, for the adequate range, undervoltage magnitudes up to 0.8 pu at the supply voltage. fig. 4. undervoltage magnitudes compensated by the evr according to the autotransformer tap for a 10 mva load. iv. dynamic relationship between undervoltages phenomena and the evr effectiveness in addition to the studies related to the previously presented operational limits, this section aims at showing the dynamic performance of the evr under specific operating conditions. hence, the system used as a case study was implemented in the matlab simulink, as shown in fig. 5. the parameters of the system are the same as those shown in tables i and ii. the two operational situations considered are the following:  case 01: initially, the 13.8 kv voltage supply (1 pu) feeds the rated load (10 mva). at t = 0.5 s, the supply voltage drops to 11.73 kv (0.85 pu). then, the evr is set to turn on at t = 1 s with tap 8 selected.  case 02: this situation is similar to the previous one, except for the fact that the voltage drop is more severe (10.35 kv or 0.75 pu). a. case 01 undervoltage of 0.85 pu fig. 6 presents the voltage profile at the load terminals. it is shown that, between t = 0 and 0.5 s, the load voltage has a value of 13.25 kv (0.96 pu). the 0.04 pu voltage drop is due to the feeder impedance. at t = 0.5 s, the load voltage reduces to 11.27 kv (0.82 pu), due to the voltage variation imposed on the supplier. next, at t = 1 s the regulator starts operating, with the autotransformer switched to tap 8. under these conditions, the load voltage is restored to 13.69 kv (0.99 pu), thus showing the effectiveness of the regulator. fig. 6. load voltage profile for case 01. b. case 02 undervoltage of 0.75 pu the results associated with a more drastic undervoltage are indicated in fig. 7. as one notes, from 0 to 0.5 s the load voltage remains at 13.25 kv (0.96 pu) and, then, the load voltage is suddenly reduced to 9.94 kv (0.72 pu). after the regulator insertion, at t = 1 s, with the autotransformer switched to tap 8, the load voltage is increased to 12.08 kv (0.87 pu). therefore, for this situation, the evr does not have sufficient characteristics to restore the load voltage to the adequate range. in fact, the specified regulator is capable of restoring undervoltages up to 0.80 pu for a 10 mva load. despite the demonstrated limitation, the load voltage increased from 0.72 pu to 0.87 pu, which lessens the voltage variation at the load bus. fig. 7. load voltage profile for case 02. fig. 5. system implemented in the matlab simulink. the main electrical quantities associated with the operating conditions of the evr, for the two cases analyzed, are summarized in table iii. comparing those to the rated currents of the series transformer (105 a) and the autotransformer (523 a), the values obtained are within the rated characteristics of these electromagnetic components. the same applies to the voltage on the primary side of the series transformer. table iii currents and voltages associated with the studied cases electrical quantity case 01 case 02 il 415.1 a 366.2 a is 518.9 a 457.8 a i1at 103.8 a 91.6 a v1st 2.61 kv 2.3 kv finally, fig. 8 highlights the evr performance for other load power values, previously fixed at 10 mva. the figure shows the undervoltage magnitudes that the regulator is capable of restoring to the adequate range at the load bus. as expected, as the load power is reduced, the evr can restore more severe undervoltage magnitudes. fig. 8. undervoltage magnitudes compensated by the evr for the adequate range according to load power. v. conclusions this paper presented an electromagnetic device for compensating voltage variations. the proposal acts directly on the load voltage by inserting a reinforcement – additive or subtractive – to restore the voltage magnitude to the standards established by legislation. the device has an attractive operating strategy given the use of components that offer constructive and operational simplicity, reliability, attractive costs, versatility of installation in uncontrolled environments, among other attributes. in order to contextualize the theme, general information concerning the physical arrangement and mathematical modeling of the regulation process was synthesized. as exposed in the introduction, the evr was initially proposed by [11] and a similar device was materialized as a commercial product by [14] [15]. once the device physical and mathematical model were presented, studies related to the effectiveness of the device when faced with typical influence quantities of electrical networks were carried out. the results clearly showed that the evr, in the terms designed and defined by its basic characteristics, also presents a strong dependence on the feeder and load parameters. its efficacy may be full, partial or insufficient, depending on the variables involved in the voltage regulation process. in general, the results obtained are encouraging for the diffusion of the technology contemplated herein. regarding the discrete response and the use of the evr for restoring short-duration voltage variations, it can be implemented with fine-tunes based on the use, for example, of electronic techniques for controlling switches continuously, which is a subject for future works. finally, it should be noted that, although the study presented has explored phenomena associated with undervoltages in the supply network, the regulator can also be used when overvoltages occur on the network. references [1] m. h. bollen, understanding power quality problems: voltage sags and interruptions. punta gorda: wiley-ieee press, 1999. [2] aneel, “procedimentos de distribuição de energia elétrica no sistema elétrico nacional prodist módulo 8,” 2018. [3] t. v. da silva, “uma proposta para o controle eletrônico de reguladores eletromagnéticos através do reforço série de tensão,” m.s. thesis, universidade federal de uberlândia, 2012. [4] s. haffner, l. a. pereira, l. v. gasperin, and l. barreto, “alocação de bancos de capacitores em redes de distribuição de energia visando eliminar violaçöes de tensão,” control. automação, vol. 20, no. 4, pp. 546–563, 2009. [5] f. o. igbinovia, g. fandi, z. muller, j. svec, and j. tlusty, “optimal location of the synchronous condenser in electric-power system networks,” proc. 2016 17th int. sci. conf. electr. power eng. (epe ), 2016. [6] k. r. padiyar, facts controllers in power transmission and distribution. new delhi: new age international (p) limited, publishers, 2007. [7] h. zhou, x. yan, and g. liu, “a review on voltage control using on-load voltage transformer for the power grid,” iop conf. ser. earth environ. sci., vol. 252, no. 3, 2019. [8] l. a. kojovic, “modern techniques to study voltage regulator dg interactions in distribution systems,” 2008 ieee/pes transm. distrib. conf. expo., pp. 1–6, 2008. [9] s. a. taher, h. t. fard, and e. b. kashani, “new switching approach for dvr using one cycle control method,” ain shams eng. j., vol. 9, no. 4, pp. 2227–2254, 2018. [10] p. r. p. sarathy, “analysis and optimization of medium volta ge – line voltage regulator,” m.s. thesis, norwegian university of science and technology department, 2018. [11] f. p. santilio, “proposta, modelagem e validação de uma nova concepção de regulador eletromagnético através do reforço série de tensão,” ph.d. dissertation, universidade federal de uberlândia, 2013. [12] l. e. vasconcelos, “modelagem no domínio da frequência de um regulador eletromagnético de tensão baseado na compensação série,” m.s. thesis, universidade federal de uberlândia, 2014. [13] g. ram, v. prasanth, p. bauer, and e. m. barthlein, “comparative analysis of on-load tap changing (oltc) transformer topologies,” 2014 16th int. power electron. motion control conf. expo., pp. 918–923, 2014. [14] m. carlen et al., “line voltage regulator for voltage adjustment in mv-grids,” cired 23rd int. conf. electr. distrib., 2015. [15] g. leci and f. cornelius, “increasing grid capacity to connect renewable energies,” cigre, pp. 1–9, 2016.  transactions on environment and electrical engineering issn 2450-5730 vol 1, no 3 (2016) © dan jigoria-oprea, gheorghe vuc, marcela litcanu  abstract— deregulation of energy market led to the development of flexible and efficient framework for energy trading by energy companies in a competitive environment. both deregulation and the concern towards environment issues increased the number of small and medium renewable power plants distributed in the network. the variability of renewable energy sources and the lack of their central monitoring led to new challenges concerning power system operation. the idea of aggregation for distributed energy sources led to the concept of virtual power plant, which determines a better control of production units but also a better visibility for the system operator. in this paper, the authors propose an optimal management solution which can offer a virtual power plant the capability to sell complete services, both for production and demand side management, by decreasing the necessary reserve for balance. index terms—energy market, optimal management, renewable energy sources, virtual power plant. i. introduction he increased share of renewable energy sources in the electricity production brings issues concerning power balance in the power system. generating electricity from renewable sources is influenced by weather conditions and by the availability of source – wind or sun. using the power reserve of centralized sources is justified only form economic point of view. this reserve is used to compensate the shortage of energy determined by the unpredictable nature of renewable sources power generation. as a result, it seems more intelligent to transfer the balance load to another level of structure in the network. this structure should include different types of distributed resources, energy storage units and to have control and command rights. all these can be combined in a structure like the virtual power plant (vpp) that can operate like a classical power plant. all operations for each unit can be programed in advance. the concept of vpp has already a history of over two decades, experimental projects being tested in several parts of the world [1-4]. for the romanian power system, using the vpp as a solution for the management of renewable energy is not applicable yet; the solution used now considers including the renewable energy this manuscript was submitted on 29 july 2016 and accepted on 18 september 2016. d. jigoria-oprea is with the power systems department at politehnica university timisoara, v. parvan 2 blvd., timisoara, timis, 300223, romania (e-mail: dan.jigoria@upt.ro). sources (res) in a large and diverse portfolio of a strong actor on the energy market. ii. mathematic model the optimization problem is actually a problem of maximizing the profit of the vpp [5], [6]. the aim is to maximize the profit for each and every one of the 24 hours:     24 1r t disp tt vpp tt dr tt iso t startyedrbidprofit  (1) with constrains concerning:  limits of the dispatchable generator: 2 ttt disp t gcgbxae  (2) maxmin gxggx ttt  (3) rampggramp tt  1 (4) ttt yxx  1 (5)  energy balance equation: ttttt biddswg  (6)  constrains concerning energy delivery:       24 1 24 1t t ititt rld (7)     n t ititt rld 1 9.0 (8)     n t ititt rld 1 1.1 (9)  demand response (dr) constrains: g. vuc is with the power systems department at politehnica university timisoara, v. parvan 2 blvd., timisoara, timis, 300223, romania (e-mail: gheorghe.vuc@upt.ro). m. litcanu was with politehnica university timisoara. she is now with qmb energ srl, eduard benes 6, timisoara, timis, romania (e-mail: marcela.litcanu@upt.ro). optimal management of a virtual power plant dan jigoria-oprea, gheorghe vuc, marcela litcanu t    vpp t dr tititit lr   exp1 (10) iso t dr t vpp t   (11)    24 1 min t it rr (12) where: iso t – forecasted price on the day-ahead market in the t period (€/mwh/h); vpp t – contracted price inside the vpp during t period (€/mwh/h); gmax – maximum production of dispatchable generator (mw); gmin – minimum production of dispatchable generator (mw); ramp – maximum ramp rate of the dispatchable generator (mw/min); start –dispatchable generator starting costs (€); wt – forecasted wind production for t period (mw); st – forecasted pv production for t period (mw); lit – forecasted load for consumer i for t period (mw); dt – forecasted demand for t period (mw); it – elasticity price factor for consumer i in t period; rmin – minimum acceptable level for total load reduction (mw); disp te – generation costs for dispatchable generator during t period (€/mwh/h); gt – output of the dispatchable generator during t period (mw); dr t – price for demand side reduction during t period (€/mw); profit – corresponding profit considering demand side reduction (€); rit – forecasted load reduction for consumer i during t period (mw); bidt – hourly bid on energy market during t period (mw); xt – binary variable that indicate the state (operational/shutdown) of the dispatchable generator during t period; yt binary variable that indicate if the dispatchable generator started during t period. the objective function takes into consideration the vpp offers on the market, which can be positive or negative. domestic consumers pay a fixed price according to bilateral agreements, vpp t equal to the levelized cost of energy (lcoe). the constrains for the dispatchable generator (2-5) include the square cost function (2), minimum and maximum generation levels (3), ramp up/down limits (4) and starting elements (5). the energy balance constraint (6) imposes the balance between the dispatchable generator production, the renewable sources production and consumption. an excess of generated power or stored energy gives the sign of the demand side on the market. the energy delivery constrains grant the necessary power covering all the demand. some deviations, positive or negative, are included in (8) and (9). also, for the delivery constrains, the demand reduction using dr are subtracted from the entire demand quantity. this model also presents the minimum aggregated demand reduction which can be accepted by the vpp (12). the reductions are not accepted when the entire quantity is smaller than the minimum acceptable level. iii. case studies all the presented case studies were conducted considering the entire romanian power system as a vpp, more exactly like a bulk vpp (bvpp) [7], but with an arbitrary separation of each constituents of the vpp, in order to test and use all the integrated facilities of the optimacev application [6], including individual influence of each member of the bvpp. the objective of the case studies was minimizing the financial losses (13) of the bvpp on the balancing market during 24 hours, losses which are determined by the errors between forecasted values for generation and real values of energy generation, errors which cannot be compensated by the dr.     24 1t bm ttconsprod wwfloss  (13) where: bm t is the difference of energy price between the balancing market and day-ahead market during t period (€/mw); wprod – deviation of real energy production from forecasted value; wcons – deviation of real energy consumption from forecasted value. several case studies were conducted for different characteristic days, from different seasons and with different structure of the history used for the forecast. the first case study is based on history data from 25-30 august 2014 and the focus day is 31 august. the second case study uses data from 5, 12, 19, 26 june, 3, 10 and 17 july and the focus day is 24 july. all the case studies are using real data from 2014 in order to compare the forecast results to real evolution of consumption and generation for each considered source. meanwhile, real market prices were used, both from the day-ahead market and balancing market. values for load category elasticity factors were used from literature. the data used to model the vpp were obtained from the romanian transmission system operator – transelectrica [8]. to compute the deviations of the forecast for each vpp component and the vpp imbalance, the real value of the consumption was used. the computation relations are:  %100 _    realcons realprod w vv ab (14)  %100 _    realcons consprod w ww dez (15) where: ab – corresponding deviation for the considered value (consumption, classical power plant production, etc.); vprog – forecasted value (consumption, classical generation, wind generation, pv generation, etc.); vreal – real value for the considered time period (consumption, classical generation, wind generation, etc.); wcons_real – real consumption value; dez – value of unbalance for the vpp as the result of deviation of forecasted values from the real ones. iv. results and discussions a. case study #1 the results of interest for the forecasted (focus) day (on hourly intervals) are presented in table 1. the results for case study #1 show high values of deviation for all hourly intervals for the classical power plant generation (22%), while for pv the values are practically null. this is due to the fact that the installed power in these plants is smaller compared to other energy sources (also see (14)). for the consumption component of the vpp, the deviations values are big. it is important to be noted that by aggregating the generation and consumption components in the vpp, the deviations per component are compensated and for all the vpp the maximum value of deviation is 21% and only for a few hourly intervals. another fact to note is that for 1 to 8 hour interval, the positive deviation of vpp is a “consumption excess” type and requires a reaction to decrease the value of the load. to minimize the unbalances of vpp, the r reaction of demand reduction (completely or partially), is used as necessary. reducing the value of the unbalances have a positive influence for the vpp profit due to the fact that any unbalance must be covered using the balancing market, where prices are least favorable than on the day-ahead market. the unbalance on vpp can reach the value 0 after considering the demand response for hourly interval 9, 11 and 16 to 20, because the demand response was greater or at least equal with the necessary needed to bring the unbalance to 0. table i vpp unbalance for case study #1 without and with dr hourly interval classical power plant deviation wind deviation pv deviation consumption deviation unbalance vpp  dr r iso unbalance vpp with dr 1 8.2% 13.3% 0.0% 3.7% 17.8% 141.30 370.00 179.00 10.5% 2 11.8% 15.8% 0.0% 7.8% 19.7% 140.80 360.00 178.00 12.2% 3 9.8% 17.1% 0.0% 5.7% 21.2% 138.90 326.00 174.00 14.3% 4 12.2% 15.3% 0.0% 10.3% 17.2% 130.70 257.50 157.00 11.7% 5 12.7% 11.8% 0.0% 10.3% 14.2% 129.20 243.50 154.00 9.0% 6 12.0% 10.7% 0.0% 9.6% 13.1% 126.30 211.30 148.00 8.5% 7 18.0% 10.5% 0.0% 18.6% 9.8% 126.30 224.30 148.00 4.9% 8 8.2% 6.6% -0.1% 8.0% 6.6% 126.80 249.70 149.00 2.0% 9 20.6% 5.5% -0.6% 24.8% 0.7% 130.70 305.40 157.00 0.0% 10 22.6% 6.0% -0.6% 30.4% -2.4% 95.00 60.70 174.00 -2.4% 11 21.5% 7.4% -0.4% 28.4% 0.0% 134.60 371.70 165.00 0.0% 12 20.0% 6.1% 0.0% 28.0% -1.9% 95.00 64.50 168.00 -1.9% 13 17.4% 6.4% 0.2% 24.3% -0.3% 95.00 63.00 157.00 -0.3% 14 19.7% 6.9% 0.2% 28.3% -1.5% 95.00 65.40 154.00 -1.5% 15 20.8% 7.8% -0.2% 29.7% -1.4% 95.00 64.10 148.00 -1.4% 16 20.9% 7.2% 0.3% 27.5% 0.8% 121.80 213.40 139.00 0.0% 17 17.6% 7.9% -0.5% 22.5% 2.5% 121.90 206.50 139.00 0.0% 18 2.8% 5.7% -1.0% 3.7% 3.8% 129.20 295.00 154.00 0.0% 19 17.4% 6.6% 0.3% 22.4% 1.9% 134.60 367.60 165.00 0.0% 20 15.7% 4.0% 0.4% 19.4% 0.7% 135.90 422.60 174.00 0.0% 21 12.6% 2.4% 0.1% 17.6% -2.6% 95.00 69.00 194.00 -2.6% 22 11.9% 2.7% 0.0% 16.9% -2.4% 95.00 70.30 229.00 -2.4% 23 10.9% 2.3% 0.0% 14.9% -1.7% 95.00 63.50 188.00 -1.7% 24 12.7% 4.0% 0.0% 15.9% 0.7% 141.80 425.00 180.00 0.0% the balance price for the demand response, dr for intervals 10 and 21-23 is smaller than the internal energy price for consumption, vpp, and signals the necessity to stimulate the growth of internal consumption to reduce the unbalance. for all occurrences, the demand response was used to reduce the unbalance. another thing that is noted is the negative value of the deviation of vpp unbalance for interval 10, 12-15 and 21-23. this is a “low consumption” type, which means that is necessary a demand response in order to increase the consumption, a signal given even by the demand response balance price. regarding minimizing the financial losses, from table 2 it can be seen very clearly that using dr reduces supplementary costs determined by acquiring or selling energy to the balancing market during unbalance hourly intervals. the economy is about 844398.9 ron from 1448759.3 to 604360.4 ron. the presented results also emphasize the fact that deviation of vpp when using demand response is the one who better attenuates the individual unbalance of their components. table ii dr influence on financial losses for case study #1 hourly interval unbalance  witihout dr unbalance costs [ron/mwh] supplementary cost unbalance influence without dr unbalance  with dr dr supplementary cost unbalance influence with dr [mw] excess deficit [ron] [ron/mwh] [%] [ron] [mw] [ron [ron/mwh] [%] 1 -906.91 5.38 198.00 118450.91 23.25 14.5% -536.91 -370.00 83011.4 16.29 10.18% 2 -950.73 39.57 243.49 92860.78 19.23 12.9% -590.73 -360.00 64643.4 13.39 8.98% 3 -999.23 0.10 213.78 128521.93 27.30 18.3% -673.23 -326.00 100244.5 21.29 14.29% 4 -801.29 0.10 214.62 64196.18 13.76 9.2% -543.79 -257.50 80970.0 17.35 11.64% 5 -662.18 0.10 213.12 18747.80 4.01 2.7% -418.68 -243.50 63178.8 13.52 8.95% 6 -605.25 0.10 221.68 1799.65 0.39 0.2% -393.95 -211.30 66927.6 14.45 8.50% 7 -446.25 0.10 251.69 11680.67 2.57 1.3% -221.95 -224.30 44028.3 9.69 4.88% 8 -360.72 0.10 258.83 14157.89 2.61 1.1% -111.02 -249.70 25412.0 4.68 2.04% 9 -32.54 0.10 306.95 50763.69 10.38 4.1% 0.00 -32.54 0.0 0.00 0.00% 10 118.20 0.10 317.93 80914.05 16.38 7.0% 118.20 0.00 9920.8 2.01 0.86% 11 -2.37 0.10 310.29 63125.76 12.52 6.0% 0.00 -2.37 0.0 0.00 0.00% 12 97.05 0.10 290.83 54392.55 10.75 5.2% 97.05 0.00 8233.2 1.63 0.79% 13 14.49 0.10 292.78 36625.46 7.19 3.5% 14.49 0.00 1230.2 0.24 0.12% 14 76.41 0.10 304.85 58899.04 11.50 6.1% 76.41 0.00 8852.5 1.73 0.91% 15 67.90 0.10 299.80 53363.86 10.65 5.6% 67.90 0.00 7523.2 1.50 0.79% 16 -42.38 0.10 281.81 31384.22 6.22 3.3% 0.00 -42.38 0.0 0.00 0.00% 17 -127.81 0.10 278.55 19984.04 3.93 2.1% 0.00 -127.81 0.0 0.00 0.00% 18 -229.42 0.10 278.07 20381.66 3.38 1.8% 0.00 -229.42 0.0 0.00 0.00% 19 -101.35 0.10 260.70 61196.76 11.70 7.7% 0.00 -101.35 0.0 0.00 0.00% 20 -36.50 0.10 255.97 58790.76 10.89 5.9% 0.00 -36.50 0.0 0.00 0.00% 21 150.53 0.10 259.25 91217.77 15.47 8.6% 150.53 0.00 11929.4 2.02 1.12% 22 142.29 0.10 269.88 133859.31 22.16 14.9% 142.29 0.00 17200.5 2.85 1.91% 23 94.73 0.10 235.70 106233.32 19.13 16.1% 94.73 0.00 11054.5 1.99 1.67% 24 -37.83 0.10 232.78 77211.25 14.95 12.6% 0.00 -37.83 0.0 0.00 0.00% b. case study #2 the results of interest for the forecasted (focus) day (on hourly intervals) are presented in table 3. the results for case study #2 show deviation values smaller than 10% for the majority of hourly intervals, while for the pv the values are practically null. for the consumption component of the vpp the deviation values are also smaller (with a maximum value of 14.6%), the large values appearing only in three intervals. again, by aggregating the generation and consumption components in the vpp, the deviations per component are compensated and for all the vpp the maximum value of deviation is 12% and only for a few hourly intervals. the positive values for deviation during 3-7, 9-13 and 15-24 hourly intervals is a “consumption excess” type and requires a reaction to decrease the value of the load. to minimize the unbalances of vpp, the r reaction of demand reduction (completely or partially), is used as necessary. also for this case, reducing the value of the unbalances have a positive influence for the vpp profit due to the fact that any unbalance must be covered using the balancing market, where prices are least favorable than on the day-ahead market. for interval 2, 8 and 14, using dr the unbalance can be eliminated (the final value of the unbalance is 0) due to dr contribution – its availability was greater or at least equal with the necessary value to compensate the unbalance. for this case study, only for the first hourly interval the deviation has a negative value – low consumption unbalance. this signals the fact that an increase of consumption reaction is needed. table iii vpp unbalance for case study #2 without and with dr hourly interval classical power plant deviation wind deviation pv deviation consumption deviation unbalance vpp  dr r iso unbalance vpp with dr 1 0.2% 3.7% 0.0% 4.4% -0.5% 149.0 -0.5% 2 0.0% 4.6% 0.0% 3.6% 1.0% 116.9 131.9 129.0 0.0% 3 -12.7% 3.9% 0.0% -13.1% 4.3% 112.0 73.5 119.0 3.1% 4 2.4% 2.2% 0.0% 4.1% 0.6% 107.0 21.3 109.0 0.1% 5 0.0% 1.8% 0.0% 1.0% 0.8% 107.0 20.9 109.0 0.4% 6 -0.3% 5.2% 0.0% 0.7% 4.2% 109.3 44.7 113.6 3.3% 7 1.3% 4.9% 0.0% 3.1% 3.1% 116.9 131.7 129.0 0.7% 8 1.4% 3.4% 0.2% 1.7% 3.2% 129.2 288.3 154.0 0.0% 9 0.3% 5.0% 0.5% -1.3% 7.1% 139.4 437.9 175.0 0.4% 10 -1.6% 6.2% 0.9% -3.7% 9.2% 141.3 463.7 179.0 2.3% 11 0.3% 6.3% 0.8% -0.8% 8.2% 141.3 479.1 179.0 1.1% 12 1.6% 7.2% 1.0% 1.0% 8.7% 141.3 489.3 179.0 1.5% 13 -10.1% 6.2% -0.4% -14.6% 10.3% 141.3 482.6 179.0 4.1% 14 3.8% 6.8% -0.4% 3.8% 6.3% 141.3 507.9 179.0 0.0% 15 -0.1% 5.7% -1.1% -1.9% 6.3% 136.5 417.7 169.0 0.2% 16 1.6% 5.1% -0.1% 0.4% 6.3% 131.1 344 158.0 1.1% 17 -0.3% 8.5% -0.4% -2.3% 10.0% 131.1 340.2 158.0 4.9% 18 0.2% 10.6% -0.4% -1.2% 11.5% 136.5 408.5 169.0 5.4% 19 -0.3% 10.5% -0.1% -1.5% 11.7% 136.5 403.2 169.0 5.5% 20 0.2% 9.3% -0.2% -0.9% 10.2% 136.5 405.7 169.0 4.0% 21 5.7% 10.3% 0.0% 6.3% 9.8% 136.5 441 169.0 3.2% 22 2.7% 8.6% 0.0% 2.0% 9.3% 136.5 438.8 169.0 2.9% 23 -9.5% 5.7% 0.0% -14.3% 10.6% 129.2 321.3 154.0 6.5% 24 2.5% 8.6% 0.0% 2.7% 8.4% 129.2 299.9 154.0 3.5% table iv dr influence on financial losses for case study #2 hourly interval unbalance  witihout dr unbalance costs [ron/mwh] supplementary cost unbalance influence without dr unbalance  with dr dr supplementary cost unbalance influence with dr [mw] excess deficit [ron] [ron/mwh] [%] [ron] [mw] [ron [ron/mwh] [%] 1 29.4 30 344 4369.5 32.74 16.8% 29.38 0.00 4369.5 0.78 0.40% 2 -51.4 34.26 356.78 7858.5 32.25 17.2% 0.00 -51.36 0.0 0.00 0.00% 3 -262.5 75.79 394 24435.9 27.99 16.6% -188.97 -73.50 17593.0 2.90 1.72% 4 -28.8 50.22 384 3448.8 41.21 24.2% -7.50 -21.30 897.7 0.18 0.10% 5 -42.0 77.04 390 4026.3 40.32 23.3% -21.10 -20.90 2022.8 0.39 0.23% 6 -217.3 67.4 390 26899.0 26.65 13.9% -172.60 -44.70 21365.6 4.12 2.16% 7 -168.8 67.51 400 23682.3 29.36 14.1% -37.11 -131.70 5206.2 0.97 0.47% 8 -187.2 78.1 431 30863.5 29.69 12.2% 0.00 -187.15 0.0 0.00 0.00% 9 -460.8 30 407 101829.6 16.29 6.5% -22.87 -437.90 5053.7 0.78 0.31% 10 -616.7 30 415 136285.6 9.73 3.9% -152.98 -463.70 33807.9 5.05 2.01% 11 -551.8 30 413 118630.7 9.98 4.1% -72.67 -479.10 15624.2 2.33 0.95% 12 -588.9 30 408 122490.1 6.82 2.9% -99.59 -489.30 20715.7 3.08 1.29% 13 -806.7 30 400.28 169403.4 1.23 0.5% -324.12 -482.60 68062.2 8.67 3.61% 14 -430.4 30 401 83920.1 9.66 4.3% 0.00 -430.36 0.0 0.00 0.00% 15 -431.9 30 380 81626.1 6.66 3.0% -14.18 -417.70 2680.8 0.39 0.18% 16 -415.7 30 382 73169.7 6.10 3.0% -71.74 -344.00 12625.7 1.91 0.93% 17 -670.3 30 378 117976.4 2.42 1.2% -330.12 -340.20 58101.2 8.67 4.21% 18 -764.0 30 381 148974.0 4.79 2.1% -355.47 -408.50 69316.5 10.48 4.66% 19 -765.0 30 388 152235.1 1.42 0.6% -361.80 -403.20 71998.3 10.98 4.80% 20 -669.6 30 399 140612.3 6.08 2.5% -263.91 -405.70 55419.3 8.46 3.52% 21 -651.0 30 405 143865.5 8.57 3.4% -209.98 -441.00 46404.5 6.99 2.78% 22 -636.8 30 413 140731.2 10.11 4.0% -197.99 -438.80 43756.4 6.36 2.53% 23 -826.0 30 411 161060.4 2.92 1.3% -504.65 -321.30 98406.9 12.62 5.61% 24 -512.9 30 333 92329.3 3.72 1.8% -213.04 -299.90 38347.3 6.31 3.01% regarding minimizing the financial losses, from table 4 it can be seen very clearly that using dr reduces supplementary costs determined by acquiring or selling energy to the balancing market during unbalance hourly intervals. the economy is about 1418947.9 ron from 2110723.3 to 691775.4 ron. v. conclusion implementing the concept of vpp, determine the growth of power system benefits, due to using more efficient the distributed generation units, hence a greater operation efficiency. in this case, distributed generation can become more visible, can have a better access to energy markets and also can maximize the opportunities regarding incomes from selling the energy and reducing the environment pollution by using fewer classical power plants. the vpp can be considered an observable instrument for optimal solving of renewable energy sources integration and the case studies presented emphasize this aspect. the results from the case studies prove that including the original elements in the vpp management meaning considering the level of bvpp and the level of additional optimization by minimizing the financial losses due to acquiring energy from the balancing market, is justified and leads to better performance for the vpp. even more, it is again confirmed the fact that the vpp can be the favorable element in the power grid evolution towards “smart grid”. references [1] l. nikonowicz, j. milewski, (2012). virtual power plants – general review: structure, application and optimization, journal of power technologies. 92 (3), pp. 135-149. [2] p. andersen, b. poulsen, m. decker, c. traeholt, j. ostergaard, “evaluation of a generic virtual power plant framework using service oriented architecture”, 2nd ieee international conference on power energy (pecon 08), johor baharu, malaysia, december 1-3, 2008, pp. 1212-1217. [3] g. kaestle, “virtual power plants as real chp-clusters: a new approach to coordinate the feeding in the low voltage grid”, 2nd international conference on integration of renewable and distributed energy resources, napa, ca, usa, 4-8 december, 2006. [4] s. you, developing virtual power plant for optimized distributed energy resources operation and integration, phd. thesis, department of electrical engineering, technical university of denmark, september 2010. [5] a. mnatsakanyan, s. kennedy, “optimal demand response bidding and pricing mechanism: application for a virtual power plant”, 1st ieee conference on technologies for sustainability (sustech), portland, or, usa, 1 – 2 august 2013, pp. 167-174,. [6] l. marcela, virtual power plant management optimization optimacev, phd. thesis, ed. politehnica, timisoara, romania, 2015 (in romanian). [7] r.a. ahangar, a. sheykholeslami (2014). bulk virtual power plant, a novel concept for improving frequency control and stability in presence of large scale res, international journal of mechatronics, electrical and computer technology, 4 (10), pp. 1017-1044. [8] cntee transelectrica s.a., romanian trasmission system operator, productie, consum, sold. available: http://www.transelectrica.ro dan jigoria-oprea (m’08) was born in caransebes, romania on the 31st of october 1983. he received the b.s. (2007) and phd. (2010) in power systems engineering from politehnica university timisoara, romania. from 2007 to 2010 he was a phd. student and research assistant at the power systems department of politehnica university timisoara. from 2010 to 2012 was a teaching assistant and since 2012 is an assistant professor/lecturer at the power systems department of politehnica university timisoara. he is the author of over 45 articles, 5 books and member in 10 research contracts. his research interests include power systems optimization and operation, renewable energy sources integration and energy market. mr jigoria-oprea is also a member of sier, the romanian society of power systems engineers. gheorghe vuc (m’05) was born in timisoara, romania, in 1958. he received the b.s. (1984) and the ph.d. degree (1998) in power systems engineering from politehnica university timisoara, romania. from 1984 to 1990 he was operational engineer at romanian tso (transelectrica). from 1990 to 1998 he was a phd. student and assistant professor at the power systems department of politehnica university timisoara. from 1998 to 2006 was a lecturer and since 2006 is an associate professor at the power systems department of politehnica university timisoara. he is the author of over 68 articles, 4 books and member in 12 research contracts. his research interests include power systems optimization and operation, renewable energy sources integration and energy market. mr. vuc is also a member of sier, the romanian society of power systems engineers. marcela litcanu was born in otelu rosu, romania on the 30th september 1986, received the phd. in power systems engineering from politehnica university timisoara, romania, in 2015. previously, she recieved a bachelor and master degree in engineering and management from faculty of management in production and transportation, at politehnica university of timisoara. from 2011 to 2015, he was a research assistant and phd. student with the power systems department at politehnica university of timisoara. currently, she is with qmb energy srl from timisoara, romania. she is an author of 8 articles. her research interest includes energy market and power system operation.  transactions on environment and electrical engineering issn 2450-5730 vol 1, no 3 (2016) © sobhy s. dessouky, ahmed e. kalas, r.a.abd el-aal & abdel moneim m. hassan  abstract— dissolved gas-in-oil analysis (dga) is a sensitive and dependable technique for the detection of incipient fault condition within oil-immersed transformers. when the mineral oil is subjected to high thermal or/and electrical stresses, it decomposes and, as a result, gases are generated. this paper presents modification of duval triangle dga diagnostic graph to numerical method that is easy to use for diagnosing and a matlab program. to study such as the following evaluation. this evaluation is carried out on dga data obtained from three different groups of transformers each group are two identical transformers. a matlab program was developed to automate the evaluation of duval triangle graph to numerical modification, also the fault gases can be generated due to oil decomposing effected by transformer over excitation which increasing the transformer exciting current lead to rising the temperature inside transformer core beside the other causes. index terms— dissolved gas analysis ) dga), mineral oil, decomposition, degradation, and transformer condition. i. introduction issolved gas analysis (dga) is a popular diagnostic technique that is used to detect incipient faults in oil-filled power transformers [1]. by using dga data, transformer criticality can be identified with proposing the proper maintenance action [2]. several methods were proposed to diagnose incipient faults based on dga. these methods are key gas method, rogers's ratio methods, duval triangle method, doernenburg ratio method, basic gas ratio, and artificial intelligence based methods. the key gas method identifies the key gas for each type of fault and uses the percent of this gas to diagnose the fault as suggested by ieee standard c57.104 [3]. the percent amount of gas is obtained in terms of the total combustible gases (tcg). the main disadvantage of this method is that the interpretation sobhy s. dessouky, electrical engineering dept. faculty of engineering, port-said university. port said, egypt (e-mail: sobhyserry@yahoo.com). ahmed e.kalas , electrical engineering dept. faculty of engineering, portsaid university. port said, egypt (e-mail: kalas_14@yahoo.com). by the individual gases is difficult in practice since each incipient fault produces traces of other gases in addition to the key gas of such fault. the ratio methods for fault diagnosis use certain ratios of dissolved gas concentrations according to combinations of codes [4, 5]. an incipient fault is detected when a code combination matches with the code pattern of the fault. the most widely used ratio methods are the doernenburg ratio method, rogers ratio method, and iec standard. six gas ratios have been used by different methods. the major drawback of ratio methods is the “no decision” problem associated with some cases that lie out of the specified codes. in recent years, many researchers have studied the application of artificial intelligence based techniques for transformer fault diagnosis. these techniques include expert systems, fuzzy logic, artificial neural networks or mixed techniques [6, 7]. however, these methods are too complicated to be implemented practically on a wide range. this paper investigates the new aspects, accuracy and consistency of these methods in interpreting the transformer condition. ii. dga to diagnose transformer faults when an incipient fault occurs, either thermal or/and electrical, a number of gases are generated and dissolved into the oil. these gases are mainly h2, ch4, c2h2, c2h4 and c2h6. in addition co and co2 will exist if cellulose degradation is involved, based on the type and amount of generated gases [1, 8-9]. a. duval triangle (dga) diagnostic graph method m. duval. proposed another diagnostic method to overcome this limitation, well known as duval triangle. this method is based on a triangle graphical representation to visualize the different cases for oil-insulated high-voltage equipment (mainly transformers), fig. (i) provides a graphical method of identifying a fault. it uses a three-axis coordinate r.a.abd el-aal, electrical engineering dept. faculty of engineering, portsaid university. port said, egypt (e-mail: ramadanhv@yahoo.com). abdel moneim m. hassan, abo-sultan steam power plant, ismailia. egypt (e-mail: abdelmoname333@yahoo.com). modification of duval triangle for diagnostic transformer fault through a procedure of dissolved gases analysis sobhy s. dessouky a, ahmed e.kalas b, r.a.abd el-aal c, abdel moneim m. hassan d a, b, c electrical engineering dept. faculty of engineering, port-said university. port said, egypt d abo-sultan steam power plant, ismailia. egypt d mailto:sobhyserry@yahoo.com mailto:1@yahoo.com mailto:ramadanhv@yahoo.com mailto:abdelmoname333@yahoo.com system, where concentrations of ch4, c2h4 and c2h2 are used as coordinates, and the likely fault falls within one of the fault regions of the triangle. the various regions within the duval triangle are given in table (i) [10-13]. for example if c2h2 = 0.07, ch4 = 0.2 and c2h4 = 0.73. the fault diagnostic is t3 (thermal fault t > 700 °c), and if c2h2 = 0.36, ch4 = 0.32 and c2h4 = 0.32, the fault diagnostic is d2 (high-energy electrical discharge), as shown in fig (i). fig. 1. duval triangle table i. fault code a. duval triangle graph to numerical method in this paper, we developed a matlab program to automate the evaluation of duval triangle graph to numerical modification. table (ii) shows the modification of duval triangle dga diagnostic graph to numerical method. for example if c2h2 = 0.1, ch4 = 0.3 and c2h4 = 0.6. we can use table (ii) easy to determine the fault diagnostic (thermal fault t > 700 °c), and if c2h2 = 0.36, ch4 = 0.32 and c2h4 = 0.32, the fault diagnostic is (high-energy electrical discharge), the same results as in the previous example. table ii. modification of duval triangle (dga) diagnostic graph to numerical method iii. case study dissolved gas analysis the case study carried out from three different groups of transformers each group are identical in abu-sultan steam power plant. fig. (2) shows the schematic diagram configuration for transformers under testing. the first group of transformers are three single phase 192 mva, 15/220 kv, off l.t.c. the second group of transformers are three phase 16 mva, 220/6.3kv, on.l.t.c, and the third group of transformers are three phase 16 mva, 15/6.3/6.3 kv, on.l.t.c. the rating and (dga) testing results for the abovementioned power transformer are shown in tables (iii, iv). fig. 2. schematic diagram for transformers under evaluation pd partial discharge t1 low-range thermal fault (below 300 °c) t2 medium-range thermal fault (300-700 °c) t3 high-range thermal fault (above 700 °c) d1 low-energy electrical discharge d2 high-energy electrical discharge dt indeterminate thermal fault or electrical discharge. c2h2% ch4% c2h4% fault 0.00 0.02 0.98 1.00 0.00 0.02 partial discharge (electrical fault) 0.00 0.04 0.46 0.80 0.20 0.50 thermal fault 300 < t < 700 °c 0.76 0.98 0.02 0.20 thermal fault t < 300 °c 0.00 0.15 0.00 0.50 0.50 1.00 thermal fault t > 700 °c 0.04 0.13 0.47 0.96 0.00 0.40 mixtures of thermal and electrical faults 0.13 0.29 0.21 0.56 0.40 0.50 0.15 0.29 0.00 0.35 0.50 0.85 0.13 0.29 0.31 0.64 0.23 0.40 discharge of high energy (electrical fault) 0.29 0.77 0.00 0.48 0.23 0.71 table iii. rating of power transformer under testing iv. diagnostic method used by modification system. the diagnostic methods for dga are used by a numerical method, the matlab program diagnoses output for the under testing transformers. table (v) shows application of the faults diagnosed by various methods, which indicate that all transformers are thermal faults. v. results and discussion comparison of various methods as shown in the table (v), a thermal fault in oil within all transformers is diagnosed for all five methods. where winding temperature do not exceed 95°c and oil temperature do not exceed 85°c for all transformers during normal operation. moreover, the possible collapse of cooling system during operation in this case is too small and there is no increase in the viscosity of the oil, as it is clear in the results of chemical analysis of samples oil and no wax materials. however, there is an important factor is the increased over excitation due to reduction of generator speed when some of the generating units from the network goes out during normal operation or the frequency disturbances that occur when large loads are connected to the electrical network system. over-excitation or/and under frequency protection may be or may be not operate depends on the response of power system control. the under frequency relay operate at 47.5 hz with time lag 0.5 sec and over excitation relay operate at v/hz = 1.1pu for 45 sec time lag or v/hz =1.18 pu for 2 sec time lag at generators. table iv. (dga) testing results t ra n sf o rm e r n a m e o p e ra ti n g d a te r a te d p o w e r m v a r a te d v o lt a g e k v n u m b e r o f p h a se s o il t y p e main transformer unit no. 1 ( tr1) 19/3/1983 192 15/220 3 s in g le p h a se m in e ra l o il n a p h th e n ic main transformer unit no. 2 ( tr2) 15/8/1983 start up transformer a ( tr3) 19/3/1983 16 220/6.3 3 p h a se s start up transformer b ( tr4) 15/10/1984 aux. transformer unit no. 1 ( tr5) 19/3/1983 16 15/6.3/6.3 3 p h a se s aux. transformer unit no. 2 ( tr6) 15/8/1983 t ra n sf o rm e r & s a m p le s d a te m a in t ra n sf o rm e r u n it n o . 1 p h ( b ) fr o m 0 8 /0 5 // 2 0 1 3 t o 2 7 /1 1 /2 0 1 3 m a in t ra n sf o rm e r u n it n o . 2 p h ( b ) fr o m 0 8 /0 5 /2 0 1 3 t o 0 5 /1 1 /2 0 1 4 s ta rt u p t ra n sf o rm e r a fr o m 0 8 /0 5 /2 0 1 3 to 0 6 /0 5 /2 0 1 4 s ta rt u p t ra n sf o rm e r b fr o m 0 7 /0 4 /2 0 1 3 to 2 7 /1 1 /2 0 1 3 a u x . tr a n sf o rm e r u n it n o . 1 fr o m 0 8 /0 5 /2 0 1 3 to 2 9 /0 3 /2 0 1 5 a u x . tr a n sf o rm e r u n it n o . 2 f ro m 0 7 /0 4 /2 0 1 3 to 0 2 /0 4 /2 0 1 4 t o ta l c o m b u st ib le g a se s ( t .c .g ) w it h o u t c 3 h 6 & c 3 h 8 274 477 164 592 98 249 219 426 246 429 193 400 h y d ro g e n c o m b u st ib le g a se s h2 9 7 3 16 1 19 5 6 14 28 7 35 h y d ro c a rb o n s ch4 25 48 15 37 2 4 19 61 48 49 9 12 c2h2 0 0 0 0 0 0 0 0 0 0 0 0 c2h4 5 2 1 12 2 8 5 6 3 10 2 3 c2h6 12 29 10 50 1 3 57 142 28 45 2 3 c3h6 & c3h8 14 26 5 2 3 30 81 14 2 3 c a rb o n o x id e s co 223 392 135 477 91 215 132 212 154 297 173 348 n o n -c o m b u st ib le g a se s co2 2877 6052 775 4854 482 1324 848 1772 1632 3787 439 2581 n o n -f a u lt o r a tm o sp h e ri c g a se s o2 2042 2664 1633 3758 3432 5766 991 1911 1420 13615 1118 3300 n2 31551 38801 45633 90526 39302 56161 74493 88856 82762 137375 \ 30606 119152 if frequency decreases and the voltage is constant, the transformer core is heated. fig. (3) shown voltage, current and frequency of generating unit transformer number one at abusultan steam power plant from 17/5/2015 to18/5/2015, which indicate that frequency, reduced to 49.2 hz at voltage 14.85. kv. the rated generator voltage and frequency is 15 kv and 50hz respectively. so generator is over excitation =1.0061 pu. at unit, start up the voltage may be built to 15kv at generator frequency 48 hz then 1.042 pu over-excitations. disturbance in frequency is repeated from 18/5/2015 to 20/5/2015 in power system as shown in fig. (4) and affect all network transformers in this moment and there is an instantaneous decrease in power system frequency to 45.36 hz without operate under frequency or/and over-excitation relays because disturbance duration less than 0.5 sec as shown in fig. (5). transformers require an internal magnetic field to operate. the core of a transformer is designed to provide the magnetic flux necessary for rated load. an over-excitation condition occurs when this equipment is operated such that flux levels exceed design values. the voltage output of a transformer is a function of the rate of change of the flux and the number of turns in the output winding. e = n dφ/dt during normal power system operation. the voltage is sinusoidal and the rate of change is determined by the frequency, which is in turn determined by generator speed [14]. the equation shows core flux to be directly proportional to voltage and inversely proportional to frequency φ α v/f. the actual magnitude of flux in transformer core is can be quantified in terms of per unit volts / hertz. a generator or transformer operating at no load with rated voltage and frequency would have one per unit excitation. the same equipment operating at rated voltage and 95% frequency would have 1.0/0.95 = 1.05 pu flux or 1.05 pu excitation. over-excitation will result from high voltage at rated frequency and from rated voltage with low frequency. because over excitation is a function of voltage and frequency, it can occur without notice. transformers and generators can be subject to repeated over excitation by inappropriate operating. practices or operator error without a disruption to operations. the resulting thermal faults lead to oil decomposing to generate fault the practices or operator error without a disruption to operations. the resulting thermal faults lead to oil decomposing to generate fault gases h2, ch4 at temperature 120°c, c2h6 at temperature 150°c, c2h4 at temperature 300°c, and c2h2 at temperature 700°c. in addition, degradation of insulating material is cumulative. a transformer or generator that survives a serious over excitation event or many small events may fail because of a moderate event during normal service as all transformers under study. in addition, if voltage increased, at rated frequency, the exciting current increases, as shown in fig. (6). so tr1 through tr6 are effected by over excitation due to network normal operation but tr1, tr2,tr5, tr6 are effected by over excitation damage usually occurs during periods of off-frequency operation such as start up or shut down for unit transformer as shown in fig.(2) and table (vi). t ra n sf o rm e r n o . duval's triangle numerical modified p(96/4) basic gas ratio p(77/8) doernenburg ratio p(71/3) rogers ratio p(62/5) kay gas p(42/58) tr1 thermal fault t < 300°c thermal fault t < 300°c thermal decomposition slight overheating t <150 °c pyrolysis in cellulose tr2 thermal fault 300 < t < 700 °c thermal fault t < 300°c thermal decomposition slight overheating 150-200 °c pyrolysis in cellulose tr3 thermal fault t > 700 °c thermal fault of low temperature t <150°c cannot be applicable general conductor overheating pyrolysis in cellulose tr4 thermal fault t < 300 °c thermal fault t < 300°c thermal decomposition slight overheating 150-200 °c pyrolysis in cellulose tr5 thermal fault t > 700 °c thermal fault t < 300°c thermal decomposition cannot be applicable pyrolysis in cellulose tr6 thermal fault 300 < t < 700 °c cannot be applicable cannot be applicable general conductor overheating pyrolysis in cellulose table v. application of the fault diagnosed by various methods fig 3. voltage current and frequency for unit no.1 generator system frequency generator current generator voltage generator current fig 4. repeating disturbances in power system frequency fig 5. instantaneous decrease in power system frequency system frequency generator voltage system frequency voltage/hertz increased frequency reduction vi. conclusion. modification of duval triangle dga diagnostic graph to numerical method is easy to use for diagnoses and a matlab program. transformer thermal faults during dynamic load cycle due to temperature increase from over load, cooling system failure or trouble, fault currents and /or over excitation condition. over excitation, damage usually occurs during periods of off-frequency operation such as start up or shut down for unit transformer. in addition, the fault gases can be generated due to oil decomposing effected by transformer over excitation. transformers and generators can be subject to repeated over excitation by inappropriate operating practices or operator error without a disruption to operations. it's can be concluded also, the resulting thermal faults lead to oil decomposing to generate fault gases h2, ch4 at temperature 120°c, c2h6 at temperature 150°c, c2h4 at temperature 300°c, and c2h2 at temperature 700°c. the gas type and gas quantity depends on the intensity and duration of over-excitation. transformer diagnostic thereby results depends on the events inside evaluation interval or before evaluation time. tr1 tr2 tr3 tr4 tr5 tr6 n o rm a l a g in g d u e t o d y n a m ic l o a d c y c le o v e r t e m p e ra tu re fault currents overload &unbalanced load cooling system failure increased oil viscosity o v e r e x c it a ti o n unit startup maintain the set point voltage at low frequency χ χ χ χ unit shutdown field breaker fails to open when the generator trips χ χ χ χ over voltage at rated frequency the charging current for a highvoltage transmission line. χ χ χ χ χ χ power system disturbance loss of some units during operation or suddenly heavy load χ χ χ χ χ χ a c c e le ra ti n g a g in g n o rm a l o p e ra ti n g t e m p e ra tu re 8 0 1 2 0 ° c . moisture oxidation of the insulation and oils forms acids, acid attacks cellulose and accelerates insulation degradation, with moisture (pd) electrical stress can occur and more insulation degradation χ oxygen χ acidity χ fig 6. voltage increased, at rated frequency exciting current increase table vi. causes of thermal faults, normal and accelerated aging references [1] t. k. saha, “review of modern diagnostic techniques for assessing insulation condition in aged transformers”, ieee transactions on dielectrics and electrical insulation, vol. 10, pp. 903-917, 2003. [2] a. abu-siada and s. islam, “a new approach to identify power transformer criticality and asset management decision based on dissolved gas-in-oil analysis”, ieee transactions on dielectrics and electrical insulation, vol. 19, pp. 1007-1012, 2012. [3] “ieee guide for the interpretation of gases generated in oilimmersed transformers”, ieee standard c57.104-2008, 2009. [4] m. j. heathcote, the j & p transformer book, twelfth edition, reed educational and professional publishing ltd, 1998. [5] s. m. islam, t. wu and g. ledwich, “a novel fuzzy logic approach to transformer fault diagnosis”, ieee transactions on dielectrics and electrical insulation, vol. 7, pp. 177-186, 2000. [6] m. a. izzularab, g. e. m. aly and d. a. mansour, “on-line diagnosis of incipient faults and cellulose degradation based on artificial intelligence methods”, ieee international conference on solid dielectrics (icsd), pp. 767-770, 2004. [7] md umar farooque, shufali awani,shakeb akan "artificial neural network (ann) based implementation of duval pentagon"2015 international conference on condition assesment techniques in electrical systems (catcon) pp 46-50, 2015. [8] diaa-eldin a.monsour "-development of a new graphical technique for dissolved gas analysis in power transformers based on the five combustible gases"ieee transactions on dielectrics and electrical insulation, vol. 22, pp. 2507 2512, 2015. [9] alamuru vani and pessapaty sree rama chandra murthy" hybrid diagnosing techniques for analyzing dissolved gases in power transformers " issn 2006 9790, pp 33-34, 2015. [10] m. duval, “a review of faults detectable by gas-in-oil analysis in transformers”, ieee electrical insulation magazine, vol. 18, pp. 8-17, 2002. [11] nitin k. dhote1 and jagdish b. helonde" fuzzy algorithm for power transformer diagnostics" academic editor: m. onder efe, pp 1-2, 2013. [12] stefan tenbohlen , sebastian coenen , mohammad djamali " andreas müller diagnostic measurements for power transformers" academic editor: issouf fofana, energies, pp 2, 2016. [13] sherif s.m.ghoneim , ibrahim b.m.taha , nagy i.elkalashy "integrated ann-based proactive fault diagnostic scheme for power transformers using dissolved gas analysis" ieee transactions on dielectrics and electrical insulation, vol.23,no 3 , pp1838-1845, 2016 [14] l.g hewitson “protective relaying for power generation systems” book taylor & francis group, publishing ltd 2006. sobhy serry dessouky was born in dakahlie of egypt in 1946. he received the b.sc. degree (1970) and m.sc. (1977) in electrical engineering from suez canal university in helwan university respectively. dr. dessouky received the ph. d. degree from tu, dresden, german in 1982. from oct. 1970 to 1975, he was joined faculty of engineering, suez canal university, as demonstrator. he worked as demonstrator from 19751977 in faculty of engineering, helwan university. in 1977, he worked as lecturer assistant in electrical engineering department, faculty of engineering, suez canal university. from 1983 to 1987, he worked as assistant professor (lecturer), in electrical engineering department, faculty of engineering, suez canal university, port said campus. in 1987, he promoted as associate professor in the same department. in 1991, dr. dessouky became a full professor of electrical power and h.v engineering. he was a member in ieee from 1996. in parallel, he worked as a department chair, vice dean for community affairs and environment, and director of engineering research center for developing and technological planning in suez canal university. ahmed e.kalas received the b.sc. degree in electrical engineering from the suez canal university with honor first rank in egypt 1982, m.sc. degree (power electronic and electrical drives), from the suez canal university, egypt 1987, ph. d. degree (power electronic and electrical drives) from gdansk university, poland 1994 from 1994 up to 2010 he worked as a lecture in electrical engineering at suez canal university, from 2010 up to now he worked as a lecture in electrical engineering at port said university research contributions, as well as his on-going efforts/investigations in the area of ac drives and power electronics, can be classified into the following topics: control of electric machines; vector control, nonlinear control, adaptive control, model predictive control, double feed induction motors ,dtc -power electronic converters, two-level and multilevel, matrix converter, zs-artificial intelligence in machines and power electronics control, fuzzy logic, neural networks -renewable energy conversion for pv and wind systems, maximum power point tracking -fault detection diagnosis in electrical machines and drives. r.a.abd el-aal was born in egypt, 1971. he received the b.sc. degree (1996) and m.sc. (2002) in electrical engineering from suez canal university. he received the ph. d. degree in h.v engineering from port said university in 2008. he works as lecture in electrical engineering dept., port said university, egypt. his research interests are h.v engineering and power system protection. abdel moneim m. hassan. was born in ismailia of egypt in 1963. he received the b.sc. degree (1986) in electrical engineering from helwan university. he works as general manager in abu sultan steam power plant 4*150 mw, from 1988 to 1998, he worked in operation department as operation engineer in abu sultan power plant, from 1998 to 2014, he worked as electrical maintenance, measuring and protection engineer in the same plant.  transactions on environment and electrical engineering issn 2450-5730 vol 1, no 3 (2016) © pan yunying, gu danzhen  abstract—wind energy is well known as a renewable energy because its clean and less polluted characteristic, which is the foundation of development modern wind electricity. to find more efficient wind turbine is the focus of scientists around the world. compared from conventional wind turbines, superconducting wind turbine generators have advantages at zero resistance, smaller size and lighter weight. superconducting wind turbine will inevitably become the main trends in this area. this paper intends to introduce the basic concept and principle of superconductivity, and compare form traditional wind turbine to obtain superiority, then to summary three proposed machine concept.while superconductivity have difficulty in modern technology and we also have proposed some challenges in achieving superconducting wind turbine finally. index terms—superconducting, wind turbines, torque, types i. introduction a. the background of wind turbine generator wind power generation technology originated in europe, dan wheat, holland, germany and other countries where have proposed it more than 20 years. the united states, canada and other countries also focus on wind power in recent years. with the development of the global economy, the wind power market also develops rapidly. in the past 5 years, the world wind power market annual growth at the rate of 40%. and the cost of wind power decrease increasingly. european wind energy association estimated that until 2020, the cost of wind power generation will decrease to 3cent/kwh. on the end of 2006, the world wind power installed capacity was 74 wg, while, until 2010, the value have reached 150gw. global wind power installed capacity in 2020 will reach 12.31 million kilowatts, which will account for the global power generation of 12%. therefore, wind power will be the mainstream for generation and become mature technology and emerging industries. wind power is a clean, renewable energy to generate the electricity. although its reliability is not very well, wind turbine market has been growing rapidly. while, due to the interference of surrounding residents and low wind speed, the number of the wind power sites development are limited. on the contrary, the off shore wind power has high speeds and low interference. therefore, it is more competitive. nowadays, 1000mw wind power is under development. the key of wind power is the wind turbine generator. currently, the common used wind turbine generators are double-fed induction generator (dfig) and direct-driven wind turbine generator. dfig is most common in the market. the generators consist of the motor and the cooling system. the motor is mainly composed by the stator, rotor and bearing system. stator winding is directly connected to the grid, the rotor windings are connected to the grid through a converter. the frequency, voltage, amplitude and phase of rotor windings’ power are adjusted by the inverter according to operational requirements automatically. the unit can achieve constant-frequency generation at different speed to meeting the electricity grid and load requirements. compared with dfig, direct-driven wind turbine generator has no gearbox. it is a generator that direct driven by wind, also known as the wind turbine generator with no gearbox. it uses a multi-pole motor directly connected with the impeller to drive, replacing the traditional parts of the gear box. [1] the conventional wind turbine generators have some drawbacks such as the low capacity and low economy. there are two methods to improve the wind turbine generator technology, one is improving the capacity, the other is using superconductors. therefore the scientists put forward an idea about superconducting wind turbine generators to overcome the problems on the conventional wind turbine generators. b. superconductor and its superconductivity superconductivity is a character of certain substances that are reduced to zero at a certain temperature (generally low temperature). in 1911 dutch physicist heike kamerlingh onnes found that the mercury will suddenly be a new state when the temperature down to 4.2k and its resistance is too low to measure. he called this new state of mercury as the superconducting state. and later, he found that many other metals also have superconductivity. a substance that occurs at a temperature below a certain temperature is called a superconductor. the direct current resistivity of the superconductor is suddenly disappeared at a certain low temperature, and is called the zero resistance effect. conductor had no resistance, current flows through the superconductor has no heat losses, therefore the current can be without resistance in a wire to form a strong current and generating a strong magnetic field. superconducting materials and superconducting technologies have a broad application prospect. due to the characteristics of the superconducting, more and more scientists try to applicate it in the turbine in order to improve the efficiency and advantage superconducting wind turbine generators pan yunying, gu danzhen shanghai university of electric power of wind power.[2] c. merits of superconducting wind turbine generators the emergence of superconducting technology and its rapid development make the large-scale offshore become possible. the characteristic of the superconducting wind turbine generators is that the power generation efficiency is high, the volume is small and the lightweight. the low resistance and even zero resistance of hts materials can effectively improve the power generation efficiency. in theory, the weight of the high temperature superconducting generators can be decreased to 1 / 2-1 / 3 of the conventional motor with the same capacity, which greatly reduce wind power construction and installation cost when the high power generators placed to high meters. if the size and the weight of the generator is maintained, the capacity of the high-temperature superconducting generator can be increased several times, effectively reduce the cost of wind power generation.[2]the reason is that superconducting wind turbine generators air gap magnetic field intensity is 2 times more than the traditional permanent magnet wind turbine air gap magnetic field intensity. improving the power density by increasing the magnetic flux density, thus the capacity of wind turbine generators can improve.[3] additionally, the superconducting wind turbine generators also have the advantages of low synchronous impedance, low noise, low harmonic content, simple maintenance and so on. ii. superconducting wind turbine generators a. the introduction of superconducting wind turbine the wind turbines rely on the input of mechanical energy. it is a machine improving gas pressure to guide the gas flow. and it is a driven fluid machine. the working principle of wind turbine is the same as the working principle of the turbine compressor. because of the low gas flow rate and the pressure has little change, it is unnecessary to consider the changes in the specific volume of gas, so handling the gas as incompressible fluid[3]. high temperature superconducting and its application technology were developing from 1986 which has been a mature technology. generally speaking, high temperature superconducting generators use high temperature superconducting magnets instead of ordinary copper wire coil as the excitation winding of the generators. due to the high current carrying capacity of hts materials (100-200 times as the carrying capacity of the same section of the copper conducting wire). additionally, hts wind turbine generators allow cancelling total or part of the magnetic circuit of the generators. the effective work magnetic field of the generators can reach a few t or more which is conducive to design a more compact, smaller generators. the geometry of superconducting wind turbine generators depend on the comparison of operational magnetic field strength r b and saturation field strength of the magnet steel in the generator. the generator with r b lower than the saturation limit is similar to the traditional wind turbines, where the magnetic flux path is shaped by the magnetic teeth in rotor and stator. however, the effect of steel is reduced above the saturation limit and resulting the sever hysteretic losses in the steel. therefore, when designing the superconducting wind turbine generators operating at field strength above the saturation limit of the steel, it is necessary to remove the steel in the rotor and the stator. for the magnetic field of the stator will rotate synchronously with the superconducting rotor coils, which ideally experience the time of magnetic field is constant and therefore only suffer from dc loss, so the superconducting wind turbine generators is operated as a synchronous machine[4]. b. the structure of superconducting wind turbine generators the most common topology of a superconducting wind turbine generator is illustrated in fig.1,which showing a multi-pole rotor based on superconducting race track coils[5]. fig.1 a superconducting multi-pole generator. reproduced from courtesy of converteam. it is a direct-drive high temperature superconducting wind turbine generator system with low speed. the machine consists of the stator back iron, stator copper winding, hts field coils, rotor core, rotor support structure, rotor cooling system, cryostat and external refrigerator, electromagnetic shield and damper, bearing, shaft and housing. the key of these all components is the arrangement of stator, rotor, cooling and gearbox. the stator back iron is holding the three phase stator cu winding. the coils are mounted on the rotor support structure, might holding an inner steel tube to confine the magnetic flux between the rotor and the stator and transfer the torque to the turbine shaft at room temperature. the rotor generally contains steel laminate, concentrating the magnetic flux to the air gap and a cryostat to maintain the good thermal insulation of the superconducting coils. c. design of superconducting wind turbine generators conventional wind turbine has resistive windings and the iron cores are highly-multipolar machines. but, in general, the superconducting wind turbine is different. superconducting machines requires thermal insulation between the very low temperature components and the room-temperature components. thus, it is inevitable produce a large air gap between the superconducting field coils and air-gap armature windings. and need a pole-pitch larger than the traditional machines. if the pole pitch is very small, the magnetic field generated by superconducting field magnets can not reach the armature windings sufficiently. here introducing a size and design of superconducting wind turbine generator introduced in [2]. fig. 2 displays a wind turbine generator designed from the finite element method analysis of the magnetic field. the rotor has 12 superconducting field coils generating 12-pole magnetic field. the coils have a simple racetrack shape with a rectangular cross-section. the stator has air-gap armature windings, which are double layer, distributed three-phase windings. fig 3 shows the outer and inner diameters of the armature windings which is 4.32 m and 3.84 m and the outer diameter of the superconducting field coil set is about 3.6 m. the racetrack superconducting coil is as large as 1.8 m (length)× 0.8 m (width) and its cross section is 0.18 m × 0.18 m as shown in fig. 4. fig.2a wind turbine generator designed from the finite element method analysis of the magnetic field. reproduced from reference [2] fig.3 dimensions of the designed superconducting wind turbine generators. reproduced from reference [2] fig.4 superconducting racetrack coil for field magnets. reproduced from reference[2] d. torque of synchronous superconducting generator the torque t of the superconducting generators can be determined by integrating the force f acting on each stator wire at air gap radius r, frdt ~ .moreover, the lorentz force on each wire bif s ~ .in addition, assuming the magnetic flux density in the air gap is harmonic, )cos()( 0  pbb  (1) 0 b is the peak flux density and p is the pole number. assuming a harmonic current distribution of the stator in the coordinate system of the rotor, ))(cos(   pai ss （2)  is the angular displacement of the current distribution relative to the rotor field distribution and s a is the rms armature loading in the units of [a/m]therefore, we can get the torque that )cos( 2 0 plrabt s  （3） l is the length of the machine.[1] iii. different types a. wind turbine with superconducting field windings recently, the most common superconducting wind turbine generator is the generator with superconducting field winding two concepts of superconducting wind turbine generator have been proposed. first is that only field windings are superconducting. second is that both field windings and armature windings are superconducting. fig.5 shows the axial-radial cross-section of superconducting synchronous generators, in which the field winding is superconducting and the armature winding is copper. the superconductor is thermal insulated from ambient and cold section is equipped with torque transfer element, which can not only insulate but also transfer the torque from cold area to warm area.[1] fig.5 axial-radial cross section of a superconducting synchronous machine. reproduced from reference [1] b. fully superconducting wind turbine generators fully superconducting wind turbine generators can be divided into two types. one is using lts which used widely. the other is using hts that only little projects use it.[1]. compared to conventional superconducting wind turbine, the merits of fully superconducting generator is that the air gap flux density and the electric loading is increased. moreover, if both armature and field windings operate at the same temperature, the air gap can be much smaller. because there is no thermal insulation between these two, the whole magnetic design of the machine can be more efficient [1]. however, the biggest problem of this concept is the ac losses that generated in the armature windings. nowadays, because of the poor cooling efficiency, the amount of ac losses in high field ac application is considered as prohibitive. c. high temperature superconducting wind turbine generators (htswtg) in the past, using low temperature superconductors (lts) to improve the traditional design of electric machines had been tried in many projects. and all these have shown the successful and advantages of superconducting wind turbine generators. however, there are still some challenges such as the cost of the refrigeration system and the temperature control of 2k-4k of lts. nowadays, with the development and discovery of hit materials, superconducting machines become more competitive from the commercial view. the temperature rang of hts is 30k-80k which is much higher than lts. moreover, a device that using hts has a wider temperature window, which has profound influences on the refrigeration complexity and efficiency, as shown by fig.6.by using hts, the refrigeration system and thermal insulation is more simple and efficiency than using lts and mgb2. fig.6 operation boundaries of different superconductors, where the critical current vanishes. the top bar indicated the typical efficiency of a cryogenic cooling system. reproduced from[6]and[7] actually, superconductors are rarely used in the stator because of the ac losses and it is difficult to build stator winding by hts coils. therefore, a radical design of a superconducting arrangement is not realistic. as a result, the htswtg is a synchronous generator consists of copper stator and superconductor rotor. hts machines can be divided to four different types[8]: type1. conventional stator and hts rotor with magnetic pole bodies. type2. conventional stator and hts rotor with non-magnetic pole bodies type3. airgap stator winding and hts rotor with magnetic pole bodies type4. airgap stator winding and hts rotor with non-magnetic pole bodies different types of the machines have different merits and characteristics. due to the minimized rotor loss and does not offer substantial reduction in weight and dimension,type1 has high efficiency. and type 2 can also get high efficiency but it needs more hts wires to build necessary flux density. compared with conventional stator,type3 has higher flux density at the airgap which reduce the mass and size of machine. however, the efficiency is reduced due to the rotor iron can operate highly saturated.type4 allows significant reduction in weight and dimension, and also minimizes potential high cost cold magnetic materials. but it requires more hts wires in use. wind power is a cost sensitive market and the capital cost is important in producing large machines. for offshore wind turbines, saving cost is more important than size reduction. therefore, the actual compromise is made between the low cost, low mass and high efficiency. for the direct-drive wind turbine generator, types 3 and 4 both can offer the lowest mass. type 4 may be a more cost-effective module for iron and 2g hts wires. converteam use type4 to design an 8mw 12rpm high temperature superconducting wind turbine generators. iv. produced wind turbine generators a. amsc——seatitan several machines were constructed in projects led by american superconductor (amsc). currently, amsc announced intentions to construct a direct drive wind turbine generator under the brand named seatitan. based on the ship propulsion motor technology and superconducting technology of united states navy projects, the output power of seatitan from amsc can be double than the largest wind turbine generator, up to 10mw.the key to get success is the high temperature superconducting direct drive engine with a diameter of 5m and the weight is 160t.in contrast, the diameter of a permanent magnet direct drive motor that producing the same output will reach 10m and weight is also more than 200t.moreover,the power density of superconducting wind turbine generators can achieve high power output and high economic.fig.7 shows 10mw seatitan offered by amsc. fig.7 10mw seatitan offered by amsc. reproduced from[1] amsc also claims that, although the focus is on the 10mw seatitan system, but it is also feasible to design 20mw wind turbine generator with the high temperature superconducting technology. b. ecoswing last year, ecoswing superconducting wind turbine generators produced by envision and funded 1 billion yuan by eu and its member state. significantly different from the other superconducting technologies, the power of superconducting wind turbine generator is only 3.6mw. the purpose of ecoswing is not the large capacity, but to promote the reduction of the cost of wind power. compared with the traditional wind turbine generator, the advantage of ecoswing is that under the same torque, the weight can reduce more than 40% which can reduce the entire cabin weight 25%.and the weight of other materials can reduce in proportion. at the same time, the use of rare earth will be reduced the amount of at least two orders of magnitude. due to the advantages of ecoswing in torque density, the economy of wind power generation will be greatly improved by the ecoswing superconducting wind turbine generator. according to the prospect of global innovation center, the use of ecoswing superconducting wind turbine technology are expected to decline the wind power cost by at least 30% or more. in order to make the technique applied successfully, ecoswing project team will carry out comprehensive risk research and evaluation. after the laboratory test for ecoswing, envision plans to use it on the large megawatt wind turbine generators in denmark at least one year and maintain it regularly. "a lightweight and highly competitive superconducting wind turbine has a very exciting development potential." anders rebsdorf, the president of envision in denmark global innovation center said that ecoswing superconducting wind turbine will be an important progress in the way of pursuit of reducing the cost of renewable energy." anders rebsdorf also said that ecoswing superconducting wind turbine technology can reduce the cost of wind power more than 30%. and envision hopes to use their own top technology to make the wind and solar power become the mainstream energy of the global.fig.8 the concept of ecoswing.[9] fig.8 the concept of ecoswing. reproduced from [9] c. suprapower on 20th april 2015, suprapower held a workshop about high power electric generators for cost reduction of offshore wind. the workshop included presentations about trends on high power generators and about new developments of superconducting machines. suprapower is funded by the european union (eu) and it is a project that aims to a high power, lightweight and reliable superconducting generator for large offshore wind turbines. the goal is using superconductor to develop the wind power. there are nine industrial and scientific partners participate in this project. the technical physics research of university kit will build rotating cryostat for this project. by pure heat conduction, the superconducting coil cooled to -253.15 ℃ . at this temperature, the superconductor does not appear the resistance phenomenon. therefore it can realize the lossless conduction. a generator with a superconductor can be increased to 10 mw, reducing the volume and weight as well. in addition, compared to the current wide use of permanent magnet wind turbine, the superconducting wind turbine generators only need less than 1% of the rare earth. therefore superconductors can make the wind turbine higher, more stable, simpler, saving a great deal of raw materials and reduce the construction, operation and maintenance costs, improving the service life of the turbine.[10] other companies also have designed different superconducting generators. however, comparing these enterprises, sway started earliest. and the advantage of american superconductor is that they have the core technology. tecnalia cooperate with kit to make the product better. superconducting wind turbine generator is superior than permanent magnet generator. although sway started earliest, its development is slow and in wrong direction. american and tecnalia ＆ kit have advanced technologies. while, their technologies in wind power still need to explore. otherwise the design and manufacture of the large-scale wind turbine generator still have some difficulties. the 3.6mw superconducting wind turbine generator produced by envision is improved by the basis of 3mw wind turbine generator. therefore, the design and production are easier. each company has its own preponderance, the key to success is the new technology. v. challenges superconducting wind turbine generator is only a concept now. the application of superconducting technology in other areas has been realized. but the application in the wind turbine generators is not quite feasible currently. maybe there will be 5-10 years to produce a prototype or make a major technical breakthrough. for the large-scale commercial production, it is possible require 30-50 years. a. the materials of hts wires the material is the key to manufacture the superconducting wind turbine. currently, most of use is called first generation (1g) wires. the ceramic superconductor is bscco. this kind of wires can bear the rang of the current from 100a to 180a with the cross-section of 4mm*0.3mm.the manufacture process is very complex, therefore the cost is too high for application. recently, the second generation (2g) wires are also available. the ceramic superconductor is ybco. compared with bscco, its price is lower and have more merits. however these materials still have some problems like the supply and the price. whether these can be in use widely is the main to consider.[13] b. the maintenance of superconducting wind turbine generators there will be high vibration when the superconducting wind turbine generators in operation. and with different wind speeds, the frequency resonant is different. therefore, the requirement of sealing is very high. the interior of superconducting wind turbine generator is liquid nitrogen refrigeration. once it leak, it will cause the irreparable fault on other device. on the other hand, thunder and lightning may cause the wires of wind turbine some faults. once a failure occurs, the maintenance cost is very high because of the high technology used in the superconducting wind turbine generators. additionally, the general staff may cannot do it for its complexity. it will affect the utilization. c. excitation system the coils have negligible ohmic resistance. under normal operation, there will be a high current with a low voltage. however, when ramping the hts coils, a large excitation voltage is required. the rang of the excitation voltage is 100v or higher to achieve the field current changes within an acceptable periods of time. therefore, a control scheme using excitation current is required. and the system must be able to operate in merely inductive load. on the other hand, hts can not disperse the energy stored in the rotor inductance. thus, the exciter is the key of superconducting wires protection. currently, two methods are in use. first is using brushes and slip rings to transfer the current. the other one is that via rotating transfer to achieve brushless excitation. this require a complex control and protection system. the first one is widely used in synchronous generators. however, the maintenance requirement is high and normally hard to tolerate. in addition, the second solution is a complex and high requirement system. its design is difficult. although there have been some designed hts, the utilization is not common. in conclusion, the excitation system of superconducting wind turbine generators still have a lot to improve and develop. [13] d. refrigeration for rotor cooling system refrigeration for rotor cooling system is an important part of superconducting power equipment. the refrigeration to cryogenic temperature for hts is different with conventional wind turbine generator. the temperature of hts is 30k in normal operation and requiring a few hundred of watt refrigeration power. recently, there have been no commercial applications that close to those parameters. the best fit refrigeration is gm (gifford-mcmahon), which is regenerative refrigeration, generally developed from the cooling thermal shield of lts mri system, which reduce the boiloff of fluid helium. gm used widely and many industrially established technologies provide very attractive price. however, those parameters have been designed and the operation condition is a little rough now. gm refrigerator has several drawbacks:[19] 1) gm refrigerator relies on the rotary valve to control the air flow to the cold head or return to the low pressure side of the compressor. air flow through the valve will produce pressure head loss. therefore, the efficiency of gm is relatively low. 2) the compressor of gm pressure is relatively stable and it can use the activated carbon to remove oil. therefore, using commercial oil-lubricated compressor, which is reliable and having lower cost. but the compressor is large and requiring maintenance annually. 3) in addition, the cooling head includes a moving device, also need to maintain. currently, refrigeration power of models with the temperature of 25k and 100w is limited. thus, gms have to be operated in parallel. the operation of a conventional oil-lubricated compressor depends on the orientation with respect to gravity. thus, in some applications, it also requires other techniques. 4) pulse tube refrigerator is also regenerative refrigeration. in large-scale cooling capacity, the actual efficiency is low, especially cooling capacity of pulse tube refrigerators are met the problem of flow angle distribution is not uniform and interior of heat exchanger flow unsteadily, which leads to lower efficiency. and it has not been widely used in the field of superconducting machines. the refrigerator has its own advantages and disadvantages, also affect the development of hts. its technology, defects and manufacturing conditions are the key factors that result the slow development of hts. e. stator cooling effective stator cooling is the basic to achieve the effective current density of stator winding. to realize a more compact generator, a more intensive stator cooling method is required. recently, two cooling schemes are in use. the first one is air-cooled stator which is more conservative. the other one is fluid-cooled stator and it is more innovative. the stator winding is completely immersed in the dielectric insulating oil. this method has been used commonly in large power generators.[13] f. manufacturing processes and acceptance tests nowadays, the acceptance tests for superconducting generators have been produced as products. to reduce the production costs, developing a standardized manufacture is necessary. compared with conventional generators, the factory acceptance tests need to be modified. special tests are designed for special components of superconducting generators. however, this kind of tests and manufacturing process still need to be improved and develop. vi. conclusion . in recently years, due to more environment pollution and high carbon emission, the implementation of clean energy become popular and wind power has been the dominant position. however, conventional wind turbine generators still have some drawbacks, such as high cost and low efficiency. therefore, many companies began to implement the superconducting wind turbine. superconducting wind turbine generators have the advantages of high efficiency, lighter and volume is smaller. this greatly improves the preponderance of wind power. however, there are still many challenges and problems to overcome in superconducting wind turbines such as design and capacity. when these problems are overcome, it is believed that the use of superconducting wind turbine generators will become more and more common, and become a mainstream. references [1] bogi b. jensen, nenad mijatovic, asger b. abrahamsen: "development of superconducting wind turbine generators", journal of renewable and sustainable energy, 2013, 5(2):347-355 [2] hiroyuki ohsaki, yutaka terao, rashidul m. quddes, masaki sekino, “electromagnetic characteristics of 10 mw class superconducting wind turbine generators”, international conference on electrical machines and systems, 2010:1303-1306 [3] introduction of wind turbine. [online]. available: http://www.baike.baidu.com/link?url=wcmy4guxtaut_annpo5mw ca1uij05gav9ztz7p6ofvo6mzjyziiz8gaezivmynskkkgg7avge---o wast-kuza [4] a.b.abrahamsen, n. mijatovic, e. seiler, t. zirngibl, c. traeholt ”superconducting wind turbine generators” superconductor science and technology, vol. 23, no. 3, 2010 [5] a.b.abtahamsen, n.magnusson,b.b.jensen and m.runde “large superconducting wind turbine generators” energy procedia 2012, 24(24): 60-67 [6] h. rogalla and p.h .kes, ”100years of superconductivity”, crc press, 2012, 24(suppl 1):s147-s148 [7] j.bray, ”superconductor in application; some practical aspects.” ieee transactions on applied superconductivity,vol.19,no.3,pp.2533-2539,2009. [8] wenping cao, high-temperature superconducting wind turbine generators, [online]. available: http://www.intechopen.com [9] da keke (2015,step), what is ecoswing, [online]. available: http://www.china-nengyuan.com/news/83418.html [10] suprapower, [online]. avaliable: http://www.suprapower-fp7.eu/ [11] g. snitchler, b. gamble, s.s. kalsi, “the performance of a 5 mw high temperature superconductor ship propulsion motor”, ieee transactions on applied superconductivity, 2005, 15(2): 2206-2209 [12] h. ohsaki, m. sekino, t. suzuki, and y. terao, “design study of wind turbine generators using superconducting coils and bulks”, international conference on clean electrical power, 2009, 234(3): 479-484 [13] j. frauen hofer, j. grundmann, g. klaus, w. nick. “basic concepts, status, opportunities ,and challenges of electrical machines utilizing high-temperature superconducting (hts) windings”, 8th european conference on applied superconductivity, 2008, 97(1): 182-189 [14] zhang hongjie, sun jianfeng. “application prospect of high temperature superconducting generator in wind power generation”. new material industry, 2008, preface (5): 55-59 [15] ]zheng hailu, jin jianxun. “development and research status of high temperature superconducting generators”, journal of electronic science and technology university, 2007, 1673(6530):1-6 [16] zhu yingna (2015,april), analysis of 10mw superconducting generator connected to grid. [online]. available: http://xueshu.baidu.com/s?wd=paperuri%3a%280946fa26130bcaa8776 47c268ee6b6c1%29&filter=sc_long_sign&tn=se_xueshusource_2kduw 22v&sc_vurl=http%3a%2f%2fcdmd.cnki.com.cn%2farticle%2fcd md-10614-1015714516.htm&ie=utf-8&sc_us=8883208649839666566 [17] [clive lewis, jens muller. “a direct drive wind turbine hts generators”.wind power monthly magazine, 2005, 13(2): pp 53-60 [18] xu hongling, wang huiling, wang jian, shi ling, rao rongshui, chen jin, tang yuejin, “low temperature technology in high temperature superconducting system”, cryogenics, 2003, 2:20-24 bi yanfang, hong hui, xin ying, “cryogenic cooling system and refrigerating machine for high temperature superconducting”, science china press, 2013, 42(10):pp1001-1111 pan yunying was born in shanghai, china, in 1992. she received the b.s. degrees in power engineering from shanghai university of electric power in 2014. and now, she is a double m.s. student in shanghai university of electricity power and brandenburg university of technology. gu danzhen is an associate professor in shanghai university of electric power. she mainly engaged in modeling and simulation of power system, power market and so on. she has participated in the study of baosteel power grid energy management system, the transient stability study of 500kv yangcheng--huaiyin after its in operation, the research of large impact load of jiangsu power grid and its influence and so on. she has also participated in the major program of the national natural science foundation of china, such as wide area modeling and simulation theory and methods of dynamic security analysis of power system (project no. 50595412). she has published more than 10 papers in the chinese journal of electrical engineering, power grid technology, and other journals and conferences. http://xueshu.baidu.com/usercenter/data/journal?cmd=jump&wd=journaluri%3a%28cb9d8826ce89275f%29%20%e3%80%8ajournal%20of%20renewable%20%26%20sustainable%20energy%e3%80%8b&tn=se_baiduxueshu_c1gjeupa&ie=utf-8&sc_f_para=sc_hilight%3dpublish&sort=sc_cited http://xueshu.baidu.com/usercenter/data/journal?cmd=jump&wd=journaluri%3a%28cb9d8826ce89275f%29%20%e3%80%8ajournal%20of%20renewable%20%26%20sustainable%20energy%e3%80%8b&tn=se_baiduxueshu_c1gjeupa&ie=utf-8&sc_f_para=sc_hilight%3dpublish&sort=sc_cited http://xueshu.baidu.com/usercenter/data/journal?cmd=jump&wd=journaluri%3a%286d8f90df550d9d13%29%20%e3%80%8aieee%20transactions%20on%20applied%20superconductivity%e3%80%8b&tn=se_baiduxueshu_c1gjeupa&ie=utf-8&sc_f_para=sc_hilight%3dpublish&sort=sc_cited http://xueshu.baidu.com/usercenter/data/journal?cmd=jump&wd=journaluri%3a%286d8f90df550d9d13%29%20%e3%80%8aieee%20transactions%20on%20applied%20superconductivity%e3%80%8b&tn=se_baiduxueshu_c1gjeupa&ie=utf-8&sc_f_para=sc_hilight%3dpublish&sort=sc_cited http://xueshu.baidu.com/usercenter/data/journal?cmd=jump&wd=confuri%3a%28e1a95c6c959f07f5%29%20international%20conference%20on%20clean%20electrical%20power&tn=se_baiduxueshu_c1gjeupa&ie=utf-8&sc_f_para=sc_hilight%3dpublish&sort=sc_cited  transactions on environment and electrical engineering issn 2450-5730 vol 1, no 3 (2016) © raivo melsas, argo rosin and imre drovtar abstract— demand side response enables cost optimization for energy systems and industrial consumers. in many countries, it is not widely used because of implementation complexity. one of the solutions for applying demand side response is industrial process scheduling according to the energy market needs. from the energy system point of view, process scheduling implies load scheduling. the aim of this paper is to provide a solution for load scheduling by implementing value stream mapping, which is a straightforward enough for production management. decision makers in the industry should have a clear understanding about positive effect from load scheduling and its effect to production outcome and process availability. value stream mapping is a wellknown process optimization tool from lean production philosophy. the aim of value stream mapping is to shorten the lead time of industrial processes and to reduce the intermediate stock amounts. by complementing value stream map with process energy intensity and energy stored in intermediate stocks, we can promote load scheduling possibilities. our methodology provides a tool that is understandable and traceable for industry-minded decision makers. finally, we present a real life test example for the new methodology, which is based on the production process of a district heating plant. index terms— demand response, energy storage, load management, load scheduling, value stream mapping i. introduction oad scheduling (ls) as part of demand side response (dsr) must meet the needs of industry. one of the effects for the industry appears when load consumption is shifted from periods of high electricity price to those of low price. as a result, cost savings can be achieved by means of reduced consumer demand in high price periods. this requires better production planning, which is related to production management. previous dsr studies have resulted in the following: static stackelberg game theory for voluntary load curtailment programs [1]; numerical calculation method for dsr when a battery energy storage system (bess) is utilized [2]; solutions for dsr by means of automatic lighting [3]; dsr with microchp systems [4]; an overview for dsr methods in high this research was supported by the estonian centre of excellence in zero energy and resource efficient smart buildings and districts, zebe, grant 2014-2020.4.01.15-0016 funded by the european regional development fund. consumption industries and examples of market tools that support dsr [5]. in [6] an automated complex system for ls in industry is described, which takes into account stock restrictions, maintenance schedules, and crew management. all the necessary inputs are analyzed with a fuzzy/expert-based system combined with an optimization module. as a result, the system is able to identify whether and how much the industrial plant can participate in a dsr event. in [7] and [8] dsr is addressed as a part of the following main load shaping strategies: a) conservation energy saving is achieved through static methods; b) load growth energy consumption is increased when an energy system has surplus energy production; c) valley filling load is increased through the off-peak periods or keeping stable consumption; d) peak clipping energy consumption is decreased in peak periods; e) load shifting peak consumption is shifted from peak periods to non-peak periods; f) seasonal load reduction annual energy peaks are reduced. ls is used mainly in “load shifting” strategy e; however, in some cases, “valley filling” strategy can be utilized as well. to use the ls, an industry must have the following one or several dsr options [8]: cooling equipment with cooling storage, heating equipment with heat storage, dual fuel systems that can operate either on electricity or on an alternative fuel, discretionary loads and process equipment that can be shifted during a short period or material handling equipment with storage possibilities (silos, stock, etc.). this paper focuses on the last option by looking at process as a whole in order to find ls solutions. in addition, it provides a method applicable in industry for outlining the possibilities with ls as a part of dsr by using value stream mapping (vsm). vsm is applicable in r. melsas, a. rosin and i. drovtar are with the department of electrical engineering, tallinn university of technology, ehitajate tee 5, 19086 tallinn, estonia (e-mail: raivomelsas@gmail.com, argo.rosin@ttu.ee, imre.drovtar@gmail.com). value stream mapping for evaluation of load scheduling possibilities in a district heating plant raivo melsas, argo rosin and imre drovtar l various ways. originating from toyota production systems [9], it was further elaborated and adjusted to find solutions for different problems in the production process. for example, vsm is used to solve quality problems [10]. this paper elaborates on vsm. our proposal is to use it for indicating a possibility to shift an electrical load from a high price period to a low price period and utilize an intermediate stock for energy storage [11], [12], [13]. this paper will provide a straightforward solution for the industry in order to apply ls effectively and which is easily applicable. ii. ii. energy price as a driver for load shifting ls can yield an economic effect under rational consideration. cost reduction can be achieved by taking advantage of energy price fluctuations during a day. it is reasonable to have demand response implemented in countries where an electricity pool exists and hourly based spot prices are known for a short period ahead. as a result, industries can plan their production according to the spot price. fig. 1 shows an average 24-hour electricity market spot price in estonia in 2014 [14]. as can be seen, electricity spot price is typically higher from 07:00 a.m. to 8:00 p.m. in general, there is at least 10-euro price difference during a day and a night. considering the price peak and dip approximately 20-euro difference per mwh exists during a day. on average, price difference during a day is 15-euro mwh. however the consideration above includes only electricity price; in addition, there are some fluctuations in grid service price as well. from 12:00 a.m. to 8:00 a.m. (11:00 p.m. to 7:00 a.m. in winter time), the grid service price is lower. this period is not overlapping 100% with a low spot price period. we will call this period (12:00 a.m. to 8:00 a.m.) as the low price period (lpp) and the other period during a day as the high price period (hpp). grid tariffs depend on grid connection voltage level and connection capacity (amps). in the following, we will describe one example to examine the daily price difference for the industry in estonia. at substantial electricity consumption, an industry is usually connected into a middle voltage (mv) grid. in that case, estonian grid service price is 14.5 euros per mwh from 8:00 a.m. to 24:00 a.m. and during lpp 8.3 euros per mwh [15]. for some customers, no time difference is applied; the price for grid services is constant in time12 euros per mwh. fig. 1 shows both the grid price and the spot price fluctuations. in 2014, an average electricity spot price for lpp was 29.5 euros per mwh and with grid price fluctuations it amounted to 37.8 euros per mwh, which we call as the low price period price (lppp). in 2014, an average hpp spot price was 40.9 and together with grid tariff, the average number was 55.1 euros per mwh, which we call as the high price period price (hppp). for clarity, 55.1 is an average found on hourly bases, i.e. spot price + grid price. based on the data provided, we can calculate potential savings for industry under ls. potential cost saving is 31%, which is calculated by (1): . hppp lppphppp cs − = (1) in general, it can be concluded that grid tariff fluctuation has an important role in ls in estonia, as the average difference in the spot price between hpp and lpp was 11.4 euros per mwh and grid tariff will add extra for the difference between hppp and lppp, according to [15], tariff depends on the grid connection parameters. the shape of the electricity spot price in fig. 1 can be considered as a typical shape of the daily demand of electricity as well; a similar shape of demand can be found in various places, e.g., even in south africa [16]. thus, dsr has a positive impact to overall efficiency to energy systems, not only to an industry itself. iii. improvement of value stream mapping methodology for evaluation of load scheduling possibilities a. load scheduling at intermediate storage use in production in the following, ls possibilities are examined for intermediate storage use as an energy saving unit, to enable shifting of energy intense production from hpp to lpp. in ls, it is important to understand energy intensive production units and their overall role in the production. methods from lean philosophy [lp] can be used here. we propose to improve the value stream mapping (vsm) methodology with ls principles. today lp is a leading production management philosophy. vsm is used to plan production as efficiently as reasonably possible. energy intensity can be added in a process as additional information for a vsm, which will give a good overview about the possibilities in energy saving. fig. 2 gives an overall picture of a typical vsm, elaborated with process energy intensity and amount of energy stored in the production intermediate storage. fig. 1. average estonian 2014 electricity spot price during a day. 20 25 30 35 40 45 50 55 60 65 1 3 5 7 9 11 13 15 17 19 21 23 pr ic e (e u r /m w h) hours in a day spot+grid price on hv line spot price as vsm gives an overview of the production planning and can additionally give information about energy intensity, the possibility of the ls should be studied in detail. from vsm, we will know if an energy intense process is at the same time a bottleneck in production. if it is not true, then the conclusion is that this process is not 100% utilized in time and ls can be implemented without increasing process capacity. alternatively costs for increasing process capacity (cpc) need to be calculated. next, we focus on the storage. if production is shifted in time, storage volume can increase as compared to the state without ls utilized. storage increase may need additional investments, which should also be considered as costs for storage (cs). also, costs of the process (cp) itself for ls should be taken into account. as an example cp related to ls can be an increase in labor costs due to night shifts. finally, if the costs related to ls are lower than a possible income from ls (ils), the ls can be implemented in the industrial process. this criterion is given in the following (2): ils<cp+cs+cpc . (2) b. methodology for ls with vsm. the main initial goal of lp and vsm was to reduce lead time (l/t), i.e, the time it takes to move a produced piece all the way in the process or production from start until to the end. as a result, many costs or wastes, as defined in lp, will be reduced. as shown in (2), income from ls should be higher than costs related to it because ls can increase the intermediate stock and due to that it has a negative effect on the l/t as well. vsm is a tool that will help to find the processes which have a reasonable effect and income. to achieve that, cost saving potential must be estimated. in section ii, we showed a possible gain from ls in estonia. the next step is to define the energy consumption of the processes. it is useful to combine that with vsm. if a company already has vsm, then energy consumption should be added into the vsm. vsm will highlight important information to be considered for ls. most importantly, the following information must be taken into account: a) identify if the process is a bottleneck in production. if the process is a bottleneck, then load shifting is usually impossible without investment into the process output increase. b) process cycle time (c/t) is slower or faster than its next process c/t. c/t in vsm describes how often a part or a product is completed by a process. if the process c/t is slower than the process coming next, it is possible to increase intermediate stock at the end of the process coming after rather than at the end of the first process. in case the first process is faster than the next one, the ls applied will increase the intermediate stock. c) weather process uptime (uptime) is high or low, shows important information about process reliability. a low reliability process has a negative effect on ls. d) if stock between processes is high or low, the reasons should be found out before applying load scheduling. e) change over time (c/o) is long or short. c/o is the time period for switching the process unit from one product to another. it can show also time for cold start of the process. c/o can highlight important information about the process start up time. the best way to start defining the energy consumption of a process is to make the consumer list. consumer list should be based on processes described with vsm. the consumer list may be available in the facility electrical department. in that case, consumer list is usually based on the power cabinets and it has to be made process by process. if the consumer list is not available, it should be done from scratch. as the purpose is to find an initial energy intensive process, the consumer list can be composed without actual measurements. the consumer list should contain the following information: process name; device fig. 2. vsm with energy consumption and storage overview. supplier customer receiving twice a week daily shiping process a one shift process b one shift process c one shift storage storage storage storage c/t= 2 min 10 c/t= 4 min 15 c/t= 7 min 30 5 days c/o 2 hr units c/o 3 hr units c/o 4 hr units uptime 94% uptime 91% uptime 98% nom. p 30 kw nom. p 50 kw nom. p 2 kw kwh/unit 2 20 kwh/unit 5 75 kwh/unit 0,3 9 kwh kwh kwh 5 days 40 minutes 105 minutes 1 day 2 minutes 4 minutes 7 minutes ii i i label; rated power; cosᵩ; nominal current; nominal voltage and consumption type that can be either continuous, intermittent or stand by. upon completion of the consumer list, traditional vsm should be elaborated and total nominal loads in the process added. as a result, elaborated vsm will show which processes have sufficiently high energy intensity to gain benefits from ls and on the other hand, to estimate the particular potential to the process scheduling. after completing the estimation in conjunction with experts from production, the vsm should be elaborated further. based on the measurements, energy intensity or the production unit is to be found. as the energy usage was previously estimated, measurements can be done in the production process where energy usage is estimated to be high or vsm shows that ls can be implemented without investing to process capacities or intermediate stocks. as vsm describes product flow, the energy intensity should also be given per product or production unit. the unit energy consumption in the production process should be calculated based on the actual measurement. fig. 2 shows a theoretical example of vsm together with data from the consumer list and energy intensity in the process. energy intensity is energy consumption in each process to make one unit. as can be seen, process a c/t takes 2 minutes, i.e., one unit is completed during 2 minutes in the process. the total nominal power of the process is 30 kw and 2 kwh is consumed to make one unit. first intermediate stock capacity (after process a) is 10 pieces. it consists of 20kwh energy which is available for scheduling. process b c/t is 4 minutes and the total nominal power in that process is 50 kw. as the second intermediate stock (after process b) capacity is 15 pieces, it has 75 kwh energy available for scheduling. process c cycle time is 7 minutes, total nominal power is 2 kw and 0.3 kwh is consumed for making one unit. we can see that process a c/t is 2 times faster than process b and process c cycle time is 3.5 times faster than process a. as a result, we can conclude that the last process will dictate the whole process time and previous processes can be scheduled taking into account the possibilities of the last process. for that reason, we need to know how much time it takes to empty the intermediate stock before the slowest process, which we call buffer time (bt). in order to find bt, process c/t must be multiplied with an available intermediate stock capacity (asc) (3) and (4). sscmscasc −= . (3) asctcbt ⋅= / . (4) asc is a difference between maximum stock capacity (msc) and safety stock capacity (ssc). ssc is to be defined by production management. bt shows the maximum load scheduling period in the process. neglecting safety stock capacity, in our example bt is 105 minutes. the total energy we can shift during 105 minutes is 95 kwh, which is the sum of stored energy in two intermediate stocks with 75 and 20 kwh accordingly. iv. experimental load scheduling analysis in a district heating plant with improved vsm methodology an example of the method above is described here. this example covers a district heating plant in paide, estonia. the company has one 8 mw woodchip boiler, several boilers fuelled with shale-oil and one chp plant based on woodchips and with 8 mw thermal, 2 mw electrical output. fig. 3 shows the process of the plant. woodchip boiler stock (moving floor) is filled by a conveyor from the main storage (moving floor). the main storage is filled by the incoming tucks or a wheel loader. the company has 5-day storage available on site. woodchip boiler stock can contain woodchips a day with boiler nominal load i.e. 180 m³. the conveyor between the main storage and the woodchip boiler stock has a maximum output of 50 m³/h. the conveyor from the woodchip boiler stock to the boiler has a strict limitation for processing output from the woodchip boiler. as it has no stock or daily silo available between the boiler and the conveyor, the output is 7.5 m³/h with the nominal boiler load. we start the cost reduction estimation from ls by modeling the process using vsm as a basis. one cubic meter of woodchips is taken here as one unit in vsm. according to the conveyor output parameters, the total output time is 6 days and 15.2 minutes and the value creating time is 17.2 minutes (time when one cubic meter of woodchips is actually processed). also, the vsm shows that the conveyor from the main to the boiler stock has excessive capacity and is able to process one cubic meter of woodchips approximately 6.67 times faster than the process bottleneck, i.e. woodchip boiler. the boiler will process one cubic meter of woodchips into heat in 8 minutes. it can be concluded from vsm that there is a possibility of ls of the conveyor from the main storage to the boiler stock. it must be emphasized that we are dealing with a biomass conveyor; therefore, the following simplifications are used: a) woodchip processing by a boiler is calculated based on the nominal load, and boiler efficiency parameter is 0.8; b) woodchip energy intensity is approximated at 1.3 mw/m³, in real life it can be different, based on the fuel type and humidity level. the next task is to estimate process energy intensity by making a consumer list behind the process. in our example, it is evident from vsm that the only process that can be scheduled is the process called “conveyor from main stock to boiler stock”. thus, we need to make a consumer list behind this process, which is given in table i table i shows that the total nominal power behind the process is 73.5 kw. the consumer list is necessary to estimate of the process energy demand. it is possible that some consumer’s nominal power is much greater than the actual absorbed power. therefore it is necessary to measure the process energy use. in our example, fluke 1735 is used for measurements of the process energy use and current measurements for some continuous consumption. there is calculated absorbed power with the following (5). ϕcos3 ⋅⋅⋅= ff uip , (5) where p is absorbed power, ifphase current, ufphase voltage and cosᵩ is power factor. by using absorbed power for a continuous load, energy consumption was estimated as well. total energy consumption for processing 1 m³ of woodchips was 0.96 kwh. complete vsm for the process examined is given in fig. 3. the vsm shows that by processing one cubic meter of woodchips there is consumed 0.96 kwh of electrical energy. as there is storage available with 180 m3 and ssc is 7.5 m3, asc is 172.5 m3 using (3). by knowing process c/t behind the intermediate stockit is 8 minutes per m3. there can be found out that bt is 23 hours by using (4). we can conclude that bt is sufficiently long for considering ls from hpp to lpp given in section ii. paide boiler plant is connected to a grid at low voltage line, so the grid tariffs applicable will be different from those given in section ii. hppp for the current example is 83.4 euros per mwh with 43.5 euros per mwh grid tariff and lppp is 55.2 euros per mwh with 25.7 euros per mwh grid tariff. based on (1), we can calculate potential saving which is 33.8%. as stored energy in the process is 0.166 mwh, by multiplying that with hppp, the cost for the company will be 13.81 euros per day and saving from ls will be 4.67 euros per day. in the previous example, the vsm process is not complicated; however, vsm is essential in case production is more complicated and processes are more complex and dependent on each other. in that case, vsm enables cost savings with ls. moreover, ls is well understandable with vsm for staff involved in production management and planning. v. conclusion various aspects of dsr were studied. focus was on the price difference in the electricity stock market in estonia. if the loads are shifted from hpp to lpp ls will enable a cost reduction of 31%. the methodology based on vsm was introduced where cost savings were achieved with ls. our analysis of experimental load scheduling was based on the district heating company in estonia. the proposed methodology was found to be applicable as a straightforward tool to achieve cost savings. it can serve as first step for industry when implementing ls because of its simplicity. with minor costs, reduced time and lower complexity, a major cost saving can be achieved with simple solutions such as the method proposed. table i consumer list for conveyor from main stock to boiler stock description pn cos ϕ in un consumption conv. screen to dist. conv. 15 0.8 46.9 0.4 continuous distr. conveyor 7.5 0.8 23.4 0.4 continuous floor to conveyor motor 4 0.8 8.6 0.4 continuous floor to conveyor motor 4 0.8 8.6 0.4 continuous hydro pack 15 0.85 28.9 0.4 intermittent hydro pack 15 0.85 28.9 0.4 intermittent screen 5.5 0.81 11.4 0.4 continuous conv. floor to screen 7.5 0.8 23.4 0.4 continuous fig. 3 vsm for ls for paide district heating company.. supplier paide boiler plant customer receiving daily continuous heat supply storage storage c/t (min) 1.2 c/t (min) 8 c/t (min) 8 5 days c/o 0 1 day c/o 0 c/o 4 hr uptime 95% uptime 95% uptime 95% 600 (m³) nom. p 73.5 (kw) 180 (m³) kwh/m³ 0.96 173 (kwh) kwh/m³ 0.3 5 days 1 day 1.2 min 8 min 8 min conveyer from storage to boiler stock conveyer from boiler stock to boiler boiler ii references [1] p. kontogiorgos and g. p. papavassilopoulos, "subsidized power interruption for reducing peaks in energy demand: a stackelberg game," communications, control and signal processing (isccsp), 2014 6th international symposium on, athens, 2014, pp. 421-424. doi: 10.1109/isccsp.2014.6877903 [2] g. carpinelli, s. khormali, f. mottola and d. proto, "demand response and energy storage systems: an industrial application for reducing electricity costs. part ii: numerical application," power electronics, electrical drives, automation and motion (speedam), 2014 international symposium on, ischia, 2014, pp. 1038-1042. doi: 10.1109/speedam.2014.6872040 [3] s. a. raziei and h. mohscnian-had, "optimal demand response capacity of automatic lighting control," innovative smart grid technologies (isgt), 2013 ieee pes, washington, dc, 2013, pp. 1-6. doi: 10.1109/isgt.2013.6497854 [4] m. houwing, r. r. negenborn and b. de schutter, "demand response with micro-chp systems," in proceedings of the ieee, vol. 99, no. 1, pp. 200-213, jan. 2011. doi: 10.1109/jproc.2010.2053831 [5] a. rosin, i. drovtar, s. link, h. hõimoja, h. mõlder and t. möller, "analysis of large electricity consumers demand profiles to determine implementation of demand-side management measures”, “tarbimise juhtimine ‒ suurtarbijate koormusgraafikute salvestamine ning analüüs juhtimise rakendamise võimaluste tuvastamiseks," eleringi toimetised, pp. 1-57, 20 october 2014. [6] s. mohagheghi and n. raji, "dynamic demand response solution for industrial customers," industry applications society annual meeting, 2013 ieee, lake buena vista, fl, 2013, pp. 1-9. doi: 10.1109/ias.2013.6682583 [7] a. mohamed and m. t. khan, "a review of electrical energy management techniques:supply and consumer side (industries)," journal of energy in southern africa, vol. 20, no. 3, 2009. [8] a. kuhi thalfeldt and p. raesaar, "possibilities and feasibility of demandside management in the estonian industrial sector," in 5 th international symposium "topical problems in the field of electrical and power engineering”, kuressaare, 2008. [9] t. ohno, “toyota production system: beyond large-scale production”, portland, 1988. [10] b. haefnera, a. kraemera, t. staussa and g. lanzaa, "quality value stream mapping," in 47th cirp conference on manufacturing systemss, windsor, 2014. [11] p. uuemaa, j. valtin, j. kilter, a. puusepp, i. drovtar and a. rosin, "load control implementation in the energy intensive industry," mediterranean electrotechnical conference (melecon), 2014 17th ieee, beirut, 2014, pp. 213-218. doi: 10.1109/melcon.2014.6820534 [12] i. drovtar, p. uuemaa, a. rosin, j. kilter and j. valtin, "using demand side management in energy-intensive industries for providing balancing power the estonian case study," power and energy society general meeting (pes), 2013 ieee, vancouver, bc, 2013, pp. 1-5. doi: 10.1109/pesmg.2013.6672418 [13] p. uuemaa, j. kilter, j. valtin, i. drovtar, a. rosin and a. puusepp, "cost-effective optimization of load shifting in the industry by using intermediate storages," innovative smart grid technologies europe (isgt europe), 2013 4th ieee/pes, lyngby, 2013, pp. 1-5. doi: 10.1109/isgteurope.2013.6695404 [14] as, nord pool spot, "elspot_prices_2014_hourly_eur," 2014. [online]. available: http://www.nordpoolspot.com/historical-market-data. [accessed 20 june 2015]. [15] elektrilevi oü, "price list fo elektrilevi from 01.04.2014”, “elektrilevi hinnakiri alates 01.04.2014," 1th april 2014. [online]. available: www.elektrlilevi.ee/hinnakiri. [accessed 19th june 2015] i. introduction a) conservation energy saving is achieved through static methods; b) load growth energy consumption is increased when an energy system has surplus energy production; c) valley filling load is increased through the off-peak periods or keeping stable consumption; d) peak clipping energy consumption is decreased in peak periods; e) load shifting peak consumption is shifted from peak periods to non-peak periods; f) seasonal load reduction annual energy peaks are reduced. ii. ii. energy price as a driver for load shifting iii. improvement of value stream mapping methodology for evaluation of load scheduling possibilities a. load scheduling at intermediate storage use in production b. methodology for ls with vsm. a) identify if the process is a bottleneck in production. if the process is a bottleneck, then load shifting is usually impossible without investment into the process output increase. b) process cycle time (c/t) is slower or faster than its next process c/t. c/t in vsm describes how often a part or a product is completed by a process. if the process c/t is slower than the process coming next, it is possible to increase intermediate... c) weather process uptime (uptime) is high or low, shows important information about process reliability. a low reliability process has a negative effect on ls. d) if stock between processes is high or low, the reasons should be found out before applying load scheduling. e) change over time (c/o) is long or short. c/o is the time period for switching the process unit from one product to another. it can show also time for cold start of the process. c/o can highlight important information about the process start up time. iv. experimental load scheduling analysis in a district heating plant with improved vsm methodology a) woodchip processing by a boiler is calculated based on the nominal load, and boiler efficiency parameter is 0.8; b) woodchip energy intensity is approximated at 1.3 mw/m³, in real life it can be different, based on the fuel type and humidity level. v. conclusion references  transactions on environment and electrical engineering issn 2450-5730 vol 3, no 1 (2018) © ulrich lühring, daniel wienold, frank jenau  abstract—partial discharges reduce the residual lifespan of operating equipment and can indicate serious deficiencies of electrical insulations at an early stage. regarding ac voltage stress, the phase resolved pattern is an established diagnostic tool, which allows conclusions about the location and the type of fault. despite the increasing importance of dc transmission systems, a comparable proceeding for dc voltage stress does not exist. this implies the development and evaluation of alternative and suitable basic approaches. although diverse promising approaches are identified, recent research is focused on standard atmospheric conditions. due to the fact that this is just partly consistent to real operating conditions, additional research is required. focusing on the time domain analysis of corona discharges, occurring under positive dc voltage stress in air, a measurement method for investigating the influence of varying atmospheric quantities is presented. measurements are carried out for five different relative humidity levels in the range of 20 % to 95 % and for four different temperature levels in the range of 20 °c to 65 °c. as characterizing pulse shape parameters, the rise time, the pulse width and the fall time are determined as well as the apparent charge. the gained values are compared to each other and reconciled with physical processes. index terms—corona discharge, dc voltage, humidity level, pulse shape parameter, temperature level i. introduction artial discharges are the result of a local electrical stress concentration and are defined as localized discharges, which partially bridge the electrical insulation between this manuscript was submitted on june 8 th , 2018 and revised on june 25 th , 2018 as an extended version of the paper “influence of humidity on pulse shape parameters of positive corona discharges in air at dc voltage” presented at the 17 th international conference on environment and electrical engineering, milan/italy, june 2017 [1]. ulrich lühring, daniel wienold and frank jenau are with the institute of high voltage engineering, tu dortmund university, dortmund, germany (e-mail: ulrich.luehring@ tu-dortmund.de). this work has been funded by the german federal ministry for economic affairs and energy (bmwi) as a part of the e²hgü project under grant number 03et7514. conductors [2]. due to the damaging effect on the condition of isolation, the partial discharge (pd) diagnosis is a key component of electrical diagnostic procedures. in addition to a registration of local weak points, the focus of measurement and analysis techniques is in particular on the identification of the location and the type of fault. regarding the latter, it is taken advantage of the fact that the appearance of pds depends on various factors, which are influenced by the type of fault. under ac voltage stress the phase resolved pattern is the best established and most commonly used diagnostic tool to distinguish between different types of fault [3]. this diagnostic procedure analyzes the link between discharge pulses and the phase angle of their occurrence with respect to the test voltage. the application is consequently restricted to ac voltage and not transferable to dc voltage. a similar recognized diagnostic tool for dc voltage stress does not exist. the increasing importance of dc transmission systems requires the development and evaluation of alternative and suitable basic approaches. therefore, the pd diagnosis at dc voltage stress is subject of current research, whereby diverse promising basic approaches could be identified. regarding measurements based on the conventional pd measuring circuit, experimental tests indicate that an estimation of the magnitude of pds [4], an analysis of the time lag between subsequent pd impulses [5] and the time domain analysis [6] are appropriated for a precise defect classification. in addition to that, unconventional measuring methods, such as the capturing of the electromagnetic spectrum by using a broadband antenna, are increasingly moving to the fore. with respect to dc applications it is e.g. shown in [7] and [8], that an evaluation of acquired data of the electromagnetic spectrum of pds in the frequency domain respectively in the time domain is suited to distinguish between different types of fault. an advantage of the time domain analysis is the direct relationship between the pulse shape of the pd impulse current and the causing processes. furthermore, it is well known that the electromagnetic spectrum of pds depends on the pulse shape of the discharge current [9]. the time domain analysis focusses on parameters, which are directly obtained from the pulse shape. the rise time, the pulse width and the fall time as primary parameters are considered as well as about the influence of temperature and humidity level on pulse shape parameters of positive dc corona discharges in air ulrich lühring, daniel wienold and frank jenau p secondary parameters such as the coefficient of variation, the kurtosis or the skewness. a comparative analysis of these parameters, gained from pds of idealized and artificial defects under dc voltage stress, outlines the possibility to differentiate between different types of fault [10] and defects in varying insulating media [11], [12]. extensive experimental investigations and statistical evaluations, presented in [6], verify that the time domain analysis of pds enables a precise defect classification. comparative investigations of corona discharges occurring under ac and dc voltage stress point out, that the associated pulse shape parameters are not affected by the voltage waveform [13]. based on these findings [14] demonstrates the general suitability of the time domain analysis to identify the type of fault at ac voltage as well. the abovementioned investigations have in common that they are comprehensively realized under ambient atmosphere, which approximately corresponds to the reference atmosphere, and under a voltage stress which complies with the partial discharge inception voltage or a predefined voltage level above it. a variation of these parameters is not taken into consideration, which is just partly consistent with real operating conditions. whereas detailed information regarding an influence of the test voltage level are outlined in [15], additional research concerning the influence of varying atmospheric parameters, such as the air pressure, the humidity level and the temperature level, on the appearance of pds has to be carried out. studies in relation to an influence of the gas pressure are outlined in [16] and [17]. although detailed information from the presented measurement data are not derivable, these investigations indicate increasing discharge amplitudes and decreasing pulse widths as well as decreasing fall times if the gas pressure is raised. a consideration of a varying temperature and humidity level focuses on investigations regarding the pd intensity, the discharge inception voltage and the breakdown voltage. for example, [18] respectively [19] evince a slight decrease of the pd inception voltage and the breakdown voltage, if the temperature level is raised. concerning a rising humidity level, a decreasing pd inception voltage and an increasing breakdown voltage is pointed out in [20] respectively [21]. this indicates a dependency between the partial discharge impulse current and the temperature level as well as the humidity level. the purpose of this contribution is to investigate the influence of the temperature and the humidity level on the pulse shape by focusing on positive corona discharges in air. therefore, corona discharges are generated by using a needleto-plate set-up. the latter is part and parcel of an optimized measuring circuit. an appropriate measuring method is presented and applied to capture the pulse shape of corona discharges. to characterize the pulse shapes the rise time, the pulse width and the fall time are determined as well as the apparent charge for pds occurring at temperature levels of 20 °c, 35 °c, 50 °c and 65 °c and for relative humidity levels of 20 %, 40 %, 60 %, 80 % and 95 %. ii. experimental set-up an accurate determination of pulse shape parameters presupposes a broadband measuring system, consisting of a measuring circuit and a measuring instrument. in [22] an overall bandwidth of at least 1 ghz is demanded to capture initial pulse shapes, which are not impaired by reflection and refraction processes. if this requirement is fulfilled, detailed information regarding the causing and influencing processes can be gained. a. pd measuring circuit the measurements are implemented by using a basic pd measuring circuit, which is in accordance to iec 60270. as shown in fig. 1, the pd measuring circuit consists of a coupling capacitor cc, a measuring cell with a needle-to-plate set-up (dut), a measuring impedances zm1 respectively zm2 and a decoupling impedance zd. the latter decouples the discharge current circuit from a high voltage power supply u0, which provides a positive dc voltage with a low residual ripple. to determine the pd inception voltage an external quadripole of a pd measuring system is set as zm1 in series to cc. the pulse shapes are captured with a broadband oscilloscope, whose input resistor is set as zm2 in series to the dut. b. measuring cell a decoupling of the partial discharge impulse current from the weak point to the broadband oscilloscope devoid of distortions is enabled by using a measuring cell optimized from a high voltage and high frequency point of view. core component of the measuring cell, whose sectional view is shown in fig. 2, is a conical transmission line with a constant line impedance of approximately 50 ω. the suitability of such a conical transmission line for an accurate measurement of the pulse shape is proven in [11] and [23]. in total four air inlets, which are arranged radially in the same height, ensure that the temperature and the humidity level within the measuring cell can comply with the external levels. c. test device to generate corona discharges a sharp needle is attached to the cylindrical high voltage electrode. the air clearance between the ground electrode and the needle tip is set to 15 mm. fig. 1. pd measuring circuit according to iec 60270 to determine the pd inception voltage and to capture the pulse shape. zm1 zd cc ipd u0 dut climatic chamber zm2 iii. experimental procedure the influence of the temperature level and the humidity level on the pulse shape of positive corona discharges is investigated by considering different climatic conditions. as shown in in fig. 3 13 different climatic conditions are considered, whereby the temperature level is varied in the range of 20 °c to 65 °c and the relative humidity level is varied in the range of 20 % to 95 %. a variation of these is made possible by placing the measuring cell in a climatic chamber, which is presented in [24]. the climatic chamber is characterized by a high long-term stability and an accurate regulation of the temperature and the relative humidity level. to exclude an influence of the test voltage level, which occurs if the test voltage is raised above the inception voltage [15], all measurements are executed at the respective pd inception voltage. the latter is ascertained in advance for each climatic condition. a. determination of the inception voltage up to now, the term of the pd inception voltage under dc voltage stress is not clearly defined. in the framework of this investigation the pd inception voltage is determined by increasing the test voltage in steps of 0.5 kv starting from zero. as shown in fig. 4 the test voltage is enhanced by additional 0.5 kv, if pds are not detected within one minute. if sufficient pds are measurable at a constant test voltage level over a period of five minutes, the pd inception voltage is assumed to be reached. b. capturing of the pulse shapes the climatic conditions are successively provided by the climatic chamber, whereby special attention is paid in order to prevent the formation of condensate. in the case of a sole variation of the humidity level, an additional residence time of at least two hours is taken into consideration before performing a measurement after having reached a stationary state in the climatic chamber. if the temperature level is changed, the additional residence time is set to at least 12 hours. this ensures the desired climatic conditions within the measuring cell. previous investigations demonstrate that the first determinable pulse shapes reach larger amplitudes than the following [25], [26]. due to these findings, the capturing of 100 sequent pulse shapes for each climatic condition is started after a stress duration of ten minutes. furthermore, this ensures that the pds are not mainly energized by the capacitive field [5]. iv. data analysis and presentation the captured pulse shapes are characterized by the determination of the rise time (20 % to 80 %), the pulse width (50 % to 50 %), the fall time (80 % to 20 %) and the apparent charge. to present and compare the location and the distribution of the determined pulse shape parameters, box plots are used. this kind of graphic representation, which is among others shown in fig. 5, divides the entire range of values into four subsets. the median is represented by the bar within the box, which itself contains the middle 50 % of all determined values. the boundaries of the box are the 25 % quartile and the 75 % quartile. their distance delineates the interquartile range. values, having a greater distance to the box than 1.5 times of the interquartile range, are designated as outliers and are characterized by a plus symbol. fig. 3. overview of the investigated climatic conditions. fig. 2. sectional view of the measuring cell optimized from a high voltage and high frequency point of view. pvc insulator conical transmission line hv-connection ground electrode cylindrical hv-electrode air inlet fig. 4. flow diagram for the determination of the pd inception voltage. 0 kv ui test voltage 0.5 kv 1.0 kv 1.5 kv test device without voltage stress determination of the inception voltage ui inception voltage ui yes sufficient discharges over a period of five minutes? detection of a discharge within one minute? increase test voltage by 0.5 kv yes no no v. results and discussion a. appropriateness of the measurement method the appropriateness of the experimental procedure presented in section iii is evaluated by investigating selected climatic conditions more than once. to realize the different measurement series the climatic condition within the measuring cell is adjusted anew and the needle, which is attached to the high voltage electrode, is exchanged. in addition to that a long-term study is executed to examine the temporal development of the considered pulse shape parameters. in dependence on the reference atmosphere, these measurements are comprehensively carried out at a temperature level of 20 °c and an absolute humidity level of 11 g/m³. in terms of two different measurement series, the box plots for 100 values of the determined pulse shape parameters are exemplarily shown in fig. 5. regarding the median the rise time evinces with approximately 10 % the largest deviation of all considered pulse shape parameters. in contrast, the deviation of the associated median values of the pulse width and the fall time are smaller than 5 % and therefore very low. the temporal development of the pulse shape parameters, determined from the captured pulse shapes in a long-term study over a period of 360 minutes, is shown in fig. 6. in the case of the rise time, the maximum deviation of all identified medians from that after a stress duration of ten minutes amounts to 5 %. for the pulse width and the fall time the maximum deviation is smaller than 1 %. a consideration of the illustrated temporal development clarifies that a significant decrease of the medians after a stress duration of 10 minutes is not ascertainable. accordingly, the presented measurement method is suited to investigate the influence of the temperature and the humidity level on pulse shape parameters with a high repeatability, if the defined stress duration is obeyed before starting the data acquisition. b. influence of the absolute humidity level a consideration of the influence of a varying absolute humidity on the pulse shape parameters enables a simultaneous consideration of the temperature level and the relative humidity level. as stated in [27] a conversion is feasible according to equation (1). 𝐴𝐻 = 6.11 ∙ 𝑅𝐻 ∙ 𝑒 17.6∙𝑇 243+𝑇 0.4615 ∙ (273 + 𝑇) (1) ah: absolute humidity level in g/m³ rh: relative humidity level in percent t: temperature level of ambient air in °c fig. 7 exemplarily shows the medians for the pulse width and the fall time in dependence on the absolute humidity. it should be noted that the fundamental curve progression is identical for all considered temperature levels. this indicates that the absolute humidity level is the decisive influencing factor, whereby an additional influence of the temperature level on the pulse shape parameters cannot be excluded. c. influence of the temperature level in fig. 8 the box plots for 100 values of the determined pulse shape parameters rise time, pulse width and fall time are depicted as well as those for the apparent charge for positive corona discharges occurring at a varying temperature level. it is shown, that a variation of the temperature level affects the pulse shape for both considered relative humidity levels. all determined pulse shape parameters decrease with an increasing temperature level. this is particularly apparent in the case of the fall time and the pulse width. focusing on the latter and on the measurements at a relative humidity level of 40 %, the median for a temperature level of 20 °c is in the range of 130 ns whereas the median for a temperature level of 65 °c is in the range of 22 ns. this corresponds to a reduction of almost one sixth. fig. 5. box plots for the pulse shape parameters rise time, pulse width and fall time of two different measurement series carried out at a temperature level of 20 °c and a absolute humidity level of 11 g/m³. fig. 6. temporal development of the the pulse shape parameters rise time, pulse width and fall time as part of a long-term study carried out at a temperature level of 20 °c and an absolute humidity level of 11 g/m³. fig. 8. box plots for the pulse shape parameters rise time (top left), pulse width (top right), fall time (bottom left) and the apparent charge (bottom right) of positive corona discharges occuring under a varying temperature level for relative humidity levels of 40 % and 80 %. fig. 7. medians of the pulse width (left) and the fall time (right) for positive corona discharges in dependence of the absolute humdity. fig. 9. box plots for the pulse shape parameters rise time (top left), pulse width (top right), fall time (bottom left) and the apparent charge (bottom right) of positive corona discharges occuring under a varying relative humidity level for temperature levels of 20 °c and 50 °c. d. influence of the relative humidity level the box plots for the determined values concerning an influence of the relative humidity level are illustrated in fig. 9. the distribution and the location of the 100 gained values for the rise time, the pulse width, the fall time and the apparent charge clarifies that the pulse shape is affected by a varying relative humidity level. with the exception of the rise time, all considered parameters decrease with an increasing relative humidity level. in relation to the pulse shape parameters this is particularly apparent in the case of the pulse width and the fall time. focusing on the fall time and on the measurements at 50 °c, the median for a relative humidity level of 20 % is in the range of 77 ns whereas the median for a relative humidity level of 95 % is in the range of 17 ns. this corresponds to a reduction of almost one fifth. besides the fundamental influence of the relative humidity level on the pulse shape parameters, the box plots, illustrated in fig. 9, suggest a linear dependency between the location of the pulse shape parameters and the relative humidity level for the measurements at 20 °c. for the measurements at 50 °c the change in value suspects a non-linear relationship. this indicates that the relative humidity level is not the decisive influencing factor. e. reconciliation with physical processes independent of the considered climatic condition, the captured pulse shapes are characterized by a steep ascent and a slower descent. relating to [28] this fundamental curve progression is directly traceable to involved physical processes. the needle-to-plate setup generates a strongly inhomogeneous electric field. prerequisite for the development of a partial discharge process is the availability of a free electron in the area of the highest electric field strength. in the current case of a needle, appearing as positive high voltage electrode, the free electron is accelerated into the area of increasing field strength. this is associated with an energy absorption. collisions with gas molecules lead to the emergence of secondary electrons and positive ions if the ionization energy is reached. a continued existence of the necessary ionization condition results in an avalanche effect, which causes the steep ascent of the pulse shape. an increase of the mean free path, e.g. by reducing the air density as a result of a rising temperature level, favors this effect. whereas the electrons move very rapidly towards the needle tip and are accepted by the anode or attached to molecules with a high electron affinity, the positive ions form a positive space charge. because of the low mobility of the positive ions, the space charge drifts very slowly in the direction of the cathode. this entails a reduction of the effective field strength near the needle tip, which hampers the ionization process, favors the attachment process and leads to a decreasing discharge current. therefore, the characteristic of the pulse shape is significantly dependent on the ionization and the attachment coefficient experimental tests prove that a surge of the humidity level goes along with an increasing attachment coefficient [29] and an increasing ionization coefficient [30]. the same applies to a rise of the temperature level [31]. this knowledge is already utilized to explain the impact of the temperature and the humidity level on the breakdown voltage and the pd inception voltage. coupled with the fact that the mean free path is magnified, which favors relevant collision processes, a declaration of the ascertained influence of a varying temperature and humidity level on the pulse shape parameters is enabled. considering the temporal progression of the pd process described above, an increasing ionization coefficient results in a steeper ascent of the pulse shape and a decreasing rise time. apart from a growing number of free electrons, the hampering positive space charge is build up in a shorter time span. this is accompanied by a reduction of the pulse width and the amplitude. the minimization of the effective field strength near the anode enhances the possibility of attachment processes. these are favored by an upward attachment coefficient, which results in a decreasing fall time. consequently, the apparent charge decreases as well, if all abovementioned parameters decline. vi. conclusions to obtain detailed information regarding the influence of the temperature and the humidity level, the pulse shapes of corona discharges are captured for 13 different climatic conditions. by performing a long-term study and carrying out investigations at one climatic condition more than once, it is proven that the proposed measuring method is suited to investigate the influence of the temperature and the humidity level. regarding the latter, a determination of the pulse shape parameters results in comparable values, so that a high repeatability is achieved. this applies also if the needle, which emulates corona discharges, is substituted. a comparative analysis of pulse shape parameters, gained from measurements at different climatic conditions, illustrates that an increasing temperature level at a constant relative humidity level as well as an increasing humidity level results in decreasing pulse shape parameters. therefore, the interpretation of pd impulses requires a consideration of the climatic condition. the dependency can be reconciled with the reliance of the ionization coefficient and the attachment coefficient on the varying atmospheric parameter. regarding the influence of the humidity level it is assumed that the absolute humidity is the decisive influencing factor. to verify this assumption, it is recommended to carry out supplementary investigations. this should entail an investigation of a varying temperature level at a constant absolute humidity. references [1] u. lühring, d. wienold and f. jenau, “influence of humidity on pulse shape parameters of positive corona discharges in air at dc voltage”, 17 th international conference on environment and electrical engineering, milan/italy, june 2017. [2] iec 60270:2000, “high-voltage test techniques – partial discharge measurement (iec 60270:2000); german version en 60270:2001”, din deutsches institut für normung e.v. und vde verband der elektrotechnik elektronik informationstechnik e.v., berlin, 2001. [3] d. könig and y. n. rao, “partial discharges in electrical power apparatus”, berlin: vde-verlag, 1993. [4] r. s. bever and j. l. westrom, “partial discharge testing under direct voltage conditions”, ieee transactions on aerospace and electronic systems, vol. aes-18, no. 1, january 1982. [5] u. fromm, “interpretation of partial discharges at dc voltages”, ieee transactions on dielectrics and electrical insulation, vol. 2, no. 5, october 1995. [6] t. vogt, “teilentladungsdiagnose bei gleichspannung”, dissertation technische universität dortmund, 2015. [7] d. wienold, u. lühring and f. jenau, “detection and distinction of partial discharges in air at dc voltage by using a non-conventional approach in the high-frequency range“, 17 th international conference on environment and electrical engineering, milan/italy, june 2017. [8] d. wienold, u. lühring and f. jenau, “data analysis and verification for dc pd measurements in the high frequency range”, 2 nd ieee international conference on dielectrics, budapest/hungary, 2018, in press. [9] t. mutakamihigashi, r. sakurai, s. okada and h. ueno, “relationship between electric field strength and characteristic frequency in partial discharge”, 18 th international symposium on high voltage engineering, seoul/korea, august 2013. [10] t. klueter, j. wulff, f. jenau and d. wienold, “evaluation of surface and corona discharges at dc voltage”, 13 th international conference on environment and electrical engineering, wroclaw/poland, november 2013. [11] t. klueter, j. wulff and f. jenau, “time domain analysis of partial discharges at dc voltage in air and insulation oil”, 12 th international conference on environment and electrical engineering, wroclaw/poland, may 2013. [12] t. klueter, j. wulff and f. jenau, “measurement and statistical analysis of partial discharges at dc voltage”, 48 th international universities´ power engineering conference upec 2013, dublin/ireland, september 2013. [13] u. lühring, d. wienold and f. jenau, “comparative investigation on pule shape parameters of partial discharges in air under ac and dc voltage stress”, 51 st international universities´ power engineering conference upec 2016, coimbra/portugal, september 2016. [14] u. lühring, d. wienold and f. jenau, “investigation on the applicability of the time domain analysis of discharges in gases for the defect identification at ac voltage”, transactions on environment and electrical engineering, vol. 2, no. 1, january 2017. [15] u. lühring, d. wienold and f. jenau, “investigation on the pulse shape of dc corona discharges in air under varying test voltage level”, 2 nd ieee international conference on dielectrics, budapest/hungary, july 2018, in press. [16] d. a. scott and g. n. haddad, “negative point-to-plane corona pulses in oxygen”, journal of physics d: applied physics 19, january 1986. [17] x. liu, d. g. kasten and s. a. sebo, “partial discharge measurements for a twisted pair of insulated conductors at low pressures in air, argon and helium“, ieee international symposium on electrical insulation, june 2006. [18] m. abdel-salam and n. l. allen, “current-voltage characteristics of corona in rod-plane gaps as influenced by temperature”, iee proceedings – science, measurement and technology, vol. 150, no 3, may 2003. [19] n. l. allen, m. abdel-salam and i. cotton, “effects of temperature and pressure change on positive corona and sparkover under direct voltage in short airgaps”, iet science, measurement & technology, vol. 1, no. 4, july 2007. [20] x. bian, l. wang, j. m. k. macalpine and z. guan, “positive corona inception voltages and corona currents for air at various pressures and humidities”, ieee transaction on dielectrics and electrical insulation, vol. 17, no. 1, february 2010. [21] e. kuffel, “influence of humidity on the breakdown voltage of spheregaps and uniform-field gaps”, proceedings of the iee, paper 3322m, february 1961. [22] p. morshuis, “assessment of dielectric degradation by ultrawide-band pd detection”, ieee transactions on dielectrics and electrical insulation, vol. 2, no. 5, october 1995. [23] m. kurrat and d. peier, “wideband measurement of partial discharges for fundamental diagnostics”, 7 th international symposium on high voltage engineering, dresden/germany, august 1991. [24] r. schmerling, m. freiburg, f. jenau, t. weissgerber and f. pohlmann, „development and application of a climate chamber for testing large dimensioned high voltage components up to 100 kv”, 18 th international symposium on high voltage engineering, seoul/korea, august 2013. [25] h. okubo and n. hayakawa, “a novel technique for partial discharge and breakdown investigation based on current pulse waveform analysis”, ieee transactions on dielectrics and electrical insulation, vol. 12, no. 4, august 2005. [26] k. nanao, y. murakami and m. nagao, “analysis of internal partial discharge based on pd current waveform”, annual report: conference on electrical insulation and dielectric phenomena, october 2008. [27] iec 60060-1:2010, “high-voltage test techniques – part 1: general definitions and test requirements (iec 60060-1:2010); german version en 60060-1:2010”, din deutsches institut für normung e.v. und vde verband der elektrotechnik elektronik informationstechnik e.v., berlin, 2011. [28] l. b. loeb, “electrical coronas – their basic physical mechanism”, berkeley and los angeles: university of california press, 1965. [29] e. kuffel, “electron attachment coefficients in oxygen, dry air, humid air and water vapour”, proceedings of the physical society, vol. 74, no. 3, april 1959. [30] w. bauke, “über den einfluß des wasserdampfgehaltes der luft auf den elektrischen durchschlag im inhomogenen feld bei wechselspannung”, dissertation technische universität berlin, 1968. [31] w. s. zaengl, s. yimvuthikul and g. friedrich, “the temperature dependence of homogeneous field breakdown in synthetic air”, ieee transactions on electrical insulation, vol. 26, no. 3, june 1991. i. introduction ii. experimental set-up a. pd measuring circuit b. measuring cell c. test device iii. experimental procedure a. determination of the inception voltage b. capturing of the pulse shapes iv. data analysis and presentation v. results and discussion a. appropriateness of the measurement method b. influence of the absolute humidity level c. influence of the temperature level d. influence of the relative humidity level e. reconciliation with physical processes vi. conclusions references iqawce.pdf an ingenious application-specific quality assessment methods for compressed wireless capsule endoscopy images kinde anlay fante, fetulhak abdurahman and mulugeta tegegn gemeda faculty of electrical and computer engineering, jimma institute of technology, jimma university, ethiopia abstract—image quality assessment methods are used in different image processing applications. among them, image compression and image super-resolution can be mentioned in wireless capsule endoscopy (wce) applications. the existing image compression algorithms for wce employ the generalpurpose image quality assessment (iqa) methods to evaluate the quality of the compressed image. due to the specific nature of the images captured by wce, the general-purpose iqa methods are not optimal and give less correlated results to that of subjective iqa (visual perception). this paper presents improved image quality assessment techniques for wireless capsule endoscopy applications. the proposed objective iqa methods are obtained by modifying the existing full-reference image quality assessment techniques. the modification is done by excluding the noninformative regions, in endoscopic images, in the computation of iqa metrics. the experimental results demonstrate that the proposed iqa method gives an improved peak signal-tonoise ratio (psnr) and structural similarity index (ssim). the proposed image quality assessment methods are more reliable for compressed endoscopic capsule images. index terms—image quality assessment, wireless capsule endoscope, compressed images, image compression. i. introduction wireless capsule endoscopy (wce) is a tablet size camera with a radio frequency transmitter which can be easily ingested by patients to diagnose gastrointestinal abnormalities [1]. it enables the non-invasive imaging of the gastrointestinal tract with a high-resolution camera operating in the visible electromagnetic spectrum. the capsule captures the image of the inner wall of the human digestive system and transmits it to a data recorder located outside the patient’s body. the transmission of high-resolution images via a wireless link consumes high power and requires a high transmission bandwidth [2]. in order to reduce the power consumption and transmission bandwidth, image compression algorithms are included inside the capsule [3]. image processing software on workstation is used to decompress and process the received images in order to improve the visual quality. among the image processing methods, super-resolution algorithms [4] are proposed to improve the spatial resolution of the images, image post-processing and error-correcting codes are proposed to suppress the distortion due to transmission channel noise [5], and computer-aided diagnosis systems [6] are used to improve the detection of abnormalities during the evaluation phase. generally, high quality images are required for accurate diagnosis of abnormalities. ideally, lossless image compression methods are needed to retain all information in medical images. however, the lossless image compression algorithms achieve a low compression rate. it cannot significantly reduce the power consumption and the bandwidth requirement of the wce transmission system. hence, lossy image compression algorithms are preferred to achieve a high compression rate. the amount of information loss due to the compression process should not affect the diagnosis outcome [7]. on the other hand, image compression algorithms that have high information loss achieve a high compression rate. a good compromise between the image quality and compression rate needs to be maintained in this particular application. in order to design an image compression algorithm that can achieve a high compression ratio without affecting the diagnosis accuracy, we need reliable image quality assessment methods. such methods can help us to optimize the parameters of the image compression algorithms, such as quantizer values and sub-sampling factors [2]. the image quality assessment methods can be categorized into two groups: the subjective and objective quality assessment methods [8]. the subjective quality assessment method is more accurate, time-consuming, and expensive. hence, the objective image quality assessment methods are widely used in practice. the objective image quality assessment methods are widely used to quantify the quality of medical images in compression algorithms [2], [3]. among the popular methods, the peak signal-to-noise ratio (psnr) [9], structure similarity index (ssim) [10], and feature similarity index (fsim) [11] were employed to evaluate the distortion introduced due to the quantization operation of the endoscopic image compression algorithms. in literature, a minimum peak signal-to-noise ratio (psnr) value of 30 db has been considered to be sufficient for accurate diagnosis of pathology in endoscopic images [12]. however, in [7], [13], a psnr value of 35 db was considered as a threshold value for an accurate diagnosis of abnormalities using medical images. a reliable objective iqa method for wce application will help to preserve the information needed for accurate diagnosis of abnormalities and optimize the parameters of image compression algorithms. some of these iqa methods, for instance, psnr and ssim, give equal importance to all parts of the image in their evaluation of the image quality assessment metrics. however, the quality of all the parts of the image may not be equally important in some applications. for example, in an endoscopic capsule image, the four corners of the image are non-informative regions. the pixels in those regions are all zero. these regions have no important 11 information, and the quality of the pixels in these regions is unaffected by the image compression algorithms. this makes the existing objective image quality assessment techniques less reliable for wce application, and their correlation with the subjective assessment techniques is low. in this work, we present an improved objective image quality assessment methods that exclude the impact of the non-informative regions. the proposed ad-hoc algorithm improves the performance of general-purpose algorithms by incorporating knowledge about the nature of the images. the incorporated knowledge helps these algorithms to completely remove the impact of noninformative regions in the iqa metrics calculation. the existing objective iqa methods can be classified into three categories based on the availability of reference images [14]. these are full reference, reduced reference, and nonreference iqa methods. the focus of this paper is on full reference iqa methods for compressed endoscopic capsule images. the full reference iqa methods can be further divided into two broad categories. the fist category uses statistical error metrics such as mean square error (mse), psnr, and visual signal-to-noise ratio (vsnr), etc. and the second category takes into account the knowledge of the human visual system (hvs), which includes fsim, ssim, visual information fidelity (vif), etc. these iqa metrics are general-purpose methods and may not be suitable for some applications. an example is an image taken using a capsule camera, as shown in fig. 1. due to the circular shape of the lens of the capsule camera and the rectangular shape of the image sensor arrays, the four corners of the wce image are occluded from the illumination source. the corner regions of this image have no important information. and, all of its pixel values are zero. it also creates a sharp transition boundary between the corner regions and the circular visual region of the image. we can put forward the following two observations from the nature of endoscopic capsule images. first, the quality of the four corner regions of endoscopic images is unaffected by lossy image compression algorithms since all pixel values are zeros. this affects the statistical error based image quality metrics such as mse and psnr. second, the sharp transition boundary between the corner regions (non-informative regions) and circular visual region of the endoscopic capsule image is highly affected by lossy image compression algorithms that involve sub-sampling and quantization such as dct-based [15] and dwt-based [16] algorithms. similarly, the quantization and subsampling operation in dpcm-based [17], [18], and error due to inadequate side information in distributed video coding algorithms [19]. however, this region has no significant information relevant to the diagnosis of abnormalities in the gastrointestinal tract. most of the iqa methods that use the knowledge about the human visual system, such as fsim, ssim, vif, etc., estimate the information content of image regions based on the variance of the image blocks. an image block that lies in the boundary between the non-informative regions and the circular visual region has a very high variance. the presence of this high variance region significantly affects the reliability of the iqa metrics for this particular domain of images. generally, we can use three types of knowledge in the design of image quality assessment methods. these are pieces of knowledge about the human visual system (hvs), statistics of the image, and the distortion type [14]. the previous works on this topic assessed the suitablity of the general-purpose image quality assessment methods for wce application. both subjective and objective image quality assessment methods were presented [20]–[22]. however, the objective image quality assessment methods for compressed wce images [20] have not taken into consideration the peculiar nature of the wce images. by using the knowledge about the wce camera, we developed more reliable image quality assessment methods. in this work, we use the knowledge about the statistics of the image acquisition system to improve the performance of the general-purpose objective iqa methods for wireless capsule endoscopy applications. the contribution of this work is that it excludes non-informative regions in the computation of objective image quality assessment methods for wce application. to the best of our knowledge, this the first work which takes into consideration the peculiar nature of wce images in the image quality assessment method for wce application. the remaining sections of the paper are organized as follows. section ii discusses the analysis of the proposed iqa method. in section iii, the result and discussions are given. finally, the conclusion is given in section iv. ii. analysis of the proposed image quality assessment methods the proposed iqa algorithm improves the performance of the general-purpose iqa methods by excluding the following two regions in the computation of the iqa metrics. these are: (1) the non-informative (corner) regions, and (2) the boundary between the corner regions and the circular visual region. this section will investigate the overall impact of excluding these two regions in the computation of general-purpose iqa metrics for wce application. the block diagram of a fullfig. 1. block diagram of a full-reference objective image quality assessment method for compressed wce images. reference image quality assessment method is depicted in fig. 1. the image is compressed to reduce its size for storage or transmission. the lossy image compression process introduces distortion to the image. in both dpcm-based [2] and dct-based [23] image compression algorithms, the chroma subsampling and quantization processes introduce distortion. the compressed image is then decompressed (decoded) for visualization. the amount of distortion in the decompressed is quantified using image quality assessment method. the image quality assessment method takes both the decoded image and the reference image to compute the objective image quality assessment metrics. a. the effect of corner regions as shown in fig. 2, the visual region of the capsule camera is the circular region. for the image of size w × w , the area of the visual region is πw 2 4 , and the area of the whole image is w 2. the total area of the non-informative regions, ani, is given by: ani = w 2(1 − π 4 ) (1) for w = 256, ani = 14064 pixels. this is about 21.46 % of the total pixels of the image, which constitutes nearly 220, 8 × 8 patches. the total number of pixels in the noninformative region is more than one-fifth of the total numbers of pixels of the image, and the impact on the image quality assessment metrics must be considered. the exclusion of the pixels in the non-informative regions in the computation of image quality assessment metrics improves the reliability of the methods. several image quality assessment methods fig. 2. wireless capsule endoscopy images [24]. were used to evaluate the performance of wireless capsule endoscopy image compression algorithms. among them, mean squared error (mse) and peak signal-to-noise ratio (psnr) can be mentioned [2]. these methods give equal importance to all pixels in the image during the computation of quality assessment metrics. hence, they are generally not correlated with subjective image quality assessment methods [25]. they are used because they are based on the energy of the error signal (difference in signal between the images being compared), which is preserved after linear unitary (orthogonal) transforms such as discrete cosine transform (dct). the mse and psnr are computed using the following equations [9]. mse = ∑m,n,3 i=1,j=1,k=1(xi,j,k − yi,j,k)2 m × n × 3 (2) psnr = 10log10( 2552 mse ) (3) where xi,j,k is the pixel value of the reference image, yi,j,k is the pixel value of the decompressed image, m, n are sizes of the image, i, j are spatial coordinates of the image array and k is the index of the color components of a color image. as described above, the pixel value error in the corner regions of an endoscopic image during the compression process is zero, but the number of pixels shown in equation (2) includes the corner pixels. this reduces the mse and increases the psnr which makes these methods less correlated with the subjective image quality assessment method. to alleviate this drawback, the modified versions of mean squared error (mse wce) and peak signal-to-noise ratio (psnr wce) are given below. mse wce = ∑m,n,2 i=0,j=0,k=0(xi,j,k − yi,j,k)2 m × n × 3 − 3 × nc (4) where nc represents the number of corner pixels. psnr wce = 10log10( 2552 mse wce ) (5) the number of corner pixels, in equation (4), includes those pixels whose values are unaltered by the image decompression process. it should be noted that the corner pixel values which are near to the boundary between the circular visual region and corner regions are affected by the reconstruction process of the lossy image decompression algorithms. furthermore, the effect of the pixels in the corner regions on the objective image quality assessment methods can also be investigated using the structural similarity index (ssim). it is given by the following equation [10]. s(x, y ) = (2μxμy + c1)(2δxy + c2) (μ2x + μ 2 y + c1)(δ 2 x + δ 2 y + c2) (6) where μx, μy are the mean values, δx, δy are the unbiased standard deviations of the reference and distorted images respectively, and c1, c2 are constants. from equation (6), we can conclude that the presence of zero-value corner pixels reduces the mean values and increases the values of variances in both the reference and distorted images. moreover, the corner regions have no structural units that can be affected by compression algorithms. hence, the structures in the decompressed and original images are similar in the corner regions of endoscopic images. this tends to increase the ssim value, which is misleading. to eliminate the effect of the corner regions, the pixel values in the circular visual regions only are included in the computation of iqa metrics. we clip the corner regions and compute the metrics. overall, the presence of corner regions in capsule endoscopic images increases both psnr and ssim values during the evaluation of the quality metrics of the image compression algorithms. hence, the exclusion of the corner regions in the image quality assessment metrics computation of wireless capsule endoscopy images makes the metrics more reliable. this is due to the fact that non-informative regions are not affected by the image compression process, and it increases the image quality scores of the iqa methods. b. the effect of boundary region unlike natural image boundaries, the boundary between corner regions and the circular visual region has a sharp pixel value transition. consequently, it contains large highfrequency components when a block-based transform image compression algorithm is applied. the high-frequency components are quantized using larger quantizer values in such algorithms. hence, this region is highly distorted by image compression algorithms. the highest distortion occurs in this region in both dpcm-based image compression algorithms that employ sub-sampling [2] and transform coding (dct and dwt) methods [15], [16]. this is due to the fact that there is little correlation between the corner pixel values and those in the visual region. the drawback of the existing objective image quality assessment methods for wce is that both the psnr and the ssim values decrease as a result of the artificial boundary. this leads to unrealistic quality metrics value. fig. 3 shows the scaled structural similarity map of the image given in fig. 2(a). as we can see from the figure, the boundary between the corner regions and the circular visual region shows a very high difference. unlike the rest of the region, the boundary region pixels are uncorrelated. this leads to significantly large highfrequency components in block-based transforms. similarly, the uncorrelated nature of the pixels leads to a large error in subsampling based image compression algorithms. however, the distortion in this region has no effect on the quality of the visual perception of the image for of gastrointestinal abnormalities. fig. 3. the structural similarity map of the image shown in fig. 2 (a). similar to the corner regions, the proposed solution in this work is the exclusion of the boundary region in the computation of iqa metrics. in this particular case, threadlike region whose width is 17 pixels (8 pixels into the corner region and 8 pixels into the visual region) perpendicular to the boundary is considered as a boundary region. the exclusion of the boundary region pixels helps us achieve reasonable image quality assessment metrics for wce. iii. results and discussions the experimental setup, the results obtained from the experiments, and the analysis of the results are discussed in this section. a. experimental setup extensive experiments were performed to compare the proposed iqa methods with general-purpose iqa methods. the experiments were designed and implemented by modifying the existing lossy image compression algorithms, which were proposed for wce application. these image compression algorithms use dct [15], dwt [16], and dpcm with subsampling [2] based image processing methods. the comparative analysis of the general-purpose and the proposed iqa methods was performed. the two selected iqa methods used in this work for demonstration purposes are psnr and ssim. these metrics are chosen here because they are widely used in the performance analysis of image compression algorithms for wce. in addition to that, the two methods are good representatives of the two categories of objective iqa algorithms. psnr is based on the statistical error, and ssim is based on the hvs perception [26], [27]. the main goal of these experiments is to prove the concept in practical scenarios using comparative analysis. it is not to compare the performance of different image compression algorithms. in order to evaluate the performance of the proposed iqa methods, 120 images from gastrolab [24] were selected. the selected images are in rgb format with different resolutions. the images from this database were used in the performance evaluation of several image compression algorithms [2], [17]. b. analysis of results the comparison of mean squared error (mse) of the pixel values in the boundary region and those inside the visual area are shown in fig. 4. as shown in the figure, the average mse values of pixels in the boundary region are 54.82, and those of the pixel values inside the visual area are 3.55 for standard jpeg image compression algorithm with quantizer value of 1.5. this shows that the pixels around the boundary region are 0 10 20 30 40 50 60 70 80 90 100 110 120 0 20 40 60 80 100 mse of pixels in the visual area mse of boundary pixels fig. 4. the mean square error value of the pixels in the boundary and those in the visual region of 120 endoscopic images for standard jpeg image compression algorithm with a quantizer value of 1.5. highly affected by the compression algorithms due to the lack of correlation of the pixel values in the two regions. this leads to iqa metrics, which has less correlation with the subjective image quality assessment method. the performance of the proposed ad-hoc iqa methods are evaluated using the implementation of eight different image compression algorithms which were proposed for wce application. both the general-purpose and the proposed iqa methods were implemented for comparative analysis. the summary of the comparison is given in table i. as compared to the conventional iqa methods, the proposed algorithm that employs only the exclusion of corner regions in the iqa metrics computation achieves lower values of psnr and ssim. when the corner region pixels, which are not affected by the image compression process, are excluded from the computation of the iaq metrics, the mse increases as given in equations 2 and 4. this, in turn, leads to a lower psnr value in all the image compression algorithms as shown in equations 3 and 5. on the other hand, the exclusion of 0 20 40 60 80 100 120 4.2 4.4 4.6 4.8 5 5.2 5.4 5.6 5.8 imageid m s e general−purpose mse mse for wce (a) mean squared-error for 120 test images [24]. 0 20 40 60 80 100 120 46 46.5 47 47.5 test image id p s n r ( db ) general−purpose psnr psnr for wce (b) psnr value for 120 test images [24]. fig. 5. comparison of the mse and psnr values for the conventional and proposed iqa methods. table i the comparison of different image quality assessment methods. image compression methods conventional iqa methods proposed iqa methods for wce psnr (db) ssim corner clipping corner clipping and boundary region exclusion psnr (db) ssim psnr (db) ssim jpeg-based [3] 30.79 0.9998 30.07 0.9997 36.38 0.9997 dpcm-based [2] 38.23 0.9996 37.50 0.9995 41.65 0.9998 dct-based [15] 38.62 0.9987 37.89 0.9985 38.94 0.9984 dwt-based [16] 39.90 0.9899 39.42 0.9894 40.56 0.9905 modified h.264 [28] 38.29 0.9990 37.57 0.9988 37.76 0.9988 dct-based [29] 43.60 0.9995 42.88 0.9994 43.66 0.9994 dct-based [30] 31.45 0.9991 30.73 0.9991 31.38 0.9995 dct and sub-sampling [23] 35.99 0.9049 35.26 0.9009 36.65 0.9061 average 37.11 0.9863 36.42 0.9857 38.37 0.9865 the boundary region which is highly affected by the image compression processes reduces the mse value and increases the psnr value. in a nutshell, the exclusion of the corner regions alone reduces the psnr value, whereas the removal of corner pixels alone increases the psnr value. for the eight image compression algorithms implemented shown in table i, the exclusion of corner regions in the computation of psnr alone reduces it from 37.11 db to 36.42 db on average. similarly, the exclusion of both corner regions and boundary region increases the psnr value to 38.37 db on average. this gives an improvement of 1.12 db average value of psnr. hence, the combined effect is dominated by the impact of the boundary region between the corner regions and the visual region because this region is highly distorted by most the image compression algorithms due to the existence of sharp transition of pixel values at the boundary which leads to large values of high frequency components. this is consistent in all the images, compression algorithms. the observed psnr value difference obtained differs from one algorithm to another since different algorithms induce a different level of distortion to the boundary region of wce image. unlike the psnr, ssim value depends on image contrast, hue, and distribution of error [26], [27]. because of this, we have not obtained a consistent pattern of changes in all image compression algorithms shown in table i. on average, it is observed that the ssim decreases from 0.9863 to 0.9857 with application of only corner clipping algorithm. it has increased to 0.9865 of average ssim value with both corner clipping and boundary region omission in the computation of objective iqa for wce images. obviously, the level of distortion of the corners pixels is zero since all the pixel values are zero. as a result, the average ssim and psnr values reduce when the corner region pixels are omitted in the computation of iqa metrics. the overall effect of corner clipping is decreasing the considered iqa metrics on average. the iqa metrics, which exclude only the boundary regions in the iqa computations, achieve a lower value of mse compared to the original value as depicted in fig. 4. hence, it results in higher values of psnr and ssim than those of the conventional methods. this is due to the fact that the boundary region is highly affected by the image compression processes. the quantization and sub-sampling algorithms introduce high error at the boundary between the informative and non-informative regions due to the abrupt transition of intensity values of pixels in wce images, as shown in fig. 3. the distortion in these regions has no significance for the detection of abnormalities in this particular application. the modified iqa metrics, which are implemented with a combination of both corner clipping and boundary region exclusion in iqa metrics computation, show more realistic metric values than the conventional methods. generally, the cumulative effect of excluding both regions is dominated by the impact of the boundary pixels. this shows that the distortion that occurs in the boundary region dominates the overall process, and the proposed iqa metrics are higher than that of the general-purpose iqa metrics. the simulation result shows consistent results for the majority of the evaluated image compression algorithms and test images, as shown in fig. 5 (b). overall, more consistent image quality assessment metrics that are more correlated with subjective image quality assessment are proposed in this work, which can be reliably used in the image compression and super-resolution algorithm optimization process. the exclusion of the informative regions in the computation of the image quality assessment methods can sightly increase the computational complexity of the objective image quality assessment methods. however, with the workstation computational power, we have now, its impact in practical scenarios is negligible in terms of delay. iv. conclusion in this paper, an application-specific objective image quality assessment method for wireless capsule endoscopy is presented. the proposed method uses knowledge about the statistics of the image acquisition system to improve the performance of the general-purpose image quality assessment methods. specifically, the corner regions, which are not affected by image compression algorithms, and the boundary regions between informative and non-informative regions, which are highly distorted by the image compression algorithms, are excluded in the computation of the objective image quality assessment metrics. the distortion in these two regions has no significance for the detection of abnormalities in this particular application. hence, the proposed algorithm has a higher correlation to subjective quality assessment than the general-purpose iqa metrics. the overall impact of excluding the two regions in the computation of iqa metrics is that the general-purpose image quality measures are lower than the iqa measures proposed for wce particular application. the proposed iqa method can be used in the optimization of image compression and image super-resolution algorithms. the proposed algorithm can easily be applied to other image quality assessment methods such as fsim [11] etc. references [1] g. iddan, g. meron, a. glukhovsky, and p. swain, “wireless capsule endoscopy,” nature, vol. 405, no. 6785, pp. 417–417, 2000. [2] k. a. fante, b. bhaumik, and s. chatterjee, “design and implementation of computationally efficient image compressor for wireless capsule endoscopy,” circuits, systems, and signal processing, vol. 35, no. 5, pp. 1677–1703, 2016. [3] c. cheng, z. liu, c. hu, and m. q.-h. meng, “a novel wireless capsule endoscope with jpeg compression engine,” in 2010 ieee international conference on automation and logistics. ieee, 2010, pp. 553–558. [4] y. wang, c. cai, and y. zou, “single image super-resolution via adaptive dictionary pair learning for wireless capsule endoscopy image,” in 2015 ieee international conference on digital signal processing (dsp). ieee, 2015, pp. 595–599. [5] p. a. floor, i. farup, m. pedersen, and ø. hovde, “error reduction through post processing for wireless capsule endoscope video,” eurasip journal on image and video processing, vol. 2020, pp. 1–15, 2020. [6] s. charfi and m. el ansari, “computer-aided diagnosis system for colon abnormalities detection in wireless capsule endoscopy images,” multimedia tools and applications, vol. 77, no. 3, pp. 4047–4064, 2018. [7] r. istepanian, n. philip, m. martini, n. amso, and p. shorvon, “subjective and objective quality assessment in wireless teleultrasonography imaging,” in 2008 30th annual international conference of the ieee engineering in medicine and biology society. ieee, 2008, pp. 5346– 5349. [8] z. wang and a. c. bovik, “modern image quality assessment,” synthesis lectures on image, video, and multimedia processing, vol. 2, no. 1, pp. 1–156, 2006. [9] j. korhonen and j. you, “peak signal-to-noise ratio revisited: is simple beautiful?” in 2012 fourth international workshop on quality of multimedia experience. ieee, 2012, pp. 37–38. [10] z. wang, a. c. bovik, h. r. sheikh, and e. p. simoncelli, “image quality assessment: from error visibility to structural similarity,” ieee transactions on image processing, vol. 13, no. 4, pp. 600–612, 2004. [11] l. zhang, l. zhang, x. mou, and d. zhang, “fsim: a feature similarity index for image quality assessment,” ieee transactions on image processing, vol. 20, no. 8, pp. 2378–2386, 2011. [12] m.-c. lin, l.-r. dung, and p.-k. weng, “an ultra-low-power image compressor for capsule endoscope,” biomedical engineering online, vol. 5, no. 1, pp. 1–8, 2006. [13] p. c. cosman, r. m. gray, and r. a. olshen, “evaluating quality of compressed medical images: snr, subjective rating, and diagnostic accuracy,” proceedings of the ieee, vol. 82, no. 6, pp. 919–932, 1994. [14] l. he, f. gao, w. hou, and l. hao, “objective image quality assessment: a survey,” international journal of computer mathematics, vol. 91, no. 11, pp. 2374–2388, 2014. [15] p. turcza and m. duplaga, “low power fpga-based image processing core for wireless capsule endoscopy,” sensors and actuators a: physical, vol. 172, no. 2, pp. 552–560, 2011. [16] k. a. fante, b. bhaumik, and s. chatterjee, “a low-power color mosaic image compressor based on optimal combination of 1-d discrete wavelet packet transform and dpcm for wireless capsule endoscopy.” in biodevices, 2015, pp. 190–197. [17] i. intzes, h. meng, and j. cosmas, “an ingenious design of a high performance-low complexity image compressor for wireless capsule endoscopy,” sensors, vol. 20, no. 6, p. 1617, 2020. [18] c. babu, d. a. chandy, and p. karthigaikumar, “novel chroma subsampling patterns for wireless capsule endoscopy compression,” neural computing and applications, vol. 32, no. 10, pp. 6353–6362, 2020. [19] b. sushma and p. aparna, “distributed video coding based on classification of frequency bands with block texture conditioned key frame encoder for wireless capsule endoscopy,” biomedical signal processing and control, vol. 60, p. 101940, 2020. [20] m. a. usman, m. r. usman, and s. y. shin, “quality assessment for wireless capsule endoscopy videos compressed via hevc: from diagnostic quality to visual perception,” computers in biology and medicine, vol. 91, pp. 112–134, 2017. [21] m. a. usman and m. g. martini, “on the suitability of vmaf for quality assessment of medical videos: medical ultrasound & wireless capsule endoscopy,” computers in biology and medicine, vol. 113, p. 103383, 2019. [22] l. lévêque, h. liu, s. baraković, j. b. husić, m. martini, m. outtas, l. zhang, a. kumcu, l. platisa, r. rodrigues et al., “on the subjective assessment of the perceived quality of medical images and videos,” in 2018 tenth international conference on quality of multimedia experience (qomex). ieee, 2018, pp. 1–6. [23] m.-c. lin and l.-r. dung, “a subsample-based low-power image compressor for capsule gastrointestinal endoscopy,” eurasip journal on advances in signal processing, vol. 2011, no. 1, p. 257095, 2011. [24] “gastrolab,” date last accessed january 10,2020. [online]. available: http://gastrolab.net/ [25] q. huynh-thu and m. ghanbari, “the accuracy of psnr in predicting video quality for different video scenes and frame rates,” telecommunication systems, vol. 49, no. 1, pp. 35–48, 2012. [26] a. hore and d. ziou, “image quality metrics: psnr vs. ssim,” in 2010 20th international conference on pattern recognition. ieee, 2010, pp. 2366–2369. [27] z. kotevski and p. mitrevski, “experimental comparison of psnr and ssim metrics for video quality estimation,” in international conference on ict innovations. springer, 2009, pp. 357–366. [28] l.-r. dung, y.-y. wu, h.-c. lai, and p.-k. weng, “a modified h. 264 intra-frame video encoder for capsule endoscope,” in 2008 ieee biomedical circuits and systems conference. ieee, 2008, pp. 61–64. [29] p. turcza and m. duplaga, “hardware-efficient low-power image processing system for wireless capsule endoscopy,” ieee journal of biomedical and health informatics, vol. 17, no. 6, pp. 1046–1056, 2013. [30] k. wahid, s.-b. ko, and d. teng, “efficient hardware implementation of an image compressor for wireless capsule endoscopy applications,” in 2008 ieee international joint conference on neural networks (ieee world congress on computational intelligence). ieee, 2008, pp. 2761– 2765. kinde a. fante received the bsc degree in electrical engineering from bahir dar university in 2008, the m.tech degree in electronics and computer engineering from addis ababa university, ethiopia, in 2010,ethiopia and the ph.d. degree in electrical engineering from the indian institute of technology delhi, india, in 2016. currently, he is an assistant professor in the faculty of electrical and computer engineering, jimma institute of technology, jimma university. his research interests include biomedical signal processing, low-power integrated circuits, vlsi design for real-time image and video processing, and machine learning. fetulhak abdurahman received the bsc in electrical engineering in 2009 and the msc in computer engineering in 2014 from addis ababa university. from 20092020 working as a lecturer in jimma university, ethiopia. his research interest includes computer vision, pattern recognition and natural language processing. mulugeta t. gemeda received a bsc degree in electrical and computer engineering in 2017 and m.sc. degree in electronics communication engineering from jimma university, ethiopia, in 2019. currently, he is working as lecturer at jimma institute of technology in the faculty of electrical and computer engineering, jimma university. his research interests include antenna design, mobile communication, and computer networking. untitled 1 system identification and tuning of wireless power transfer systems with multiple magnetically coupled resonators johan winges, thomas rylander, carl petersson, christian ekman, lars-åke johansson and tomas mckelvey abstract—we present a procedure for system identification and tuning of a wireless power transfer (wpt) system with four magnetically coupled resonators, where each resonator consists of a coil and a capacitor bank. the system-identification procedure involves three main steps: 1) individual measurement of the capacitor banks in the system; 2) measurement of the frequency-dependent two-port impedance matrix of the magnetically coupled resonators; and 3) determining the inductance of all coils and their corresponding coupling coefficients using a bayesian approach. the bayesian approach involves solving an optimization problem where we minimize the mismatch between the measured and simulated impedance matrix together with a penalization term that incorporates information from a direct measurement procedure of the inductance and losses of the coils. this identification procedure yields an accurate system model which we use to tune the four capacitance values to recover high system-performance and account for, e.g., manufacturing tolerances and coil displacement. for a prototype wpt system, we achieve 3.3 kw power transfer with 91% system efficiency over an air-gap distance of approximately 20 cm. index terms—bayesian estimation, charging electric vehicles, impedance matching, magnetically coupled resonators, system identification, tuning, wireless power transfer (wpt). i. introduction magnetically coupled resonators are becoming increasingly popular for mid-range wireless power transfer (wpt) that is suitable for charging electric vehicles [1], [2], [3]. given the low magnetic coupling that is typical in such wpt systems, resonant circuits with low losses are used to achieve high power transfer at high efficiency. however, it is crucial that such resonators are tuned in order to achieve satisfactory system performance, as the power transfer and efficiency rapidly decays should the system become de-tuned [4], [5]. aldhaher et al. [6] studied a wpt system with two resonators, where they mitigated de-tuning issues associated with coil misalignment and displacement by means of a current-controlled inductor and a variable switching frequency. however, in automotive applications, it may not be possible manuscript accepted januari 12, 2018. this work was funded by the swedish energy agency in the project “säker induktiv energiöverföring för elfordon – safe wireless energy (sawe)”, which has the project number 38577-1. j. winges, t. rylander and t. mckelvey are with the department of electrical engineering, chalmers university of technology, göteborg, 41296 sweden (e-mail: winges@chalmers.se; rylander@chalmers.se). c. petersson, c. ekman and l.-å. johansson are with qrtech ab, flöjelbergsgatan 1c, mölndal, 43135 sweden. fig. 1. photo of the air-wound coils used in the wpt system prototype with four magnetically coupled resonators. to change the switching frequency due to regulations and standards [7]. beh et al. [8] considered a wpt system that operates at a fixed frequency (13.56 mhz), where they used an automatized impedance matching system with three extra components (an l-match network) to tune the resonance frequency of the resonator pair to that of the power source. similarly, li et al. [9] used four tunable components for a wpt system with two resonators, which allows for some advantages, such as making the resonant network inductive and achieving zero-voltage switching. halpern et al. [10] analyzed the limits of capacitor tuning for a wpt system with two resonators and present the optimal selection of the capacitance values with respect to maximizing a weighted sum of both system efficiency and power transfer. in an effort to increase the power transfer distance and efficiency, wpt systems with more than two magnetically coupled resonators have been considered in the literature [11], [12], [13]. zhong et al. [14] presents a wide variety of socalled domino wpt systems where a large number of magnetic resonators are placed in a domino pattern to increase the power transfer range. liu et al. [15] show how two additional resonant circuits can be located at an off-center location about midway between the primary and secondary resonators to increase the range and performance of a wpt system. however, the total size of such domino-like wpt systems may be unpractical for some mid-range high power transfer applications such as electric vehicle charging. lin et al. [16] show how it is possible to identify the capacitance values and inter coil distances for a 3-coil domino system from measurements of the input voltage transactions on environment and electrical engineering issn 2450-5730 vol 2, no 2 (2017) © johan winges, thomas rylander, carl petersson, christian ekman, lars-åke johansson, tomas mckelvey 2 and current over a frequency band using a genetic optimization algorithm. however, in their work they assume that the coil losses and inductance can be accurately calculated, which is not necessarily the case for more involved coil designs. in this extended article based on the conference submission [17], we present a system identification procedure combined with tuning for a wpt system with four magnetically coupled resonators, where two resonators are located at the primary side and two on the secondary side. here, the four resonators can improve the efficiency and power transfer for applications with low magnetic coupling between the primary and secondary side [11], [12]. moreover, the extra resonant circuits yields additional degrees of freedom in the wpt system design, and we show that that the capacitance values can be used for system tuning. in the system identification step, we estimate the coil inductances, mutual couplings and losses using a bayesian approach [18], where we minimize the weighted sum of two terms: 1) the mismatch between the coil parameters and an approximate measurement of these parameters; and 2) the mismatch between the simulated and measured impedance matrix for the magnetically coupled resonators over a frequency band. here, the system identifications is used to find deviations from a nominal system design due to e.g. manufacturing tolerances and coil misalignment. furthermore, we demonstrate that high power transfer and efficiency can be recovered for an experimentally realized (non-nominal) design at a fixed operating frequency using only one tunable capacitor per resonant circuit, where the capacitance of each capacitor can be adjusted in a few discrete steps. in addition, the realized design complies with a set of circuit constraints which prevent over-heating or voltage breakdown of the components in the wpt system. ii. wpt system design a wpt system consisting of four magnetically coupled resonators is considered and fig. 1 shows a photograph of the coils in a wpt system prototype. the coil geometry of the prototype is described in table i in terms of the number of turns nm, the side length sm and the placement along the z-axis, which is perpendicular to the plane of each coil and coincides with their respective centers. here, the air gap between the two groups of coils is 18.5 cm. we use litz wire with an outer diameter of 10 mm, consisting of 1080 copper strands with a diameter of 0.1 mm. for all coils, the minimum inner bend radius is 4.5 cm. the different coil sizes in the prototype system were chosen to illustrate that different geometrical constraints may be applicable to the primary and secondary side in a wpt system depending on the application. further, the small additional coils may in theory be incorporated into the same plane as the large coils, which yields a wpt system that is relatively thin, which is important for e.g. charging electric vehicles. we use comsol multiphysics® [19] to compute the selfinductance lm of each coil and the mutual coupling coefficient kmn between each coil, and the results are presented in table i. the individual litz wires of the coils are not explicitly modeled in the comsol model, which instead uses a multiturn coil domain applied to a simplified coil geometry where we approximate the coil cross-section as a rectangle of height equal to the wire diameter and length equal to the number of turns times the wire diameter. table i geometry description with number of turns nm of litz wire (� = 10 mm), outer side length sm and axial position zm of the coils shown in fig. 1 enumerated by m = 1, . . . , 4 and corresponding inductance and coupling coefficients computed using a comsol model coil m 1 2 3 4 nm (-) 11 16 8 6 sm (cm) 51.0 29.0 34.0 25.0 zm (cm) 0 2.50 24.5 22.0 lm (µh) 120 66.4 41.3 15.6 k1m (%) 21.3 9.09 6.87 k2m (%) 21.3 6.99 6.77 k3m (%) 9.09 6.99 37.9 a. circuit model a circuit diagram model of the magnetically coupled resonators are shown in fig. 2, which also includes capacitor banks and series resistances that model losses in the system. our nominal system design for the prototype is described by the component values in table i and the capacitance values c1 = 27.0 nf, c2 = 42.7 nf, c3 = 67.7 nf and c4 = 140 nf. at the terminals a1-a2, the magnetically coupled resonators are fed by a power inverter, which is modeled by a squarewave voltage source uin(t) = u0 sgn[cos(ωpt)] in series with an internal resistance rg = 0.25 ω. it operates at the frequency fp = 85 khz with a maximum output voltage of u max 0 = 450 v. at the terminals b1-b2, the magnetically coupled resonators are loaded by a rectifier, smoothing filter and resistive load rl = 27 ω as shown in fig. 3. the current-voltage curve of the diodes is modeled as piece-wise linear with a forward voltage drop of 1.3 v and a forward resistance of 0.08 ω. we define the efficiency η of the system as the power pload transferred to the load rl divided by the input power pin supplied by the generator. the power levels are found after the system has reached steady-state operation, and are time-averaged over one period. the complete circuit is simulated in time-domain using the circuit simulation software ltspice and an in-house circuit simulation implementation in matlab. iii. system identification in simulation, the nominal system design presented in section ii achieves both high power transfer and efficiency. however, our experimental realization suffers from various manufacturing tolerances and, consequently, the actual component values of the magnetically coupled resonators differ from the nominal design. such deviations are common in practical situations where misalignment and incorrect distance between 3 r1 l1 i1 c1 + − u1 a1 a2 r2 l2 i2 c2 − + u2 r3 l3 i3 c3 + − u3 r4 l4 i4 c4 + − u4 b1 b2 primary side a ir -g a p secondary side fig. 2. circuit diagram for the magnetically coupled resonators with four coils and capacitors. d1 d2 d3 d4 b1 b2 30 µh 20 µf 2.2 µh 10 µf rl fig. 3. circuit diagram for rectifier, filter and load resistance. the primary and secondary side are typical disturbances. thus, we propose the following procedure for identifying the component parameters of wpt systems. first, the capacitance of the capacitor banks is measured to high accuracy when they are disconnected from the rest of the circuit. here, the capacitor banks are assumed to be tunable to a few discrete values, where we measure each discrete capacitance value to high accuracy. next, the realized system parameters p = [l1, . . . , l4, k12, . . . , k34, r1, . . . , r4] of the magnetically coupled resonators are identified. initially, the coil inductance, coupling coefficients and losses are measured approximately using a direct procedure where we disconnect the individual coils from the rest of the circuit. the approximate measurement result is denoted p0. next, we connect the capacitor banks (which are set to known values) to form the system of the magnetically coupled resonators. the corresponding model is shown in fig. 2 and we introduce the two-port impedance matrix z = [ zaa zab zba zbb ] , (1) where the index a corresponds to the terminals a1-a2 and the index b corresponds to the terminals b1-b2 and, due to reciprocity, we have zba = zab. the impedance matrix z(fν) is then measured for nν frequency points in a frequency band around the operating frequency of the wpt system. to simplify the notation, we vectorize the three unique elements of the impedance matrix with respect to frequency, i.e. z = vec{zaa(fν), zab(fν), zbb(fν)}. (here the bar notation indicates that the quantity is measured.) to identify the system parameters p, we use a bayesian approach [18] to formulate a weighted maximum a posteriori estimator that combines: 1) the mismatch between the measured impedance matrix z and the simulated impedance matrix z(p) = vec{zaa(fν, p), zab(fν, p), zbb(fν, p)}; and 2) a penalization term that corresponds to the deviation between the parameters p subject to optimization and the measured parameters p0 from the direct procedure. this gives the minimization problem min p 1 nz ‖z(p) − z‖ 2 p + γ np ‖p − p0‖ 2 q . (2) we assume independent gaussian distributions for the measurement errors and compute the norms with respect to the diagonal matrices pkk = 1/|δzkzk| 2 and qll = 1/(δplpl) 2, where δzk and δpl are estimates of the relative standard deviation of the respective measurement uncertainties. in this work, we use δzk = 1/3 % for z, and for the optimization parameters we use δpl = 1/3 % for lm, δpl = 2% for kmn and δpl = 5% for rm. furthermore, nz = 3nν and np = 10 are the number of elements in the vectors z and p, respectively. from (2), we can interpret that the regularization parameter γ penalizes deviation of the estimate p from the measurement p0, weighted by the corresponding measurement uncertainties of the direct measurement procedure. we solve the minimization problem (2) using the gradientbased optimization algorithm fmincon implemented in matlab [20] with the initial guess p0, and an optimum is typically found in less than 70 iterations. an important feature of the proposed optimization problem (2) is that the impedance matrix model only involves linear components and can be solved efficiently in frequency domain. consequently, the objective function is associated with a relatively low computational cost per frequency point and it is therefore feasible to resolve rapid frequency variations over a relatively large frequency band, which is beneficial for real-time applications. iv. tuning the system identification procedure in section iii yields accurate estimates of the circuit parameters, shown in fig. 2, that deviate from the nominal design due to manufacturing tolerances and coil misalignment. consequently, the realized system performance may have deteriorated considerably when compared to the nominal system. however, by tuning the four capacitance values, we can recover high power transfer and efficiency for rather significant perturbations due to, e.g., coil misalignment. to recover high performance, we simulate the wpt system for a large set of different combinations of the discrete values of the capacitor banks and compare the final power transfer and 4 table ii constraints on rms values of the currents and voltages, which are enforced for the components listed within the parentheses for m = 1, . . . , 4 component quantity unit max coil (lm, rm) i a 60 capacitor (cm) u kv 5 i a 40 u × i kva 50 power inverter (rg) i a 30 rectifier (dm) u v 850 i a 15 efficiency after the system has reached steady-state operation. for each of these solutions, we limit the generator voltage u0 such that none of the constraints listed in table ii are violated. these constraints prevent over-heating or voltage breakdown for the particular components in the realized wpt prototype system. here, we require that the power inverter is connected to an inductive load zloadedaa when the terminals b1-b2 of the magnetically coupled resonators are connected to the load circuit shown in fig. 3. thus, we require that 0◦ < 6 zloadedaa < 90 ◦ for the operational frequency fp and remove all solutions that yield a non-inductive load. the remaining solutions are sorted with respect to power transferred to the load rl and a suitable candidate is selected. v. results we present results for the system identification and tuning of the wpt system shown in fig. 1, which is a prototype system intended to transfer 3.3 kw power over a 20 cm air-gap at high efficiency. prior to assembly, we measure all components with an omicron lab bode100 network analyzer using a direct procedure where each component is disconnected from the rest of circuit and measured. a rather limited set of discrete capacitance values for the capacitor banks are realized for the prototype system and the measured results are shown in table iii. the different capacitance values are implemented by a set of discrete low loss capacitors (kemet r73 series) in a parallel configuration connected by switches. table iii discrete values for each cm in capacitor banks parameter values c1 (nf) 17.6, 21.7, 26.4, 30.8, 35.2 c2 (nf) 23.5, 31.1, 39.2, 47.0, 54.8, 62.7 c3 (nf) 47.0, 53.7, 62.0, 66.4, 73.9, 80.6 c4 (nf) 110, 122, 130, 139, 145, 152 next, the coils and capacitors are assembled with the discrete values marked in bold in table iii to form the system of magnetically coupled resonators. we measure the impedance matrix z(fν) for nν = 600 uniformly distributed frequency points fν from 50 khz to 170 khz. in fig. 4, we show the absolute value and phase of the measured impedance matrix zpq(fν) with solid curves and the computed result using the coil parameters from the direct procedure, i.e. zpq(fν, p0), with dashed curves. it should be noted that there are substantial differences between the measured and computed results in fig. 4, with an average relative error of about 40%. this error indicates that we have an unsatisfactory identification of the component values in our system of magnetically coupled resonators. here, this deviation is in part due to the measurement uncertainty of the direct measurement procedure in combination with incorrect inter-coil spacing which may occur as the wpt system is assembled prior to the impedance matrix measurement. it should be noted that the resonant behavior of the magnetically coupled resonators results in that the wpt system is relatively sensitive to rather small perturbations of the circuit parameters. 60 80 100 120 140 160 10 -2 10 0 10 2 10 4 relative error 39.7% (a) magnitude 60 80 100 120 140 160 -300 -250 -200 -150 -100 -50 0 50 100 (b) phase fig. 4. magnitude and phase of impedance matrix entries as a function of frequency: 1) solid curves – measurements zpq(fν); and 2) dashed curves – model zpq(fν, p0) for the initial parameter vector p0. the vertical dashed line show the operating frequency 85 khz. 5 10 -2 10 -1 10 0 10 1 10 2 10 3 0 2000 4000 6000 8000 10000 12000 0 10 20 30 40 50 fig. 5. magnitude of the two different norms in the minimization problem (2) versus penalization. we solve the optimization problem (2) for 10−2 ≤ γ ≤ 103. fig. 5 shows the corresponding magnitude of the two norms in (2) as a function of the penalization parameter γ. we note that the impedance matrix norm increases significantly for γ > 30, which indicates that γ = 30 is a suitable penalization value that achieves good model agreement, with low deviation between the optimization parameters and the measured parameters. selecting the result with γ = 30 yields the optimized parameter vector p∗ and the corresponding impedance matrix entries zpq(fν, p ∗) which are shown with dashed curves in fig. 6. for reference, we also show zpq(fν) with solid curves in fig. 6. it should be noted that the agreement is excellent after the system identification and that the average relative error is less than 4%. table iv shows the initial parameter vector p0 and the optimized result p∗. here, we note a rather large reduction in the coupling coefficients associated with the air-gap when we compare the initial and optimized results. these results indicate that the distance between the primary and secondary side may have been increased by approximately 5-10 mm when the system was reassembled after the direct measurement of the individual components. in realistic usage scenarios of wpt systems, air-gap deviations on the order of at least 10-20 mm are expected due to, among other factors, parking misalignment, variations in the ground clearance between different car models and installation differences between different charging stations. it is reassuring that the system identification procedure presented here significantly reduces the deviation between the model and the measurement. thus, the complete wpt model presented in section ii can be used for accurate system simulation, which in turn can be used to tune the wpt system for high power transmission and efficiency. given the identified model parameters, we tune the four capacitor banks by comparing the simulated performance of the 1080 different discrete capacitance combinations given by table iii using an exhaustive search. for each combination in the exhaustive search, the generator voltage is reduced from umax0 = 450 v until all constraints in table ii hold and, consequently, the maximum power transfer can vary 60 80 100 120 140 160 10 -2 10 0 10 2 10 4 relative error 3.57% (a) magnitude 60 80 100 120 140 160 -300 -250 -200 -150 -100 -50 0 50 100 (b) phase fig. 6. magnitude and phase of impedance matrix entries as a function of frequency: 1) solid curves – measurements zpq(fν); and 2) dashed curves – model zpq(fν, p ∗) for the optimized parameter vector p∗. significantly between different realizations. in the search, we also identify and disregard capacitance combinations which yield non-inductive loading of the power inverter. for the prototype, more than 100 different capacitance combinations yielded a power transfer of at least 3.0 kw with system efficiency of above 87 % in simulation. the top ten capacitance combinations with respect to maximum power transfer are presented in table v. the capacitance combination c1 = 26.4 nf, c2 = 39.2 nf, c3 = 62.0 nf and c4 = 139 nf marked in bold in table v is selected as an appropriate tuning for the prototype system as it has good performance and only required that the first capacitance bank was switched from the previously used value of c1 = 21.7 nf. in an experiment that aim at a power transfer of 3.3 kw, our prototype system transfers 3.35 kw at 91.4% efficiency with this capacitance selection. this results compares well with the expected power transfer of 3.49 kw at 91.1% efficiency predicted by the model at the lower generator input power used in the experiment. 6 table iv initial component values when measured individually for each component by a direct procedure and optimized parameter values using system identification procedure for wpt system shown in fig. 1 after reassembly parameter initial p0 optimized p ∗ change (%) l1 (µh) 126. 126. 0.079 l2 (µh) 68.9 68.6 -0.38 l3 (µh) 45.3 45.4 0.21 l4 (µh) 17.8 17.9 0.67 k12 (%) 20.5 20.1 -2.2 k13 (%) 9.90 8.12 -18. k14 (%) 8.18 6.40 -22. k23 (%) 7.72 6.67 -14. k24 (%) 7.59 7.13 -6.1 k34 (%) 34.7 33.6 -3.4 r1 (mω) 142. 142. 0.31 r2 (mω) 81.3 82.9 2.0 r3 (mω) 66.1 71.7 8.4 r4 (mω) 43.6 46.2 5.9 table v time averaged maximum power transfer and efficiency for the top ten capacitance combinations of the capacitor banks pload η c1 c2 c3 c4 kw % – nf – 6.51 92.3 26.4 39.2 67.1 145 6.48 92.2 26.4 39.2 67.1 152 6.11 92.4 26.4 39.2 67.1 139 5.93 92.3 30.8 39.2 67.1 152 5.91 91.3 30.8 31.3 67.1 139 5.84 91.8 26.4 39.2 62 152 5.75 91.6 26.4 39.2 62 145 5.74 91.3 30.8 31.3 67.1 145 5.57 91.3 26.4 39.2 62 139 5.54 92.3 26.4 39.2 67.1 130 vi. conclusion we have presented a system identification and tuning procedure for a wireless power transfer (wpt) system that consists of four magnetically coupled resonators. in the system identification procedure, we employ a bayesian approach to estimate the self-inductance and losses of the four coils in combination with their magnetic coupling. after the system identification procedure, the wpt model is used to predict appropriate capacitance values for the four capacitor banks to recover high system performance, where the generator voltage is limited by a set of circuit state constraints that prevent component overheating and breakdown. we find that the system identification procedure combined with the four tunable capacitors can account for and take care of rather significant perturbations due to, e.g., manufacturing tolerances and coil displacement. further, a large set of different capacitance selections for the four tunable capacitor banks achieved high system performance for an experimentally realized wpt system design with magnetic coupling coefficients on the order of 6% to 8% between the primary and secondary side. this feature may be of key importance for future high performance wpt systems in applications where rather low magnetic coupling in combination with large variations in coil positions are expected, such as for electric vehicle charging. references [1] s. li and c. c. mi, “wireless power transfer for electric vehicle applications,” ieee j. emerg. sel. top. power electron., vol. 3, no. 1, pp. 4–17, mar 2015. [online]. available: http://ieeexplore.ieee.org/document/6804648/ [2] f. musavi and w. eberle, “overview of wireless power transfer technologies for electric vehicle battery charging,” iet power electron., vol. 7, no. 1, pp. 60–66, jan 2014. [online]. available: http://ieeexplore.ieee.org/document/6715805/ [3] b. esteban, m. sid-ahmed, and n. c. kar, “a comparative study of power supply architectures in wireless ev charging systems,” ieee trans. power electron., vol. 30, no. 11, pp. 6408–6422, nov 2015. [online]. available: http://ieeexplore.ieee.org/document/7119591/ [4] a. kurs, a. karalis, r. moffatt, j. d. joannopoulos, p. fisher, and m. soljacic, “wireless power transfer via strongly coupled magnetic resonances,” science, vol. 317, no. 5834, pp. 83–86, jul 2007. [online]. available: http://www.sciencemag.org/cgi/doi/10.1126/science.1143254 [5] a. p. sample, d. a. meyer, and j. r. smith, “analysis, experimental results, and range adaptation of magnetically coupled resonators for wireless power transfer,” ieee trans. ind. electron., vol. 58, no. 2, pp. 544–554, feb 2011. [online]. available: http://ieeexplore.ieee.org/document/5437250/ [6] s. aldhaher, p. c.-k. luk, and j. f. whidborne, “electronic tuning of misaligned coils in wireless power transfer systems,” ieee trans. power electron., vol. 29, no. 11, pp. 5975–5982, jan 2014. [online]. available: http://ieeexplore.ieee.org/document/6702433/ [7] sae standard, “j2954, wireless power transfer for light-duty plug-in/electric vehicles and alignment methodology,” 2016. [online]. available: http://standards.sae.org/j2954 201605/ [8] t. c. beh, m. kato, t. imura, s. oh, and y. hori, “automated impedance matching system for robust wireless power transfer via magnetic resonance coupling,” ieee trans. ind. electron., vol. 60, no. 9, pp. 3689–3698, sep 2013. [online]. available: http://ieeexplore.ieee.org/document/6226848/ [9] s. li, w. li, j. deng, t. d. nguyen, and c. c. mi, “a double-sided lcc compensation network and its tuning method for wireless power transfer,” ieee trans. veh. technol., vol. 64, no. 6, pp. 2261–2273, jun 2015. [online]. available: http://ieeexplore.ieee.org/document/6876154/ [10] m. e. halpern and d. c. ng, “optimal tuning of inductive wireless power links: limits of performance,” ieee trans. circuits syst. i regul. pap., vol. 62, no. 3, pp. 725–732, jan 2015. [online]. available: http://ieeexplore.ieee.org/document/7024943/ [11] m. kiani and m. ghovanloo, “the circuit theory behind coupledmode magnetic resonance-based wireless power transmission,” ieee trans. circuits syst. i regul. pap., vol. 59, no. 9, pp. 2065–2074, sep 2012. [online]. available: http://ieeexplore.ieee.org/document/6138883/ [12] s. y. r. hui, w. zhong, and c. k. lee, “a critical review of recent progress in mid-range wireless power transfer,” ieee trans. power electron., vol. 29, no. 9, pp. 4500–4511, sep 2014. [online]. available: http://ieeexplore.ieee.org/document/6472081/ [13] k. lee and s. h. chae, “power transfer efficiency analysis of intermediate-resonator for wireless power transfer,” ieee trans. power electron., vol. 8993, no. c, pp. 1–1, apr 2017. [online]. available: http://ieeexplore.ieee.org/document/7913710/ [14] w. zhong, c. k. lee, and s. y. ron hui, “general analysis on the use of tesla’s resonators in domino forms for wireless power transfer,” ieee trans. ind. electron., vol. 60, no. 1, pp. 261–270, oct 2013. [online]. available: http://ieeexplore.ieee.org/document/6041026/ [15] x. liu and g. wang, “a novel wireless power transfer system with double intermediate resonant coils,” ieee trans. ind. electron., vol. 63, no. 4, pp. 2174–2180, dec 2016. [online]. available: http://ieeexplore.ieee.org/document/7362019/ 7 [16] d. lin, j. yin, and s. y. r. hui, “parameter identification of wireless power transfer systems using input voltage and current,” energy convers. congr. expo. (ecce), 2014 ieee, pp. 832–836, nov 2014. [online]. available: http://ieeexplore.ieee.org/document/6953483/ [17] j. winges, t. rylander, t. mckelvey, c. petersson, c. ekman, and l. å. johansson, “system identification and tuning of wpt systems,” in proc. eeeic/i&cps conf., jun 2017, pp. 1–5. [online]. available: http://ieeexplore.ieee.org/document/7977544/ [18] s. m. kay and m. k. steven, fundamentals of statistical signal processing: estimation theory. englewood cliffs, nj, usa: prenticehall, inc., 1993. [19] comsol ab, “comsol multiphysics® v. 5.2,” stockholm, sweden, 2016. [online]. available: www.comsol.com [20] the mathworks inc., “matlab®,” natick, massachusetts, united states, 2017. [online]. available: www.mathworks.com johan winges received the b.s. degree in engineering physics from chalmers university of technology, gothenburg, sweden, in 2009, the m.s. degree in applied physics from chalmers university of technology in 2011, and the ph.d. degree in electrical engineering from chalmers university of technology in 2016. he is currently a researcher at chalmers university of technology in the signal processing group at the department of electrical engineering. his research interests are computational electromagnetics, inverse scattering problems and wireless power transfer. thomas rylander received the m.s. degree in electrical engineering from the kth royal institute of technology, stockholm, sweden, in 1997, and the ph.d. degree in electrical engineering from chalmers university of technology, göteborg, sweden, in 2002. he was a post-doctoral associate with the center for computational electromagnetics, university of illinois at urbana-champaign, champaign, il, usa, from 2002 to 2004. he was an assistant professor and an associate professor with chalmers university of technology in 2004 and 2007, respectively. his research interests are electromagnetic field theory with a broad range of applications. in particular, his research is based on computational electromagnetics with a focus on finiteelement-based methods. tomas mckelvey received the m.s. degree in electrical engineering from lund university, lund, sweden, in 1991, and the ph.d. degree in automatic control from linköping university, linköping, sweden, in 1995. he has held research and teaching positions with linköping university from 1995 to 1999, where he became a docent in 1999. from 1999 to 2000, he was a visiting researcher with the university of newcastle, newcastle, nsw, australia. since 2000, he has been with the chalmers university of technology, gothenburg, sweden, where he has been a full professor since 2006. he has been the head of the signal processing group, chalmers university of technology, since 2011. his current research interests include modelbased and statistical signal processing, system identification, machine learning, image processing and control with applications to biomedical engineering, and combustion engines. microsoft word 141-641-1-ce.docx transactions on environment and electrical engineering issn 2450-5730 vol 4, no 1 (2021) © luís fabiano barone martins, ricardo breganon, uiliam nelson lendzion tomaz alves, joão paulo lima silva de almeida and elenilson de vargas fortes abstract— in this work, a state-space model and control of a dc-dc buck converter, considering a continuous operating, are presented. a pid controller is considered in the strategy, that considers a low pass filter in the derivative term. the proposed model is validated by comparing it with a switched model. the pid gains are obtained by the ziegler-nichols method. in order to improve the system’s performance considering an environment containing high-frequency noises, the modified pid controller is implemented with several configurations. index terms—choppers, step down converter, pid controller, ziegler-nichols method, average model. i. introduction ower electronics is an applied science that aims the static converters study in order to control and convert electrical energy into a compatible signal (voltage level and/or frequency) to determined devices. these converters operate by switching semiconductor devices and they can classified as: ac-dc, ac-ac, dc-ac and dc-dc [1], [2]. dc-dc converters, also known as choppers, are power electronic circuits with linear and non-linear characteristics, that are composed by resistors, capacitors and inductors and semiconductor switches, respectively [3]. their applications encompass from power sources for electronic devices to photovoltaic generation plants. by controlling the input voltage switching, dc-dc converters can adjust the magnitude of the output voltage analogously to a ac transformer, adding the possibility of adjusting the output voltage [4]. the mentioned converters operate by the pulse width modulation (pwm) principle, in which a semiconductor switch, actuated by a pulse generator, is used to control the amperage level from the input to the output of the electrical system [2], [4]. this technique aims to reach the highest efficiency as possible (100%), however, due to the real characteristics (non-ideal) of these semiconductor devices, a typical efficiency related to the converters ranges from 70% to 95% [5]. the four classical topologies for non-isolated dc-dc converters are: cuk, buck (step-down); boost (step-up); and e buck-boost [6]. each one of them has exclusive properties that include the steady-state voltage gain, the source of the input and output currents and the oscillatory characteristics of the output voltage. the most common topology, and probably the simplest, is the buck (step-down), possibly because its output voltage is always less than the input voltage, with the same polarity and not isolated. the first step to project an efficient controller to the choppers is to obtain a complete system’s model, including the most part of the non-linearities. a linear model with small signals, by means of a state-space average model around an adequate operation point, is a valid option [7]. in general, the most control techniques applied on choppers represents a challenging field due to their non-linear and timevarying characteristics [8]. these techniques perform the switching control of the semiconductor devices aiming the maximization of the efficiency related to the power transfer and the good tracking of the output voltage. several control techniques are used in these converters, although there is a preference for controllers with simple and low cost architecture in industrial applications [9], [10]. in this way, proportional-integrative-derivative (pid) controllers are highlighted [11], [12]. one of the most popular method to adjust the pid gains is the proposed by ziegler e nichols (zn) [13], due to its simple rules and satisfactory performances when it is applied on first order systems [14]. in general, zn method is used to tune pid controllers under ideal conditions without significant noises and disturbances, which often does not represent a real situation. therefore, it is necessary to consider their effects in the control algorithm [15]. on the other hand, a filter system is necessary when there are significant noise levels in the control plant. traditionally, filters are applied on derivative term, due to its characteristic in amplification high-frequency noises. this work approaches the high-frequency noise effects in dc-dc buck converter coupled to a pid controller, with a low-pass filter in the derivative term and tuned by zn method. to do this, a linear model to small disturbances of the buck converter is considered, by means of the average state-space a comparative study among pid structures applied on a buck converter control luís fabiano barone martins1, ricardo breganon2, uiliam nelson lendzion tomaz alves3, joão paulo lima silva de almeida4, and elenilson de vargas fortes5 p model and also considering the dynamic of the pid controller. finally, the system’s output response for two different pid gains with the low-pass filter in the derivative term is compared with a standard pid architecture. the rest of the paper is divided as follows: the pid control structures considered in this work are presented in section ii; in section iii, the mathematical description of the pid controller considering the average model of the buck converter is addressed; the zn method is discussed in section iv; the main results and conclusions are presented in the sections v and vi, respectively. ii. pid controller the feedback loop in control systems aims to provide the real output signal (controlled variable added to noises, disturbances, and others) of the plant to the control system’s input and then to generate a signal to be applied on the manipulated variable, according to the considered control law and aiming to reach a desired reference. the pid controller uses this feedback concept with the proportional (p), integrative (i) and derivative (d) gains, all of them related to the error (difference between the desired reference and the real output signal), which present specific contributions in the control signal calculation [16]. the well-known pid topology, the isa algorithm [17], is presented in eq. (1).         0 1 t p d i de t u t k e t e d t t dt            (1) where u is the control variable; e r y  is the error calculation; r is the reference value (set point); y is the controlled variable (system’s output); while i t and d t are constants integral and derivative times, respectively, and p k is the proportional gain of the pid controller. although simple, the control output computed by the eq. (1) presents severe implications in real applications, mainly due to the derivative part (related to d t constant time), that carries out noise amplification and then the control variable ( u ) may be unfeasible. let n a sinusoidal noise with amplitude and frequency(  ) given by (2), and n u is its contribution to the derivative term in the control signal, given by (3).  sinn a t (2)       sin cos cos n p d p d hf d u k t a t ak t t dt k a t          (3) according to eq. (3), when    , hfk   . in order to avoid unfeasible values (very high values) from derivative part in high frequencies, a common procedure is to limit its bandwidth using a low-pass filter with the transfer function (tf) defined in (4) [18]. 1 1 f d g t s   (4) according to [16], acceptable values for  ranges from 0.05 to 0.125. converting isa algorithm presented in eq. (1) into frequency domain and applying the tf of eq. (4) to derivative term, eq. (5) can be defined.       1 1 1 d p i d st u s k e s st s t          (5) in addition to limiting gains of the high frequency components from the error, by the relation p k  , the lowpass filter solves the problem of the non-causality of the ideal pid controller (eq. (1)) adding a pole in its transfer function. iii. ziegler-nichols method ziegler and nichols publish in 1942 a work [19] in which two they describe two strategies to tune p, pi and pid gains. these strategies contemplate a step and frequency response methods. in this work, the step response method is adopted. ziegler and nichols defined as acceptable that the ratio between the amplitude peaks (due to a disturbance in the operation point) in the closed-loop response ( 1a and 2a in fig. 1 (a)) is about 4. however, there is no guarantee that this ratio corresponds to a real system, after tunning process. fig. 1. (a) impulse response of a system with transport delay; (b) step response method. an illustrative example of the step response method is shown in fig. 1 (b), considering an open-loop system. this response represents a first-order system with delay transport and its transfer function  h s is presented in eq. (6). 0 t 1 t 2 0 k time (b) open loop system fit line z-n method 2 4 6 8 -0.5 0 0.5 1.0 angle (rad) (a) a1 = 0.742 a2 = 0.185     1 2 1 1 stk h s e t t s    (6) the first-order response (fig 1 (b)) is composed by two parameters: the time delay ( 1t ) and the time constant 2 1t t . they can be determined with a tangent line to the inflection point and by observing its intersections between the axis related to the time and the state-steady value, k . in real time control systems, a wide variety of process plant can be modeled as (6) and from k , 1t and 2t it is possible to obtain the controller gains as described in table i. table i controller’s gain s o btained from the step response method. tipo do controlador kp ti td p 2 1 1 t t t  pi 2 1 1 9 10 t t t  1 10 3 t pid 2 1 1 6 5 t t kt  12t 1 1 2 t iv. pid controller added to the average state space model in dc-dc buck converter, shown in fig. 2, the input voltage is represented by an ideal dc source inv , while the switch s is a mosfet transistor. li is the current through inductor l , c v is the capacitor’s voltage ( c ) and d ∈ �0,1� is the pwm’s duty cycle and it is the control signal of the switch s . r and d are resistor and diode, respectively. fig. 2. dc-dc buck converter. in order to determine the differential equations of the buck converter, with switching period t , the kirchhoff’s laws are applied to each functioning state, that is, when s is closed at  0, dt and when s is open at  ,dt t . in this way, the average state space model for the buck converter [7] is given by eq. (7)-(8). in1 l c v i v d l l   & (7) 1 1 c l c v i v c rc  & (8) the transfer function in (9) is related to the capacitor voltage under a small duty cycle variation around the operation point of the buck converter.       2 1 1 1 c in v s lc g s v d s s s rc lc       (9) from the buck converter definition (eq. (7)-(8)), an expanded state-space model is proposed, including the modified pid dynamics (eq. (5)). the buck converter dynamic with a pid controller can be described by (10).       , , 0 , , , , x f x w u h x w u y g x w u    & (10) where f , h and g represent differential, algebraic and output equations, respectively, while x , w and u are the following vectors, respectively: state variables, algebraic vector and system’s input. the use of linear techniques for stability analysis under small disturbances in a dynamic system is a valid approach due to its linearization around an equilibrium point. in this way, assuming small variations around the operation point  0 0 0, ,x w u , the linearized form of (10) can be described in matrix form as (11).                             1 2 1 3 4 2 5 6 3 δx j j b δx 0 j j b δw δy j j b δu & (11) in (11), j matrices are the system’s jacobians at the equilibrium point  0 0 0, ,x w u . assuming that 4j is nonsingular, δw (vector of linearized algebraic variables) can be excluded and then the linear and time-invariant in (11) can be rewritten as (12).                 e ee e e e a bδx δx c dδy δu & (12) in (12), the state e a , input e b , output e c and feedback e d matrices can be defined, respectively by eq. (13)-(16). 1  e 1 2 4 3 a j j j j (13) vin + il vc 1  e 1 2 4 2 b b j j b (14) 1  e 5 6 4 3 c j j j j (15) 1  e 3 6 4 2 d b j j b (16) for the system addressed in this paper, the output variables are not related to the system’s input which implies  e d 0 . in order to represent the buck converter with a control about the capacitor voltage c v , as in (12), it is necessary to consider each term of the pid dynamic equation (eq. (5)), individually. from that, fig. 3 illustrates the considered control structure through block diagram with the algebraic and states variables from (7) and (8). fig. 3. block diagram of the system with the pid controller. from fig. 3, it is possible to obtain the new state variables, i  and d   , in addiction to the equations that define the algebraic variables p  , d  and  d s , as shown in eq. (17)(21), respectively. 1 i p i t  & (17)  1d p d d t       & (18)  refp s c ck v v    (19)  1d p d      (20)  p p i dd k      (21) applying the linearization by first order taylor approximation to eq. (7)-(8) and (17)-(21) and considering the linearized system’s output as c y v   , it is possible to obtain the matrix formulation for the complete system pid-buck, as shown in (22), in which the vector of the expanded state variables  tl c i di v       eδx . in10 0 0 0 0 0 1 1 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 1 0 0 0 0 0 00 0 0 0 0 1 0 0 0 1 1 0 0 0 1 0 0 0 0 0 1 0 0 1 0 0 0 0 0 0 ll cc ii i dd p d d s s p p p v l l ii c r c vv t t t k k y k k k                                                                       & & & & ref d c d v                          (22) the submatrices j and b are used to determine e a , e b and e c , from (22) and (13)-(16), resulting (23)-(25).  11 0 1 1 0 0 0 0 0 1 0 0 i p s i p i p s i s d d v k k v k v k l l l l c rc k t k t t                                   e a (23)  1 0 t i p s s s i d v k k k k l t t            e b (24)  0 1 0 0ec (25) from eqs. (23)-(25), it is possible to represent the dynamic model of the buck converter in matrix form and compact, when it is under small disturbances around an operating point  ref, cvex . the transfer function (  pg s ), aiming the control of voltage c v , is obtained according to the model parameters in state space, assuming that the disturbance inputs are null, as in eq. (26).         1 ref c p c v s g s s v s      e e e c i a b (26) in this work, results obtained considering the both pid (standard) and modified, according to eq. (1) and (5), respectively, are compared, in order to validate the approach addressed in this paper. for that, the transfer function of each compared controller, defined in (27) and (28), with the constants 0a , 1a , 2a and 3a , is defined in eq. (29)-(32). 1 1dt s  pk d st pk ref c v c v p i d  d   d s1 ist pk  p s k    20 1 3 2 2 1 0 1 i d i p a t t s t s g s s a s a s a       (27)           2 2 0 4 3 2 3 2 1 1 0 0 1 1 1 1 1 1 1 p i d i d d d d d d d d g s a t t s t t s t s a s a t a s a t a s a t t t t                       (28) 0 1 p s i i k k v a lc t  (29)  1 1 1 p s i a k k v lc   (30) 2 1 p s i d l a k k v t lc r       (31) 3 1 d l a lc t lc r         (32) v. results the proposed system, based on the transfer functions of buck converter,  g s , and pid controller,  cg s , and considering the voltage sensor gain, s k , can be represented through the diagram of fig. 4. fig. 4. system’s diagram. in fig. 4,  n s represents the measurement noises of cv and  cg s encompasses eq. (27)-(32) and the characteristics of the buck converter shown in table ii. table ii parameters of the contro lled sy stem parameter value input voltage ( inv ) 12 v output voltage ( c v ) 5 v inductance ( l ) 2.4 mh output capacitance ( c ) 5.6 µf load resistance ( r ) 10 ω pwm frequency ( f ) 10 khz with the output power of 10 w, the ripples of the output voltage, c v , and of the inductor’s current, l i , are 5.39% and 6.04%, respectively, which leads the converter to operate in continuous mode. in order to validate the average state-space model of the buck converter, represented by eq. (7) and (8), the simulation software psim® was used. to do that, all parameters of table ii were considered to obtain the switched model response. furthermore, a 5v step was applied in the reference voltage of the switched models, aiming to obtain the openloop response. the respective responses under disturbances are presented in fig. 5. fig. 5. transfer functions’ response related to the state-space and switched model. it is important to highlight in fig. 5 that the state-space average model is a satisfactory representation of the converter’s dynamic under low frequency. the suppression of high frequency components of the switched model, due to the discontinuity of the switching process, occurs with the use of average values of the instantaneous variables for a switching period of the power switch s of fig. 2. after validation of the average model, the response of the open-loop transfer function not compensated was obtained (fig. 6), considering   1cg s  and   0n s  . in this way, a step of 0.415 was applied in the duty-cycle of pwm reference ( refd ). that response is presented in fig. 6 and it can be represented by (6) with a line tangent to its inflection point. the constants of static gain, transport delay and apparent time (eq. (6)) can be determined by inspection of fig. 6, that is, 0.415k  , 1 32t  µs and 2 1 322t t  µs. based on these values and according to the table i, the pid parameters are presented in table iii. table iii gain and time constants of pid controller by means of zieglernichols method. controller kp ti (µs) td (µs) pid 29 64 16  g s cg s s k s k   cv srefcv  p s  d s  n s  ref d  outd s fig. 6. response for ziegler-nichols method. applying the constants of table i and iii in the closed-loop transfer function presented in (27), that is, considering a standard pid ( 0  ), it is possible to determine the constants about damping ( 0.391   ) and the damped frequency ( 36.4 d   krad/s) of the oscillatory mode caused by the pair of conjugated complex poles. the time response with ref 5 c v  v is presented in fig. 7. fig. 7. response of the capacitor voltage ( c v ) with changes in the reference voltage ( ref c v ). from the response presented in fig. 7, it is possible to conclude that the controlled variable presents low accommodation time for starting ( 1 a t  ms) and for disturbances around the equilibrium point ( 0.7 a t  ms), however the percentual overshoot is higher than 60%. therefore, if the designer aims a better performance for overshoot characteristic, a fine tuning would be necessary in the pid gains. according to the literature [16]–[18], a standard pid controller requires a special attention in its derivative term when it is applied in real process. in order to mitigate this situation, a modified structure is considered in this work, in which a low-pass filter is added to derivative term, as presented in (5). to compare the performance between the standard and modified pid architectures, a gaussian noise (average value equal to zero, components at frequencies up to 2.5 mhz and standard deviation  = 0.1) is inserted into the system. fig. 8. (a) density and probability function; (b) gaussian noise. the response of the density and probability function is shown in fig. 8 (a), while in fig. 8 (b) is the gaussian noise along the time (applied in  n s ). to compare the performance between the considered standard and modified pid controllers, 3 levels of  (eq. (28) ) are considered: 0  (pid standard in (27)); 0.05  ; and 0.125  . the last two values are chosen based on [16]. the system’s responses to small disturbances around the operation point, considering the two mentioned controllers are presented in fig. 9. in the first experiment ( 0  ), the obtained response is the dotted line (fig. 9), which suggests the difficulty of the control action to follow the reference quickly, when this characteristic is compared to the other cases. in the following experiments, with 0.05  and 0.125  , the responses are presented in blue and black colors, respectively. it is reasonable to conclude that the voltage c v is faster when 0.125  , in which there is a greater rejection of high-frequency disturbances due to the modified pid. this effect can be explained due to the lower cutoff frequency of the low-pass filter with 0.125  (79 khz), when it is compared to 0.05  (198 khz). fig. 9. response of the capacitor voltage ( c v ) to the changes in the reference voltage ( ref c v ). 0 1.00 3.25 5.50 7.75 10.00 time (ms) 0 3 6 9 c ap ac it o r v o lt ag e (v ) reference state-space average model it is important to notice that the increase of  causes a higher overshoot in the system’s response related to the reference voltage, what is expected due to the decrease in the derivative term. vi. conclusions a modified pid controller (a low-pass filter is added to the derivative term) was considered in the state-space average model of a dc-dc buck converter. from the expanded state equations, that is, adding the controller dynamics, two closedloop transfer functions were obtained, related to the use of standard and the modified pid controllers, respectively. in order to validate the proposed model, its step response was compared to the switching model available on psim® software and the obtained results show equivalent dynamics. the pid controller was tuned by means of ziegler-nichols method. is can be observed that, from a step input reference, the output voltage reaches acceptable performance, according to the project requirements adopted in this work. aiming the minimization of a high frequency noises effects, a low-pass filter was considered in the derivative term of the pid controller. its performance was compared to a standard pid, according to the literature. from the presented results, it can be concluded that the system’s output shows faster response, when it is compared to the other addressed cases. in future works, we intend to consider other control techniques with optimization methods, aiming to minimize the overshoot and the stabilization time. references [1] n. mohan and m. p. e. r. \& education, first course on power electronics and drives. mnpere, 2003. [2] b. k. bose, “power electronics-a technology review,” proc. ieee, 1992, doi: 10.1109/5.158603. [3] r. b. ridley, “a new, continuous-time model for current-mode control,” ieee trans. power electron., 1991, doi: 10.1109/63.76813. [4] n. mohan, t. m. undeland, and w. p. robbins, power electronics: converters, applications, and design, no. v. 1. john wiley \& sons, 2003. [5] r. w. erickson, “dc–dc power converters,” in wiley encyclopedia of electrical and electronics engineering, american cancer society, 2007. [6] e. van dijk, j. n. spruijt, d. m. o’sullivan, and j. b. klaassens, “pwm-switch modeling of dc-dc converters,” ieee trans. power electron., vol. 10, no. 6, pp. 659–665, 1995, doi: 10.1109/63.471285. [7] r. d. middlebrook and s. cuk, “a general unified approach to modelling switching-converter power stages,” in 1976 ieee power electronics specialists conference, 1976, pp. 18–34, doi: 10.1109/pesc.1976.7072895. [8] s. buso, “design of a robust voltage controller for a buck-boost converter using /spl mu/-synthesis,” ieee trans. control syst. technol., vol. 7, no. 2, pp. 222–229, 1999, doi: 10.1109/87.748148. [9] c. k. tse and k. m. adams, “quasi-linear modeling and control of dc-dc converters,” ieee trans. power electron., vol. 7, no. 2, pp. 315–323, 1992, doi: 10.1109/63.136248. [10] a. j. forsyth and s. v. mollov, “modelling and control of dc-dc converters,” power eng. j., vol. 12, no. 5, pp. 229–236, 1998, doi: 10.1049/pe:19980507. [11] s. seshagiri, e. block, i. larrea, and l. soares, “optimal pid design for voltage mode control of dc-dc buck converters,” in 2016 indian control conference (icc), jan. 2016, pp. 99–104, doi: 10.1109/indiancc.2016.7441112. [12] d. e. rivera, m. morari, and s. skogestad, “internal model control: pid controller design,” ind. eng. chem. process des. dev., vol. 25, no. 1, pp. 252–265, jan. 1986, doi: 10.1021/i200032a041. [13] j. g. ziegler and n. b. nichols, “optimum settings for automatic controllers,” j. dyn. syst. meas. control, vol. 115, no. 2b, pp. 220– 222, 1993, doi: 10.1115/1.2899060. [14] c. c. hang, k. j. astrom, and w. k. ho, “refinements of the ziegler-nichols tuning formula,” iee proc. d control theory appl., vol. 138, no. 2, pp. 111–118, 1991, doi: 10.1049/ipd.1991.0015. [15] s. agrawal, v. kumar, k. p. s. rana, and p. mishra, “optimization of pid controller with first order noise filter,” in 2015 international conference on futuristic trends on computational analysis and knowledge management (ablaze), 2015, pp. 226–231, doi: 10.1109/ablaze.2015.7154996. [16] k. j. äström and t. hägglund, advanced pid control. research triangle park, nc: isa-the instrumentation, systems, and automation society, 2006. [17] a. o’dwyer, handbook of pi and pid controller tuning rules, 3rd ed. covent garden, london: imperial college press, 2009. [18] m. a. johnson et al., pid control: new identification and design methods. london: springer-verlag, 2005. [19] j. g. ziegler and n. b. nichols, “optimum settings for automatic controllers,” the american society of mechanical engineers (asme), vol. 64, pp. 759–768, 1942. luís fabiano barone martins is professor at the federal institute of paraná, jacarezinho, is graduated in electrical engineering from são paulo state university (unesp), campus ilha solteira, msc. degree in electrical engineering and phd. degree in electrical engineering from unesp campus bauru and ilha solteira, respectively. his research interests include electrical power systems, power electronics, bio-inspired optimization metaheuristics, industrial automation and dynamic systems control. ricardo breganon is professor at the federal institute of paraná, jacarezinho, holds degree in production engineering from estacio de sa university and in mechanical technology from federal technological university of paraná, msc. degree in mechanical engineering and phd. degree in mechanical engineering with a concentration in aeronautical from school of engineering of são carlos university of são paulo. his research interests include industrial automation and dynamic systems control. uiliam nelson l. t. alves is graduated in control and automation engineering from unicesumar, maringá, pr, brazil (2011). msc. and phd. in electrical engineering from unesp (universidade estadual paulista), campus of ilha solteira, sp, brazil in 2014 and 2017, respectively. at this moment, he is control and industrial processes professor in ifpr (instituto federal de educação, ciência e tecnologia do paraná), campus jacarezinho. his research interests include fuzzy modeling and control, robust and nonlinear control. joão paulo lima silva de almeida received the b.sc. degree in industrial automation (2011), the m.sc. degree in electrical engineering (2014) and the ph.d. degree in electrical engineering (2019) from federal university of technology paraná. he has been with the federal institute of paraná, since 2012. his research interests include intelligent systems to model and control dynamic systems. elenilson de vargas fortes received the degree in mathematics in 2004 from the federal university of espírito santo, são mateus, brazil, the m.sc. degree in mathematics in 2007 from the federal university of brasília, brasília, brazil, and the ph.d. degree in electrical engineering in 2016 from the são paulo state university, ilha solteira, brazil. he is currently a professor at the goiás federal institute of education, science, and technology, jataí, brazil. his current research interests include small-signal stability analysis in power systems. preparacion of ful paper for the international conference on renewable energies and power quality transactions on environment and electrical engineering issn 2450-5730 vol 1, no 3 (2016) © igor bolvashenkov and hans-georg herzog degree of fault tolerance as a comprehensive parameter for reliability evaluation of fault tolerant electric traction drives igor bolvashenkov and hans-georg herzog abstract – this paper describes a new approach and methodology of quantitative assessment of the fault tolerance of electric power drive consisting of the multi-phase traction electric motor and multilevel electric inverter. it is suggested to consider such traction drive as a system with several degraded states. as a comprehensive parameter for evaluating of the fault tolerance, it is proposed to use the criterion of degree of the fault tolerance. for the approbation of the proposed method, the authors carried out research and obtained results of its practical application for evaluating the fault tolerance of the power train of an electrical helicopter. keywords: reliability, degree of fault tolerance, multi-phase electric motor, multilevel inverter, traction vehicle drive. i. introduction with the constant growth of complexity of modern engineering systems it becomes more complicated to achieve the required level of sustainable and safety operation. the task of implementing the specified requirements is closely related to the problem of the most accurate assessment of indicators of sustainable operation of the system, shown in fig.1. particularly important is to assess the required reliability and security for the safety-critical systems correctly. in safetycritical applications such as vehicle propulsion systems, the fault tolerance of all the equipment is obligatory. according to the plans for the electrification of various types of vehicles based on the electric energy generated by renewable sources, the tasks of a quantitative estimation of fault tolerance in creating of the safety-critical systems have now become very topical. fig.1. indicators for sustainable operation one of the most important requirements for the vehicle’s propulsion system is the level of fault tolerance. in other words, the vehicle’s propulsion system should operate and continue its sustainable functioning even if one or more of its components have failed. to implement this requirement, all components included in the system should be fault tolerant. as an example of practical use of the proposed method the fault tolerance of a two main parts of the vehicle’s propulsion system traction multi-phase permanent magnet synchronous motor (psm) of electrical helicopter and electric inverter in conventional and multilevel versions were evaluated. in this case, according to the specified requirements of a safe flight of the designed electrical helicopter, the total failure rate for the entire traction drive of designed electrical helicopter should be less than 10-9/h [1]. due to the constant need of traction electrical machines for special application, there are many publications describing the comparison or analysis of different fault tolerant electric machines and electric inverter topologies for different vehicle applications [1]-[7] and [8]-[12]. thus, it should be noted that the authors generally have studied various aspects of the fault tolerance, and in most cases only a qualitative assessment was performed. for example in [1] a qualitative method for the fault tolerance evaluation is proposed, which results are given in table i. table i. comparative evaluation of the fault tolerance [1] phase number parameter 5 phase 6 phase 7 phase 9 phase overload capacity 8 9 9.5 10 partial load mode 7 9 8 10 torque ripple 7 8 10 9 total 22 26 27.5 29 modern methods for determining the degree of fault tolerance of electrical machines, power electronics, and the computer network topology are presented in [13]-[15] and [28]-[31]. the proposed methods have one typical common disadvantage the lack of universality. each of them allows solving a local specific problem for a particular object. the authors have proposed a new universal approach and developed the methodology of assessing the degree of fault tolerance (doft) of safety-critical systems as a whole, as well as the doft of their components. ii. approach and methodology a. degree of fault tolerance considering the definition of the fault tolerance of a technical system as an ability to maintain the required functional level of the system in case of one or more failures of its components, doft can be defined as the amount of time which the system may remain in a degraded state without irreversible changes in its functionality. mathematically, in general form this can be written by (1): (1) where wr and wn – reduced and nominal values of performance of technical system; ti , tn – respectively, duration of functioning after i-number of failures and without failure. as the value of the performance w can be considered the productivity, power, energy, quantity of information, etc. the value of ti is defined by overload capability of the system after i-number of failures. in the case when the required level of wr has been predetermined by project requirements, it is useful to determine doft for each corresponding level of performance in accordance with (2): (2) tn is determined depending on the type of electric vehicles and is largely depending on the specifics of its operation (aircraft, ships, trains, cars). for example, for the electric helicopter 30 minutes are needed for sustainable and safety completion of its function “search and rescue”. ti is determined by the overload capacity of system and its thermal stability. ti indicates the duration time during which the system may operate in a critical failure mode without irreversible changes in the quality and functional inability to further use. based on the considerations above, the procedure for determining the doft is following:  determination of all possible failures on the basis of fault-tree analysis (or failure mode and effects analysis);  classification of all failures on critical dangerous and non-dangerous failures;  definition of opportunities and complexity (if recovery is possible) of the failure elimination;  definition of development and the consequences of each non-repairable failure (if the recovery is not possible);  evaluation of degree of reduction of the system’s functionality;  estimation of an acceptable level of performance reducing in accordance with the requirements;  evaluation of the ratio between the performance and demand for each failure operational mode;  in the cases when the demand is greater than the required performance, the analysis and evaluation of system’s overloading state is carried out, and the value of ti is calculated for the appropriate level of performance reduction;  based on the calculated data the doft can be computed. on the figure 2 and figure 3 the values of doft, respectively of traction multi-phase electric motor and electric power inverter are equal to the area below the degradation curves. the “grey” area above the curve of the degradation of 9-phase motor is equal to probability of the transition of this motor in the next degraded state following the critical failure. the number of "steps" in fig.2 corresponds to the number of degraded states of the electric machine after each critical failure until the moment of time when the motor losses completely their functional ability. for multi-phase electric motor, this corresponds to the next critical phase failure. thus, multi-phase traction motor can be regarded as multi-state-system, which will be given special attention in the section ii.b. fig.2. graphical doft representation of different multi-phase motors as multi-state systems fig.3. graphical doft representation of conventional vs multilevel inverter as multi-state systems quantitative assessment of the total duration of the electric motor operations in all degraded states allows a quantitative comparative analysis of various possible topologies and design of the machine, considering their lifetime. figure 3 is a graphical representation of the qualitative comparison’s results of the doft of conventional and multilevel inverter for 6-phase traction motors. the figure shows that doft of the multilevel inverter is much superior to the conventional version. additionally to the critical failures, the effect of small (noncritical) failures on the fault tolerance value of traction electric motor and electric power inverter, leading to partial or temporary loss of system functionality, can also be investigated using doft, as can be seen in figure 4. fig.4. graphical doft representation of different types of failure a distinctive feature of the proposed methodology of quantitative assessment of technical systems in comparison with existing techniques is its universality, i.e. the possibility of its use for different types of technical systems. as an example of its practical use, figure 5 shows the evaluation of the fault tolerance and transition probabilities of electric helicopter traction drive with multi-phase motors and multilevel inverters. b. mss markov models and transition probabilities considering the above requirements on the probability of total failure of electrical helicopter, as well as statistical data on the reliability of traction electric motor and electric inverter it was determined that the optimal model for an analysis of fault tolerance in such conditions is a mss markov model (mssmm), with k states, as shown in general form in fig.5. theoretical base of this method is well known and described in [16]-[18] and examples of application in [25]-[27]. fig.5. multi-state markov model of electric traction drive [16] the first state corresponds to a completely failure-free operation of the system. the second and other states (before state k) are the states of degradation and correspond to failure cases – phase open-circuit failure, respectively, of the one, two or more phases. the state k of the model corresponds to the completely failed system when the helicopter’s drive completely loses the ability to operate. thus, every following state of the degradation corresponds to one critical failure with a corresponding partial functionality loss of the traction drive. at the same time, the number of the degraded state of the mss-mm determined in accordance with the requirements of the project on the fault tolerance defines the technical capabilities of the system to continue functioning with reduced performance level as a result of the critical failure. the most important and most difficult point for simulations using markov models is to determine the transition probabilities and the number of states with a reduced level of functionality. the values of the transition probabilities , kare derived from the results of calculations of degree of fault tolerance (doft) for the states 1, 2 and k respectively: (3) here r is the value of reduced performance level according to the project requirements and i number of critical failures. the values of transition probabilities are affected by a large number of different factors, from design and environmental parameters to the using types of maintenance strategies, monitoring, and diagnostics. as can be seen from (3), for the calculation of the transition probabilities the main importance is the correct calculation of doft for a given project required performance levels. application cases of proposed methodology for the estimation reliability features of electric traction drive of helicopter will be presented in the next section. iii. fault tolerant electric traction drive a. electric traction drive the power part of the traction drive of electric vehicles includes a source of electrical energy (battery, fuel cell, etc.), an electric energy converter/inverter, and a traction electric motor, as can be seen from figure 6. in the present study, the electric power source was not considered, but will be considered at the next stage of research. (a) (b) fig.6. schematic graphical system definition for a multi-phase traction motor (a) and for the whole traction drive (b) b. critical failures in the paper [6] has been demonstrated that the vast majority of elements failures of the system "power invertertraction electric motor" can be reduced to four basic types of failures of electric traction drive:  open-circuit and short-circuit of the electric motor’s phase;  open-circuit and short-circuit of the inverter submodule. in the development of an electric motor scheme is usually provided such a connection of the each inverter submodule with the protection system, which disconnects the electrical circuit of the failed submodule. this solution allows the failure "short-circuit" of the inverter’s submodule to lead to the failure "open-circuit" of submodule. the failure type "phase short-circuit" of electric motor can be reduced to the failure of "phase open-circuit" based on special design options of the stator winding performance, improve the quality of insulation materials and advanced production technology. excluding the possibility of failure type "short-circuit" on the basis of their reducing to the failures of the type "opencircuit" is one way to keep a functionality of multi-phase electric motors in failure cases. thus, in the simulation of functioning of electrical traction drive in failure conditions, as the critical failures will be considered the motor’s "phase opencircuit" and inverter’s “submodule open-circuit". c. multi-phase electric motor the results of a previous study by the authors [1] indicate that for the 3-phase psm the specified requirements on the reliability for the entire power drive of a designed helicopter is not achievable without a functional and/or structural redundancy, and/or other activities that improve the indicators of reliability to the required value. one way to create the fault tolerant traction electric motors of high reliability is to increase the number of motor phases without changing the value of the motor’s power. this allows to reduce the current value in each phase’s open winding and to perform the power electronics unit integrally fabricated. based on the diversity of various schemes, connection methods of power supplies and inverters as well as switching algorithms of the stator phases of electric motor, it is possible to implement the system using automatically changed structures and parameters, resp., depending on the purpose and functioning conditions. at the same time, independent performance of each phase’s switching channels provides increased reliability of the electric machine based on the principle of functional redundancy. unlike an electrical machine with a small number of phases in which the majority of critical elements failures leads to a complete failure of the machine, the multi-phase electrical machine remains in operation up to a certain level of degradation and a corresponding change in the output characteristics. this allows realizing so called functional redundancy. thus, on the one hand, increasing the number of phases of the electric motor reduces the impact of each failure in power electronics control channel or in the phase of the motor, on the characteristics of the whole traction drive. on the other hand, increasing the number of phases leads to an increase of the failure probability in one of phases. in this context, it is necessary to find an optimal compromise solution, based on the design requirements, the possibility of redundancy, and the reliability indices of the electrical machine and power electronics. on the basis of the known values for the failure probability of each phase of the electric motor the optimal number of phases can be calculated, at which the required reliability and fault tolerance indicators of electric motor can be implemented, taking into account the possibility of one or more critical failures. the use of this redundancy method has its own features that must be considered in the study of physical processes and the design of the traction electric drive as a whole. as shown in [1] the optimal electric machine for the safety-critical electric drives, considering system approach techniques, is a multi-phase psm with distributed stator windings and internal v-shaped arrangement of permanent magnets on the rotor. for a detailed study 5-, 6-, 7and 9phase psm were selected. for the traction electric motor of the helicopter considering high requirements on the drive safety and fault tolerance, the overload capability in the fail operational modes is especially important. in such operating conditions, a stable operation for a specified time on the modes of reduced power without critical asymmetry of psm’s parameters is also extremely important. considering a normal, i.e. failure-free, operational mode, the electric motor can endure a short-term overload because the thermal capacity is sufficiently large, whereas for failure cases the situation is changing dramatically. as known from operating experience, the largest numbers of operational failures are caused by technological overloads [19] [21]. most of the total failures of traction electric motors (more than 80%) occur because of stator windings faults and bearings failures, so that these components play a crucial role in the overall reliability value of the motor. their lifetime and fault tolerance significantly affect the operating temperature, developed inside the motor. the overheating causes quickly deterioration of the motor winding insulation and bearing. the causes for overheating of electrical machines can be technological overload of the motor as well as the occurrence of different failure modes. the main of them are the various types of short circuits, unbalanced work at the loss of one or several phases, jamming of the rotor of electric machine. technological overload leads to an increase in temperature of the motor windings to a higher level, a gradual deterioration and finally to its total failure. in the case of a short circuit in the stator windings, the current value exceeds the nominal value to a multiple. thus, the rate of rise of the stator winding temperature reaches 7-10°c per second and after 10-15 seconds the motor temperature is out of limits. the unbalanced mode (phase loss) occurs in the case of burned out fuse in the phase, wire breakage, faults in contacts or in the case of protection operation. thus there is a redistribution of the currents, the magnitude of current in the remaining phase’s increases, which leads to overheating and total failure of the electrical machine. according to recent research [20], long-term operation of the electric motor with an overcurrent by only 5% of the nominal reduces its lifetime by 10 times. most of the overcurrent failures of electric motors are associated primarily with damages inside the motor, leading to an asymmetric overcurrent and, as a consequence, to overheating. the main types of failures that lead to dangerous overheating of the stator windings and to a total failure of the electric motor (without system of protection) are the short circuit faults:  between turns;  between coils;  between phases;  between wires and the housing of the motor. their effects are described in detail in the literature. these effects lead to dramatic increasing of the current in one or more phases of the motor and ultimately to the motor’s overheating. at an effective system of protection against emergency situations, each of the above-discussed failures can be reduced to an embodiment of the loss of failed phases (or automatic shutdown). when it is required to maintain the load at a given level, which is a common requirement for safety-critical systems, such as a helicopter traction drive, it is necessary to increase the phase currents in the remaining phases of the electric motor. this will result in a certain level of the motor’s overheating and severely limited, in terms of reliability, duration of operation in this mode of load. for the traction drives of the electric helicopter the given levels of maintain load in the failed operation are: 65% of the nominal load (long time operation) and 85% of the nominal load (short time operation). (a) (b) fig.7. overheating at the phase loss, (a) 65%, (b) 85% of nominal load to estimate the level of overheating of the windings of multi-phase traction motors and respective conclusions on its thermal stability in the case of failure in one or two phases, it is proposed to use the overheating factor kt, which shows how many times the motor windings temperature exceeds the nominal value: (3) where ti and tn –, respectively, describe the windings temperature in failed operational mode in i-phases and in the nominal failure-free operational mode. overheating factor is graphically presented in fig.7. table ii shows the results of a preliminary assessment of the overload capability of multi-phase traction motors in case of an open-circuit failure of one or two phases. table ii. comparison of overload capability table ii has the following designations: ++ there is no overload; + overload is less than 15%; 0 overload is greater than 15%. the main goal of the preliminary analysis of the presence (or absence) of overload modes of a traction electric motor at various critical failures, is to find the critical points in terms of thermal stability and overload capacity. conditional separation of considering multi-phase motors on the level of overload capability into three groups was carried out based on the analysis of the thermal stability of the motor windings for different thermal load conditions. from the operating experience of electrical machines with appropriate power [20] it is known that when the phase current has insignificant excess (10-15% of the nominal value), the electric motor can be operated a certain time without critical deterioration of insulation of the stator windings. therefore, such failure operational overloads modes can be considered as partially acceptable. considering the low rates of overload capability and thermal stability of the 5-phase traction motor, as can be seen from fig.7, this version was excluded from further analysis. when current overloads are larger than the above mentioned values, this operational mode has been regarded as critical. the consequences of the large overload in failure operational mode are overcurrent and overheating of psm, which lead to a reduction of reliability indices and decrease lifetime of the motor, as can be seen in fig.8. fig.8. lifetime of the parts of electric drive [22] the main characteristic of the load modes of psm for evaluation of doft is the thermal characteristics, estimated by formulas [20]: tn = {ln k 2 – ln (k2 – 1)}/(a/c) (4) where tn – describes the time of achievement of acceptable motor temperature value, k – the rate of exceeding the nominal value of the phase current, a – the heat irradiation of the motor, and c – the thermal capacity of the electrical machine. figure 9 shows graphically the results of calculating the thermal behavior of the electric motor in overload conditions, as a result of one or two critical failures considering thermal stability of the stator windings. the values of ndeg and nnom in fig.9 correspond to the values of traction drive performance (driving power) in degraded and nominal modes, respectively. (a) (b) fig.9. the duration of the safe motor operation at 113% load at a one (a) and two (b) critical failures on the basis of the analysis of the thermal behavior of the motor the values ti for the different versions of the traction motor and the various failure modes can be determined. based on the value ti, the transition probabilities of markov models were calculated. schedule of doft at 65%, 85% and 113% of nominal load, on the number of critical dangerous failures is shown in fig.10, a, b, c, respectively. (a) (b) (c) fig.10. doft of multi-phase psm in fail operations, (a) 65%, (b) 85% and (c) – 113% of nominal load based on the constructed graphs a comparative analysis of doft for considered variants of the electric motors can be carried out quite easily. however, according to the authors, more informative and convenient for practical use are the dependencies of doft on the value of a given load maintenance level, as shown in fig.11. thus, it is possible to assess the compliance of parameters of fault tolerance for each compared alternative to design requirements. (a) (b) fig.11. doft of a given load level in fail operations, (a) – one failed phase, (b) – two failed phases d. electric inverter considering the current developments in the field of power electronics, for the comparative analysis of the fault tolerance has been considered two options of electrical power inverter: conventional and multilevel. voltage source inverter (vsi) is the most commonly used power converter between the seriesparallel-configured batteries and the electric motor in electrical mobility applications. it is well known that multilevel inverters offer several advantages compared to their two-level counterparts [12]: smaller power filters, smaller voltage ratings for semiconductors, lower switching frequencies and less power losses. they offer also more modularity and are more reliable. currently, mainly three types of electric multilevel inverters are used: neutral point clamping (npc), flying capacitors (fc), and cascaded h-bridge (chb). considering [8], [10], [11], and [12], chb inverters need the lowest number of components. npc and fc inverters need more components than chb and have less modularity, due to the central storage unit. thus, chb inverters are constructed on a series connection of single-phase inverters supplied by isolated dc electric energy sources. this kind of inverters gives a high modularity degree and consequently high reliability and fault tolerance. an approach based on the full inverter’s power control could optimize the implementation and the reliability of the inverter while offering optimized operational behavior. it should also be noted that chb is an interesting inverter’s topology for the electrical vehicle because of its possibility to work with unbalanced or faulty dc sources. this in turn can increase the life time of battery packs and the autonomy of the vehicle. in the case of using chb inverters with appropriate control strategies, state of charge unbalances can be managed as well as state of health allowing to optimize the access to the electrical energy and electric vehicle autonomy. based on the reqired design parameters of the traction drive of electric helicopter, the preferred inverter option is a topology 17-level chb inverter. so, in each phase there are 8 submodules, as shown in fig.12. figure 12 shows the basic topology of one phase of a chb using batteries on the dc side. each battery module is connected to a h-bridge with 4 mosfets. the use of mosfets enhances the efficiency of the chb inverter, because of the low conduction losses. fig.12. one submodule of the proposed 17-level chb inverter for the fault tolerant operation in failure case either the inverter module must be shut down or switched to a possibly existing redundant inverter leg/module. each of these modes leads to a rapid increase of the current at the emergency site and an overload of the semiconductors. at overload, power losses occur in the p-n-transition and its temperature increases dramatically, due to the low heat capacity. in the case of exceeding a certain critical temperature of p-n-transition, the semiconductor device fails. hence, overheating temperature is the main parameter characterizing the overload capability of semiconductor devices, as it is shown in fig.8. the graph presented in fig.13 has been plotted based on the data regarding the standard overload capability of the power electronics. fig.13. overload capacity of inverter based on the analysis of the thermal modes of inverters in the failure cases, critical operational point considering the thermal stability of semiconductor devices have been identified as shown in table iii and table iv. table iii. overload of conventional inverter in failure modes table iv. overload of multilevel inverter in failure modes table iii and iv have the following designations: + + + overload is less than 10 %; ++ overload is less than 25 %; + overload is less than 40%; mf motor failure; phf phase failure. qualitative analysis of the data presented in the tables shows that the 17-level inverter significantly exceeds the level of fault tolerance of the inverter with a conventional topology. iv. simulation results on the markov model for more accurate assessment of fault tolerance of each comparable traction motor and electric inverter, and for compliance (or non-compliance) with design requirements, mss markov models of reliability were built, which theoretical base is described in [16] and [17]. to construct such models the multi-phase psm as well as the multilevel inverter can be considered as a system with a loaded functional redundancy and consequently, with an appropriate reserve of fault tolerance. the transition probabilities for mss markov models were calculated using the above mentioned doft method. as was discussed above, the phase open-circuit failure has been considered as the main dangerous failure for electric motor’s stator, i.e. it is the most severe kind of failure, to which less dangerous failures can be summarized. for the electric inverter, the mosfet’s submodule open-circuit failure was considered to be the worst case. considering the above requirements on the fault tolerance of safety-critical drives as well as statistical data on the reliability of the multi-phase psm and electric power inverter from [1], [23], [24], and [27], it was determined that the optimal model structure for the analysis of fault tolerance in such conditions is a mss markov models with four states for traction motor and five states for electric power inverter, respectively, as shown in fig.14 and fig.15. the number of states of the mss markov model is determined by the probability of failure-free operation of the system in the failure-free mode and fault tolerance requirements established by the design documentation. fig.14. multi-state markov model of traction multi-phase psm fig.15. multi-state markov model of multilevel inverter the first state corresponds to a completely failure-free operation of the motor and inverter. the second and third states of the motor and the second, third, and fourth of the inverter are the states of degradation and correspond to failure cases. for the electric motor this means phase opencircuit failure of one and two phases, respectively. the fourth state of the motor’s model (fig.14) corresponds to the completely failed electric motor when the helicopter’s drive completely loses the ability to operate. thus, every following state of the degradation corresponds to one emergency phase loss with a corresponding loss of the part of functionality of the motor. the fifth state of the inverter’s model (fig.15) corresponds to the completely failed conventional inverter when the traction motor does not receive the necessary power for a safe flight. thus, every following state of the degradation corresponds to the worst case – a loss of an inverter’s submodule. for the multilevel chb inverter the fifth state of the model means the loss of one complete phase of the electric motor. the proposed assessment approach for transition probabilities in the mss markov reliability model is well formalized and suitable for practical application in reliability engineering to assess fault tolerance indices of multi-phase traction motors considering the aging process under the influence of operating conditions. regarding the design requirements on fault tolerance of an electric helicopter, the reliability of the traction electric motor at reduced power of 65% or 85% of the nominal value as well as at 113% of nominal value was analyzed using the mss mm. corresponding graph for 6-, 7and 9-phase psm at the 113% load level are presented in fig.16. the horizontal axis indicates operational time in hours and vertical axis probability of total failure of traction motor. fig.16. probability function of total failure of psm at the 113% load table iii. probabilities of complete failure of psm the simulation results of two consecutive critically dangerous failures allow of quantifying the reliability indices degree of the fault tolerant multi-phase electric motor, which is one important part in the traction drive of electric helicopters, and of estimating the compliance of calculated values with design requirements. the obtained results confirm the results of the studies presented in [5] and [6], that 7-phase electric traction motors can be operated after the loss of two phases during a limited time, at nominal load (limitation because of thermal stability), and for a long time with a reduced load. for the simulation the worst case and critically dangerous variant of failure has been considered, i.e. the submodule failures occur in the same phase. in case of a simulation of a not safety-critical failure, as well as the possibility of partial recovery of the power drive’s operating ability in the degraded state, the value of the fault tolerance will be significantly higher. regarding the design requirements on fault tolerance, the reliability of the inverter was analyzed using the above mss markov models, in case of emergency reducing the power to 113% of the nominal value. the corresponding graphs for a different number of phases and inverter topologies are presented in fig.17 and fig. 18, respectively. fig.17. probability function of a total motor failure with conventional inverter at the 113% load fig.18. probability function of a total failure of one motor phase with multilevel inverter at the 113% load the results of the probability calculations of a failurefree operation of the electric inverter during one operational hour at 113% of nominal load for two inverter’s topologies are summarized in table iv. table iv. probabilities of complete failure of inverter the results of simulation of three consecutive critically dangerous failures allow quantifying the degree of fault tolerance of a 17-level inverter, which is one of the important parts of the traction drive of helicopters. the 7 and 9-phase options have shown the maximum compliance with the requirements relating to the safety-critical drives. v. conclusion the paper presents a new approach and methodology for assessing the degree of fault tolerance of a safety-critical technical system, such as a vehicle’s electric traction drive. the suggested assessment of the reliability of fault tolerant topologies of electric traction drives is well formalized and suitable for practical application in reliability engineering to assess fault tolerance indices of multi-phase traction motors as well as an electric power inverter, considering the aging process under the influence of operating conditions. as an example of practical use of the proposed method the fault tolerance of two important parts of a vehicle’s electric propulsion system traction motor and electric inverter were evaluated. results of comparative analysis allow to conclude that for given project requirements on the level of reliability and fault tolerance of helicopter’s electric traction drive in the real flying conditions only 9-phase motors in combination with 17-level inverters completely fulfill the design requirements without any restriction. more broadly, the proposed method can be used as a universal tool for evaluation and optimization of the degree of fault tolerance in safety-critical technical systems, considering all the possibilities of its increasing, such as redundancy, monitoring, predictive control and type of maintenance strategy. references [1] i. bolvashenkov, j. kammermann, s. willerich and h.-g. herzog, “comparative study of reliability and fault tolerance of multi-phase permanent magnet synchronous motors for safety-critical drive trains”, in proceedings of the international conference on renewable energies and power quality (icrepq’16), 4 th to 6th may, madrid, spain, 2016, pp.1-6. [2] e. levi, “multiphase electric machines for variable-speed applications”, ieee transactions on industrial electronics, vol.55, no 5, 2008, pp.1893-1909. [3] m. villani, m. tursini, g. fabri and l. castellini, “multi-phase permanent magnet motor drives for fault-tolerant applications”, in proc. of ieee international electric machines & drives conference (iemdc), 5-18 may 2011, niagara falls, canada, 2011, pp.1351-1356. [4] f. scuiller, j.-f. charpentier and e. semail, “multi-star multi-phase winding for a high power naval propulsion machine with low ripple torques and high fault tolerant ability”, in proc. of the ieee vehicle power and propulsion conference (vppc), 1-3 sept. 2010, lille, france, 2010, pp.1-5. [5] e. semail, x. kestelyn and f. locment, “fault tolerant multiphase electrical drives: the impact of design”, european physical journal applied physics (epjap), vol.43, iss.2, 2008, pp.159-162. [6] p.g. vigrianov, “assessment the impact of different failures on the power characteristics of the low power 7-phase permanent magnet synchronous motors”, moscow, journal “questions to electromechanics”, moscow, vol.128, iss.3, 2012, pp.3-7. (in russian) [7] d. fodorean, m. ruba, l. szabo and a. miraoui, “comparison of the main types of fault-tolerant electrical drives used in vehicle applications”, in proc. of international symposium on power electronics, electrical drives, automation and motion, (speedam), june 11-13, ischia, italy, 2008, pp. 895-900. [8] o. josefsson, t. thiringer, s. lundmark and h. zelaya, “evaluation and comparison of a two-level and a multilevel inverter for an ev using a modulized battery topology”, in proc.of ieee 38th annual conference on industrial electronics society (iecon), oct. 25-28, montreal, canada, 2012, pp. 2949-2956. [9] a. v. brazhnikov and i. r. belozyorov,”prospects for use of multiphase phase-pole-controlled ac inverter drives in traction systems”, european journal of natural history, russia, vol.2, 2011, pp.47-49. [10] b. sarrazin, n. rouger, j. p. ferrieux and j. c. crebier, “cascaded inverters for electric vehicles: towards a better management of traction chain from the battery to the motor?”, in proc. of ieee international symposium on industrial electronics, june 27-30, gdansk, poland, 2011, pp. 153-158. [11] s. fazel, s. bernet, d. krug and k. jalili, “design and comparison of 4-kv neutral-point-clamped, flying-capacitor, and series-connected h-bridge multilevel converters”, ieee transactions on industry applications, vol.43, no.4, jul.-aug. 2007, pp. 1032-1040. [12] m. malinowski, k. gopakumar, j. rodriguez and m. perez, “a survey on cascaded multilevel inverters”, ieee transaction on industrial electronics, vol. 57, no. 7, july 2010, pp. 2197-2206. [13] b. a. welchko, t. a. lipo, t. m. jahns and s. e. schulz, “fault tolerant three-phase ac motor drive topologies: a comparison of features, cost, and limitations”, in: ieee transactions on power electronics, vol.19, no 4, 2004, pp.1108-1116. [14] u. de pra, d. baert and h. kuyken, “analysis of the degree of reliability of a redundant modular inverter structure”, in proc. of ieee 12th international telecommunications energy conference, 0408 oct. 1998, san francisco, ca, 1998, pp.543548. [15] s. krivoi, m. hajder, p. dymora and m. mazurek, “the matrix method of determining the fault tolerance degree of a computer network topology”, sofia, bulgaria, publisher: ithea, vol.13, no 3, 2006, pp.221-227. [16] a. lisnianski, i. frenkel and y. ding, “multi-state system reliability analysis and optimization for engineers and industrial managers”, berlin, new york, springer, 2010, 393 p. [17] b. natvig, “multi-state systems reliability theory with applications”, john wiley & sons, new york, 2011, 232 p. [18] i. bolvashenkov and h.-g. herzog, “use of stochastic models for operational efficiency analysis of multi power source traction drives”, in proc. of ieee of international symposium on stochastic models in reliability engineering, life science and operations management (smrlo’16), 15-18 feb. 2016, beer sheva, israel, 2016, pp.124-130. [19] d. hann,“a combined electromagnetic and thermal optimisation of an aerospace electric motor”, int. conference on electrical machines, icem, 6-8 sept. 2010, rome, italy, 2010, pp.1-6. [20] m. m. katzman, “electrical machines“, akademia, moscow, russia, 2001, 463 p. (in russian) [21] s. mahdavi, t. herold and k. hameyer, “thermal modeling as a tool to determine the overload capability of electrical machines”, international conference on electrical machines and systems (icems), 26-29 oct. 2013, busan, korea, 2013, pp.454–458. [22] i. bolvashenkov and h.-g. herzog, “approach to predictive evaluation of the reliability of electric drive train based on a stochastic model”, in proc. of ieee 5th international conference on clean electric power (iccep’15), 16-18 june 2015, taormina, italy, 2015, pp.1-7. [23] e. lauger, “reliability in electrical and electronic components and systems”, north holland publ. co., amsterdam, 1982, 1171p. [24] n. p. ermolin and i. p. zerichin, “zuverlässigkeit elektrischer maschinen“, berlin, verlag technik, 1981, 227 p. (in german) [25] a. h. ranjbar, m. kiani and b. fahimi, “dynamic markov model for reliability evaluation of power electronic systems”, in proc. of ieee international conference on power engineering, energy and electrical drives (powereng), malaga, spain, may 2011, pp.1-6. [26] m. molaei, h. oraee and m. fotuhi-firuzabad, “markov model of drive-motor systems for reliability calculation”, in proc. of ieee international symposium on industrial electronics, 9-13 july 2006, montreal, canada, pp.2286-2291. [27] t. geyer and s. schroder, “reliability considerations and faulthandling strategies for multi-mw modular drive systems”, in: ieee transactions on industry applications, vol.46, no.6, nov.dec. 2010, pp. 2442-2451. [28] k. s. trivedi, “probability and statistics with reliability, queuing, and computer science applications”, second edition, wiley, 2002, 848 p. [29] s. j. bavuso, j. b. dugan, k. s. trivedi, e. m. rothmann and w. e. smith, “analysis of typical fault-tolerant architectures using harp”, in: ieee transactions on reliability, vol.r-36, iss.2, june 1987, pp.176-185. [30] n. muellner and o. thee, “the degree of masking fault tolerance vs. temporal redundancy”, in: ieee workshops of international conference on advanced information networking and applications (waina), 22-25 march 2011, biopolis, singapore, 2011, pp.21-28. [31] r. ubar, s. devadze, m. jenihhin, j. raik, g. jervan and p. ellervee, “hierarchical calculation of malicious faults for evaluating the fault-tolerance”, in proc. of 4th ieee international symposium on electronic design, test & applications (delta), 23-25 jan. 2008, hong kong, 2008, pp.222-227. igor bolvashenkov, dr.-eng., senior researcher, institute of energy conversion technology, technical university of munich (tum), munich, germany. igor bolvashenkov obtained his diploma (1981) and doctoral degree (1989) in electrical engineering from admiral makarov state university of maritime and inland shipping, leningrad, ussr. from 1987 till 1993 he worked as associate professor at the murmansk state technical university, russia. since 2004 he works at the institute of energy conversion technology of technical university of munich (tum), germany. he has specialized in development and simulation of electric propulsion safety-critical systems for ships, cars, aircraft with analysis of their reliability, survivability, and fault tolerance. he published more than 75 scientific articles, book chapters and patents in the fields of development of traction drives for the various type of electric vehicle and system analysis of efficiency, reliability and fault tolerance. hans-georg herzog, prof. dr.-eng., institute of energy conversion technology, technical university of munich (tum), munich, germany. hans-georg herzog holds a diploma and a doctoral degree (with distinction) from technical university of munich (tum). after his time as a research associate, he joined robert bosch gmbh, leinfeldenechterdingen, germany. since 2002, he is head of the institute of energy conversion technology at tum. his main research interests are energy efficiency of hybridelectric, full-electric vehicles and electric aircrafts, reliability of drive trains and their components, energy and power management and advanced design methods for electrical machines hans-georg herzog is senior member of ieee and member of vdi and vde.  transactions on environment and electrical engineering issn 2450-5730 vol 1, no 3 (2016) © oluwafemi e. oni, kamati i. mbangula, and innocent e. davidson  abstract—high voltage direct current (hvdc) systems has been an alternative method of transmitting electric power from one location to another with some inherent advantages over ac transmission systems. the efficiency and rated power carrying capacity of direct current transmission lines highly depends on the converter used in transforming the current from one form to another (ac to dc and vice versa). a well-configured converter reduces harmonics, increases power transfer capabilities, and reliability in that it offers high tolerance to fault along the line. different hvdc converter topologies have been proposed, built and utilised all over the world. the two dominant types are the line commutated converter lcc and the voltage source converter vsc. this review paper evaluates these two types of converters, their operational characteristics, power rating capability, control capability and losses. the balance of the paper addresses their applications, advantages, limitations and latest developments with these technologies. index terms—commutation failure, hvdc, line commutated converter, modular multilevel converter, voltage source converter. i. introduction he first electric generator was dc machine, as well as the first electric power transmission system by thomas edison [1, 2]. in the last few decades, high voltage direct current (hvdc) technology has been used, due to some of its inherent benefits in long distance transmission application. it is widely used all over the world for bulk power delivery over long distances, interconnections of asynchronous systems, stability of ac lines, power control, long submarine transmission and renewable energy integration. reduction in the right of way (row) is another edge over ac systems [3]. hvdc transmission system involve the use of converter for the conversion of ac to dc (rectifier) at the transmitting end, and converting the dc back to ac at the receiving end (inverter), [4]. this converter usually has a 12-pulse arrangement, of valves connected in a star-delta, star-star formation to the ac networks. a reactor, dc capacitor and ac filters are also part of the converter circuitry. the two ends of this paper was submitted for review on july 28, 2016 and accepted on september, 9, 2016. this work was supported by eskom power plant engineering institute, eskom centre of excellence in university of kwazulunatal. westville campus, south africa. o. e oni is with electrical engineering department, university of kwazulu-natal. durban 4041, south africa (e-mail: maxiphem@yahoo.com). k. n. i. mbangula was with department electrical engineering, university of kwazulu-natal. durban 4041, south africa. he is now with the department the converters are connected via dc transmission lines which can be either overhead cable or submarine cable or directly in the same location as in the case of back-to-back configuration. continuous progress in hvdc systems is linked to advances in the power electronics technologies for the fabrication of highly efficient semiconductor devices for hvdc converter topology [5]. there are two dominant methods used in converting ac to dc and vice versa. these methods are the line commutated converter lcc and the voltage source converter vsc. the success of these two technologies became possible with the development of power electronics devices [6, 7]. before the power electronics was the transverter, electrolytic and the atmospheric converter, all these are part of the several attempts made for ac/dc conversion. these entire attempts failed due to some technical reasons and safety measures inherent in using them [8]. the invention of mercury-arc valves brought temporary success to ac/dc conversion which later became outdated. the mercury arc valve which operated then have either been scrapped or upgraded to semiconductor converter technology [9]. semiconductors devices have been in used since 1970s and are still a growing technology because of the high switching capacity and ability to withstand high current rating. examples are the diode, diac, triac, thyristors, mos-controlled thyristors (mcts) [10], insulated-gate bipolar transistors (igbt) and integrated gate-commutated thyristors (igcts) etc. [11]. this paper looks critically into the two dominant hvdc converter technologies taking into consideration their operational characteristic and their output ac waveform when subjected to three-phase short circuit as well as dc line fault. the simulation is carried out on digsilent powerfactory and the results of each technology are compared alongside each other. ii. converter configuration and topology hvdc interconnections can be configured in different forms to suit different desired performance and operational of electrical engineering, university of namibia. ongwediva 3624, namibia (e-mail: imbangula@unam.na). i. e. davidson was with department of electrical engineering, university of kwazulu-natal. durban 4041, south africa. he is now with the department of electrical power engineering, durban university of technology. durban 4001, south africa (e-mail: innocentd@dut.ac.za). a review of lcc-hvdc and vsc-hvdc technologies and applications oluwafemi e. oni, kamati i. mbangula, and innocent e. davidson t requirements, namely: 1) back to back connection this has both the inverter and the rectifier in the same location, and the valves are normally in the same building. it therefore has a short dc line of few meters located inside the same environment. 2) monopolar connection this has both converters separated by a single dc pole line, either positive or negative voltage. the ground is used as a current return path. most submarine cable connections use monopolar systems. 3) homopolar connection this has two or more dc line of the same polarity connected to the converters. negate polarity is normally used for less corona and reactive power loss. ground is used as the return path. it works as a monopole when one pole develop a fault. the disadvantage of high cost make it unpopular and seldom used. 4) bipolar connection this is the most popular method in hvdc interconnection of converters. it is similar to the homopolar connection, but it has different polarities. each pole is independent, that is, it can operate with a single pole with ground used as return path [3]. 5) multi-terminal connection this has more than two sets of converters operating independently. each converters can operate as a rectifier or an inverter [12]. b. components of a converter station hvdc converter stations comprises of different interconnected system working together for efficient power transmission. predominantly, any hvdc converter station comprises of the converter circuits itself, then the converter transformer, smoothening reactors, harmonics filters, and other peripherals devices. few definitions of the most important parts are explained below. 1) converter transformer lcc hvdc uses special type of transformer different from the ac transformer in that it has special features such as on load tap changes and follow different configuration. for example, the 12-pulse converter can follow six single-phase two windings, three single-phase three winding or two threephase two windings configuration to suit specification and operational performance [13, 14]. but the vsc hvdc uses same transformer as the normal ac transformer. 2) smoothing reactors this is used for removal of ripples of the dc current. it is also used to limit the rate of rise of the fault current on the dc line. 3) harmonic filters these are connected to the converter terminals to provide a low impedance path to ground for removal of harmonics current. filter used also provide the ac line with the reactive power compensation. iii. line commutated converter (lcc-hvdc) lcc, also known as a current source converter csc uses a thyristors base technology for its converter. the thyristors are silicon semiconductor devices with four layers of n and p type material acting as bi-stable switches, triggered on with a gate pulse and stayed in that on condition until the next current zero crossing. in other for lcc to commutate, the converters require a very high synchronous voltage source, thereby hindering it use for a black start operation. with lcc current rating reaching up to 6250a and blocking voltage of 10kv, this make lcc to have the highest voltage and power rating level of all the hvdc converter technologies [15-17]. lcc achieves its control by regulating the firing angle ᾱ on both rectifier and inverting side. it has an approach that utilizes a uni-directional line commutated flow of dc current which is inject into a receiving ac network, thereby termed csc because the output current is kept at a constant level [18]. table i recent lcc –hvdc projects project name location characteristic (mw) (kv) year (km) uk netherlands 1000 ±400 2011 260 jinpin – sunan china 7200 ±800 2012 2093 mundra – haryana india 2500 ±500 2012 960 rio – madeira brazil 800 100 2012 b-b rio – madeira brazil 2x3150 ±600 2013 2375 xiluodu – guangdong china 6400 ±500 2013 1251 nuozhadu – guangdong china 5000 ±800 2013 1451 southern hami – zhengzhou china 8000 ±800 2014 2200 biswanath – agra india 6000 ±800 2014 1728 xiluodu zhejiang china 8000 ±800 2014 1688 zhundong – sichuan* china 10000 ±1100 2015 2600 power reversal from one station to another is carried out by inverting the dc voltage polarity in both stations but the current direction remains constant. the technology operates with good reliability and minimal maintenance. it is the most suitable way of transmitting bulk power using high voltage transmission lines. these features make lcc technology the most popular among hvdc schemes [19]. table 1 shows few of the recent lcc-based hvdc around the globe. zhundong-sichuan scheme has the highest voltage and power, and the longest distance, project in china [20]. iv. voltage source converter (vsc-hvdc) voltage source converter uses insulated gate bipolar transistor igbt technology. the current in this technology can both be switched on and off at any time independent of the ac voltage, that is, it creates its own ac voltages in case of blackstart [21]. its converters operate at a high frequency with pulse width modulation pwm which allows simultaneous adjustment of the amplitude and phase angle of converter while keeping the voltage constant [22]. vsc has high degree of flexibility with inbuilt capability to control both its active and reactive power as shown in fig. 1, which makes it more useful in urban power network area [23]. this technology was develop in the 1990’s with the first project commissioned by abb, 1997 [9]. but due to its capacity limits, vsc-hvdc has not been able to make much edge over its contemporary lcc scheme due to low device rating, high power losses and high dielectric stress on equipment insulation. its application is approaching 1800mw, 500kv. an example is the 1400mw, ±525kv nordlink that interconnect the grid of statnett in norway and tennet in germany over a distance of 623km [24]. a lot of research is ongoing to override this limitation [25] and to have the ability to ride through fault [26]. ref. [27] explain vsc control, modelling, simulation and stability analysis in power systems. the basic building block of vsc-hvdc topology start with two-level converter [28, 29]. it is like a six-pulse bridge in which igbt with inverse-parallel diodes replaced the thyristors, and the dc smoothing reactor of lcc is replaced by dc capacitor as shown in fig. 2. it derives its name from the fact that it has a switching device which are complementarily operated to generate two levels of voltage (+vdc/2 and –vdc/2) at the ac output terminal of the converter. this complementary operation only allows one switching devices to operate at a time, and the other is turned off. simultaneous turning on of both two switching devices will lead to short circuit of the capacitor across the dc link which may destroy the converter switches due to over-current. with this topology, each semiconductors switch withstands the full voltage stress that is flowing in the link [30]. prevention of the dc voltage from changing polarity is done by the diode that is connected in parallel to the igbt, since the diode can only conduct when forward biased, thereby discharging the dc circuits. but the current flows in both direction, passing through either the igbt or the diode [31]. it adopt the pulse width modulation (pwm) techniques to control the gate switching frequency of the igbt, and to reduce the harmonic distortion generated by the converter. due to high switching losses in the igbts as a result of the pwm which is switched on and off many times in a cycles, the overall transmission efficiency of a two-level converter is very poor compared to the lcc converter. another major setback is that a high level of electromagnetic interference occur when twolevel converter is used for high voltage dc systems [32]. an attempt to reduce the poor harmonic distortion and to have a high efficient vsc converter brings about the multi-level converter (mmc, which start from the three-level converter with three discrete voltage level). it synthesizes more than two voltage level at the ac terminal of each phase as shown in fig. 3. several types of multilevel converter have been mention and analyzed in the literature [3335], such as the diode clamped, where diodes are used as clamped and the dc output is subdivided into switches by a capacitors. fig. 1. vsc-hvdc scheme design fig. 2. block diagram of a two-level vsc-hvdc with n-levels, there will be n+1 capacitors, and n-1 switch pair are required to work in a complementary manner to generate the output dc voltage. high efficiency for switching at fundamental frequency, low cost and a lesser number of components are some of its merits. however it suffer setback as its less attractive for high voltage transmission due to difficulty in charging and discharging of its dc capacitor, lack of modular index and large inductance stray in the clamping path which have effect on the converter switching characteristics [33, 34]. flying capacitors multilevel converter is another type which made use of a pre-charged capacitor. unlike the diode clamped, two or more switch can synthesize an output voltage at the ac output terminal of the converter, and has a phase redundancies which allows specific choice of capacitor to be charge or discharged for voltage balancing across different levels. it has the ability to control real and reactive power flow, and to ride through fault and voltage sag because of its large number of capacitors [35]. nevertheless, as the level increases, so does the size of the capacitors, as it becomes bulky. also, the control to track the voltage for all the capacitor becomes complicated, as it requires high frequency switches. the single-phase full bridge is the building block for the cascaded h-bridge multilevel link. it has four switches connected to an isolated capacitor (separate dc source). each h-link generate three voltage levels. easy modularized layout package for the series h-bridge, makes it cheap and quickly to fabricate. it also have more possible output voltage-levels than the dc source. good for reactive power compensation. with good voltage balancing capability through adaptive control action. however, cascaded h-bridge conversion is not suitable for hvdc application because it h-bridge requires the use of many isolated dc sources in series [36]. recently, a new alternative of vsc-hvdc circuits was proposed in 2003, at the university of bundeswehr in munich, germany, by prof. rainer marquardt [37, 38]. these converter topologies is based on series-connection of several sub-modules of two semiconductor switches and a capacitor. this topology is known as modular multilevel converter (mmc or m2c) as shown in the (fig 3). the converter can either adopt the half bridge cascaded or full-bridge connections for the arrangement of each sub modules. the half-bridge modular multilevel (hbmmc) addresses some of the limitation encountered in the convectional vsc converter. namely; the reduction in the magnitude of the transient dc fault current, converter scalable to the highest transmission voltage through addition of more levels, great reduction in the harmonic content and elimination of low-order harmonics which usually requires large filters, and losses reduced to approximately 1% per converter. all these features made hb-mmc to be widely adopted in recent years. but the hb-mmc freewheeling diodes is unable to stop ac grid contribution to the dc fault current which makes it in need of fast acting dc circuit breaker, else the excessive current stresses may damage the freewheeling diode. the recent technology that overrides the overcurrent fault condition of the hb-mmc is the full bridge multilevel converter (fb-mmc). though, this technology increases semiconductor losses but the important feature of dc fault reverse blocking capability was achieved by the converter by blocking current flow in the converter switches during dc faults, thereby disallowing both active and reactive power exchange that may want to occur between the dc systems and the ac grid [39, 40]. other recent hvdc converter topologies with intrinsic dc fault ride-through capabilities are alternative arm modular multilevel (aa-mmc) converters and hybrid cascaded multilevel converter with ac side h-bridge cells. these converters achieve dc fault reverse blocking capability in order to eliminate ac grid contribution to dc side faults, but has little footprint and conversion losses compared to the h-bridge modular multilevel converter [39-41]. independent control of power at each converter is possible, with one converter controlling the dc voltage at the link to match the nominal level and the other converter sets the amount of active power through the link. with the help of the phase reactor from the series inductance between the converter and the ac grid (fig 4), active and reactive power control was achieved as depict in (1) and (2). x uu p convac sin  (1) x uuu q acconvconv )sin(   (2) x-represent the series reactance of the phase reactor and the transformer in the converter station. ability of vsc-hvdc to absorb and inject active and reactive power is shown in the p-q-capability chart below (fig 4). this p-q capability chart characteristic can be termed to a circle with a radius equal to the maximum mva rating of the converters. available reactive power depends on the active power transmitted which directly fall between the operating ranges of the converter mva rating. the converters are restricted by the power electronics switches current rating and the capability circles. vac is raised above the ac grid voltage to inject reactive power. the converter voltage however suffers restriction to the maximum rating of the power electronics which limit the capability chart for higher ac voltage. nevertheless, vsc remains the most suitable choice in transmitting renewable energy (such as wind power and solar power) either offshore or onshore systems. table ii shows some existing vsc-hvdc installations. fig. 3. modular multilevel converter topology fig. 4. simplified pq characteristic of a vsc hvdc terminal [42] table ii some vsc-hvdc installations project name location characteristics (kv) year (mw) (km) borwin 1 germany ±150 2009 400 200 caprivi link namibia ±350 2010 300 951 transbay usa ±200 2010 400 85 ewic uk ±200 2012 500 261 inelfe france ±320 2013 1000 65 skagerrak 4 norway ±500 2014 700 244 table iii a comparison of lcc and vsc schemes lcc vsc thyristor base technology igbt base technology the semiconductor can with-stand voltage in either polarity withstand current in either direction constant current direction current direction changes with power energy is stored inductively store energy capacitively turned on by a gate pulse but rely on external circuit for its turn off both turn on and off is carried out without the help of an external circuit high power capability lower power capability good overload capability has weak overload capability requires stronger ac systems for excellent performance operate well in a weak ac systems requires additional equipment for black start operation possesses black start capability requires ac and dc harmonic filters for removal of distortion and harmonics requires no filter because it generates an insignificant level of harmonics poor in reactive power control good reactive power control large site area, dominated by harmonic filters a more compact site area requires converter transformer conventional transformer is used lower station losses higher station losses more mature technology still at its infancy reversal of power is done by reversing the voltage polarity power is reverse by changing the current direction higher voltage capability of over 1000kv lower voltage capability of almost 600kv mostly used to transmit bulk power for a long distance used for transmitting power from remote area with renewable energy suffers commutation failures as a result of a sudden drop in the amplitude or phase shift in the ac voltage, which result in dc temporal over-current though, the effect has no significant impact on the ac systems as it’s a self-clearing effect within a few power frequency cycles. ability to turn on as well to be turned off makes it immune to any voltage dips or transient ac disturbance; therefore, it does not suffer commutation failure. commutation failures, need for change in dc polarity when converter want to change from rectifier to inverter mode make lcc hvdc more problematic to adopt in a multi-terminal hvdc system. reason for low number of lcc base technology for multiterminal hvdc. suitable for multi-terminal hvdc systems because it does suffer from commutation failures, has independent, multidirectional power flow, and operate with the same voltage polarity. during short circuits on the dc line, control of the firing angle of the thyristors valves stops the increase of dc fault current. this converter control and protections reduces the damage caused by the fault current. incased of overhead lines fault, power transmission is stopped for arc de-ionization, after which power transmission resumed. continuous conduction in the diode will cause an increase in dc fault current even when the igbts are turned off. the ac circuit breakers at both vsc hvdc ends must be opened to stop the diode conduction. the converter link must be re-started after fault has been removed. table iv a comparison of thyristor and igbt features thyristors igbt max. voltage rating (v) 8000 1700 voltage blocking sync/async async. voltage blocking sync/async async. gating pulse voltage conduction drop (v) 1.2 3 switching frequency (khz) 1 20 development target maximum voltage rating (kv) 10 3.5 development target maximum current rating (ka) 8 2 fig. 5 shows an overview of hvdc projects around the world and fig. 6 depict hvdc available ratings for different transmission medium. fig. 5. overview of hvdc projects around the world fig. 6. available ratings of hvdc systems (udc refers to voltage per pole, and idc is the current rating, in a bipolar setup, p=2udc idc)[13] v. fault characteristic of lcc and vsc digsilent powerfactory was used for the modelling of both technologies. this is to explain briefly the transient response of lcc and vsc hvdc to faults in the ac network on the two side of the converters’ end. fig 7 show the lcc hvdc setup. it is a bipolar hvdc system, each pole consisting of twelve-pulse thyristors on both the inverter and rectifier side, with 1000mw of power at ±600kv transmitted per pole via 1000km overhead dc lines. an external grid is connected at both the rectifier and inverter end of the converter station to serves as power generation and load respectively at both ends. vsc hvdc setup is modelled as shown in fig. 8. each converter rated at 1000mw, 600kv via 1000km overhead lines. to study these responses, the ac busbar at the inverter side is subjected to a three-phase short circuit of 10ω fault impedance for 200ms using the time domain simulation (emt). 0 5 10 15 20 25 30 35 40 45 50 thyristor igbt fig. 7. lcc modelling on digsilent during emt simulation of both the lcc and vsc model, fig. 9 shows the graphic subplots for the current waveforms for lcc hvdc converter scheme. during the fault period, each converter controller helps in alleviating the effect of the fault on the converter. like in the case of lcc, the voltage current order limiter (vdcol) in the rectifier controller help to reduce the dc current, which in turn aid the inverter side to regain fast from commutation problem. fig. 10 shows subplot for vsc hvdc, during three-phase ac fault on the inverter busbar with little or no impact of the ac fault on the converter operation. this subplot shows that vsc hvdc system is immune to ac fault fig. 8. vsc hvdc model on digsilent different fault analysis which has been carried out on both two technologies on ability to reduce switching surge overvoltage and power systems restoration after blackout was discussed in [43]. the use of lcc-hvdc for different purposes, such as to improved voltage stability, transient and rotor angle stability was discussed in [44-47], while [48-50] talks on the new hybrid multi-level converter (alternate-arm multi-level) with half-bridge multi-level benefit of low distortion, losses and full h-bridge converter benefit of dc-side fault blocking capability. the alternate-arm multi-level discussed also have the ability to supply reactive during severe abnormal operation. this makes it more useful for ac grid during fault since it can provide reactive power support during voltage instability fig. 9. lcc hvdc converter current during fault. fig. 10. monopolar hvdc model. vi. future trend vsc hvdc has more technical advantages than the contemporary lcc hvdc being a new method of hvdc transmission technology. the future trend in the development of this technology is likely to lead to a more efficient and cheaper use of vsc-hvdc. with an ongoing, never ceasing improvement, research and development on vsc-hvdc technology, especially in the area of converter design and topology, such breakthrough will surely contribute to the spread of vsc-hvdc transmission systems with overhead lines and accelerate the practical realization of hvdc networks that use vsc technology. future trends also include the manufacturing of better power cable with higher voltage rating for vsc hvdc transmission. 320xlpe hvdc are still the maximum rating in service. but with ongoing research on power cables with high power rating and reduced cost, this will bring about more attraction to vschvdc technology. the use of vsc hvdc will continue to increase and apply to different power system interconnections at higher dc voltage and power rating. fig. 11 shows the earlier stage of vsc based hvdc converter technology with much power losses. but due to the development in the converter and control technology, the present vsc-based hvdc technology is of lower magnitude. but with the introduction of the multi-level vsc configuration, this has significantly narrow the gap between lcc and vsc hvdc schemes. fig. 11. lcc and vsc power loss vii. conclusion the two dominant hvdc transmitting technology have been reviewed in this paper. power electronics being the building block of any converter station, and the efficiency of these two technologies depend in the converter topology and the switches (semiconductors) used in fabricating them. lcc has the highest power rating and can sustain better during faults. however, for power control, flexibility and high converter efficiency, the vsc is superior. though with this trend, lcc may remain the more utilized of these technologies in the near future due to its high reliability and well-established thyristors base technology that it utilizes. however, with the improvement in vsc technology and the advantages which it offers over lcc, vsc is bound to grow, and gain more recognition and market share, especially with the large-scale renewable energy integration into traditional ac power grids going on worldwide. references [1] s. shah, r. hassan, and s. jian, "hvdc transmission system architectures and control a review," in control and modeling for power electronics, 2013 ieee 14th workshop on, 2013, pp. 1-8. [2] n. m. kirby, "hvdc system solutions," in transmission and distribution conference and exposition t&d, ieee pes, 2012, pp. 1-3. [3] m. h. okba, m. h. saied, m. z. mostafa, and t. m. abdel-moneim, "high voltage direct current transmission a review, part i," in energytech, 2012 ieee, 2012, pp. 1-7. [4] r. radzuan, m. a. a. raop, m. k. m. salleh, m. k. hamzah, and r. a. zawawi, "the designs of low power ac-dc converter for power electronics system applications," in computer applications and industrial electronics, ieee symposium on, 2012, pp. 113-117. [5] n. flourentzou, v. g. agelidis, and g. d. demetriades, "vsc-based hvdc power transmission systems: an overview," power electronics, ieee transactions on, vol. 24, pp. 592-602, 2009. [6] m. p. bahrman, "hvdc transmission overview," in transmission and distribution conference and exposition, 2008; ieee/pes, 2008, pp. 1-7. [7] h. k. müller, s. s. torbaghan, m. gibescu, m. m. roggenkamp, and m. a. m. m. van der meijden, "the need for a common standard for voltage levels of hvdc vsc technology," energy policy, vol. 63, pp. 244-251, 12// 2013. [8] l. de andrade and t. p. de leao, "a brief history of direct current in electrical power systems," in history of electro-technology conference (histelcon), 2012 third ieee, 2012, pp. 1-6. [9] s. m. yousuf and m. s. subramaniyan, "hvdc and facts in power system," international journal of science and research, vol. 2, 2013. [10] b. k. bose, "evaluation of modern power semiconductor devices and future trends of converters," industry applications, ieee transactions on, vol. 28, pp. 403-413, 1992. [11] s. tamai, "high power converter technologies for saving and sustaining energy," in power semiconductor devices & ic's (ispsd), 2014 ieee 26th international symposium on, 2014, pp. 12-18. [12] e. kontos, r. t. pinto, s. rodrigues, and p. bauer, "impact of hvdc transmission system topology on multiterminal dc network faults," power delivery, ieee transactions on, vol. 30, pp. 844-852, 2015. [13] m. jafar and m. molinas, "a transformerless series reactive/harmonic compensator for line-commutated hvdc for grid integration of offshore wind power," industrial electronics, ieee transactions on, vol. 60, pp. 2410-2419, 2013. [14] t. n. tran, l. luo, j. xu, s. dong, z. zhang, z. zhao, et al., "analysis of the characteristics of the new converter transformer based on the matrix model," power delivery, ieee transactions on, vol. 27, pp. 821-830, 2012. [15] j. vobecky, "the current status of power semiconductors," facta universitatis, series: electronics and energetics, vol. 28, pp. 193-203, 2015. [16] m. schenk, j. przybilla, u. kellner-werdehausen, r. barthelmess, j. dorn, g. sachs, et al., "state of the art of bipolar semiconductors for very high power applications,"; proceedings of international exhibition and conference for power electronics, intelligent motion, renewable energy and energy management;, europe 2015, pp. 1-8. [17] j. vobecky, v. botan, k. stiegler, u. meier, and m. bellini, "a novel ultra-low loss four inch thyristor for uhvdc," in power semiconductor devices & ic's,ieee 27th international symposium, 2015, pp. 413-416. [18] c. guo, y. liu, c. zhao, x. wei, and w. xu, "power component fault detection method and improved current order limiter control for [19] h. jingbo, l. mingjie, y. jun, c. qing, x. tao, and y. zhao, "research on dynamic characteristics and countermeasures of ac-dc hybrid power system with large scale hvdc transmission," in power system technology (powercon), 2014 international conference on, 2014, pp. 799-805. [20] anonymous, "hvdc projest list," prepared for hvdc and flexible ac transmission subcommitted of the ieee treansmission and distribution committee, march 2012. [21] y. jiang-hafner, h. duchen, m. karlsson, l. ronstrom, and b. abrahamsson, "hvdc with voltage source converters-a powerful standby black start facility," in transmission and distribution conference and exposition, 2008; d. ieee/pes, 2008, pp. 1-9. [22] k. friedrich, "modern hvdc plus application of vsc in modular multilevel converter topology," in industrial electronics (isie), 2010 ieee international symposium on, 2010, pp. 3807-3810. [23] j. luo, j. yao, d. wu, c. wen, s. yang, and j. liu, "application research on vsc-hvdc in urban power network," in power engineering and automation conference (peam), 2011 ieee, 2011, pp. 115-119. [24] m. callavik, p. lundberg, and o. hansson, "nordlink pioneering vsc-hvdc interconnector between norway and germany," march 2015. [25] v. gelman, "insulated-gate bipolar transistor rectifiers: why they are not used in traction power substations," vehicular technology magazine, ieee, vol. 9, pp. 86-93, 2014. [26] o. abarrategui, d. larruskain, i. zamora, v. valverde, g. buigues, and a. iturregi, "vsc-based hvdc system capability to ride through faults," international conference on renewable energy and power quality, 2015. [27] f. shewarega and i. erlich, "simplified modeling of vsc-hvdc in power system stability studies," international federation of automatic control, cape town, south africa, 2014. [28] c. c. davidson and d. trainer, "innovative concepts for hybrid multilevel converters for hvdc power transmission," in ac and dc power transmission. 9th iet international conference on, 2010, pp. 1-5. [29] r. marquardt, "modular multilevel converter: an universal concept for hvdc-networks and extended dc-bus-applications," in power electronics conference (ipec), 2010 international, 2010, pp. 502-507. [30] t. guangfu, h. zhiyuan, and p. hui, "r&d and application of voltage sourced converter based high voltage direct current engineering technology in china," journal of modern power systems and clean energy, vol. 2, pp. 1-15, 2014. [31] p. hurtuk, r. radvan, and m. frivaldský, "investigation of possibilities to increasing efficiency of full bridge converter designed for low output voltage and high output current applications," in elektro, 2012, 2012, pp. 129-132. [32] r. marquardt, "modular multilevel converter topologies with dc-short circuit current limitation," in power electronics and ecce asia, 2011 ieee 8th international conference on, 2011, pp. 1425-1431. [33] h. abu-rub, j. holtz, j. rodriguez, and g. baoming, "medium-voltage multilevel converters—state of the art, challenges, and requirements in industrial applications," industrial electronics, ieee transactions on, vol. 57, pp. 2581-2596, 2010. [34] g. p. adam, s. j. finney, a. m. massoud, and b. w. williams, "capacitor balance issues of the diode-clamped multilevel inverter operated in a quasi two-state mode," ieee transactions on industrial electronics, vol. 55, pp. 3088-3099, aug 2008. [35] e. najafi and a. h. m. yatim, "design and implementation of a new multilevel inverter topology," industrial electronics, ieee transactions on, vol. 59, pp. 4148-4154, 2012. [36] y. zhang, g. p. adam, t. c. lim, s. j. finney, and b. w. williams, "analysis of modular multilevel converter capacitor voltage balancing based on phase voltage redundant states," iet power electronics, vol. 5, pp. 726-738, 2012. [37] m. glinka and r. marquardt, "a new ac/ac multilevel converter family," industrial electronics, ieee transactions on, vol. 52, pp. 662669, 2005. [38] a. lesnicar and r. marquardt, "an innovative modular multilevel converter topology suitable for a wide power range," in power tech conference proceedings, 2003, p. 6. [39] g. p. adam, o. anaya-lara, g. m. burt, d. telford, b. w. williams, and j. r. mcdonald, "modular multilevel inverter: pulse width modulation and capacitor balancing technique," iet power electronics, vol. 3, pp. 702-715, 2010. [40] g. adam and i. davidson, "robust and generic control of full-bridge modular multilevel converter high-voltage dc transmission systems," in ieee power electronic transaction, 2015. [41] s. qiang, l. wenhua, l. xiaoqian, r. hong, x. shukai, and l. licheng, "a steady-state analysis method for a modular multilevel converter," power electronics, ieee transactions on, vol. 28, pp. 3702-3713, 2013.m. glinka and r. marquardt, "a new ac/ac multilevel converter family," industrial electronics, ieee transactions on, vol. 52, pp. 662669, 2005. [42] eskom, hvdc power transmission: basic principles, planning and converter technology (part 1): crown publication cc., 2012. [43] g. p. adam, s. finney, k. bell, and b. williams, "transient capability assessments of hvdc voltage source converters," in power and energy conference at illinois (peci), 2012 ieee, 2012, pp. 1-8. [44] k n i mbangula, o. e. oni and i e davidson, "the impact of hvdc schemes on network transient rotor angle stability," 24th southern african universities power engineering conference, south africa january 2016. [45] o. e. oni, k n i mbangula, and i. e davidson "voltage stability improvement of a multi-machine system using hvdc", in press, ieee power system conference, usa, march 2016. [46] k n i mbangula and i e davidson, “detailed power system transient stability analysis using expert system concepts and stability improvement of a large multi-machine hvac network using hvdc technologies,” proceedings of the 23rd south african universities power engineering conference, south africa, january 2015. [47] k n i mbangula, i e davidson and r tiako, “improving power system stability of south africa’s hvac network using strategic placement of hvdc links”. proceedings of the cigre international symposium 2015 development of electricity infrastructures for sub-saharan africa, cape town, south africa, october 26-30, 2015. [48] m. merlin, t. green, p. d. mitcheson, d. trainer, d. critchley, and r. crookes, "a new hybrid multi-level voltage-source converter with dc fault blocking capability," in ac and dc power transmission, 2010. acdc. 9th iet international conference on, 2010, pp. 1-5. [49] g. p. adam, k. h. ahmed, s. j. finney, k. bell, and b. w. williams, "new breed of network fault-tolerant voltage-source-converter hvdc transmission system," ieee transactions on power systems, vol. 28, pp. 335-346, feb 2013. [50] n. nayak, s. k. routray, and p. k. rout, "a robust control strategies to improve transient stability in vsc-hvdc based interconnected power systems," in energy, automation, and signal (iceas), 2011 international conference on, 2011, pp. 1-8. oluwafemi e. oni was born in nigeria on 17 september 1988. he received his bsc (honours) degree in electrical and electronic engineering from ekiti state university, ado ekiti, nigeria, in 2013. he then proceed to university of kwazulunatal, durban, south africa in 2015 for his msc degree in electrical engineering (currently handed in his thesis for examination). he was a system and maintenance engineer at egbin power thermal plant, lagos, nigeria, in 2012 and omotosho power plant, ore, nigeria, in 2013/2014. he is currently a research assistance with department of electrical engineering, university of kwazulu-natal. his research includes power systems stability analysis using high voltage direct current transmission scheme, integration of renewable energy into the grid using multi-terminal hvdc scheme, and smart grid systems using facts. mr. oni’s award and honors include mtn foundation scholarships, etisalat scholarship and ekiti state scholarship. kamati n.i. mbangula was born in namibia on 20 august 1989. he graduated with a bsc. (honours) degree in electrical engineering from the university of namibia (unam). he pursued his postgraduate studies in south africa at the university of kwazulu-natal (rsa), and carried out his research at the eskom centre of excellence in high voltage direct current (hvdc). his work experience includes working as a staff development fellow at unam, and working as a research assistant and lab technician at the eskom centre of excellence in hvdc. he is currently employed as a lecturer at unam. his fields of interests include power systems stability analysis, and low voltage reticulation systems design and analysis. innocent e. davidson (m’92–sm’02) received the bsc (hons) and msc degrees in electrical engineering from university of ilorin in 1984, and 1987 respectively. phd in electrical engineering from the university of cape town, rondebosch, south africa1998; and postgraduate diploma in business management from the university of kwazulu-natal, south africa, 2004; associate certificate, sustainable energy management, british columbia institute of technology, burnaby, canada, 2011. from 1994-1995, he was engineering inspector, rainbow energy project at easigas (pty) ltd, cape town; senior lecturer, university of pretoria (1999-2001); senior lecturer, department of electrical engineering, university of kwazulunatal (ukzn), 2001–2006; part-time instructor, graduate engineering program (power & energy), ukzn high voltage dc centre (2000-2008) a program co-offered by ukzn and eskom south africa’s electric utility. from 2005–2006, he was a visiting professor, powertech labs inc., surrey, bc, a world leading consortium in clean energy technologies, independent testing services, power system solutions and smart utility services. from 2007-2011 he was energy consultant in surrey, bc, implementing energy efficiency (electricity/gas) measures, british columbia provincial government’s mandate on climate change. he has been an invited guest writer for the ieee power and energy technical magazine as an expert on africa: “energizing africa’s emerging economy”, ieee power and energy, vol. 3, no 4, july/august 2005. he was associate professor of electrical engineering and research coordinator, university of namibia (2012-2014); director, eskom centre of excellence in hvdc engineering, ukzn (2014-2016). currently, he is a full professor of electrical engineering, durban university of technology, south africa. he is the author/co-author of over 150-refereed journal and conference papers. his research focus is on grid integration of renewable energy using smart technologies and innovation for smart cities. prof davidson is a member, western canada group of chartered engineers (wcgce); the institute of engineering and technology (iet canada) british columbia chapter; a chartered engineer, c.eng. united kingdom. he is a fellow of the south african institute of electrical engineers and a registered professional engineer, p. eng. (ecsa), south africa.  transactions on environment and electrical engineering issn 2450-5730 vol 3, no 1 (2019) © vittorio guida, damien guilbert, bruno douine abstract—recently, the use of electrolyzers for hydrogen production through water electrolysis is of great interest in the industrial field to replace current hydrogen production pathways based on fossil fuels (e.g. oil, coal). in order to reduce the emission of pollutants into the atmosphere and minimize the cost of electricity, it is preferable to use renewable energy sources (e.g. solar, wind, hydraulic). the electrolyzers must be supplied with a very low dc voltage in order to produce hydrogen from the deionized water. for this reason, dc-dc step-down converters are generally used. however, these topologies present several drawbacks from output current ripple and voltage gain point of view. in order to meet these expectations, interleaved dc-dc stepdown converters are considered as promising and interesting candidates to supply proton exchange membrane (pem) electrolyzers. indeed, these converters offer some advantages including output current ripple reduction and reliability in case of power switch failures. in addition, over the last decade, many improvements have been brought to these topologies with the aim to enhance their conversion gain. hence, the main goal of this paper is to carry out a thorough state-of-the-art of different interleaved step-down dc-dc topologies featuring a high voltage gain, needed for pem electrolyzer applications. furthermore, a comparison of candidate interleaved step-down converters not only from the voltage ratio point of view but also from the phase and/or output current ripple point of view. index terms— electrolyzer, interleaved converters, renewable sources, conversion ratio, current ripple, energy efficiency, power switch faults, reliability. i. introduction he random behavior of the renewable energy sources (res) makes the hydrogen production and storage an engaging and efficient solution. this is because hydrogen has much higher specific energy than the classical storage devices such as batteries [1]. on planet earth, there are several resources available for hydrogen production such as fossil fuels (e.g. natural gas and coal), and res (e.g. biomass and water). however, from an environmental point of view, hydrogen production from fossil fuels (although it does save money) contributes considerably to the release of greenhouse gases and other pollutants into the atmosphere [2]. in this perspective, v. guida, d. guilbert, b. douine are with the group of research in electrical engineering of nancy (green), université de lorraine, 54500, vandoeuvre-lès-nancy, france (e-mail: vittorio.guida@univ-lorraine.fr , damien.guilbert@univ-lorraine.fr , bruno.douine@univ-lorraine.fr ). water is considered an attractive raw material for hydrogen production (having two atoms of hydrogen and one of oxygen, as is well known). being free of nitrogenous, carbonaceous or sulfured species, water is ideal for hydrogen production, contributing to the reduction of polluting emissions. among the different hydrogen production processes, starting from water, the most consolidated is electrolysis. water electrolysis allows obtaining practically pure hydrogen. this process, for which electricity currently has a cost up to three or four times higher than the methane used for steam reforming, becomes economically acceptable as a result of technological innovations and under extremely low-cost conditions of electricity (if electricity is produced from res) [3]. water electrolysis is based on an electrochemical reaction using electricity to split water into hydrogen and oxygen; it is carried out by means of an electrolyzer (el). there are three types of els in the literature: proton exchange membrane (pem) el, alkaline el, and solid oxide el (the latter exists only in the field of research and development) [4]. in order to produce hydrogen from deionized water, the el must be supplied with a very low dc voltage. hence, the use of dc-dc converters is decisive to adjust the voltage levels between the el and the dc bus. generally, classic dc-dc step-down converters are used for this purpose due to their simplicity and low cost [5,6]. unfortunately, these converters have several drawbacks from availability in case of electrical failures, output current ripple, conversion ratio, and energy efficiency point of view for el applications. the same issues have been highlighted regarding classic step-up converters for fuel cell applications [7,8]. over the last decade, a family of dc-dc step-down converters called interleaved has spread particularly in the research field. indeed, many interleaved step-down topologies have been proposed in the scientific literature [9-16], bringing improvements (e.g. energy efficiency optimization, output current ripple minimization, and availability in case of electrical failures) compared to the conventional interleaved step-down converter. as it has been mentioned earlier, els must be supplied with a very low dc voltage; so interleaved dc-dc step-down converters are suitable for this type of applications. literature survey of interleaved dc-dc stepdown converters for proton exchange membrane electrolyzer applications vittorio guida, damien guilbert, and bruno douine t mailto:vittorio.guida@univ-lorraine.fr mailto:vittorio.guida@univ-lorraine.fr mailto:damien.guilbert@univ-lorraine.fr mailto:damien.guilbert@univ-lorraine.fr mailto:bruno.douine@univ-lorraine.fr mailto:bruno.douine@univ-lorraine.fr starting from these observations, the main purpose of this work is to carry out a thorough literature survey focused on the family of interleaved dc-dc step-down converters featuring a high voltage gain. this article is divided into six sections. after this introduction providing the current state-of-the-art and issues, section ii compares the three existing technologies of els with the aim to select the most suitable technology for this study. then, section iii presents the main requirements of dc-dc converters for el applications. afterward, in section iv, candidate interleaved step-down topologies for el applications are presented including their advantages and drawbacks. after that, in section v, a comparison is carried out between candidate interleaved converters, especially from voltage gain and phase and/or output current ripple point of view. finally, in section vi, conclusions and perspective of the work are given. ii. proton exchange membrane technology currently, different types of el can be distinguished by their electrolyte and the charge carrier: (1) alkaline el; (2) proton exchange membrane (pem) el; and (3) solid oxide (so) el [2,3]. table i provides the main features of each technology; while table ii introduces the advantages and drawbacks of each technology. from tables i and ii, alkaline and pem els are currently the two main technologies, which are commercially available. alkaline els are the most mature and widespread compared to pem els (still under development). as highlighted in tables i and ii, alkaline els have a higher durability and gas purities, and cheaper catalysts than pem els. however, pem els have several advantages over alkaline els, such as compactness, fast system response, wide partial load range and high flexibility in terms of operation. as a result, this technology is an attractive option for integration into the grid including renewable power generating systems [3]. for this reason, pem els are considered within hybrid renewable energy systems and hydrogen production pathways based on renewable energy sources. iii. main requirement for dc/dc converters like for fuel cells, dc/dc converters are needed to interface the dc voltage grid and the el. these converters can be used both for hybrid renewable energy systems (fig. 1) and hydrogen production pathways based on renewable energy sources (fig. 2). generally, a pem el needs a very low dc voltage in order to produce hydrogen. indeed, at rated power, the cell voltage range of a pem el is included between 1.75 and 2.2 v [2]. a higher input el voltage can be obtained by stacking more cells. however, the number of the cells has to be limited in order to guarantee a high reliability of the pem el. currently, this compromise between the el reliability and its stack voltage (which is the sum of each cell voltage) is a challenging issue for el applications [4]. generally, step-down dc/dc converters are used to supply pem els; whereas for fuel cell applications, step-up converters are preferred. in any systems including a hydrogen buffer storage, dc/dc converter must meet a certain number of requirements, table i main features of each electrolyzer technology alkaline pem so maturity commercial commercial medium and small-scale applications research and development current density 0.2-0.4 acm-2 0.6-2 acm-2 0.3-0.6 acm-2 cell area <4m2 <0.3 m 2 / cell voltage 1.8-2.40 v 1.75-2.20 v / hydrogen output pressure 0.05-30 bar 10-30 bar 50 bar operating temperature 60-80 °c 50-80 °c 700-800 °c system efficiency 52-69 % 57-69 % >90 % (heat and hydrogen) <80% (only for hydrogen) indicative system cost 1-1.2 €/w 1.9-2.3 €/w 1.2 €/w system size range 0.25-760 nm3h-1 1.8-5300 kw 0.01-240 nm3h-1 0.2-1150 kw laboratory scale lifetime stack <90000 h <60 000 h ≈1000 h provided below [4]: 1) high energy efficiency. 2) low electromagnetic disturbances. 3) reduced cost. 4) high voltage ratio. 5) low output current ripple (to optimize el performances). 6) ability to operate in case of electrical failures. among these requirements, the most important feature expected from the dc/dc converter is a high conversion ratio. indeed, for electrical systems including wind turbines, the dc bus voltage is very high (i.e. between a hundred and a thousand volts) [4]. besides, current hybrid renewable energy systems with hydrogen storage based on a dc bus configuration are limited to low-power applications due to the use of classic dc/dc converters (buck for pem el) [17]. hence, in order to move towards medium and high-power applications, dc/dc converters must feature high conversion ratio ability [18]. interleaved step-down dc-dc converter topologies have much to offer for pem els. some improvements have been reported in the literature to enhance the conversion ratio ability while benefiting low output current ripple, high energy efficiency, and availability in case of electrical failures. over the last years, many interleaved dc-dc buck converters proposed in the literature [9-16] can be suitable for pem el applications. some candidate topologies with their advantages and drawbacks are presented in the following section. table ii comparison of electrolyzer technologies alkaline pem so advantages mature technology long-term stability high durability and gas purities cheaper catalyst stacks in the mw range high current densities high voltage efficiency fast system response compactness high gas purity dynamic operation high gas purity high efficiency possible reversibility: operation in fuel cell mode drawbacks low current densities crossover gases low partial load range load dynamics low operational pressures corrosive liquid electrolyte low tolerance to impurities in the water high cost of components technology relatively new acidic corrosive environment limited durability low tolerance to impurities in the water not commercially available (under research and development) fragility of materials need for a significant heat input limited lifetime of ceramics long start-up time fig. 1. hybrid renewable energy system with a hydrogen buffer storage based on a dc bus configuration. fig. 2. hydrogen generation pathways from wind turbines. iv. candidate interleaved dc/dc step-down converters a. interleaved buck converter based on the classic buck converter, interleaved buck converters can be achieved. these topologies are built by connecting in parallel n buck converters (from n=2 to n=6) with a common dc bus [4]. they present several benefits compared to the classic buck converter, especially from energy efficiency, output current ripple reduction and reliability point of view [4]. generally, a three-leg interleaved buck converter (ibc) is preferred for optimization reasons (i.e. magnetic component size, output current ripple, energy efficiency) as shown in fig. 3. however, ibc topologies present the following drawbacks [4]: 1) large voltage stresses at the terminals of power switches and diodes (limited energy efficiency). 2) medium conversion ratio (not suitable for electrolyzers requiring a high voltage ratio). over the last decade, many improvements have been brought to the classic ibc, especially from voltage ratio and energy efficiency point of view. these important issues can be solved by modifying the architecture and/or using coupled inductors. in the next subsections, several candidates interleaved buck topologies are presented with the improvements brought to the classic ibc. fig. 3. ibc connected with the electrolyzer. b. interleaved buck converters with a single-capacitor snubber the first topology (fig. 4) differs from the conventional ibc topology for two aspects [9]: 1. a single-capacitor snubber that consists of a resonant capacitor c1 and either inductor l1 or l2; 2. an ei core thanks to which the two coupled inductors (l1 and l2) are designed. the snubber circuit is employed to minimize turn-off losses, switching losses and number of components as well. in addition, it allows limiting the rising rate of the voltage at the terminals of the power switch. the magnetic core (i.e. ei) is employed to decrease the volume of the converter. besides, to optimize energy efficiency, the inverse coupling method is used for l1 and l2 that leads to better stationary and dynamic performance. on the one hand, this converter features the same dynamic performance of the classic ibc. on the other hand, ibcs with a single-capacitor snubber can lead up to higher efficiency than conventional ibc for applications requiring a low voltage ratio; whereas for high voltage ratio, the two converters produce approximately the same efficiency. the voltage ratio of the converter according to the duty cycle d is given by the following equation [9]: 𝑉𝑜 𝑉𝑖 = (1−𝐷+𝑘𝐷)3𝐷2 (7𝑘−3)𝐷2−(5𝑘−3)𝐷+𝑘 (1) where: ▪ the coupling coefficient k (k = m / l) of the coupled inductors l1 and l2 is considered equal to 0.33 [9]; ▪ m is the mutual inductance; the coupled inductors are made with a symmetric structure (l1 = l2 = l). fig. 4. ibc with a single-capacitor snubber [9]. c. interleaved buck converter with coupled windings compared to the previous topology, the second topology (fig. 5) is composed of the following elements [10]: 1. two windings coupled with a transformation ratio n, connected to each phase of the converter. each winding is coupled with the inductance of the corresponding phase; 2. a synchronous ibc composed of two phases. the addition of the two windings situated before the classic interleaved structure leads to a new topology. it significantly enhances energy efficiency without deteriorating the dynamic response of the converter. furthermore, it proposes an improved voltage ratio, given by the following expression [10]: 𝑉𝑜 𝑉𝑖𝑛 = 𝐷 𝑛+1 (2) fig. 5. ibc with coupled windings [10]. d. interleaved three-level buck converter the interleaved architecture (fig. 6) is a three-level dc-dc converter. in multilevel dc-dc converters, each power switch must withstand only a part of the input voltage and this allows operation with input voltages that are higher than the ratings of the power switches [11]. this topology consists of two interleaved buck converters, each of which includes [11]: 1. the main inductor l0/2; 2. two commutation inductors (l1, l2 or l3, l4). the four auxiliary inductors (l1, l2, l3, l4) allow an important decrease of the power losses related to diode reverse recovery and turn-on transitions at no current. the interleaved zct tl topology is addressed to high-power, high-voltage applications. furthermore, it can be observed that [11]: ▪ it can operate at high switching frequencies and this makes easier the design of the output filter; ▪ all power switches play a part in the power management of the topology and equally divide the electrical power; ▪ the volume of the converter can be minimized by using coupled inductors. the converter must operate at duty cycle values smaller than 0.5 permitting the diodes to switch. otherwise, if the duty cycle is higher than 0.5, the soft switching feature will not be ensured. the conversion gain of the converter is obtained by the following equation [11]: 𝑉𝑜 𝑉𝑖𝑛 = 2𝐷 1+ 2𝑅𝑒 𝑅𝑜 (3) where: ▪ re is the lossless resistance (re = 2lc / ts); ▪ ts is the switching period; ▪ lc is the commutation inductance (if l1 = l2 = l3 = l4 = l: without coupled inductors, lc = 2l, instead of with coupled inductors, lc = 4l); ▪ ro is the output load resistance. fig. 6. interleaved zero current transition (zct) three-level (tl) buck converter [11]. e. interleaved zero-current-transition buck converter the topology (fig. 7) is an interleaved zct buck converter. it differs from the conventional topology since there is an output inductance lo. the auxiliary inductors l1 and l2 set the current slopes during the switching phases. as the result, these inductors impact the losses related to diode reverse recovery issues. furthermore, the additional turn-on losses, related to the amount of the leakage diode current, can be guided by the appropriate choice of the auxiliary inductors. the larger the magnetic components (l1 and l2), the smaller the reverse recovery and leakage currents. however, the switching times last longer and therefore it is needed to find a compromise [12]. moreover, these auxiliary magnetic components enable zct turn-on. the output inductor lo, which is larger than the inductors l1 and l2, allows operating at a continuous conduction mode with low output current ripple. the two power switches contribute towards the power management of the converter. hence, it makes easier the thermal design and leads to a significant reduction of losses. finally, the conversion gain of the converter is provided by the following equation [12]: 𝑉𝑜 𝑉𝑖𝑛 ≈ 2𝐷 (4) equation (4) clearly emphasizes that the complete range of conversion gains (0 < vo/vin < 1) can be achieved by operating each power switch with a duty cycle value included between 0 and 0.5 [12]. fig. 7. interleaved zero-current-transition (zct) buck converter [12]. f. stacked interleaved converter the converter (fig. 8) is the stacked interleaved topology and it differs from the conventional topology since there is a capacitor (cs) in the secondary phase. the capacitor located in the second phase (cs) stops from flowing the continuous load current from the second phase, making the continuous current for the first phase. this aspect is useful for practical applications where the magnetic components have various parasitic resistances leading up to increasing losses in the secondary phase [13]. the first advantage of the stacked interleaved topology is that it allows a full suppression of the output current ripple whatever the duty cycle values, reducing the needed phases to two (unlike conventional ibc topologies where the number of cancellations strongly depends on the duty cycle and the number of phases). this current ripple cancellation is achieved through the following components and operation [13]: ▪ the first phase (sp, lp, cp) connected with the load and operating with a duty cycle d; ▪ the second phase (ss, ls, cs) no connected with the load and operating with a duty cycle 1-d; and with the timing chart shown in fig. 8. eliminating the output current ripple, it allows removing the relation between the current ripple of the inductors (lp and ls) and the output voltage ripple. as a result, the inductors are smaller than inductors met in ibc. additionally, the volume reduction of inductors brings more compactness and enhances the dynamic response of the topology. connecting the two phases together through a capacitor (cs), it allows obtaining two different voltages. indeed, the voltage ratios of the converter are given by the following expressions, respectively for the first and second phase [13]: 𝑉𝑂𝑈𝑇 𝑉𝐼𝑁 = 𝐷 (5) and 𝑉𝑂𝑈𝑇 𝑉𝐼𝑁 = 1 − 𝐷 (6) the stacked interleaved topology allows coupling the two inductors lp and ls. in this case, this coupling permits the reduction of the volume of the inductance, and the attenuation of the current ripple flowing through each inductor. therefore, by reducing the current ripple, energy efficiency is improved. in addition, another advantage of using magnetic coupling is the area reduction by stacking the inductors [13]. finally, as highlighted in [13], all the process effects occurring in a practical implementation, the non-idealities of the converter bring about a delay between the switching transitions. any overlaps lead up to high current ripples depending on the gain of the magnetic coupling and the duration of the overlaps. the higher the current ripple, the lower the energy efficiency of the converter. in summary, the gain of magnetic coupling has to be chosen judiciously to reduce the effects of time errors [13]. fig. 8. stacked interleaved topology and timing chart [13]. g. interleaved buck converter with winding-cross-coupled inductors and passive-lossless clamp scheme the topology depicted in fig. 9 differs from the conventional ibc topology for two aspects: 1. a basic cell with wccis and interleaved architecture; 2. a passive-lossless clamp circuit. the basic cell has two wccis (l1 and l2). each wcci has three windings (l1a, l1b, l2c and l2a, l2b, l1c). the second winding with n2 turns is linked with the winding in its phase with n1 turns (l1b versus l1a and l2b versus l2a) and the third winding with n2 turns is linked with the windings in another phase (l1c versus l1a and l1b, l2c versus l2a and l2b) [14]. the first windings l1a and l2a have similar features as the magnetic components in the basic ibc. the second and the third windings (l1b, l1c, and l2b, l2c) are used as continuous voltage sources and are in series in the circuit to alleviate the power switch voltage stress [14]. moreover, the use of these windings allows achieving high step-down voltage ratios [14]. the basic cell takes advantage of: ▪ the basic interleaved structure to decrease the current ripple, which reduces the inductor size, increases the power level and enhances the dynamic response; ▪ the coupled inductors to obtain a high conversion gain. they aim also at reducing the power switch voltage stress and at avoiding the reduced turnoff pulse operation, which decreases the conduction losses and the current ripple. on the other side, the fact of using wccis leads up to leakage inductances (llk1 and llk2), which result in large switching losses, high voltage spikes, and serious electromagnetic interference (emi) issues [14]. the drawbacks caused by wccis can be solved by means of the passive-lossless clamp circuit. the passive-lossless circuit, consisting of two clamp capacitors (cc1 and cc2) and four clamp diodes (dc11, dc12, dc21, dc22), absorbs the voltage spikes on the power switch and reuses the leakage energy [14]. as a result, the energy efficiency of the topology is enhanced, and the electromagnetic disturbances noise is canceled [14]. compared with the classic ibc, this converter allows decreasing the power switch voltage stress due to the features of the wccis. furthermore, high-performance power semiconductors with low on-state resistances can be used to decrease the conduction losses [14]. the reverse-recovery issue of the output diode (do1, do2) is mitigated and the reverserecovery losses are minimized given that the output diode current falling rate is imposed by the leakage inductance [14]. in summary, this converter is fit for high power applications, high current, high step-down conversion, and to operate at a high switching frequency. finally, the conversion gain of the converter is obtained by using this following equation [14]: 𝑉𝑜𝑢𝑡 𝑉𝑖𝑛 = 𝐷 𝑁+1 (7) where: ▪ d ≤ 0.5; ▪ n is the turns ratio (n = n2/n1). fig. 9. interleaved dc–dc high step-down buck converter with winding-crosscoupled inductors (wccis) and passive-lossless clamp scheme [14]. h. interleaved coupled-buck converter with active-clamp circuits by comparison, this topology (fig. 10) differs from the conventional topology for these three aspects [15]: 1. two coupled windings on each phase (l11 and l1, l22 and l2 with transformation ratios, respectively indicated n1 and n2); 2. a resonance inductance per phase (lr1 and lr2); 3. an active-clamp circuit per phase (m11 and cr1, m22 and cr2). on the one hand, resonant inductors are used to achieve zero voltage switching for the main and auxiliary power switches, and to limit transient reverse currents of freewheeling diodes. hence, it allows reducing significantly reverse-recovery losses. on the other hand, the active-clamp circuits allow recovering the dispersion energy and limiting the voltage spikes [15]. like the previous topology, the use of coupled windings allows improving the voltage gain of the converter, provided by the equation (8) [15]: 𝑉𝑜 𝑉𝑖 = 𝐷 𝐷+𝑛(1−𝐷) (8) fig. 10. interleaved coupled-buck converter with active-clamp circuits [15]. i. interleaved buck converter with extended duty cycle the interleaved architecture of fig. 11 is similar to the conventional ibc, but it differs for two aspects [16]: 1. two active switches, q1 and q2, are connected in series; 2. a coupling capacitor (cb) is employed in the power path (it is quite large to be regarded as a voltage source). the ibc topology with extended duty cycle is particularly suitable for high input voltage applications where the operating duty cycle must be less than or equal to 0.5. the converter of fig. 11 presents the following advantages than the conventional ibc [16]: ▪ a higher step-down conversion ratio; ▪ a smaller output current ripple (therefore, the inductors with a smaller inductance can be used). moreover, the main advantage of this topology is that since the voltage stress across active switches (q1 and q2) is half of vs before turn-on or after turn-off when the operating duty cycle is below 50%, the capacitive discharging and switching losses can be reduced substantially; this allows the converter of fig. 11 to have a higher efficiency than that of the conventional ibc and operate with higher switching frequencies. the conversion gains of the ibc topology with extended duty cycle are obtained by the following equations [16]: 𝑉𝑂 𝑉𝑆 = 𝐷 2 (with d ≤ 0.5) (9) and 𝑉𝑂 𝑉𝑆 = 𝐷2 (with d > 0.5) (10) finally, we observe that the voltage stress of d1, during the cold startup, could be higher than vs. to solve this issue, an auxiliary circuit can be added to the input stage of the converter (fig. 12). this auxiliary circuit is composed of: ▪ two capacitors (cadd1, cadd2); ▪ a diode (dadd); ▪ a resistor (radd); it has the goal of absorbing transient energy generated by parasitic elements during the cold startup. fig. 11. ibc with extended duty cycle [16]. fig. 12. ibc with extended duty cycle and auxiliary circuit [16]. v. comparison of candidate interleaved step-down converters as highlighted in a previous review work [4], three types of dc-dc converters are currently used for pem el applications, such as buck, half-bridge, and full-bridge dc-dc converters. however, these classic converters are not optimized from voltage ratio, energy efficiency, output current ripple minimization, and availability point of view. in this article, only interleaved step-down converters have been considered due to their advantages for pem el applications. on the one hand, the interleaved step-down converters [9,10], [12-16] are composed of two phases. despite these topologies are fault-tolerant in case of electrical failures, if one of the phases was faulty, the converter would lose its features [4]. on the other hand, interleaved three-level step-down converter offers an enhanced availability in case of electrical failures [11]. indeed, this converter is composed of two phases in the non-floating part (upper) and two phases in the floating part (lower). if one of the phases was faulty, the converter could continue to operate without any operation. however, with the aim to improve and optimize the operation of the converter, fault-tolerant strategies must be applied after fault identification and detection. availability in the case of electrical failures is not the only requirement for pem el. indeed, one of the most important requirements is a high conversion gain since the pem el must be supplied with a very low dc voltage. furthermore, a low output current ripple (both low and high frequency) is required to optimize pem el performance, especially from energy efficiency and hydrogen production point of view. hence, a thorough analysis of the conversion gain and current ripples is provided in table iii for each interleaved step-down converter. besides, fig. 13 shows a comparison between conversion gain according to the duty cycle. table iii comparison of interleaved step-down converters from conversion gain and current ripple point of view topology conversion gain phase current ripple output current ripple ibc [4] 𝑣𝑒𝑙 𝑣𝑑𝑐 = 𝐷 for the first, second and third phase: 𝛥𝐼𝐿 = 𝑣𝑒𝑙(1−𝐷) 𝐿𝑓𝑠𝑤 with: l1 = l2 = l3 = l 𝛥𝐼 = 𝑣𝑒𝑙𝐷(1−3𝐷) 𝐿𝑓𝑠𝑤 , 0 < 𝐷 < 1 3 𝛥𝐼 = 𝑣𝑒𝑙(3𝐷−1)(2−3𝐷) 3𝐿𝑓𝑠𝑤 , 1 3 < 𝐷 < 2 3 𝛥𝐼 = 𝑣𝑒𝑙(3𝐷−2)𝐷 𝐿𝑓𝑠𝑤 , 2 3 < 𝐷 < 1 ibc with a single-capacitor snubber [9] 𝑉𝑜 𝑉𝑖 = (1−𝐷+𝑘𝐷)3𝐷2 (7𝑘−3)𝐷2−(5𝑘−3)𝐷+𝑘 for the first and second phase: 𝛥𝐼 = (𝑉𝑖 − 𝑉𝑜 ) [𝐿(1−𝐷)+𝑀∙𝐷]2 𝐿(𝐿2−𝑀2)(1−𝐷)2 ∙ 𝐷 𝑓𝑠𝑤 where: 0 < 𝐷 ≤ 1 2 m: mutual inductance l1 = l2 = l 𝛥𝐼𝐿 = 𝐷(𝐿−𝑀)(𝑉𝑖−𝑉𝑜)[𝐿(1−𝐷)+𝑀𝐷] 2−𝑉𝑜𝐿𝐷(𝐿 2−𝑀2)(1−𝐷)2 𝐿2(𝐿2−𝑀2)(1−𝐷)2𝑓𝑠𝑤 where: 0 < 𝐷 ≤ 1 2 m: mutual inductance l1 = l2 = l ibc with coupled windings [10] 𝑉𝑜 𝑉𝑖𝑛 = 𝐷 𝑛+1 n: turns of coupled windings there is not. 𝛥𝐼𝑜 = 𝑉𝑖𝑛−(𝑛+1)𝑉𝑜 𝐿𝑒𝑞 ∙ 𝐷 𝑓𝑠𝑤 = (1−𝐷)(𝑛+1)𝑉𝑜 𝐿𝑒𝑞 ∙ 1 𝑓𝑠𝑤 where: leq = l1b + l1a + 2m m: mutual inductance interleaved zct tl buck converter [11] 𝑉𝑜 𝑉𝑖𝑛 = 2𝐷 1+ 2𝑅𝑒 𝑅𝑜 there is not. 𝛥𝐼𝐿0 = 𝑉𝑜(1−4𝐷) 𝐿0𝑓𝑠𝑤 , 0 < 𝐷 < 1 4 𝛥𝐼𝐿0 = 𝑉𝑜(4𝐷−1)(2−4𝐷) 4𝐷𝐿0𝑓𝑠𝑤 , 1 4 < 𝐷 < 1 2 where: 𝐿0 ≫ 𝐿𝑐 lc: commutation inductance. l1 and l2: two small commutation inductors for the ibc connected to the positive voltage rail. l3 and l4: two small commutation inductors for the ibc connected to the negative voltage rail. lc: sum of the commutation inductors in each of the two ibcs. interleaved zct buck converter [12] 𝑉𝑜 𝑉𝑖𝑛 ≈ 2𝐷 there is not. 𝛥𝐼𝐿𝑜 = 𝑉𝑖𝑛−𝑉𝑜 𝐿+𝐿𝑜 ∙ 𝐷 𝑓𝑠𝑤 where: 0 < 𝐷 ≤ 1 2 stacked interleaved converter [13] for first phase: 𝑉𝑂𝑈𝑇 𝑉𝐼𝑁 = 𝐷 where: 0 < 𝐷 < 1 for second phase: 𝑉𝑂𝑈𝑇 𝑉𝐼𝑁 = 1 − 𝐷 where: 0 < 𝐷 < 1 for the first phase (without magnetic coupling between the inductors): 𝛥𝐼𝑃 = (1 − 𝐷)𝐷 𝑉𝐼𝑁 𝐿𝑓𝑠𝑤 where: 0 < 𝐷 < 1 ls = lp = l for the second phase (without magnetic coupling between the inductors): 𝛥𝐼𝑆 = −(1 − 𝐷)𝐷 𝑉𝐼𝑁 𝐿𝑓𝑠𝑤 where: 0 < 𝐷 < 1 ls = lp = l complete ripple cancellation across all duty cycles (0 < 𝐷 < 1) for the first phase (with magnetic coupling between the inductors): 𝛥𝐼𝑃 = = 1 𝐿(1+𝑘) 𝐷(1 − 𝐷)𝑉𝐼𝑁 1 𝑓𝑠𝑤 where: 0 < 𝐷 < 1 ls = lp = l 𝑘 = 𝑀 𝐿 k: mutual coupling factor m: mutual inductance for the second phase (with magnetic coupling between the inductors): 𝛥𝐼𝑆 = = − 1 𝐿(1+𝑘) 𝐷(1 − 𝐷)𝑉𝐼𝑁 1 𝑓𝑠𝑤 where: 0 < 𝐷 < 1 ls = lp = l 𝑘 = 𝑀 𝐿 k: mutual coupling factor m: mutual inductance complete ripple cancellation across all duty cycles (0 < 𝐷 < 1) ibc with wccis and passive-lossless clamp scheme [14] 𝑉𝑜𝑢𝑡 𝑉𝑖𝑛 = 𝐷 𝑁+1 𝑁 = 𝑛2 𝑛1 there is not. 𝛥𝐼𝑜𝑢𝑡 = 𝑉𝑖𝑛−(𝑁+1)𝑉𝑜𝑢𝑡 𝐿𝑒𝑞 ∙ 𝐷 𝑓𝑠𝑤 = (1−𝐷)(𝑁+1)𝑉𝑜𝑢𝑡 𝐿𝑒𝑞 ∙ 1 𝑓𝑠𝑤 where: 0 < 𝐷 ≤ 1 2 leq = l1b + l1a + 2m table iii (continuation) topology conversion gain phase current ripple output current ripple interleaved coupled-buck converter with activeclamp circuits [15] 𝑉𝑜 𝑉𝑖 = 𝐷 𝐷+𝑛(1−𝐷) for the first and second phase: 𝛥𝐼 = 𝑛−1 𝑛 ∙ 𝑉𝑖−𝑉𝑜 𝐿 ∙ 𝐷𝑚𝑎𝑥 𝑓𝑠𝑤 where: 0 < 𝐷 < 1 2 𝑛 = 𝑛1+𝑛2 𝑛1 n: turns ratio of coupled inductors l1 and l11 or l2 and l22. l1 = l2 = l the expression of output current ripple can only be determined experimentally. ibc with extended duty cycle [16] 𝑉𝑂 𝑉𝑆 = 𝐷 2 , 0 < 𝐷 ≤ 1 2 𝑉𝑂 𝑉𝑆 = 𝐷2 , 1 2 ≤ 𝐷 < 1 𝛥𝐼𝐿 = (𝑉𝑆−2𝑉𝑂)𝐷 2𝐿𝑓𝑠𝑤 , 0 < 𝐷 ≤ 1 2 𝛥𝐼𝐿 = 𝑉𝑂(1−𝐷) 𝐿𝑓𝑠𝑤 , 1 2 ≤ 𝐷 < 1 with: l1 = l2 = l 𝛥𝐼 = (𝑉𝑆−4𝑉𝑂)𝐷 2𝐿𝑓𝑠𝑤 , 0 < 𝐷 ≤ 1 2 𝛥𝐼 = (𝑉𝑆−𝑉𝑂)(2𝐷−1) 𝐿𝑓𝑠𝑤 , 1 2 ≤ 𝐷 < 1 with: l1 = l2 = l fig. 13. comparison of the voltage ratio according to the duty cycle. based on table iii and fig. 13, it can be observed that the classic ibc is not fit for electrolyzers requiring a high voltage gain despite the output current ripples are strongly reduced compared to a classic step-down converter. indeed, high voltage gain for an ibc leads up to operate at a very low duty cycle [4]. in addition, the most suitable converters for high voltage gain are ibc with coupled windings, ibc high stepdown with wccis and passive-lossless clamp circuit and interleaved coupled-buck converter with active-clamp circuits. these converters are very interesting for systems based on hydrogen buffer where wind turbines are used. by comparison, the stacked interleaved converter allows canceling the output current ripple whatever the duty cycle value; whereas for ibc topologies, the output current ripple can be canceled for specific duty cycle values [4]. however, this topology suffers from having a low voltage ratio like the classic step-down converter. from output current ripple and availability point of view, the three-level interleaved step-down converter is the most interesting topology for hybrid renewable energy systems with hydrogen storage based on a dc bus configuration. vi. conclusion the main objective of this paper is to carry out a thorough literature survey focused on candidate interleaved step-down converters for proton exchange membrane electrolyzer applications. based on the current literature, it was demonstrated that the classical topologies (e.g. buck, halfbridge, full-bridge) currently used for these applications present several drawbacks. hence, interleaved dc-dc step-down converters offer several advantages over classical topologies and are promising for these applications. based on the classic interleaved dc-dc step-down topology, several candidates interleaved converters were introduced in this article. each converter was thoroughly analyzed to determine current ripples and voltage gain expression. from the obtained expressions (summarized in a table and a figure), the most interesting and promising interleaved step-down converters were emphasized from output current ripple reduction and voltage gain point of view. references [1] t.s. uyar, d. beşikci, integration of hydrogen energy systems into renewable energy systems for better design of 100% renewable energy communities, international journal of hydrogen energy, vol. 42, iss. 4, 2017, pp. 2453-2456. [2] m. carmo, d.l. fritz, j. mergel, d. stolten, a comprehensive review on pem water electrolysis, international journal of hydrogen energy, volume 38, issue 12, 2013, pages 4901-4934. [3] alexander buttler, hartmut spliethoff, current status of water electrolysis for energy storage, grid balancing and sector coupling via power-to-gas and power-to-liquids: a review, renewable and sustainable energy reviews, volume 82, part 3, 2018, pages 2440-2454. [4] d. guilbert, s.m. collura, a. scipioni, dc/dc converter topologies for electrolyzers: state-of-the-art and remaining key issues, international journal of hydrogen energy, volume 42, issue 38, 2017, pages 2396623985. [5] m.e. sahin, h.i. okumus‚ m.t. aydemir, “implementation of an electrolysis system with dc/dc synchronous buck converter“, international journal of hydrogen energy, vol. 39, iss. 13, p. 6802-6812, 2014. [6] t. zhou, b. françois, m. el hadi lebbal, s. lecoeuche, “real-time emulation of a hydrogen-production process for assessment of an active wind-energy conversion system”, ieee transactions on industrial electronics, vol. 56, iss. 3, p. 737-746, 2009. [7] p. thounthong, b. davat, “study of a multiphase interleaved step-up converter for fuel cell high power applications”, energy conversion and management, vol. 51, iss. 4, p. 826-832, 2010. [8] a. kolli, a. gaillard, a. de bernardinis, o. bethoux, d. hissel, z. khatir, “a review on dc/dc converter architectures for power fuel cell applications”, energy conversion and management, vol. 105, p. 716-730, 2015. [9] yaow-ming chen, sheng-yu tseng, cheng-tao tsai and tsai-fu wu, “interleaved buck converters with a single-capacitor turn-off snubber”, ieee transactions on aerospace and electronic systems, vol. 40, no. 3, pp. 954-967, july 2004. [10] k. yao, y. qiu, m. xu and f. c. lee, “a novel winding-coupled buck converter for high-frequency, high-step-down dc-dc conversion”, ieee transactions on power electronics, vol. 20, iss. 5, p. 1017-1024, 2005. [11] m. ilic, b. hesterman and d. maksimovic, “interleaved zero current transition three-level buck converter”, in: proceedings of twenty-first annual ieee applied power electronics conference and exposition, p. 72-78, 2006. [12] m. ilic and d. maksimovic, “interleaved zero-current-transition buck converter”, ieee transactions on industry applications, vol. 43, no. 6, pp. 1619-1627, 2007. [13] j. wibben and r. harjani, “a high-efficiency dc–dc converter using 2 nh integrated inductors”, ieee journal of solid-state circuits, vol. 43, no. 4, pp. 844-854, 2008. [14] w. li and x. he, “a family of interleaved dc–dc converters deduced from a basic cell with winding-cross-coupled inductors (wccis) for high step-up or step-down conversions”, ieee transactions on power electronics, vol. 23, no. 4, pp. 1791-1801, 2008. [15] c. t. tsai and c. l. shen, “interleaved soft-switching coupled-buck converter with active-clamp circuits”, in: proceedings of 2009 international conference on power electronics and drive systems (peds), p. 1113-1118, 2009. [16] i. lee, s. cho and g. moon, “interleaved buck converter having low switching losses and improved step-down conversion ratio,” in ieee transactions on power electronics, vol. 27, no. 8, pp. 3664-3675, aug. 2012. [17] d. guilbert, b. yodwong, w. kaewmanee, p. phattanasak, power converters for hybrid renewable energy systems with hydrogen buffer storage: a short review, in: proceedings of 6th international conference on smart grid (ieee icsmartgrid), 2018, forthcoming. [18] t. arunkumari, v. indragandhi, an overview of high voltage conversion ratio dc-dc converter configurations used in dc micro-grid architectures, renewable and sustainable energy reviews, volume 77, 2017, pages 670-687. vittorio guida was born in palermo (italy) in 1984. after obtaining the high school diploma as “industrial expert technician specialization computer”, at the “istituto tecnico industriale statale” of palermo, he began his academic studies, at the university of palermo, in automation engineering. first, he obtained the bachelor’s degree with a thesis on the “automatic measurement system for the testing of electromagnetic compatibility in accordance with cei en 55014”. subsequently, he obtained the master’s degree with a thesis on the “adaptive control for robotic systems with nonsingular actuator matrix”. during the academic path, he carried out internships to obtain the respective qualifications. in particular, in 2011, he worked at the “molino e pastificio tomasello s.p.a.”, for the implementation of a program in matlab for image processing. in 2015, he carried out an internship at the university of california riverside on the “application of control-theoretic techniques to evaluate the difficulty of different cognitive control tasks, and design of advanced brain-computer interfaces to facilitate the use of eeg signals and to control different dynamical systems”. in may/june 2017, he attended a java programming course, organized by the qibit division (ict division of gi group s.p.a.), at their offices in rome. in october 2017, after passing an international competition, he was admitted to the ph.d. program in electrical engineering, at the university of lorraine in longwy (france) with a ph.d. thesis focused on “design and realization of a dc/dc converter at high conversion ratio for electrolyzers”. damien guilbert was born in paris, france, in 1987. he received the m.sc. degree in electrical engineering and control systems and the ph.d. degree in electrical engineering from the university of technology of belfortmontbéliard (utbm), france, in 2011 and 2014 respectively. his current research interests include power electronics, fuel cell, and electrolyzer system, modeling and emulation of pem electrolyzers, fault-tolerant dc/dc converters for fuel cell/photovoltaic/electrolyzer applications, and fault-tolerant control for fuel cell/electrolyzer systems. he was involved in the creation of the first ieee/ias student branch and chapter in france at utbm, where he was the president in 2015. since september 2016, he has been associate professor at université de lorraine and a permanent member of green (group of research in electrical engineering of nancy) laboratory. bruno douine was born in montereau, france, in 1966. he received his ph. d. in electrical engineering from the university of nancy, france, in 2001. he is currently full professor at the university of lorraine and a permanent member of green (group of research in electrical engineering of nancy) laboratory. his main topics of research concern characterization of superconducting material and power electronics.  transactions on environment and electrical engineering issn 2450-5730 vol 1, no 2 (2016) © muhammad taheruzzaman & przemyslaw janik abstract— this paper represents the overall electrical energy profile and access in bangladesh. in the recent past, bangladesh has been experiencing the shortage of electricity, and about 42 % of the population no access to the electricity. the electricity consumption has rapidly increased over last decade. the demand and consumption will intensify in the remote future as overall development and future growth. to set “vision 2021” of bangladesh; the government of bangladesh has devoted to ensuring access to affordable and reliable electricity for all by 2021. in the modern time, energy is the vital ingredient for socioeconomic growth in the developing country i.e., alleviating poverty. along with electricity access in bangladesh strived to become the middle-income country by 2021. bangladesh has experienced that energy consumption inclines to increase rapidly when per capita income reaches between us$ 1,000 and us$ 10,000, and a country’s growth momentum through reliable energy supply and consistent energy supply ensured by the sustainable energy. as increasing population in bangladesh, the electric energy generation is an important dispute through the sustainable way. index terms— energy profile, energy efficiency, electric power sector, electricity reformation, renewable energy access, solar home system i. introduction ccording to report 2012, bangladesh is the 134th ranked out of 144 countries on the quality of electricity supply, which suggests the most problematic obstacles to the further socioeconomic progress. the iea estimates approximately 1.5 billion people have no access to electricity in 2008 [1], which estimates more than 20 % of total population. according to undp report more than 96.2 million of people which is more than half the total population in bangladesh still remains without access to muhammad taheruzzaman (email: muhammad.taheruzzaman@tucottbus.de) department of energy distribution and high voltage, brandenburg technical university cottbus, 03046 cottbus, germany electricity city [2], furthermore, the irregular electric power supply causes load shedding. electric energy access is the far-way dream for many families in the rural area in developing countries, about 80 % of the population are living in the rural and remote areas in bangladesh where only 25 % of electricity available for people. overcoming the curse of poverty, sustainable economic growth by access energy is an essential prerequisite and major criterion. electricity access with a modern form of energy resources is promoting social and economic growth. it is also an indispensable contribution to achieving millennium development goal (mdg) and vision 2021. in the modern era, there is no country attained sustained economic growth without improving access to clean and modern energy; the modern form of energy delineates with an integration of locally available renewable energy sources. rural electrification ensuring with improved electricity is fundamental for socio-economic development. electrical energy access influences to the life standards, which affecting agricultural productivity, education, health. the government of bangladesh has set a noble vision to access electricity for all inhabitants by 2021, to comply the vision integrating solar pv and biomass sources which are richly endowed in bangladesh. in bangladesh, it is common about 4 6 h of power outage per day in rural areas, but summer season the number of hours rises to 6 8 h, mostly during 18:00 22:00 h irregular power outage causes load shedding. the demand for electricity increases with increasing with population but the generation of electricity is not increasing to meet the demand. at present, almost 52 % of total people in bangladesh are connected to the grid [3], the power supply from the grid is inadequate to meet both peak and basic demand in bangladesh. almost 75 % of people in rural areas are not connected to the main grid, and only 15 20 % of electric demand comply by the breb (bangladesh rural electrification board) supplied electricity [4]. due to life standards and social standards enhances, the consumption rate increased at 4.53 %, but the generation of electricity increased only at a rate of 5.37 % that increased przemyslaw (e-mail: janik@b-tu.de). faculty of electrical engineering, wroclaw university of science and technology, wroclaw, poland electric energy access in bangladesh muhammad taheruzzaman and przemyslaw janik* brandenburg technical university-cottbus, 03046 cottbus, germany *faculty of electrical engineering, wroclaw university of science and technology, wroclaw, poland a mailto:muhammad.taheruzzaman@tu-cottbus.de mailto:muhammad.taheruzzaman@tu-cottbus.de mailto:janik@b-tu.de the rate of 6.72 % load shedding per year [5], graphically present in figure 2-3. according to leap (long range energy alternative planning) project [6], rural households loads comprises with lighting, mobile charger, ceiling fan, tv, and refrigerators. in rural areas lighting are the main loads in the rural households. in 2010 rural households, consumes 300kwh per year for lighting solely satisfied by electricity supplies. the demand for lighting growing at constant 1.67 % per year to 350 kwh by 2020 [7]. a tropical country like bangladesh, where summer seasons comprises almost 9 months requires cooling by the ceiling fan, consumes 250 kwh per year and assume the consumption rate increase up to 1.9 % to 345 kwh in 2030. likewise, refrigeration consumption demand rate increase 0.93 %, the demand increases from 476 kwh to 565 kwh in 2030. the percentage of energy consumption has experienced promptly increasing about 2.69 % from 2012 to 2013, but still remains lowest per capita consumption. the studies of eia, the consumption has increased dramatically over 52 % within the past decade [8]. if your paper is intended for a conference, please contact your conference editor concerning acceptable word processor formats for your particular conference. ii. general country profile bangladesh is moving towards achieving the tag of developing country with an annual gdp almost 6 % over the last past decade [9]. recently population thriving dramatically nearly 158 million and annual growth rate of 1.39 % over the past decade [10]. the majority of them are living in the rural areas, and only 32 % of households have access to electricity, but the availability of electricity about 22 % [11]. bangladesh is one of the largest in population at 9th position in the world with 158 million people at the end of 2014, where total 52 % people have partially electricity access, while only 10-15 % of rural have the access to electricity demand mainly meets the light, ceiling fan, refrigeration, irrigation, productive uses loads. in bangladesh, the electricity demand of all sectors including agriculture, commercial service, industry, and domestic services. the domestic households and industry sectors are consuming of electrical power about 43 % and 44 % respectively in total of about 87 % [12]. the gdp growth rates significantly depend on the production of a country, as bangladesh is an agricultural and small size industrial production based country, and production always depends on electricity, the gdp growth and electricity generation growth present in figure 1. it is estimated that 1 % increase in per capita energy consumption causes an increase in per capita gdp by 0.23 %. figure 1: gdp growth rate with electricity access a. demand of electricity vs climate of bangladesh bangladesh is located between 20° to 26° north and 88° to 92° east. it is bordered on the west, north and east by india, on the south-east by myanmar, and on the south by the bay of bengal. the geographical location of bangladesh offers higher solar irradiation [13]. bangladesh enjoys generally a sub-tropical monsoon climate while there are six seasons in a year, with three being more prominent, namely winter, summer and monsoon season. winter begins in november and ends in february. in winter, there is not much fluctuation in temperature, which ranges from minimum of 7° 13 °c to a maximum of 24 °c – 31 °c. the maximum temperature recorded in the summer months is 37 °c although in some places this occasionally rises up to 41°c (105°f or more) [13]. as the temperature increases the demand for electricity has increased due to refrigeration, cooling, whereas the base load demand is higher than the electricity generation. bangladesh has three main seasons: the monsoon or wet season from late may to early october; the cold season from mid-november to the end of january; and the hot season from march to midseptember [15]; the imbalance between demand and supply due to high electricity demand for ceiling fan, refrigeration during march to august in each year. b. electric energy status and demand profile electric energy is one of the affable terms of energy which is the fundamental contingent for socio-economic development, which alleviate poverty. but, bangladesh has the major problem of the energy crisis that persisting poverty, conventional fossil fuel causes environmental degradation. merely, 49 % of the population have the access electricity that met by 4500 mw while peak demand 6000 mw causes the power outage. currently, 53 % electricity produced by public sectors and rest produced by several private sectors with various form of generation [16]. the existing available power generation 8,500 mw by october 2014 and vision set to 39,000 mw by 2030 [17]. the (table1), represents power generation from different organization and bangladesh power development board (bpdb) transmits and distributes across the country. natural gas and coal expected the main source of power generation in bangladesh, gob also attentive on liquid fuel based power generation. the conventional fuel consumption to generate electrical power and traditional power plant influenced to increase co2 emission, power generation sector alone contributes 40 % co2 emission [18]. the primary energy considered to consumption estimated 62% of biomass, 25 % of natural gas, 12 % imported oil, and coal and hydropower contribute 1 %. table 1: daily power generation company demand (mw) day peak (mw) evening peak (mw) power development board 4332.00 1767.00 2702.00 electricity generation company bangladesh ltd 622.00 0.00 0.00 ashuganj power station co. ltd 1617.00 723.00 896.00 independent power producer (private) 325.00 248.00 283.00 small size producers 1987.00 1269.00 1440.00 rental power producers 825.00 1101.00 1189.00 total generation 10390.0 5515.0 6987 in bangladesh, power sectors that highly dependent on conventional fossil fuel including gas and coal. the total capacity of electricity generation about 8,709 mw, and 62.9 % of electricity generation by natural gas present in figure 2. besides natural gas, 10 % high-speed diesel, 5 % of coal, and 3 % of heavy fuel oil used to produces electricity figure 2(a). besides natural gas, 10 % high speed diesel, 5 % of coal, and 3 % of heavy fuel oil used to produces electricity [20], and only 3.3 % of electricity contributes by renewable sources [21]. figure 2: installed electricity capacity (a) fuel type and (b) plant type [19] according to (bpdb) report expresses, 55 % of people have access electricity, and per capita 321 kwh electricity generation [22], which comparatively lower than other developing countries. access to power in bangladesh is limited to about 45 % – 50 % of the population and those who have access faces severe power shortages. load shedding in dhaka in 2011 and during the summer of 2012 was about 5 hours per day. power shortages have constrained the potential economic growth in bangladesh and cost of which have been estimated to be about 0.5 % of gdp. according to “vision 2021”; the government’s vision for the power sector is to ensure universal access to grid electricity by the year 2020, with an interim target to reach an access level of 68 % by year the 2015. according to government estimates, about 20,000 megawatts (mw) of new generation capacity need to be added to the system by 2020, together with matching transmission and distribution improvements to reach the universal access [23]. figure 3: electrification rate in different regions the total installed capacity was 5262 mw in fy 2007–08, which has increased to 8525 mw in fy 2012–13 with an annual increase of 10.34 %. however, the maximum generation was 4130 mw in fy 2007–08, which has increased to 6350 mw in fy 2012–13 with an annual increase of 8.96 %. the annual rise in maximum generation (8.96 %) is lower than that of the installed capacity (10.34 %) between the fy 2007–08 and 2012–13. this is mainly due to the less generation capacity of older power plants and shortage of gas supply. 0,00% 20,00% 40,00% 60,00% 80,00% 100,00% 120,00% total rural urban (a) (b) yearly electricity demand (anticipated) 0 10000 20000 30000 40000 2011 2013 2015 2017 2019 2021 2023 2025 2027 2029 table 2: different fuel consumption gas diesel hydro coal furnace 4822 mw 186 mw 230 mw 250 mw 335 mw 82.81% 3.19% 3.95% 4.29% 5.75% though attribution is difficult, this technical assistance may have played a role in supporting a ‘balanced development’ of the power sector, which during the project period (2004-2013) saw an increase in electricity access from 35 percent to about 62 %; an increase in generation capacity from 3,622 mw in 2004 to 9,500 mw; a reduction of systems losses from about 20.0 percent to 1.3 percent; and a drop in accounts receivable from 6.45 months to 2.21 months. about 40 % of electricity generated by private enterprises by april, 2010 while the number has been increased to 44 % by april 2011. currently, rental, quick rental and some others peaking plants were under taken on a first track based power generation to manage present power crisis. according to the power system master plan (psmp), the peak demand anticipated 10,283 mw in 2015, whereas total power generated about 12071 mw. the anticipated peak demand 25199 mw anticipated in 2020 and 33708 mw in 2030 show in figure 5. c. infrastructure of bangladesh power development first bangladesh power development board (bpdb), is the sole authority to delivered electricity to the national grid through a common transmission line, to meet the national demand bpdb produces and purchases electricity from independent power producers (ipps). the five authorities contributes together to produces electricity in bangladesh: (i) bangladesh power development board (bpdb) (ii) ashuganj power station company ltd. (apscl) (iii) electricity generation company of bangladesh (egcb) (iv) north west power generation company (v) independent power producers (ipps) table 3: authorities of power generation and capacities and market share name of authorities capacity (mw) market share (%) bangladesh power development board (bpdb) 4442 42.75 ashuganj power station company ltd (apscl) 682 6.56 electricity generation company of bangladesh 622 5.98 north west power generation company ltd 375 3.06 independent power producers (ipps) 4269 41.08 total 10390 100 considering country size and population, bangladesh electricity infrastructure are quite smaller than other countries which is insufficient and poorly managed by several authorities including bpdb, bpdc, desco and reb. amongst all these authorities, reb is one of the most success government company since 1977 in bangladesh, 40.10 % electricity purchased to electrifying rural areas. table 4: share of electricity distribution by authorities authority bpdb dpdc desco wzpdc reb share (%) 24.64 18.59 10.51 6.17 40.10 bangladesh power system including transmission system comprises along with 16 substations capacity of 230/132 kv besides that 103 substations dimensions of 132/33 kv substations, which total capacity of power contains 7525 mva and 11892 mva respectively. the distribution network comprises 33 kv, 11 kv, and 400 v [27]. iii. rural electrification south asia accounts for 37 % of the world's population without access to electricity [28]. such a situation continues to exist despite several initiatives and policies to support rural electrification efforts by the respective country governments including the use of renewable energy technologies including pv, wind, and biomass. the pace of rural electrification over much of the developing world is excruciatingly slow. in many countries in south asian and sub-saharan african, it is even lower than rural electrification growth in bangladesh. bringing the socioeconomic development into the development countries like bangladesh, the essential elements considers rural electrification [29], development of underprivileged rural people [30] [31]. demand for electricity with an improvement of living standard, agricultural production, community development in bangladesh. energy access through rural electrification level still not sufficient enough, but the impressive shs growth and off-grid pv system in 0 2000 4000 6000 8000 10000 2007-08 2008-09 2009-10 2010-11 2011-12 2012-13 installed capacity (mw) maximum gemeration (mw) figure 4: installed capacity and generation 2007-2013 bangladesh. development and implemented by idcol (infrastructure development company limited). figure 3: electric energy access electrification rate in rural areas still poor as only 38 % of households is electrified [11], idcol (bangladesh government owned agency) with other 30 partners organization (pos) working together for improving the access of electricity around rural areas. despite of continuous efforts from the international community and governments, the pace of rural electrification still very slow [34]. the bangladesh rural electrification program (brep) clearly expresses which benefit greatly from the involvement of local communities improve electricity access in rural areas. according to the vision 2021; gob aims at 100 % access to electricity to entire rural areas by 2020, connecting over 0.7 million consumers and only 3 % of electricity supplied by the reb, the dedicated government organization, rest of can be supplied by the including private company and partner organization (pos). the process of rural electrification in developing countries, which depends on various factors; (1) the result of pre-phase economic and social impact (2) development of pbs (local partner) (3) technically and financially power system (4)available funding from international; community there is the main process of electrical access in rural areas centralized approach and decentralized approach; centralized approached constituted by government and partner stakeholders. in bangladesh reb and bps are the main organization for rural electrification. the decentralized approach formulated by both top-down and bottom-up concept, standalone pv system, shs, and renewable integrated hybrid mini-grid the best example in bangladesh. the approach follows up and development of rural electrification in bangladesh considered; (1) extending and intensifying the central grid (2) deploying off-grid technologies (off grid mini-grid, standalone mg, bottom up swarm electrification) to implement the rural electric cooperative concept in bangladesh, a central statutory agency called the rural electrification board (reb) was formed by the government. the reb was given the responsibility of organizing the rural electric cooperatives (palli bidyut samity, pbs); it employed managers to oversee the financial and administrative activities of the cooperatives. according to the world bank manifesto, to bring most of the people electrifying under project “rural electrification and renewable energy development” which mainly deployed by pv system [43]. a. features of rural electrification before 1977, the government-owned power development board (pdb) was the sole organization providing electricity throughout the country, without there being any special emphasis on rural areas. this actually left rural areas a very little chance to get access to electricity, and so, given this situation, the country launched the rural electrification program (rep), which exclusively targets rural areas. the features of rural electricity in bangladesh characterized by low voltage loads and distributed medium voltage lines. the power supply is unreliable and about 6 to 8 hours per day and phase imbalance. average rural electric loads from 5 kw to 20 kw per village, and load factor around 0.2 to 0.3 (average demand/maximum demand). the load consumption in the households in rural areas are predominantly lighting, agricultural pumping, and mobile charge. the grids in a rural region often weak and high peak demand during evening lighting and summer agricultural pump. to implement the rural electric cooperative concept in bangladesh, a central statutory agency called the rural electrification board (reb) was formed by the government. the reb was given the responsibility of organizing the rural electric cooperatives pbs (palli bidyut samity); it employed managers to oversee the financial and administrative activities of the cooperatives. figure 4: typical household load profile b. electric energy consumption profile in the modern epoch, electricity is the fundamental infrastructural input for economic development. electricity is the flexible form of energy that drives development factors including industrialization, extensive urbanization, and intensification of living standards and modernization of agricultural sector. in bangladesh electricity is a major source of energy to meet the industrial and agricultural sector, both of these sectors contribute to 50.3 % of country’s gdp [35]. historically, bangladesh is standing at overwhelmingly electricity generation by natural gasbased. according to the estimation of iea, 1,400 mw electricity generation from 400 million cubic feet of natural in each day (iea, 2014). in bangladesh, natural gas supplied for consumption from two sources; state owned petrobangla, which contribute 99.4 % and international oil companies (iocs) which account for 0.5 % of total supply. customer category unit price (tk/kwh)* category a: residential life line: from 1 to 50 unit 3.33 first step : from 1 to 75 unit 3.80 second step : from 76 to 200 unit 5.14 third step : from 201 to 300 unit 5.36 fifth step: from 401 to 600 units 8.70 sixth step: above 600 units 9.98 category b: agricultural pumping 3.82 category-c : small industries flat rate 7.66 off-peak time 6.90 peak time 9.24 category d: non-residential 5.22 category e: commercial and office flat rate 9.80 off-peak time 8.45 peak time 11.98 80 tk= 1 us $ iv. renewable energy penetration in bangladesh according to iea energy access to comply the rural electrification, household having reliable and affordable electricity to clean cooking facilities, first electricity connection, and increasing level of electricity consumption over time as regional average. bangladesh is the most potential country for renewable energy, significantly increases the number projects to meet the electrical energy throughout the country. the most existing form of renewable energy experienced in bangladesh considering pv based off grid system including shs, nano-grid, and mini-grid, where biomass also have high portentous to integrated significantly. with increasing both life and social standards urbanization is rapidly growing in developing countries, as comply urbanization growth electricity demand also increases promptly in bangladesh. gob has set target about 90 % electricity access across the country by 2018 [36] , to meet this vision innovative rural electrification integrated renewable energy is the best solution followed by the recent experiences, and achieving the target 2018 by connecting 450,000 households per months by 66 % shs, and hybrid power system with renewable sources. although bangladesh is the seven largest natural gas producer country among asia, about 56% of gas consumption as the primary source of energy. as high dependency on natural gas, and experiences shortage of gas supply. the regular peak demands 5500 mw, but only 4000 mw of electricity produced by the conventional power generation system in 2007 that causes rolling electricity blackout. remote areas and rural villages are the major mechanisms of holistic society; the development of socioeconomy and environmental prominence in bangladesh depends on productivity, and the productivity depends on access to energy. but the true reality is the government of bangladesh not frequently involves for rural development including rural electrification due to some geographical constraints. in figure 7 represents, the electricity access increasing rapidly from 2000 to 2015. figure 5: change in access to electricity, 1990-2015 electrifying in rural areas by conventional electrification system is expensive due to households are situated scattered and remote, and consumption rate low compare to urban electrification. hence, no-electrified remote areas and poor villages electrifying by the conventional basis not promoted and focused. consequently, it is urgent for the development of social life in bangladesh by the availability of a reliable, adequate, and reasonably priced source of energy that uninterrupted balance of electricity supply. many countries and cities have already moved towards low carbon and clean energy transformations. such as in germany, for instance, is undertaking the ‘energiewende’, an economic watershed that aims to produce 80 % of its electricity from renewable by 2050 [37]. harnessing clean, renewable, and more efficient energy solutions will contribute not only to tackling a country’s or community’s energy challenges but also to the target of limiting global temperature rise to two degrees celsius. as it is, a significant amount of ghg emissions are generated from energy production, thus tying sustainable energy directly to the climate change negotiations. bangladesh today faces a different future than it did decades ago when abundant natural gas seemed to be the key to prosperity. at the same time as the centralized grid-based electrification has been the most common approach, decentralized renewable energy options especially, pv(photovoltaic) systems has also been adopted, especially for areas where it is technoeconomically not feasible to extend the electricity grid. these off-grid communities are generally small, consisting of low-income households with characteristics that may have been economically unattractive to electricity distribution companies to extend the grid. small-scale renewable energy options, such as a solar home system (shs) and biogas plants, have evolved as promising alternative for providing electricity to these disperse areas [38]. other renewable energy options, such as wind energy and hydropower, have little potential to contribute to rural electrification in bangladesh. among the renewable technologies, the shs option has accounted for the major share (80 %) of off-grid technologies in bangladesh [39] [40] [41]. bangladesh started its intensive rural electrification program in 1977 when only 10 % of its total population was connected to a grid. the country adopted a rural electric cooperative (rec) concept from the national rural electric cooperative association (nreca), which had successfully electrified rural america in the 1930s [42]. according to the world bank manifesto, to bring most of people electrifying under project “rural electrification and renewable energy development” which mainly deployed by pv system [43]. figure 6: institutional development for off-grid program amongst 49 partners’ organization, idcol has developed a competitive market for solar pv system without any geographic constraints by offering solar incentives; shs installation, pv system with battery and charge controller supplies across the country [44]. achieving quality and reliability of electricity supply is an important factor for each region, enhancement of reliability factor in integrating intermittent renewable energy like solar and the wind no choice except diesel generators, issues highlighted by (foster and steinbuks, 2009), estimates power system that generators owned compensated by 6 % of total capacity in sub-saharan africa and other low-income countries up to 20 % [45]. renewable electrification inspiring by the institutional framework in bangladesh present in figure 8. since renewable energy emerging in the power system of bangladesh, the capacity gained 78 mw until 2012 which about 95 % of solar energy [46]. to comply the master plan, targeting 30 million of population electrified by off-grid system by 2016 which is about 18 % of the total rural population, whereas the number was about 15 million in 2013. a. biomass potential it is proved that bangladesh has significant potential in biomass and biogas. bangladesh is a tropical monsoon region, and agricultural is the main income for people who are living in the rural areas. agricultural waste provides an enormous amount of biomass resources’ assimilate with animal waste, household waste, and msw which utilized to produce a large scale of electricity. biomass generation system offers a number of advantages, mainly sources in low cost but high in energy efficiency compare to other fossil fuel, which reduces fuel costs. besides electricity generation, biomass waste also affords fertilizer simultaneously. in bangladesh gas is the main source of electricity production, according [47] about 88.8 % electricity generated by domestic gas, and a big part of electricity generation from imported furnace oil. in bangladesh, from agriculture produces rice, wheat, maize, coconut, vegetables, jute, sugarcane, etc. about 46 % biomass energy sources from rice, straw, rice, husk, jute stick, sugarcane [48]. most of the households in bangladesh produces their vegetables and summer and winter accounted 48.16 % and 51.84 % respectively in the year 2011 [49]. power generation from biomass gasification is reasonably novel in bangladesh and favorable technology. electricity generation by biomass gasification can be solved our day to day problem at an immense scope. eventually, the purpose of rural electrification which is the expression of grief need of bangladesh. in addition to producing electricity, it is advantageous to the agricultural and industrial expansion and production. it is almost impossible without rural electrification to meet the bangladesh government vision of ensuring access to reliable and affordable electricity for energy security-2020. biomass and natural gas are the major sources of energy in bangladesh, whereas 70 % biomass energy consumption of total energy consumption [39]. biomass encompasses of agricultural residues in bangladesh mainly rice, maize, wheat, coconut, groundnut, bean, vegetables, jute, and sugarcane etc. about 46 % of total biomass energy has produced from agricultural crop residues. rice is the main agricultural crop, and 70 % of rice husk energy is consumed. at present, ngos are promoting small scale biomass system for clean cooking and electricity generation. there are two minor projects which supported by idcol those generating 200-300 kw by using poultry litter, moreover, the studies also suggested that up to 800 mw electricity by poultry waste litter. at present 15.00 tons of poultry litter produced each day, and a small fraction being used recycle. about 47 tons of waste expected, will be produced in 2025. in bangladesh another available but significant raw material for biomass production rice husk, several search has shown that up to 400 mw of electricity can be generated single-handedly by rice husk. b. photovoltaic potential bangladesh is blessed with enormous solar potential, as solar insolation. the average solar energy incident from 4 kwh/m2/day to 6.5 kwh/m2/day, with average 10.5 solar hours and about 300 clear sunny days. by the combination of a solar cell in pv module, under standard test condition (stc) module produces dc electricity at range 100 w to 400 w. in (figure 9) shown, clear bright sunlight, except june and july, average 7 to 9 h operates rest 10 months to produces solar energy. in figure 2-10, represents monthly average solar irradiation in different regions in bangladesh. figure 7: solar irradiation of different areas in bangladesh c. solar home system in bangladesh solar sources and shs has experienced a great success in bangladesh, particularly the improvement of rural electrification. currently, about 42 % of people have access electricity and per capita consumption of electricity is about 133 kwh in 2005 [52], which is the lower comparatively other developing countries. nevertheless, the imbalance power supply makes a big difference between demand and supply, which makes load shedding. started early 1980, pv flourished across the country and the success factors focus on; (i) rural areas electrified which are not yet accessible into the main utility grid. (ii) remote areas where electricity access is almost impossible. (iii) insufficient power supply. . figure 8: idcol shs program and growth rate [56] shs generated electricity mainly used in rural households’ loads including low power devices, cfl or led lights, tv, mobile charger [53]. bangladesh annual variation of inclination of the sun, measured from the vertical varies from 0 to 46 degrees between the summer and the winter. summer days are longer, around 14 hours, with average sunshine more than 6kw-hr/day on clear sunny day. although winter days are shorter around 10 h, still there is more than 4.5kw-hr/day of insolation on a clear sunny day. solar home system (shs) are stand-alone photovoltaic systems that offer a cost effective mode of supplying power for lighting and appliances to remote off-grid households. in remote areas, which are not connected to the grid; shs can be used to meet remote household’s energy demand. in bangladesh, shs usually at a rate of 12 v dc and provide power for low power dc appliances including lights, tv, mobile charger, for about four to five hours. in developing countries like bangladesh, where the national grid extension is not economically and technically feasible, an array of pv cells is used to build shs. the main components of shs include a solar panel, battery and a charge controller which can be operated with minimum training [54]. over the past decade, since the bangladesh government launched a rural electrification program supported by world bank and other international aid bodies, the number of off-grid installations in the country has rocketed. in 2002, installations rates stood at 7000; today the figure has exploded to nearly 2 million and continues to count, with average installation rate now topping 80,000 per month [55]. idcol with other partner organization financed by world bank 3357609 shss established until october 2014, and the numbers increase intensely present in (figure 2-11). the capacity achieved by shs about 150 mw in the year 20132014, and growth rate increases about 185 %from the previous year. in 2015 the growth rate increases to 300 % and capacity raised 234 mw electricity generation potential from shss [57]. generally distance between shs about 2 to 2.5 meter, where most of the system capacity configured with 60 wp. as shown in (figure 2-11), shs program promoted to increases more than 3.7 million by may 2015 [56], about 98 % of shs installed through idcol [58], and additional 70,000 shs being installed every month, and targeting more than 6 million more shs by 2016 [59]. v. innovation approach for rural electrification to achieved the millennium development goals (mdg), electrification across nationwide is one of the main topology widely believed contribution, renewable sources deploy to sustainable development which leads to improvement of environment and fosters of socio-economic life. in the modern time, only 11 % of people have the access electricity in the sub-saharan countries [60], whereas in bangladesh about 40 % of households have the access electricity [61] and the improvement rate of electricity through shs system and bottom-up swarm electrification successfully experienced in bangladesh past decades. the households and communities are far away from the main grid and grid extension are not always cost effective due to infrastructure and insufficient power supply. figure 11: word wide electricity access through rural electrification [62] according to the authors’ of [63] suggested, dc microgrid configured by several distributed generation such as shs and from a local grid that might connect to the main grid. a mini-grid can be configured by local distributed generation system and the distributed generation sources’ considering along with renewable resources such as pv, biomass, wind. according to swarm electrification concept, neighboring households are assimilating in an intelligent network where scheme allows sharing their information about supply, demand, and battery status within. to achieve this network by sharing electricity among participants within the scheme, consequently swarm network have the ability to integrate with legacy based where participants have the ability to produce electricity and consumption simultaneously, in order to propagate without or with limited number single centralized unit which has the ability to function independently may be called nano-grid. it is obverses that a sunny day an shs in bangladesh does not utilized their own capacity respect to their lords connected within the system, and 30% surplus electricity available for others [64]. tier based swarm concept explain in figure-13 and figure-14, tier-1 represents an shs configuration and the loads consumption, self-generated electricity from pv panel. tier -2 and tier -3 countenance shs and bhs connected and formed a dc cluster, and tier -4 cluster grid also allow to connect to the grid to sellback surplus electricity. the major strategies for rural electrification to access electricity for all, some studies expressed only about 30 % of rural areas electrified by the centralized grid, whereas 70 % people can be electrified by the small scale nano-grid or microgrid [65]. figure 2: swarm electrification concept and stepwise approaches [66] vi. reforms and policies towards renewable energy declining the fossil fuel along with natural gas, the electricity production reduces whereas demand increases day by day. gob has restricted and privatized the electricity generation sector by national energy policy (nep) in 1996. the major target of the policy to increase the power generation to meet the desires present and future demand which adopted by following policies: i. harnessing solar potential, and dissemination of ret in both urban and rural areas ii. enable and encourage facilitate public and private sector investment towards re projects iii. development of sustainable energy system to substitute non-renewable sources iv. facilitating renewable energy at every level of energy including households to commercial and industrial the national energy plan (nep) envisions 5 % pf total renewable generation from renewable sources, and by 2020 achieved by 10 % energy from renewable. bangladesh power development board (bpdb) imposed the bulk tariff for electricity consumption for distribution companies including dhaka electric supply company (desco), dhaka electric supply authority (desa), west zone power distribution company (wzpdc), dhaka power distribution company (dpdc), and rural electrification (reb). the distribution companies are working in the urban areas and reb with 77 rural electric cooperatives palli bidyut samity (pbs) working for electrification in villages and remote areas. vii. conclusion it is clear that most of the countries including low-income and developing countries gdp affected by the level of energy consumption, and per capita 0.23 % gdp increases by consuming 1 % of per capita energy consumption. the growth rate of electricity has increased by 5.5 % in the fiscal year 2006-2007, which rapidly increased to about 13.2 % in the fiscal year 2012-2013. likewise, the gdp of bangladesh has increased at the rate of 6.8 % in the fiscal year 20122013 from 2006-2007 observed at rate 6.15 %. bangladesh is the fast growing developing country, socio-economic, industrialization, other development booming while demanding of electricity increases day by day. currently, power sector of bangladesh produces 7,445 mw by 2012, and 8002 mw by 2016 along with different government entities and non-government company working together to meet the electricity demand. almost 72.42 % of total electricity generated from natural gas in the fiscal year 2013-2014, and on the other side, the renewable penetration only about 2.5 % which is the insignificant comparison to global power generation. in the present time bangladesh is one of the market leader of shs, and standalone pv system. in bangladesh average 4 to 6.5 kwh/m2 solar irradiation, and maximum amount of solar radiation is available almost each month except december-january, however, 300 high sunny days suggested solar generated system like standalone pv system, and shs. idcol and other 47 partner organizations (pos), ngo working together to installing 3 million shs by 2013 and targeting almost 7 million by mid of 2018. the conventional power system is expensive to configure and present demand is lagging behind from the continuous power supply to electrification, especially for electrifying rural and remote areas. notwithstanding, the conventional trends to generates electrical power from the top-down grid, and author convinced to follow up the concept of bottom-up swarm electrification would be the best solution for electrifying rural areas in developing countries. a robust grid can be formed amongst hybrid power system which configures with integrating distributed renewable sources and the backup diesel generator that highly efficient and reliable in the remote areas. currently, about 55.41 % of rural areas electrified by reb and cooperative organization pbs, whereas 5.05 million households connected to the grid. yet 45 % of rural areas not electrified by reb which government owned company, but idcol and others pos working together to achieve millennium development goad (mdg) and “vision 2021” simultaneously, about 94 % households decreases about 1.7 liters of fuel (kerosene) consumptions compare to those not connected to the grid, average 90,000 households connected to the grid. during summer, the number of new households slightly increased to 300,000, and to achieve 100 % of electrification about 450,000 new households need to connect to the grid by 2018. by the successful shs program along other biomass integration, and enrichment of electric power generation bangladesh has achieved almost 11000 mw electricity by 2014, but still 40 % of population living without access to electricity. viii. references [1] iea, “addressing the electricity access gap,” world energy outlook 2009, paris, 2010. [2] undp, “human development report,” 2008. [online]. available: http://hdrstats.undp.org/indicators/211.html. [accessed 29 feb 2016]. [3] h. k. alimul, r. anisur and m. moniruzzaman, “source management of a hybrid dc micro grid in rural area of bangladesh,” international conference materialsm electronics and information engineering, isbn 978-98433-8940--4, 2015. [4] h. najmul and k. d. barun, ““analysis of cost,energy and co2 emission of solar home systems in bangladesh,” international journal of renewable energy research , 2012. [5] bd-research, “electricity senario in bangladesh,” 2012. [online]. available: 〈http://www.bdresearch.org.bd/home/attachments/articl e/799/electricity%20scenario%20in%20bangladesh.pdf. [accessed 28 feb 2016]. [6] m. mondol, w. boie and m. denich, “future demand scenarios of bangladesh power sector,” energy policy report 38, 2010. [7] h. ahsan and c. supachart, “utilization of solar and biomass for rural electrification in bangladesh,” in international conference and utility exhibition , pattaya, thailand , 2014. [8] regional energy security for south asia: regional report , “south asian regional initiative for energy integration,” 2014. [online]. available: http://www.sarienergy.org/. [accessed 30 oct 2015]. [9] asian development bank, “key indicators for asia and the pacific,” asian developlment bank report, 2011. [10] jika, “bangladesh power system master plan (psmp),” tokyo electric power co. ltd., tokyo, 2010. [11] bbs, “statistical yearbook of bangladesh,” bangladesh bureau of statistics, governtment of bangaldesh, dhaka, 2003. [12] bpdb, “power generation report,” bangladesh power developmenet board , http://www.bpdb.gov.bd, 2013. [13] iaea, “country power profile edition,,” available: http://www-pub.iaea.org/, 2013, [accessed on 27 june, 2015]. [14] l. chanda and g. hossain, “a green economy in the context of sustainable development and proverty eradication: what are the implications for bangladesh?,” journal of economics and sustainable development, vol. 5, no. 3, pp. 119-131, 2014. [15] r. plunkett, a. newton and b. j. wagenhauser, bangaldesh, victoria, autralia : lonely planet publications pty ltd. , 2000. [16] n. hosnay, “acquisition of sustainable economic growth through proper utilization of renewable energy sources – a study on various aspects, challenges andprospects of re in bangladesh,” nub, et-p11, dhaka , 2013. [17] generation-report, “owner wise daily generation report,” 2014. [online]. available: http://www.bpdb.gov.bd/bpdb/index.php?option=com_c ontent&view=article&id=193&iption=com_content&vie w=article&id=193&i. [accessed 12 dec 2016]. [18] a. khairul and a. husnain, “power crisis & its solution through renewable energy in bangladesh cyber journals: multidisciplinary,” journals in science & technology. journal of selected areas in renewable and sustainable energy , 2011. [19] bari,mn; hall, do; lucas, njd; et. al., “biomass energy use at the ousehold level in two villages of bangladesh: assessment of field methods,” biomass energy , vol. 15, no. 2, p. 171–180, 1998 . [20] a. report, “annual report of 2012-2013,” bangladesh power development board (bpdb), dhaka, bangladesh . [21] asif, islam; maidul, islam; sanchita, ghose; et al, “a review on electricity generation and evacuation in bangladesh,” energy and power: academic research, 10.5923/j.ep.20150501.02, dhaka, bangladesh, 2015. [22] b. report-2014, “daily power generation report-,” bangladesh power development board, dhaka, bangladesh, 2014. [23] h. gunatilake and d. ronald-holst, “energy policy options for sustinable development in bangladesh,” adb economics working paper series , pp. 9-12, november 2013. [24] r. mustafizur, “electricity scenario in bangladesh,” unnayan onneshan-the innovators , dhaka, 2011. [25] bpdb, “power generation report,” angladesh power development board (bpdb), 2015. [online]. available: http://www.bpdb.gov.bd/bpdb/. [accessed 12 dec 2015]. [26] power-division, “bangladesh energy status report,” power division. ministry of power, energy and mineral resources. government of the peoples republic of the bangladesh, 2014. [online]. available: ttp://www.powerdivision.gov.bd/user/index. [accessed 12 dec 2015]. [27] annual-report, “power grid company of bangladesh (pgcb) 2012-2013,” pgbc, dhaka, bangladesh , 2014. [28] d. palit, “solar energy programs for rural electrification: experiences and lessons,” elsvier: energy for sustaiable development , no. 17, p. 270–279 , 2013. [29] b. douglas and f. gerald, “rural electrification in the development world: a summery of lessons from successful programs,” world bank energy sector management assistance programme (esmap) , washington dc, 2004. [30] d. palit and a. chaurey, “off-grid rural electrification experiences from south asia: status and best practice,” energy for sustainable development, pp. 266-267, december 2011. [31] w. bank, “financing for rural electrification and renewable energy development project: bangladesh (project paper number 64347-bd),” word bank , washington dc, 2010. [32] bbs, report of the household income & expenditure survey, bangladesh bureau of statistics: dhaka, bangladesh , 2010. [33] power-division, “bangladesh power sector: an overview,” bangladesh power development board , september 2015. [online]. available: http://www.bpdb.gov.bd/bpdb/. [accessed 12 feb 2016]. [34] c. paul, “infrastructure, rural electrification and development.,” energy for development , no. 15, pp. 304312, 2011. [35] m. mahedi, “electricity consumption and economic growth,” global journal of management and business research, vol. 12, no. 11, 2012. [36] srep-bangladesh, “scaling up renewable energy in low income countries (srep): investment plan for bangladesh,” sustainable & renewable energy development authority , dhaka, bangladesh , 2015. [37] m. martinez solaiman , “inter press service,” news agency , 08 july 20015. [online]. available: http://www.ipsnews.net/2015/05/opinion-clean-energyaccess-a-major-sustainable-development-goal/. [accessed 8 07 2015]. [38] world bank , “designing sustainable off-grid rural electrification projects:,” the world bank, washington dc, usa , 2008. [39] i. akm sardul, i. mazharul and r. tazmilur, “effective renewable energy activities in bangladesh,” renewable energy, no. 31, pp. 677-688, 2006. [40] m. z. rahman, “multitude of progress and unmediated problems of olar pv in bangladesh,” renewable and sustainable energy reviews, no. 15, pp. 466-473, 2012. [41] m. m. rahman and j. v. paatero, “a methodological approach for assessing potential of sustainable agricultural residues for electricity generation: south asian prospective,” biomass and bio energy , no. 47, pp. 153-163, 2012. [42] nreca, “guides for electric cooperative development and rural electrification,” nreca international, ltd , washington, dc, usa, 2004. [43] e. cecelski, “[10]. cecelski,e, “enabling equitable access to rural electrification: urrent thinking and major activities in energy,” world bank, washington dc, 2000. [44] ieg-report, “rural electrification and rnewable energy development,” workd bank, dhaak, bangladesh , 2014. [45] produse, “the impact of electricity access on economic development: a literature review,” deutsche gesellschaft für internationale zusammenarbeit (giz) gmbh, frankfurt, germnay, 2013. [46] idcol, “renewable energy status in bangladesh,” infrastructure development company limited , 2014. [online]. available: http://www.idcol.org/prjshsm2004.php. [accessed 12 dec 2016]. [47] bbs, “statistical yearbook of bangladesh 2008,” bangladesh bureau of statistics; 2009, dhaka, bangladesh , 2009. [48] m. islam, m. islam and m. beg, “renewable energy resources and technologies practice in bangladesh.,” renewable sustainable energy rev , vol. 12, p. 299–343, 2008. [49] a. huda, a. mekhilef and a. ahsan, “biomass energy in bangladesh: current status and prospects,” renewable and sustainable energy reviews, vol. 30, p. 504–517, 2015. [50] m. motin, m. zaman and m. sheikh, “energy efficient modelling of solar wind hybrid power system for a tourist island,” int j adv renew energy res (ijarer), vol. 1, no. 1, 2013. [51] hasan, am; khan, mf; dey, a; et al, “feasibility study on the available renewable sources in the island of sandwip, bangladesh for generation tion of electricity,” in: proceedings of the international conference on developments in renewable energy technology (icdret), p. 1–4, 2012. [52] psmp, “power system master plan update,” ministry of power, energy and mineral resources, dhaka, bangladesh , 2006. [53] r. paleta, a. pina and c. silva, “remote autonomous energy systems project: towards sustainability in developing countries.,” energy, vol. 48, no. 1, pp. 431-9, 2012. [54] s. islam, “photovoltaic technology for bangladesh,” in bangladesh university of engineering & technology, dhaka and center for renewable energy system technology, leicestershire, infield, 2001. [55] b. willis, “friday focus: how bangladesh become the world's biggest off-grid solar user,” pv-tech, 05 july 2013. [online]. available: http://www.pvtech.org/friday_focus/friday_focus_how_bangladesh_bec ame_the_worlds_biggest_domestic_off_grid_pla. [accessed 22 july 2015]. [56] idcol, “solar home system program and projects,” 2014. [online]. available: http://idcol.org/home/solar. [accessed 30 dec 2015]. [57] bpdb, “power divsion of bangladesh,” bangladesh, renewable energy in bangladesh , 2015. [online]. available: http://www.powerdivision.gov.bd/user/brec/49/89. [accessed 02 feb 2016]. [58] khandker, s.r; samad, h.a; sadeque, z.k; et al, “surge in solar-powerd homes experience in off-grid rural bangladeseh,” the world bank, washington dc, usa, 2014. [59] m. islam, “idcol solar home system program,” 2014. [online]. available: https://www.esmap.org/sites/esmap.org/files/esmap_s ar_eap_renewable_energy. [accessed 30 dec 2015]. [60] gunther, bensch; jochen, kluve; peters, jörg, “rural electrifi cation in rwanda–an impact assessment using matching techniques,” ruhr economic paper , dec 2010. [online]. available: http://en.rwiessen.de/media/content/pages/publikationen/ruhreconomic-papers/rep_10_231.pdf. [accessed 01 march 2016]. [61] ibrd-ida, “lighting up rural communities in bangladesh,” the world bank, 15 jan 2014. [online]. available: http://www.worldbank.org/en/news/feature/2014/01/15/li ghting-up-rural-communities-in-bangladesh. [accessed 02 mar 2016]. [62] ecowas, “ecowas renewable energy and energy efficiency status report,” ecowas, 2014. [63] sarker, m.j; asare-bediako, b; slootweg, j.gp; et. al, “dc microgrid with distributed generation for rural electrification,” in 47th international conference on universities power (upec), london, 2012. [64] k. hannes, “hidden resources in solar home systems as the base for bottom-up grids. proc. of the int. conf. innovating energy access for remote areas: discovering untapped resourc,” in proc. of the int.conf. innovating energy access for remote areas: discovering untapped resources, berkeley, usa, 2014 . [65] oecd/iea, “energy for all: financing access to the poor,” world energy outlook, paris, france , 2011. [66] martina, schafer; daniel, kammen; noara kebir; et al, “innovating energy access for remote areas: discovering untapped resources,” in proceedings of the international conference, university berkeley , berkeley, 2014. i. introduction ii. general country profile a. demand of electricity vs climate of bangladesh b. electric energy status and demand profile c. infrastructure of bangladesh power development iii. rural electrification a. features of rural electrification b. electric energy consumption profile iv. renewable energy penetration in bangladesh a. biomass potential b. photovoltaic potential c. solar home system in bangladesh v. innovation approach for rural electrification vi. reforms and policies towards renewable energy vii. conclusion viii. references  transactions on environment and electrical engineering issn 2450-5730 vol 4, no 1 (2021/2022) © umair shahzad  abstract—transient stability assessment is an integral part of dynamic security assessment of power systems. traditional methods of transient stability assessment, such as time domain simulation approach and direct methods, are appropriate for offline studies and thus, cannot be applied for online transient stability prediction, which is a major requirement in modern power systems. this motivated the requirement to apply an artificial intelligence-based approach. in this regard, supervised machine learning is beneficial for predicting transient stability status, in the presence of uncertainties. therefore, this paper examines the application of a binary support vector machinebased supervised machine learning, for predicting the transient stability status of a power system, considering uncertainties of various factors, such as load, faulted line, fault type, fault location and fault clearing time. the support vector machine is trained using a gaussian radial basis function kernel and its hyperparameters are optimized using bayesian optimization. results obtained for the ieee 14-bus test system demonstrated that the proposed method offers a fast technique for probabilistic transient stability status prediction, with an excellent accuracy. digsilent powerfactory and matlab was utilized for transient stability time-domain simulations (for obtaining training data for support vector machine) and for applying support vector machine, respectively. index terms—artificial intelligence (ai), dynamic security assessment (dsa), machine learning, probabilistic transient stability, support vector machine (svm). i. introduction he reliability of a power system is defined as the ability of the power system to provide electric energy to consumers on a continuous basis and with acceptable service quality, for both planned and unplanned outages [1]. one of the main requirements to maintain the reliability of the power is to continually operate the synchronous generators and with satisfactory capacity to satisfy the load. in the domain of power system stability, transient stability is the ability of the synchronous machines to remain in synchronism during few seconds (usually 5-10 seconds), after a large disturbance, such as a short circuit fault, occurs [2]. in addition to fault type, fault location, system inertia, and system load, fault clearing time (fct) and critical clearing time (cct) are significant parameters in assessing transient stability [3]. fct is the time u. shahzad is with the department of electrical and computer engineering, university of nebraska-lincoln, lincoln, ne, 68588-0511 usa (email: umair.shahzad@huskers.unl.edu). at which fault is cleared after fault occurrence, whereas cct is the maximum fct after which the system becomes transiently unstable [3]. transient stability is an integral component of dynamic security assessment (dsa). dsa deals with evaluation of transient performance of the system after the occurrence of contingency [2]. to evaluate the transient stability status, time-domain simulation approach is generally used to solve a set of nonlinear differential-algebraic equations, which represent the dynamics of the network [4]. although, the time-domain approach is the most accurate method and usually yields promising results, it is time consuming, as it must traverse a set of differential-algebraic equations, which can be computationally intensive, especially for large-scale systems [5]. the transient energy function (tef) method [6]-[7], and the extended equal area criterion (eeac) [8] have also been applied to transient stability evaluation; however, these approaches have some restrictions regarding modeling, and they require many computations to evaluate an index for transient stability status [9]. moreover, these methods are not appropriate for online transient stability prediction [10]. conventionally, deterministic methods, employing worstcase scenarios (seasonal peak load, three-phase fault, etc.), have been used to evaluate transient stability [11]-[13]. these approaches are very conservative and ignore the probability of various input parameters linked with transient stability, such as load, fault type, fault location, etc. with the continuous integration of renewable generation, the increasing prevalence of competitive electricity market, and the rising uncertainties in the power system, these conventional methods are becoming obsolete and unsuitable. compared with the deterministic assessment approaches, probabilistic assessment techniques can provide an inclusive and realistic measure of system stability status [14]. novel probabilistic assessment techniques are desirable and are being established. probabilistic transient stability (pts) assessment has been recognized to be a fitting approach to analyze the effect of uncertain parameters on transient stability [15]–[21]. additionally, the results from pts analysis can be associated with risk assessment, which is imperative for system operators, as economic and technical reasons can result in the power system to operate near the stability limit [21]. although, it has been long established that deterministic studies may not sufficiently characterize the full extent of system dynamic behavior, the probabilistic approach has not been extensively used in the past in power system studies, prediction of probabilistic transient stability using support vector machine umair shahzad t 2 mainly due to lack of data, limitation of computational resources, limited commercial softwares for probabilistic analysis, deterministic nature of standards enforced by regulatory authorities, such as north american electric reliability corporation (nerc), and mixed response from power utilities and planners [17], [20]. however, in recent times, there has been some research in pts. for instance, [22] presented an analytical approach to determine pts for online applications. [23]-[24] used monte-carlo simulation (mcs) approach to present a stochastic-based approach to assess the pts index. [15] proposed the inclusion of probabilistic considerations in transient stability investigation of a multimachine practical power system. some other relevant work can be found in [25]-[30]. a major drawback of these works is that these use conventional numerical and analytical methods, such as, mcs, time-domain simulation, eeac, tef, hybrid method, etc. to estimate the transient stability index. these approaches may be suitable for an offline study, however, for real-time online prediction, a faster method is required. artificial intelligence (ai)-based approaches provide a good alternative to fulfil this vital objective. among various ai approaches, machine learning (ml) is an upcoming approach for solving power system problems, including transient stability [31]-[33]. ml is generally classified into three categories: supervised, unsupervised and reinforcement [34]. in supervised learning, the goal is to learn a mapping relation between the inputs to outputs, based on a given a labeled set of input/output pairs. unsupervised learning deals with the training of an algorithm is using unlabeled data so that the algorithm may group the data based on similarity or difference. in reinforcement learning (rl), there is an interaction of an agent with its environment and consequently, the agent adapts its course, based on the reward because of its actions. the focus of this research work is supervised machine learning (sml). although, sml has various types [35], such as artificial neural network (ann), decision tree (dt), random forest, support vector machine (svm), etc., this work focuses on svm-based sml for prediction of pts. in recent years, application of ml algorithms, such as ann, to power system is an area of rising interest; the chief reason being the ability of ann to process and learn intricate nonlinear relations [36]. although ann is the most commonly used ml method for transient stability classification, it generally requires an extensive training process and an intricate design procedure. moreover, ann usually performs well for interpolation but not so well for extrapolation, which reduces its generalization ability. they are more susceptible to becoming trapped in a local minimum. although, majority of ml algorithms can overfit if there is a dearth of training samples, but anns can also overfit if training goes on for a very long duration [37]. due to these downsides, it becomes essential to develop a more efficient classifier for transient stability status prediction. svm do not suffer from these drawbacks and has the following advantages, over ann [38]: (1) less number of tuning parameters, (2) less susceptibility to overfitting, and (3) the complexity is depended on number of support vectors (svs) rather than dimensionality of transformed input space. support vector machine (svm) is an evolving ml approach that incapacitates some of the drawbacks of ann. a svm essentially is a sml algorithm that can use given data to solve certain problems by trying to convert them into linearly separable problems. recently, svm has been applied to power system transient stability classification problem. an svmbased transient stability classifier was trained in [39] and its performance was compared with a multilayer perceptron (mlp) classifier. reference [37] devised a multiclass svm classifier for static and transient stability assessment and classification. reference [40] suggested a svm classifier to predict the transient stability status using voltage variation trajectory templates. reference [38] trained a binary svm classifier, with combinatorial trajectories inputs, to predict the transient stability status. reference [41] employed the svm to rank the synchronous generators based on transient stability severity and classify them into vulnerable and nonvulnerable machines. reference [42] proposed two svms, using gaussian kernels, for classifying the post-fault stability status of the system. reference [43] presented an svm-based approach for transient stability detection, using postdisturbance signals, from the optimally located distributed generations. some other relevant work dealing with svmbased transient stability classification can be found in [44][51]. based on the detailed literature review and to the best of author’s knowledge, there exists no research work on pts which uses svm-based sml approach, considering the uncertainties of load, faulted line, fault type, fault location (on the line), and fct. moreover, [52] specifically mentions the potential of svm for online dsa, and [53]-[57] strongly indicate that ml is a promising and upcoming approach for online dsa. thus, the main contribution of this paper is to predict pts status using an svm-based sml approach. as mentioned before, although, the time-domain simulation method is one of the most accurate methods to assess transient stability; it is very computationally intensive, particularly for large scale systems. hence, this method is only suitable for offline applications. on the contrary, the direct analytical methods are comparatively fast, but requires a large number of approximations which significantly limit the model accuracy. the tef-based methods are difficult to implement, especially due to many potential function terms of the tef of the system. also, these approaches require postfault data for transient stability assessment, and hence, they are not suitable for online transient stability assessment. therefore, new approaches must be explored and applied to real-scale power systems to ensure accurate and effective online prediction of transient stability. thus, this paper proposed an artificial intelligence-based approach for this application, and this is the main novelty and contribution of this research. the rest of the paper is organized as follows. section ii describes various probabilistic factors associated with transient stability assessment. section iii discusses the pts index used in this paper. section iv provides a brief overview of svm and its application to pts classification problem. 3 section v discusses the procedure for the proposed approach. section vi and vii deals with the description of case study, and associated results and discussion, respectively. section viii describes sensitivity analysis, with respect to some important parameters/functions. finally, section ix concludes the paper, with a suggested direction for future research. ii. probabilistic factors in power system transient stability there are various factors which are involved in pts assessment of power systems, such as fault type, fault location, load, and fct. suitable probability density functions (pdfs) are used to model these factors. the modeling approaches are described below [58]. normally, shunt faults, such as three-phase (lll), double-line-to-ground (llg), lineto-line (ll) and single-line-to-ground (lg) short circuits, are considered for evaluating pts. a probability mass functions (pmf) is normally used to model the fault type. based on past system statistics, a usual practice is to select the probability of lll, ll, llg, and lg short circuits, as 0.05, 0.1, 0.15 and 0.7 respectively [22]. this paper adopts the same practice. the probability distribution of fault location on a transmission line is usually assumed to be uniform. this means that the fault can occur with equal probability at any line of the test system and at any point along the line [21]. this paper uses the same approach. the procedure of fault clearing constitutes of three stages: fault detection, relay operation and breaker operation. if the primary protection and breakers are 100% reliable, the clearing time is the only uncertain factor. a normal (gaussian) pdf is generally used to model this time [21]. in this paper, fault is applied at 1 s and it is cleared, after a mean time of 0.9 s and standard deviation of 0.1 s (based on the normal pdf). a normal pdf was used to represent the uncertainty of loads. let f(x) denote the pdf for individual bus loads, i.e., 2 2 2 2 ( ) 1 ( ) 2 x f x e       (1) where  and  denotes the mean and standard deviation of the forecasted peak load, respectively. iii. quantification index for probabilistic transient stability the transient stability index (tsi) was used to quantify the transient stability of a system consisting of synchronous machines [59]. this index is based on the maximum rotor angle separation between any two synchronous machines, after the fault has occurred. mathematically, it is given by max max 360 , where 1 1 360 i i i itsi tsi         (2) where max i is the post-fault maximum rotor angle separation (in degrees) between any two synchronous machines in the system at the same time (for a fault on i th line). a negative tsi value indicates that the power system is transiently unstable. this is a global index for a swift indication of the transient stability status of the system (for a fault on any line, at any point, for any fct and for any load). therefore, this index is used in this paper to quantify the pts status. let si represent the pts status indicator for i th iteration of mcs. mathematically, 1 , if < 0 0, if 0 i i i tsi s tsi     (3) therefore, if the system is transiently stable, for i th montecarlo (mc) sample, value of si will be 0; otherwise, it will be 1. this information will be used for training the svm model. iv. svm: brief overview and application to probabilistic transient stability prediction support vector machine (svm), which is also known as maximum margin classifier, is a type of sml, that can be used both in classification and regression problems. it was first introduced by vapnik [60]-[61] and was elaborated by schölkopf et al. [62]. svm classifiers depend on training points, which lie on the boundary of separation between different classes, where the evaluation of transient stability is important. a decent theoretical progress of the svm, due to its basics built on the statistical learning theory (slt) [60], made it possible to develop fast training methods, even with large training sets and high input dimensions [63]-[65]. this useful characteristic can be applied to tackle the issue of high input dimension and large training datasets in the pts problem. the basic implementation of an svm, commonly known as a hard margin svm, requires the binary classification problem to be linearly separable. this is frequently not the case in practical problems, and therefore, svm provides a kernel trick to resolve this issue. the forte of the svm algorithm is based on the use of this kernel trick to transform the input space into a higher dimensional feature space. this permits to define a decision boundary that linearly separates the classes. the svm algorithm attempts to determine that decision boundary or hyperplane with the highest distance from each class [38], [66]. the hyperplane can be mathematically defined as [39] ( ) 0 t w x b  (4) where w is the weight vector ( t w is its transpose), x is the sample feature vector, and b is a bias value. the samples that assist the algorithm to define the optimal hyperplane are those that lie closest to it, and they are known as svs. the kernel function plays a significant role in svm classification [67]. the kernel function is applied on each data instance to map the original non-linear data points into a higher-dimensional space in which they become linearly separable. an svm classifier minimizes the generalization error by optimizing the relation between the number of training errors and the socalled vapnik-chervonenkis (vc) dimension. this is attained using the approach of structural risk minimization (srm) which states that the classification error expectation of unseen data is bounded by the sum of a training error rate and a term that depends on the vc dimension [39]. compared to empirical risk minimization (erm)-based formulation (which is used by most ml algorithms, including ann), the srm 4 based formulation allows the svm to prevent overfitting problems, by defining an upper bound, on the expected risk. a formal theoretical bound exists for the generalization ability of an svm, which depends on the number of training errors (t), the size of the training set (n), the vc dimension associated to the resulting classifier (h), and a chosen confidence measure for the bound itself ( ) [39], [61], [68]: 2 (ln( ) 1) ln( ) 4 n h t hr n n      (5) the risk (or classification error expectation) r represents the classification error expectation over all the population of input/output pairs, even though the population is only partially known. this risk is a measure of the actual generalization error and does not require prior knowledge of the probability distribution of the data. slt derives inequality (5) to mean that the generalization ability of an svm is measured by an upper limit of the actual error given by the right-hand side of (5), and this upper limit is valid with a probability of 1- (0<  <1). as h increases, the first summand of the upper bound (5) decreases and the second summand increases, such that there is a balanced compromise between the two terms (complexity and training error), respectively [39]. the svms used for binary classification problems are based on linear hyperplanes to separate the data, as shown in fig. 1. the hyperplane (represented by dotted line in fig. 1) is determined by an orthogonal vector w and a bias b, which identify the points that satisfy ( ) 0 t w x b  . by determining a hyperplane which maximizes the margin of separation, denoted by  , it is instinctively anticipated that the classifier will have an improved generalization ability. the hyperplane having the largest margin on the training set can be completely determined by the points that lie closest to the hyperplane. two such points are x1 and x2 as shown in in fig. l (b), and they are known as svs because the hyperplane (i.e., the classifier) is completely dependent on these vectors. consequently, in their simplest form, svms learn linear decision rules as ( ) ( ) t f x sign w x b  (6) so that (w, b) are determined as to correctly classify the training examples and to maximize  . for linearly separable data, as shown in fig. 1, a linear classifier can be found such that the first summand of bound (5) is zero. fig. 1. svm (maximum margin) classifier. it is always possible to scale w and b such that 1 t w x b   (7) for the svs, with 1 t w x b   and 1 t w x b   (8) for non-svs. using the svs x1 and x2 of fig. 1 and (7), the margin  can be calculated as 2 1 2 ( ) || || || || t w x x w w     (9) where || ||w is the euclidean norm of w. for linearly separable data, the vc dimension of svm classifiers can be evaluated as   2 2 2 2 4 min , 1 min , 1 d h n n d w           (10) where n is the dimension of the training vectors and d is the minimum radius of a ball which contains the training points. thus, the risk (5) can be reduced by lessening the complexity of the svm, that is, by increasing the margin of separation  , which is equivalent to reducing || ||w . in practice, as the problems are not probable to be detachable by a linear classifier, thus, the linear svm can be extended to a nonlinear version by mapping the training data to an expanded feature space using a nonlinear transformation: 1( ) ( ( ),......, ( )) m mx x x r    (11) where m > n. then, the maximum margin classifier of the data for the new space can be determined. with this method, the data points which are non-separable in the original space may become separable in the expanded feature space. the next step is to approximate the svm by minimizing (i.e., maximizing  ) 1 ( ) . 2 t v w w w (12) subject to the constraint that all training patterns are correctly classified, i.e., { ( ) } 1, 1,..., t i iy w x b i n     (13) though, contingent on the kind of nonlinear mapping (11), the samples of training data may not be linearly separable. in this case, it is not possible to find a linear classifier that satisfies all the conditions given by (12). thus, instead of (12), a new cost function is optimized, i.e., 1 1 min ( , ) . 2 . . { ( ) } 1 for 1,..., 0 for 1,..., n t i i t i i i i v w w w c st y w x b i n i n                 (14) 5 where n non-negative slack variables i are introduced to allow training errors (i.e., training patterns for which { ( ) } 1 t i i iy w x b      and 1i  ) and allow for some misclassification. by minimizing the first summand of (14), the complexity of the svm is reduced, and by minimizing the second summand of (14), the number of training errors is decreased. c is a positive penalty factor (also known as regularization factor or soft margin parameter) which decides the tradeoff between the two terms. in case it is small, the separating hyperplane is more focused on maximizing the margin (at the expense of larger classification mistakes), while the number of misclassified points is minimized for larger c values (at the expense of keeping the margin small). the minimization of the cost function (14) leads to a quadratic optimization problem with a unique solution. the nonlinear mapping (11) is indirectly obtained by the kernel functions, which correspond to inner products of data vectors in the expanded feature space ( , ) ( ) ( ), , t n k a b a b a b r    [39], [46], [68]. common kernel functions include the linear, polynomial, sigmoid and gaussian radial basis function (rbf). in general, there is no fixed criterion for selecting these kernel functions. it majorly depends on whether the data is linearly separable or not, and how many dimensions exist when the number of features is very large (depending on the data), dimensionality reduction is applied first using principal component analysis (pca) or linear discriminant analysis (lda) (linear or nonlinear kernel variants). in general, the rbf kernel is a reasonable first choice. in short, there is no way to figure out which kernel would do the best for a particular problem. the only way to choose the best kernel is to actually try out all possible kernels, and consequently, choose the one empirically performs the best. one can empirically determine the optimal kernel via experimentation. doing so involves three major steps: (1) implementing a version of the svm model using each kernel, (2) evaluating the svm model’s performance with each kernel via cross validation, and (3) selecting the kernel that yielded optimal results. the gaussian rbf kernel generally is preferred over others because it has the ability of mapping samples nonlinearly into a higher dimensional space, and therefore, unlike linear kernel, it can tackle the scenario when the relationship between class labels and attributes is nonlinear. although, sigmoid kernel performs like a gaussian rbf kernel for certain parameters, but there are some parameters for which the sigmoid kernel is not the dot product of two vectors, thus, it is invalid. moreover, as compared to polynomial kernel, it has few hyperparameters (parameters whose values are used to control the learning process) [61]. thus, this work uses a gaussian rbf kernel, which is mathematically given by, 2 ( ) 2 1 ( , ) , 0, 2 a b k a b e          (15) where denotes the kernel parameter of the svm classifier and  is the width of the gaussian function. the hyperparameters c and  impact how sparse and easily separable the training data are in the expanded feature space. subsequently, these parameters decide the complexity and training error rate of the resulting svm classifier. these parameters must be optimized for achieving the best performance for the svm classifier. the block diagram for the proposed svm framework is shown in fig. 2. the proposed svm framework used has four inputs (system load, fault type, fault location and fct), and one output (for si). samples for training data were chosen using the mcs-based time domain simulation approach (described in section v). fig. 2. framework for the proposed svm approach (input features and corresponding output). for the pts classification task, the first step was feature extraction, i.e., to select the most relevant input and output data for the svm classification model. system load, fault type, fault location, and fct were chosen as inputs, and transient stability status, si, was selected as the output (the binary variable to be classified as transiently stable or unstable). 500 samples were used for each line to train the svm model, as shown in table i. it must be mentioned that generally, there is no accepted rule of thumb to determine the number of samples for training the ml model; this typically depends on complexity of the problem, required performance level, and the ml algorithm used. as there are 16 lines in the system, thus, the total number of samples used for svm model were 8000 ( 500 16 ). thus, the size of the input feature matrix was 8000 4 . the gaussian rbf kernel function was used for training the svm as there is ample nonlinearity amongst the data presented to the svm classifier. the hyperparameters c and  were optimized using bayesian optimization (other approaches such as grid search or random search may also be used). the optimum values of c,  , and  were found to be 210, 0.22, and 1.5, respectively. the data presented to svm is randomly divided in two subsets: training subset and testing subset. the k-fold crossvalidation approach is used to accomplish this as this prevents over fitting while training the data. in this approach, the entire data is divided into k partitions of equal size. training and testing are repeated, each time selecting a different partition for testing data, until all k partitions are utilized for testing, i.e., every data point gets to be in a test set exactly once and gets to be in a training set (k-1) times [38]. eventually, the average of these errors is taken as the expected prediction error. this work used the value of k as 5, i.e., in each fold, 20% data was used for testing and 80% for training. https://en.wikipedia.org/wiki/cross-validation_(statistics) https://en.wikipedia.org/wiki/cross-validation_(statistics) 6 v. procedure for the proposed approach the methodology for the proposed approach is described in fig. 3. the ieee-14 bus system was used to test and validate the proposed approach. this system has 16 transmission lines. for each line, 500 random mc samples were generated (the symbol i indicates the sample number for the mcs). it is assumed that pre-fault system topology (configuration) is fixed, i.e., there is no contingency before the fault occurrence. in the first step, the first line is selected. in the next step, mcs is initiated with 500 samples. in each sample, system load, fault type, fault location and fct are randomly chosen (based on the respective defined pdfs, as described in section ii). the fault is created at time t=1 s. for each mc sample, timedomain stability simulation is run for 10 s to determine the outcome (transiently stable or unstable). this is determined based on the value of si, as described in section iii. these steps are repeated and mcs is performed (for 500 samples) for all the remaining lines. when the mcs is run for all the 16 lines in the network, the resulting data obtained is used as training data for the svm classification model. a summarized workflow of sml application for online pts prediction is shown in fig. 4. as illustrated, the first step deals with the offline mode. in this mode, time-domain simulations are conducted, considering the uncertainties of input variables in the form of pdfs (generally obtained from past historical observations). in the next step, these distributions are sampled to gather enough training data. for each sample, the pts status is measured by a binary variable, say, x, which can take two labels (say, 1 for transiently unstable, and 0 for transiently stable). therefore, the final training data consists of the pts status labels and the corresponding input operating conditions. in the next step, this offline-based database is used for online pts prediction. the sml model ‘learns’ the stability rules and consequently, can be used to predict the pts status for current operating point. vi. case study the ieee 14-bus test transmission system was used to conduct the required simulations. the numerical data and parameters were taken from [69]. the single line diagram is shown in fig. 5. it should be highlighted that the proposed methodology is applicable to any test system. as mentioned before, a normal pdf is used to define the uncertainty in system loads. the active power of each load was assigned a mean equal to the original load active power value, as given in test system data in [69], and a standard deviation equal to 10% of the mean value. all time-domain simulations are rms simulations and are performed using digsilent powerfactory software [70]. for sml application, classification learner tool of matlab was used [71]. table i selected features for each line feature name number of samples system load 500 fault type 500 fault location 500 fct 500 fig. 3. flowchart for the proposed svm approach. fig. 4. proposed sml approach for online pts prediction. 7 fig. 5. ieee 14-bus test system. vii. results and discussion it is assumed that required accuracy of the classifier must be more than 95%. to quantify the performance of the trained svm classifier, the confusion matrix was used. this matrix is a graphical representation of the number of samples predicted correctly and incorrectly. the confusion matrix obtained for pts classification is shown in fig. 6 (1 and 0 represents transiently unstable and transiently stable class, respectively). classification accuracy (ca) is a commonly used classification performance metric [72]. it is calculated as the proportion of correct predictions from the total number of the data points. the ideal value of ca is 1, whereas the worst is 0. it is mathematically defined as tp tn tp tn ca tp tn fp fn n        (16) where tp (true positives), tn (true negatives) denotes the correctly predicted data, and fp (false positives), fn (false negatives) denotes the incorrectly predicted data. n denotes total data points (which are 8000 in this paper). similarly, classification error (ce) represents the number of incorrect predictions from the total number of the data points. the closer it is to zero, the better. mathematically, fp fn fp fn ce tp tn fp fn n        (17) a receiver operating characteristic (roc) curve is a graphical plot that establishes the diagnostic ability of a binary ml classifier [73]. in this plot, the true positive rate (sensitivity) is plotted against the false positive rate (1specificity). sensitivity is a measure of actual positives which are correctly identified, whereas specificity is the proportion of truly negative cases that were classified as negative [74]. a classification svm model with perfect discrimination has a roc plot that passes through the upper left corner (100% sensitivity, 100% specificity), i.e., its area under curve (auc) is equal to 1. the closer the auc is to 1, the greater the classification accuracy. the confusion matrix and the roc curve (for testing data), for the classification of si, are shown in fig. 6 and fig. 7, respectively. from fig. 6, it is evident that ca for the confusion matrix is very high, i.e., approximately 97% (59.46% + 37.23%). moreover, as evident from fig. 7, the roc curve is very accurate (auc >0.99). the values of various classification metrics are summarized in table ii. as evident, values for ca and auc are in the desired high accuracy range (>0.95), and ce is quite small (0.033). once trained, the svm classifier can be directly used to classify si. the training time for the svm classifier was only 0.03 s. thus, it can be inferred that the trained svm algorithm can rapidly classify the pts status, si, with a high accuracy (≈97%). this makes it suitable for an online application and therefore, can drastically help power system operators in the control center for decision-making tasks. thus, to sum up, the proposed svm approach can be used to predict the pts status, incorporating various uncertain factors (system load, faulted line, fault type, fault location, and fct), with a superior accuracy. this approach has an edge over the conventional approaches, as it is computationally efficient, as well as, fairly accurate. it is strongly believed that the proposed approach can drastically contribute to progressing the prevailing methods for online dsa. it must be mentioned that the proposed ml algorithm is system-specific and, although, it performed quite well for the ieee 14-bus system, it is not assured that it will perform the same for other systems. therefore, ample testing and validation of the proposed approach must be conducted on other standard test systems, before reaching a generic conclusion on the performance of ml algorithm. additional generic limitations exist for ml-based approach, for instance, the training database and ml model must be updated when the pdfs of the input random variables, and the network topology varies over time, and consequently, the number of transient stability simulations required for training may be greater than that estimated for a fixed topology. an additional limitation regarding svm is that it is sensitive to noise (target classes overlap) and outliers (target classes deviate significantly from the rest of the classes), and consequently, does not give a good performance. moreover, choosing the optimal kernel function is not straightforward and may require several optimization simulations [75]. also, the best ml approach may change depending on the application [55]. an avid reader can refer to [76] for further details. viii. sensitivity analysis as mentioned before, the value of k used in this work was 5. to verify that it is indeed the best value, a sensitivity analysis was performed. the svm classifier was trained for various values of k, and the corresponding ca values were determined. the results obtained are shown in fig. 8. as evident, increasing k beyond 5 does not alter the ca. hence, k=5 is a good choice for k-fold cross-validation, for this work. this also validates the fact that k=5 and k=10 are https://en.wikipedia.org/wiki/graph_of_a_function 8 generally the most commonly used values for a k-fold crossvalidation procedure [77]. moreover, for k=5, the values of ca, ce, and auc, for different kernel functions, are shown in table iii. as evident, gaussian rbf kernel has the highest accuracy and minimum error. this also validates the reason of gaussian rbf kernel being the most commonly used kernel function for svm classification [61]. also, table iv displays optimal values of various hyperparameters obtained for the proposed svm model. table v displays the comparison of performance metrics for the proposed approach with other related research. as evident, the results obtained by the proposed approach are comparable to similar research, and hence, this validates its effectiveness for the desired application of transient stability status prediction, in the presence of uncertainties. table vi presents a comparison of the proposed method with conventional approaches, in terms of computational performance. as evident, the proposed svm method is quite fast in predicting the transient stability status. hence, the approach is very useful for online application. moreover, various recent research [78-80] has indicated the significance of using svm for transient stability prediction. fig. 6. confusion matrix for transient stability classification performance assessment. fig. 7. roc curve for transient stability classification performance assessment. fig. 8. variation of ca with k. ix. conclusion and future work power system transient stability is an integral part of power system planning and operation. traditionally, it has been assessed using deterministic approach. with the increasing system uncertainties, environmental pressures of incorporating green energy, and widespread electricity market liberalization (deregulation), there is a strong need to incorporate probabilistic analysis in transient stability evaluation. moreover, conventional approaches (direct method, timedomain simulation method, tef approach, etc.) to assess transient stability are time consuming and hence, are not suitable for online application. ml can provide a good alternate to achieve this important goal. hence, this paper applies an svm-based approach to predict transient stability status, in the presence of uncertainty. the paper highlighted the need to consider a faster method for pts assessment and hence, proposed a binary svm approach for predicting pts status. in addition to uncertain system load conditions, various uncertain factors such as faulted line, fault type, fault location and fct were considered. time-domain simulations were used to gather the data required for training the svm model. the tsi was used as the indicator for the pts status. the proposed method was applied to the ieee 14-bus system, and promising results were obtained, indicating the significance of svm in power system pts assessment. the results indicated that the proposed approach predicted the pts status with an excellent accuracy, in a computationally efficient manner. this indicates the potential of svm for online dsa, especially for large-scale power systems. as a future work, ensemble learning, incorporating multiple learning methods, can be applied for prediction of pts. moreover, incorporating data from renewable energy generation sources (such as wind and solar) in the svm training model can prove to be very useful in online dsa procedure. table ii svm performance assessment using various classification metrics classification metric value ca 0.967 ce 0.033 auc 0.991 9 table iii variation of ca, ce, auc for different kernel functions kernel function ca ce auc linear 0.872 0.128 0.894 polynomial (order 2) 0.916 0.084 0.937 polynomial (order 3) 0.829 0.171 0.848 gaussian rbf 0.967 0.033 0.991 table iv optimal hyperparameter values for the proposed svm model hyperparameter type/value kernel function rbf penalty factor, c 210 gamma,  0.22 sigma,  1.5 table v comparison of svm performance metrics with related research approach type ca ce proposed in this paper 0.967 0.033 [38] 0.893 0.107 [45] 0.959 0.041 [72] 0.968 0.032 [81] 0.935 0.065 table vi comparison of proposed svm method with traditional methods approach type time (s) svm (proposed in this paper) 0.005 time-domain 1.26 eeac 0.91 tef 0.86 hybrid (time-domain+tef) 1.04 references [1] m. p. bhavaraju, r. billinton, r. e. brown, j. endrenyi, w. li, a. p. meliopoulos, and c. singh, “ieee tutorial on electric delivery system reliability evaluation,” in proc. ieee power energy soc. gen. meet., 2005, pp. 68-78. [2] p. kundur, j. paserba, v. ajjarapu, g. andersson, a. bose, c. canizares, n. hatziargyriou, d. hill, a. stankovic, c. taylor, t. v. cutsem, and v. vittal, “definition and classification of power system stability ieee/cigre joint task force on stability terms and definitions,” ieee trans. power syst., vol. 19, no. 3, pp. 1387-1401, aug. 2004. [3] a. m. a. haidar, m. w. mustafa, f. a. f. ibrahim, and i. a. ahmed, “transient stability evaluation of electrical power system using generalized regression neural networks,” appl. soft comp., vol. 1, no. 4, pp. 3558-3570, jun. 2011. [4] a. l. bettiol, a. souza, j. l. todesco, and j. r. tesch, “estimation of critical clearing times using neural networks,” in proc. ieee bologna pow. tech conf., 2003, pp. 1-6. [5] a. karami, “power system transient stability margin estimation using neural networks,” int. j. electr. power energy syst., vol. 33, no. 4, pp. 983-991, may 2011. [6] m. a. pai, energy function analysis for power system stability. boston, ma, usa: kluwer, 1989. [7] a. a. fouad and v. vittal, power system transient stability analysis using the transient energy function method. upper saddle river, nj, usa: prentice-hall, 1992. [8] y. xue, t. v. cutsem, and m. ribbens-pavella, “extended equal area criterion justifications, generalizations, applications,” ieee trans. power syst., vol. 4, no. 1, pp. 44-52, feb. 1989. [9] l. d. colvara, “stability analysis of power systems described with detailed models by automatic method,” int. j. electr. power energy syst., vol. 31, no. 4, pp. 139-145, may 2009. [10] h. sawhney and b. jeyasurya, “a feed-forward artificial neural network with enhanced feature selection for power system transient stability assessment,” elect. power syst. res., vol. 76, no. 12, pp. 1047-1054, aug. 2006. [11] j. g. slootweg and w. l. kling, “impacts of distributed generation on power system transient stability,” in proc. ieee power energy soc. gen. meet., 2002, pp. 862-867. [12] a. m. azmy and i. erlich, “impact of distributed generation on the stability of electrical power system,” in proc. ieee power energy soc. gen. meet., 2005, pp. 1056-1063. [13] a. khosravi, m. jazaeri, and s. a. mousavi, “transient stability evaluation of power systems with large amounts of distributed generation,” in proc. intl. uni. power eng. conf., 2010, pp. 1-5. [14] x. zhao and j. zhou, “probabilistic transient stability assessment based on distributed dsa computation tool,” in proc. pmaps., 2010, pp. 685690. [15] r. billinton and p. r. s. kuruganty, “probabilistic assessment of transient stability in a practical multimachine system,” ieee trans. power app. syst., vol. pas-100, no. 7, pp. 3634–3641, jul. 1981. [16] y.-y. hsu and c. chung-liang, “probabilistic transient stability studies using the conditional probability approach,” ieee trans. power syst., vol. 3, no. 4, pp. 1565–1572, nov. 1988. [17] e. vaahedi, w. li, t. chia, and h. dommel, “large scale probabilistic transient stability assessment using b.c. hydro’s on-line tool,” ieee trans. power syst., vol. 15, no. 2, pp. 661–667, may 2000. [18] p. m. anderson and a. bose, “a probabilistic approach to power system stability analysis,” ieee trans. power app. syst., vol. pas-102, no. 8, pp. 2430–2439, aug. 1983. [19] k. j. timko, a. bose, and p. m. anderson, “monte carlo simulation of power system stability,” ieee trans. power app. syst., vol. pas-102, no. 10, pp. 3453–3459, oct. 1983. [20] w. li, “probabilistic transient stability assessment,” in risk assessment of power systems: models, methods, and applications. new york, ny, usa: wiley-ieee press, 2014. [21] p. n. papadopoulos and j. v. milanović, “probabilistic framework for transient stability assessment of power systems with high penetration of renewable generation,” ieee trans. power syst., vol. 32, no. 4, pp. 3078-3088, jul. 2017. [22] m. abapour and m. haghifam, “probabilistic transient stability assessment for on-line applications,” int. j. electr. power energy syst., vol. 42, no. 1, pp. 627-634, nov. 2012. [23] j. fang, w. yao, j. wen, s. cheng, y. tang, and z. cheng, “probabilistic assessment of power system transient stability incorporating smes,” phy. c: supercond., vol. 484, pp. 276-281, jan. 2013. [24] s. o. faried, r. billinton, and s. aboreshaid, “probabilistic evaluation of transient stability of a power system incorporating wind farms,” iet ren. power gen., vol. 4, no. 4, pp. 299-307, jul. 2010. [25] s. o. faried, r. billinton, and s. aboreshaid, “probabilistic evaluation of transient stability of a wind farm,” ieee trans. energy convers., vol. 24, no. 3, pp. 733-739, sep. 2009. [26] r. billinton and p. r. s. kuruganty, “probabilistic considerations in transient stability assessment,” canad. electr. eng. j., vol. 4, no. 2, pp. 26-30, apr. 1979. 10 [27] l. shi, s. sun, l. yao, y. ni, and m. bazargan, “effects of wind generation intermittency and volatility on power system transient stability,” iet ren. power gen., vol. 8, no. 5, pp. 509-521, jul. 2014. [28] d. han, j. ma, a. xue, t. lin, and g. zhang, “the uncertainty and its influence of wind generated power on power system transient stability under different penetration,” in proc. int. conf. power syst. tech., 2014, pp. 675-680. [29] m. al-sarray, h. mhiesan, m. saadeh ,and r. mccann, “a probabilistic approach for transient stability analysis of power systems with solar photovoltaic energy sources,” in proc. greentech, 2016, pp. 159-163. [30] p. ju, h. li, c. gan, y. liu, y. yu, and y. liu, “analytical assessment for transient stability under stochastic continuous disturbances,” ieee trans. power app. syst., vol. 33, no. 2, pp. 2004-2014, mar. 2018. [31] d. j. sobajic and y. pao, “artificial neural-net based dynamic security assessment for electric power systems,” ieee trans. power app. syst., vol. 4, no. 1, pp. 220-228, feb. 1989. [32] m. djukanovic, d. j. sobajic, and y. pao, “neural-net based unstable machine identification using individual energy functions,” int. j. electr. power energy syst., vol. 13, no. 5, pp. 255-262, oct. 1991. [33] y. pao and d. j. sobajic, “combined use of unsupervised and supervised learning for dynamic security assessment,” ieee trans. power app. syst., vol. 7, no. 2, pp. 878-884, may 1992. [34] m. s. ibrahim, w. dong, and q. yang, “machine learning driven smart electric power systems: current trends and new perspectives,” appl. energy, vol. 272, pp. 1-19. aug. 2020. [35] r. saravanan and p. sujatha, “a state of art techniques on machine learning algorithms: a perspective of supervised learning approaches in data classification,” in proc. iciccs, 2018, pp. 945-949. [36] t. s. dillon, and d. niebur, neural networks applications in power systems. leicestershire, uk: crl publishing, 1996. [37] s. kalyani and k. s. swarup, “classification and assessment of power system security using multiclass svm,” ieee trans. syst., man, cybern a., syst., humans, vol. 41, no. 5, pp. 753-758, sep. 2011. [38] d. you, k. wang, l. ye, j. wu, and r. huang, “transient stability assessment of power system using support vector machine with generator combinatorial trajectories inputs”, int. j. electr. power energy syst., vol. 44, no. 1, pp. 318-325, jan. 2013. [39] l. s. moulin, a. p. a. da silva, m. a. el-sharkawi, and r. j. marks, “support vector machines for transient stability analysis of large-scale power systems,” ieee trans. power app. syst., vol. 19, no. 2, pp. 818825, may 2004. [40] a. d. rajapakse, f. gomez, k. nanayakkara, p. a. crossley, and v. v. terzija, “rotor angle instability prediction using post-disturbance voltage trajectories,” ieee trans. power app. syst., vol. 25, no. 2, pp. 947-956, may 2010. [41] b. p. soni, a. saxena, v. gupta, and s. l. surana, “assessment of transient stability through coherent machine identification by using least-square support vector machine”, mod. simul. eng., vol. 2018, pp. 1-18, may 2018. [42] f. r. gomez, a. d. rajapakse, u. d. annakkage, and i. t. fernando, “support vector machine-based algorithm for post-fault transient stability status prediction using synchronized measurements,” ieee trans. power app. syst., vol. 26, no. 3, pp. 1474-1483, aug. 2011. [43] p. pavani and s. n. singh, “support vector machine based transient stability identification in distribution system with distributed generation,” electr. power comp. syst., vol. 44, no. 1, pp. 60-71, nov. 2015. [44] a. e. gavoyiannis, d. g. vogiatzis, d. r. georgiadis, and n. d. hatziargyriou, “combined support vector classifiers using fuzzy clustering for dynamic security assessment,” in proc. ieee power energy soc. gen. meet., 2001, pp. 1281-1286. [45] s. ye, x. li, x. wang, and q. qian, “power system transient stability assessment based on adaboost and support vector machines,” in proc. asia-pacific power ener. eng. conf., 2012, pp. 1-4. [46] l. s. moulin, a. p. a. da silva, m . a. el-sharkawi, and r. j. marks, “neural networks and support vector machines applied to power systems transient stability analysis,” int. j. eng. intel. syst. elect. eng. comm., vol. 9, no. 4, pp. 205-211, nov. 2001. [47] n. i. a. wahab, a. mohamed, and a. hussain, “transient stability assessment of a power system using pnn and ls-svm methods,” j. appl. sci., vol. 7, no. 21, pp. 3208-3216, 2007. [48] b. d. a. selvi and n. kamaraj, “investigation of power system transient stability using clustering based support vector machines and preventive control by rescheduling generators,” in proc. ictes, 2007, pp. 137-142. [49] n. g. baltas, p. mazidi, j. ma, f. a. fernandez, and p. rodriguez, “a comparative analysis of decision trees, support vector machines and artificial neural networks for on-line transient stability assessment,” in proc sest, 2018, pp. 1-6. [50] m. arefi and b. chowdhury, “ensemble adaptive neuro fuzzy support vector machine for prediction of transient stability,” in proc. naps, 2017, pp. 1-6. [51] e. a. frimpong, p. y. okyere, and j. asumadu, “prediction of transient stability status using walsh-hadamard transform and support vector machine,” in proc. ieee pes powerafrica, 2017, pp. 301-306. [52] b. d. a. selvi and n. kamaraj, “support vector regression machine with enhanced feature selection for transient stability evaluation,” in asiapacific pow. and energy eng. conf., 2009, pp. 1-5. [53] a. sabo and n. i. a. wahab, “rotor angle transient stability methodologies of power systems: a comparison,” in ieee student conf. on res. and dev., 2019, pp. 1-6. [54] w. li and j. zhou, “probabilistic reliability assessment of power system operations,” electr. power comp. syst., vol. 36, no. 10, pp. 1102-1114, sep. 2008. [55] j. l. cremer and g. strbac, “a machine-learning based probabilistic perspective on dynamic security assessment,” int. j. electr. power energy syst., vol. 128, pp. 1-15, jun. 2021. [56] h. yuan, j. tan, and y. zhang, “machine learning-based security assessment and control for bulk electric system operation,” nrel, feb. 2020. [online]. available: https://www.nrel.gov/docs/fy21osti/76089.pdf [57] t. zhang, m. sun, j. l. cremer, n. zhang, g. strbac, and c. kang, “a confidence-aware machine learning framework for dynamic security assessment,” ieee trans. power syst., pp. 1-14, feb. 2021. [58] k. jayashree and k. s. swarup, “a distributed computing environment for probabilistic transient stability analysis,” in 16 th nat. power syst. conf., 2010, pp. 329-335. [59] p. n. papadopoulos and j. v. milanović, “impact of penetration of nonsynchronous generators on power system dynamics,” in proc. ieee eindhoven power tech., 2015, pp. 1-6. [60] v. vapnik, statistical learning theory. new york, ny, usa: john wiley & sons, 1998. [61] v. vapnik, the nature of statistical learning theory. new york, ny, usa: springer-verlag, 1995. [62] b. scholköpf, c. burges, and a. smola, advances in kernel methods— support vector learning. cambridge, ma, usa: mit press, 1999. [63] e. osuna, r. freund, and f. girosi, “training support vector machines: au application to face detection,” in proc. cvpr, 1997, pp. 1-8. [64] t. joachims, “making large-scale svm learning practical,” in advances in kernel methods support vector learning. cambridge, ma, usa: mit press, 1998. [65] n. cristianini and j. shawe-taylor, an introduction to support vector machines and other kernel-based learning methods. cambridge, uk: cambridge university press, 2000. [66] g. n. baltas, p. mazidi, f. fernandez, and p. rodríguez, “support vector machine and neural network applications in transient stability,” in proc. icrera, 2018, pp. 1010-1015. [67] c. savas and f. dovis, “the impact of different kernel functions on the performance of scintillation detection based on support vector machines,” sensors, vol. 19, no. 23, pp. 1-16, nov. 2019. [68] l. s. moulin, a. p. a. da silva, m. a. el-sharkawi, and r. j. marks, “support vector and multilayer perceptron neural networks applied to power systems transient stability analysis with input dimensionality reduction,” in proc. ieee power energy soc. gen. meet., 2002, pp. 1308-1313. [69] power systems test case archive. [online]. available: http://www.ee.washington.edu/research/pstca/pf14/pg_tca14bus.htm [70] digsilent powerfactory user manual, digsilent gmbh, 2018. [online]. available: https://www.digsilent.de/en/downloads.html [71] classification learner. [online]. available: https://www.mathworks.com/help/stats/classificationlearner-app.html http://www.ee.washington.edu/research/pstca/pf14/pg_tca14bus.htm 11 [72] n. i. a. wahab, a. mohamed, and a. hussain, “fast transient stability assessment of large power system using probabilistic neural network with feature reduction techniques,” exp. syst. appl., vol. 38, no. 9, pp. 11112-11119, sep. 2011. [73] using the roc curve to analyze a classification model. [online]. available: http://www.math.utah.edu/~gamez/files/roc-curves.pdf [74] c. sammut and g. i. webb, encyclopedia of machine learning. new york, ny, usa: springer, 2011. [75] s. karamizadeh, s. m. abdullah, m. halimi, j. shayan and m. j. rajabi, “advantage and drawback of support vector machine functionality,” in international conf. on comput., commun., and contr. tech., 2014, pp. 63-65. [76] l. duchesne, e. karangelos, and l. wehenkel, “recent developments in machine learning for energy systems reliability management,” proc. ieee, vol. 108, no. 9, pp. 1656-1676, sep. 2020. [77] j. brownlee, a gentle introduction to k-fold cross-validation, 2018. [online]. available: https://machinelearningmastery.com/k-fold-crossvalidation/ [78] c. wang and f. liu, “an improved support vector machine method for transient stability assessment in bulk power grid,” in proc. ieee sustain. power and energy conf., 2020, pp. 2423-2428. [79] q. wang, c. pang, and h. alnami, “transient stability assessment of a power system using multi-layer svm method,” in proc. ieee texas power and energy conf., 2021, pp. 1-5. [80] c. wang, s. li, c. lin, and h. chen, “transient stability assessment of power system based on enhanced projection twin svm,” in proc. ieee int. conf. on digit. twins and parallel intell., 2021, pp. 148-151. [81] e. m. voumvoulakis, a. e. gavoyiannis, and n. d. hatziargyriou, “application of machine learning on power system dynamic security assessment,” in proc. int. conf. on intell. systems appl. to power syst., 2007, pp. 1-6. umair shahzad was born in faisalabad, pakistan. in 2021, he received the ph.d. degree in electrical engineering from the university of nebraska-lincoln, usa, as a fulbright scholar. moreover, he received a b.sc. electrical engineering degree from the university of engineering and technology, lahore, pakistan, and a m.sc. electrical engineering degree from the university of nottingham, england, in 2010 and 2012, respectively. his research interests include power system security assessment, power system stability, machine learning, and probabilistic methods applied to power systems.  transactions on environment and electrical engineering issn 2450-5730 vol 1, no 3 (2016) © o. a. akinlabi and meera k. joseph  abstract— the high demand for network coverage in an indoor setting brought about the acceptance of femtocell technology as a solution using the backhaul connectivity in the existing network. the quality of signal, voice calling, internet, security and data are improved through the use femtocell at the indoor environment. here the service provider attempts to reduce their operation cost by presenting self-organizing mechanisms for optimization of the network. the remarkable part is that, femtocells improves coverage, enhances the data rate at the indoor environment. therefore, the challenges of the femtocell also known as interference deteriorates the capacity and quality performance of the whole cellular network. in this paper we simulate the bit error rate against signal behaviour at the indoor environment and we also simulate the transmitting power over signal for both macrocells and femtocells. we focus on the transmitting power that might cause interference within the cellular network. index terms— femtocells, macrocells, signal behavior, transmitting power. i. introduction ost users of mobile network demand a quality performance of service such as voice calling, data, internet and better signal within an indoor environment. the failure of macrocell in order to achieve the above mentioned has brought about femtocell technology. however, the deployment of femtocell has drawn the attention of researchers, academics and experts in telecommunication industry over macrocell for improved coverage and security purposes. thus, mobile network has become one of the daily use of the human race, such as internet, educational resource for learners, social media, advertising and business purposes. these mobile network supports voice calling and reduced the need for travel. though the signal experience poor reception, and alteration of calls because of the high demand of number of macro base station site from the service providers cannot be met due to the cost of operation. a study was carried out by the femtoforum that confirmed an increase in voice calling and data used in an indoor this manuscript was submitted in july 21, 2016 for review, accepted on 15 october 2016. this work was supported in part by the university of johannesburg, johannesburg, south africa. department of electrical and electronic engineering sciences. o. a. akinlabi is a doctoral student at the department of electrical and electronic engineering science, university of johannesburg, johannesburg, south africa, sa (e-mail: akinlabiakindeji@gmail.com). environment [1]. the deployment of femtocell technology is favorable over macrocell at the indoor environment for coverage and system capacity in a low cost manner. femtocells focuses on the quality of voice calling, data and improved signal coverage in an indoor environment for all mobile users. in line with [1], femtocell guarantees improved coverage, capacity over the internet backhaul with full operating capacity under the licensed spectrum at a low price for the end users. femtocells are small base station in an indoor environment, set up by the end user and connected through the internet, access to the mobile provide [2]-[6]. thus femtocell supports at least four to five users at the indoor environment and is applied to residential, enterprise, hot spot and metro. the values of communication such as voice calling over networks, quality of service are linked to economic stability a country. the significance of femtocells in mobile network is that it enhance quality service and improve coverage network. figure 1, illustrates the application of femtocell in a home setting and connections to the mobile provider through the broadband access. fig. 1: femtocell in a home setting [7] dr. meera k. joseph, is senior lecturer, at the school of electrical engineering, university of johannesburg, johannesburg, south africa and she leads the ict4d research group (e-mail: meeraj@uj.ac.za). optimizing signal behavior of femtocells for improved network o. a. akinlabi and meera k. joseph, university of johannesburg, south africa m the value of femtocell has contributed to the fast growth of economies and increase in mobile revenue for service providers. the market report launch by femtoforum on femtocell has shown that it increases mobile revenue [1]. in fact, it cuts down operational cost, infrastructure and maintenance cost for mobile provide. more so, it performs certain functions of which macrocell may not perform such as it guaranteed good connectivity, home security, remote control of home appliance and quality of voice experience. apart of all this, it lessens the traffic load over macrocell and increase the performance of the mobile provider. femtocell utilised the broadband for connection through the existing macrocell network for quality of voice calling, media and video streaming in which this cannot cause any problem when the provider of such broadband differs. the problem arises when they are using the same license band for the same purpose. the technical challenge is caused due to the use of the same licensed band with the existing spectrum of macrocell and ad hoc deployment of femtocell which led to interference management. but they are needed for mobile to deploy femtocell successfully for improved coverage at the indoor environment at low cost. however, the deployment of femtocells, still provides a better coverage in the network the network traffic grew widely over the last decade due to the profitable rate flat launch by femtoforum [1], and such request should be met by new mobile communication systems, as well as increasing the revenue. hence, the achievement of the wireless network will depend on the providing a broadband access for mobile user, where costs per bit error rate are low [8]. often, mobile traffic is highly demanding in homes and office environment and according to [9], more than 80% of the mobile traffic is used in an indoor environment. the new technology will offer solution for home and office, where there will be an improved reception of signals in an indoor and these promotes cost effective for the network users. the target of service provider is to satisfy the need for the high demand of mobile data in an indoor environment and also to offer an added value able service. mobile operator benefits mostly from these new technology known as femtocell in such a way that the operator has enough saving on coverage and profit, no more electricity bills and no time wasting on problems. finally, it provides broadband access point in order for a connection with a satellite backhaul, for instance, inside an airplane, complex, shopping mall, train and warship respectively. these access points are modelled to be linked to the business model. this paper is organized as follows. related work on interference in femtocell section ii. femtocell over macrocell were explained in section iii. in section iv, brief notes on the problem statement. the notation and system analysis were discussed in section v. results based on simulations as carried out in section vi. the conclusions are drawn in vii. ii. related work on interference in femtocell the related work centered on the deployment of femtocell technology and interference as a main challenge in order to achieve the desired quality of service, good voice calling and network coverage in an indoor environment. femtocell deployment used to achieve good connectivity to mobile users and improvement of signal behavior in an indoor environment. the approach in [7] has practical problems due to the architecture networks constrained by number of height of radio and power emission antenna. it is more difficult to analyse the system topologies with feasible transmitter location and to find out the optimal network. the work in [8] [9] [10] aims the optimal transmitter antenna configuration. siomina et al. [11], offers a simulated annealing due to the central algorithm used in optimizing the channel level of the power and antenna angle. the universal mobile telecommunications system (umts) networks in such a manner that the total channel level of the power is brought down. as pointed out in [7], fagen et al. carried out an algorithm that is capable to count on for the ability stage of each cell in the network in parliamentary procedure to maximize the area coverage as well as mitigate interference within a desired signal in the coverage area. in [12] the optimization of femtocell network is performed under interference in an indoor coverage using power scheme to mitigate the cause of interference among femtocell bss is presented. another proposed scheme in [13] and [14], whereby a distributed utility is offered based on the sinr method at the femtocell access point (fap). fap sets up connections to the core network over the subscriber's broadband connection and end users can enjoy improved network capacity. with the above mentioned method they provide a better output and an error estimation of sinr and intricacy of implementing the algorithm for the strength of the network. the decentralized approach strategies that allow the femtocell access point’s to sense the channel and selforganize, in conditions of resource allocation and the topology of the network systems, this give a framework on game theory approach to design decentralized mechanisms for optimization and resource allocation among each femtocell user. recently, game theory has been employed as a powerful too, which is used for the systematic analyses of the resource allocation strategies among the radio nodes [11] have been proposed for cognitive radio. the work presented in [20] focus on the signal strength of the deployment of femtocell over macrocell where they are poor reception of signals for indoor users. it proved that the signal of the femtocell is improved over macrocell but the limitation of the study is that it does not emphasis on the transmitting power to avoid interference that reduces the functioning capacity of femtocell technology. iii. femtocell over macrocell thus, femtocell technology aimed is to cognitive abilities for the purpose of mobile communication, traffic loading, capacity, and coverage optimization over others mobile cellular network at the indoor [2]. wi-fi network is generally applicable mostly to all cellular service providers due to the strength of the signal, but the femtocell technology is much better off in terms of improved signal strength, security purpose, and voice calling service at home or office environment. however, the received signal strength gets improved due to the functionality of femtocell technology as a base station in an indoor environment. a macrocell transmit in a wide range with high transmission power that cover up to about 20miles radius due to a base station. no. table 1: femtocell and other cellular networks table column subhead femtocell macrocell 1 data rate 45mbps non 2 installation customer operator 3 rent of site no site rentage rentage of site 4 operating frequency 2,6ghz 5ghz 5 power ranges 10dbm 25dbm 5 primary service quality of voice calling, multi-media, video and security data and voice calling in table 1, we provide the dissimilarity between both technologies used for communication transmission [13]. in order to mitigate interference femtocell must transmit at a lower power. what really distinguished femtocell technology is the valued added service introduced by the mobile provider for end users at the home or office environment. it is a service that a mobile provider is always enthusiastic to integrate as much as possible. iv. problem statement femtocell technology has immensely improved signal strength of the mobile network, but there are challenges that need to be controlled by the mobile provider due to the fact that it shares the same spectrum with the existing network. there are two cases of interference in two tier architecture networks such as co-tier and cross-tier interference. mostly this is caused by unwanted transmitting signal within the frequency band, which gives rise to interference. hence, the absence of interference mitigation will interfere in the quality of femtocell deployment within the network. other challenges that faced deployment of femtocell were mobility movement and handover, self-organization, access mode and synchronization and timing etc. our focus is on interference which is caused by the high transmitting power of femtocell access point of the users. this should be the main concern for the mobile operators. v. notation and system analysis here we present the system notation, and parameter for network performance results. our primary concern is to achieve an improved signal and quality of service. the estimation of sinr [14] is highly important, which is expressed as (1): fbs fbs fbs mbs g p sinr p p     (1) fbs p ……transmitting power of the femto base station mbs p …… transmitting power of the macrocell base station fbs g …… channel gain  …….. noise. in this paper, we used the path loss model [15]. these path loss models are approximations of the instability of signal behavior in an indoor environment. therefore, the path loss is given [15] in the equation (2): 10 10 2 , 2 0.46 1 1 2 ( ) max(15.3 37.6 ( )),38.4 20 ( ) 0.7 18.3 d indoor n n iw ow ow pl db log d log d d n ql l ql               (2) where, pl is the path loss model n ---number of penetration floors q ---number of walls in the flats iw l -----penetration loss of the wall that different the apartment 2 , 0.7 d indoor d -----penetration loss by the walls inside the flats d ----distance between transmitter and receiver in meter ow l --------penetration loss of outdoor wall low and liw are set to 20db and 5db respectively. then, we presumed that the capacity saved as the network throughput [16], mathematically it is given as (3): 2 log (1 sin )t r  (3) where t is the throughput and sinr is the signal. here, a number of femtocells are selected to be used in the indoor environment with equal service provider in this area and one outdoor macrocell. each femtocell act well as defined in the experiment, the parameters for the system analysis are shown in table 2 and some of the above equations as presented in [20]. nos table 2: simulators parameters 1 parameters values 2 scenario size 350x350 3 macrocell base station 1 4 femtocell as a base station 1 5 bandwidth 5mhz 6 noise -174dbm/hz 7 macro tx power 43dbm 8 femto tx power 10dbm the system simulation uses matlab as a model of operation to analysis the signal behavior of femtocell over macrocell in an indoor environment. the real life network model takes a series of events to achieve the main objective of the goal of the application of a femtocell. vi. results in this section, we observed the simulation results of deployment of femtocell for improved signal and quality of voice calling. we consider the parameters and equation in section v. the simulation results prove the signal of both femto cells and macro cells in a cellular network for better performance in an indoor environment. based on the result, figure 4 illustrates the deployment of femtocell in the residential area. here we indicate the randomness of femtocell in a network, while the colors indicate subscriber and nonsubscriber of femtocell network. here, the fap is randomly scattered around the area. fig 4: deployment of femtocell network in residential area with the simulation result shown in figure 5, bit error rate against the signal. we considered the bit error rate at the indoor environment, this is the major parameter in data transmission and communication system. from the result, we observed that the bit error rate is low and this promotes voice calling. the system throughput is obtained by a reduction of bit error rate. fig. 5: relationship between ber against the signal in figure 6 we illustrate the throughput of both femtocell and macrocell against signal in order achieve a better signal strength at the indoor environment. with the results it was observed that the throughput of femtocell improved coverage in an indoor environment which promotes quality of service, performance for femtocell user. fig 6: signal behavior of femtocell over macrocells in an indoor environment although similar simulations as in figure 5 and figure 6 were presented in [20], in this paper we also illustrate the iteration of transmitting power against a signal behavior for both femtocell and macrocell as illustrated in figure 7. this enables us to benchmark the transmitting power of both femtocell and macrocell in order to access an improved signal. thus, this also shows that the power consumption is low compared to the macrocell. however, the measures for excellent performance, and quality signal, that can be used as a benchmark for femtocell deployment. fig 7: iteration of transmitting power over signal behavior vii. conclusions in the past, macrocell have been the only means used for communication, data, media and many more at the indoor and outdoor environment. but due to poor reception of signals at the indoor environment femtocell was introduced. however, it has improved the reception of signals, particularly places like urban and rural environment. it is observed that the signal performance is achieved through the deployment of femtocell technology over macrocells. the deployment of femtocells over the existing macrocell has brought efficiency and profitable solution for mobile operators. however, the cost implication of femtocell is much less than building a macrocell site, paying rent and electricity bills which is undertaken by the mobile provider. thus, femtocell has attracted the attention of service providers due to the valuable service and decrease of related energy consumption. femtocells utilize the broadband connection which may be used for other applications such as video streaming. there can be associated problems when the provider of the broadband service differs from the mobile network provider. there will be cost drawbacks considering the installation in certain residential areas. the performance capacity is attained since of the femtocell served as the base station (bss) to the users in an indoor environment. the deployment of femtocell will continue to play a vital role in the market for mobile operator’s network. further work will be that of cost function implementation over power transmitting for all femto users in the cellular network. this will allow to reduce interference within the users since the transmitting power is bench on 10-15db for better signal used. references [1] femtocell forum http:// www.femtoforum.org.com accessed 20 october 2015 [2] 3gpp work items on self-organizing networks. (accessed 15 august 2015). [online]. available: http://www.3gpp.org/ftp/information [3] ] d. l. perez, g. d. la roche, a. valcarce, a. juttner, and j. zhang, “interference avoidance and dynamic frequency planning for wimax femtocells networks,” 11th ieee singapore international conference on communication systems, pp. 1579–1584, 19-21 nov. 2008. [4] m. yavuz, f. meshkati, s. nanda, a. pokhariyal, n. johnson, b. roghothaman, and a. richardson, “interference management and performance analysis of umts/hspa+ femtocells,” ieee commun. mag., vol. 47, no. 9, pp. 102–109, sep. 2009. [5] h. claussen, “performance of macro and co channel femtocells in a hierarchical cell structure,” ieee 18th international symposium on personal, indoor and mobile radio communications, pp. 1–5, 3-7 sep. 2007. [6] r. baines, “the need for wimax pico cell and femtocells,” wimax london, pp. 1–36, 25-26 april 2007. [7] femtocell base station compared to macrocell base station:available online http://www.analog.com/library/analogdialogue/archives/4212/ad42_12_fig-01.jpg [8] t. giles, j. markendahl, j. zander, p. zetterberg, p. karlsson, g. malmgren, j. nilsson, “ cost drivers and deployment scenarios for future broadband wireless networkskey research problems and directions for research” ieee 59th vehicular technol. conference, vtc 2004spring 4, 2042-2046 (2004) 17-19 may 2004, milan (italy). [9] fagen, p.a. vicharelli, j. weitzen, “automatedwireless coverage optimization with controlled overlap,” ieee transactions on vehiculartechnology, vol. 57, no. 4, pp. 2395–2403, jul 2008. [10] y. sun, f. gunnarsson, k. hiltunen, “cpich power settings in irregular wcdma macro cellular networks,” in proc. pimrc, beijing, china, sep 2003, pp. 1176–1180. [11] m. garcia-lozano, s. ruiz, j.j. olmos, “umts optimum cell load balancing for inhomogeneous traffic patterns,” in proc. ieee vtc, los angeles, ca, usa, sep 2004, pp. 909–913. [12] k. valkealahti, a. h¨oglund, j. parkkinen, a. hamalainen, “wcdma common pilot power control for load and coverage balancing,” in proc. pimrc, lisbon, portugal, sep 2002, pp. 1412–1416. [13] i. siomina, p. v¨arbrand, d. yuan, “automated optimization of service coverage and base station antenna configuration in umts networks,” ieee wireless communications, vol. 13, no. 6, pp. 16–25, dec 2006. [14] kwanghun han, youngkyu choi, dongmyoung kim, minsoo na, sunghyun choi, kiyoung han, "optimization of femtocell network configuration under interference constraints," 7th international symposium on modelling and optimization in mobile, ad hoc, and wireless networks, pp. 1-7, 23-27 jun. 2009. [15] v. chandrasekhar, j.g. andrews, t. muharemovic, zukang shen, a. gatherer, "power control n two-tier femtocell networks," , ieee transactions on wireless communications, vol. 8, no. 8, pp. 4316-4328, aug. 2009. [16] v. chandrasekhar, j.g. andrews, s. zukang, t. muharemovic, a. gatherer, "distributed power control in femtocell-underlay cellular networks," ieee global telecommunications conference, pp.1-6, nov. 30-dec. 4 2009. [17] d lo´pez-pe´rez, a valcarce, a´ kos lada´nyi, g de la roche, and j zhang, “femtocell versus wi-fi: a survey and comparison of architecture and performance”, ieee wireless communications, vol. pp no. 916-920, 2009. 13 [18] p. lee, t. lee, j. jeong, and j. shin, “interference management in lte femtocell systems using fractional frequency reuse,” in advanced communication technology (icact), 2010 the 12th international conference on, vol. 2, feb. 2010, pp. 1047 –1051. [19] b. s. l. castro, i. r. gomes, f. c. j. ribeiro, g. p. s. cavalcarite, “cost231-hatta and sui models performance using a lms tuning algorithm on 5.86hz in amazon region cities” proc. of ieee. eucap2010. isbn: 978-847653-472-4, pp. 1-3, jul. 2010. [20] o.a. akinlabi, m.k. joseph “signal behaviour in an indoor environment; femtocell over macrocell” 16th ieee international conference on environment and electrical engineering, pp, 7-10 june 2016. florence, italy. http://www.femtoforum.org.com/ http://www.3gpp.org/ftp/information http://www.analog.com/library/analogdialogue/archives/42-12/ad42_12_fig-01.jpg http://www.analog.com/library/analogdialogue/archives/42-12/ad42_12_fig-01.jpg akinlabi olaniyi akindeji was born in surulere, lagos state, nigeria, and west africa country in 1978. he received the b.tech in electrical and electronic engineering from rivers state polytechnic, rivers state, nigeria and his m.tech. degree in electrical engineering from the university of johannesburg (uj) in 2014. the author become a member of saiee in 2013. he is currently enrolled for dphil electrical and electronic engineering at uj. for academic excellence, he has the best paper award from imecs 2014. he has many research papers to his credit already. his research interest focuses on information and communications technology and power distribution and generation and femtocells. meera k. joseph received the degrees of dphil. engineering management from the university of johannesburg (uj) in 2013, and m.c.a in 1997 from the bangalore university. she works as a senior lecturer at uj and is a professional member of iitpsa. many post graduate students completed under her supervision and she has many ieee international conference papers, journal papers and book chapters to her credit. she runs the ict4d research group in the school of electrical, uj. her research interests include information and communication technology for development (ict4d), smart grids, femtocells, cloud computing and wireless networks. paper title (use style: paper title) transactions on environment and electrical engineering issn 2450-5730 vol 1, no 3 (2016) © turgay yalcin, muammer ozdemir pattern recognition methods for detecting voltage sag disturbances and electromagnetic interference in smart grids turgay yalcin, muammer ozdemir abstract— identification of system disturbances, detection of them guarantees smart grids power quality (pq) system reliability and provides long lasting life of the power system. the key goal of this study is to find the best accuracy of identification algorithm for non-stationary, non-linear power quality disturbances such as voltage sag, electromagnetic interference in smart grids. pqube, power quality and energy monitor, was used to acquire these distortions. ensemble empirical mode decomposition is used for electromagnetic interference reduction with first intrinsic mode function. hilbert huang transform is used for generating instantaneous amplitude and instantaneous frequency feature of real time voltage sag power signal. outputs of hilbert huang transform is intrinsic mode functions (imfs), instantaneous frequency (if), and instantaneous amplitude (ia). characteristic features are obtained from first imfs, if, and ia. the six features—, the mean, standard deviation,skewness, kurtosis of both if and ia are then calculated. these features are normalized along with the inputs classifiers. the proposed power system monitoring system is able to detect power system voltage sag disturbances and capable of recognize electromagnetic interference component. in this study based on experimental studies, hilbert huang transform based pattern recognition technique was used to investigate power signal to diagnose voltage sag and in power grid. support vector machines and c4.5 decision tree were operated and their achievements were matched for precision and cpu timing. according to the analysis, decision tree algorithm without dimensionality reduction produces the best solution. index terms— c4.5 decision trees, electromagnetic interference, feature extraction, hilbert huang transform, power quality disturbance, smart grids, support vector machines i. introduction smart grids have been constructed structure where a number of control devices are used to provide reliability, stability and efficiency in the power generation, transmission and distribution. to enhance forecasting faults and risks in addition to ensuring protection against any possible internal and external threats, the new generation, smart grids, will be supplied with communication facilities and real time measurement techniques [1, 2]. the smart grid design is mainly based on restructuring the power industry and optimizing its resources. smart grids could optimize transfer capability of transmission and distribution networks to meet the demands for higher quality and more reliable power supply [1, 2]. the main benefits of the smart grid technologies include: minimized shutdown of the distributed generation in overload conditions, power quality improvement, improved voltage profile, coordinated restoring of the power system avoid to grid blackout [1,2,3,4]. a. voltage sag voltage sags are short-duration (less than 1 sec) reduction in voltage magnitude. this kind of disturbance is presently one of main power quality problems (figure 1b.). momentary increase of current has many origins in power systems such as energizing of transformers, short circuits, earth faults and starting of induction motors [5, 6]. this scientific study is supported by tubitak. (project number: 114e919) t. yalcin is with the ondokuz mayis university, department of electrical & electronic engineering, 55139, samsun, turkey (e-mail: turgay.yalcin@omu.edu.tr ). m. ozdemir is with the ondokuz mayis university, department of electrical & electronic engineering, 55139, samsun, turkey (e-mail: ozdemirm@omu.edu.tr ). b. electromagnetic interference (emi) electromagnetic interference (emi), side-effect results of the power conversion and control devices processes, can emerge in a wide frequency range from the basic harmonic and inter-harmonics of the mains frequency. a rise in switching frequencies gives rise to the high energy obstruction, created by the realization of the energy conversion processes, to be shifted in frequency range approximately operated (9 khz-30 mhz) emi range. moreover, a new growing power quality problem especially (2 khz-150 khz) threatened the smart grid power quality [7]. emi normalized voltage signal (l1-n / phase a) generated with arbitrary function generator tektronix afg3022c is shown in figure 1c [2, 6]. fig. 1. healthy signal (1a), normalized voltage sag (1b) and emi normalized voltage signal (1c) (l1-n / phase a) generated with tektronix afg3022c 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 -1 0 1 time(sec) a m p lit u d e 1a. healthy signal 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 -1 0 1 time(sec) a m p lit u d e 1b. voltage sag 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 -1 0 1 time(sec) a m p lit u d e 1c. electromagnetic interference noise mailto:turgay.yalcin@omu.edu.tr mailto:ozdemirm@omu.edu.tr hilbert huang transform method used for recognizing and identifying real time power quality disturbances have described in section ii. ii. feature extraction a. emprical mode decomposition (emd) the algorithm [8, 9, 10] includes the steps: i. determine all the extrema of the signal, s(t). ii. find the upper and lower envelope constructed in step (i). (interpolation of the extrema analyses with the cubic spline ) iii. then, find the subtraction signal and the mean function of the upper and lower envelope (mean(t)), dif(t) = s(t)−mean(t). iv. only when the iteration stops, dif(t) becomes first imf c1(t) ; or else, branch to step (i) change s(t) with dif(t). v. find the residue signal, res(t) = s(t)−c1(t). vi. continue the operation from steps (i) to (vi) to attain second imf, c2(t). achieve cn(t), continue steps (i) – (vi) after n iterations. the routine is broken when the last imf (residual signal res(t)) is acquired as a monotonic function. this routine called sifting process. finally, we get residue res(t), gathering of m imf, from c1(t) to cn(t). the targeted signal can be expressed as: ∑ = += m i i trestcts 1 )()()( (1) we can regard res(t) as cm+1(t) [11, 12,13]. a. eemd (ensemble empirical mode decomposition) b. b. ensemble emprical mode decomposition (eemd) the eemd algorithm (fig. 2.) steps are hearunder: i. add noise, wn(t), to target signal s1(t). s2(t)=s1(t)+wn(t). ii. used emd algorithm for decomposing the final signal s2(t). iii. continue steps (i) and (ii) till the trial numbers. when new imf combination cij(t) is succeeded, predict the ensemble mean of the last imf. the aimed output: ∑ = = tn i ijj tctceemd 1 )()]([ (2) tn: trial numbers, i: iteration number and j: imf scale [13,14]. fig. 2. the representation of the eemd algorithm ∫ +∞ ∞− − = τ τ τ π d t c tv jj )(1 )( (3) cj(t) is real part and vj(t) is imaginary part of an analytic signal zj(t): )()()( tjvtctz jjj += (4) ))(exp()()( tjwtatz jjj = (5) amplitude and phase expressed with equation (6) and (7): 22 )()()( tvtcta jjj += (6) ) )( )( arctan()( tc tv t j j j =θ (7) thus, the instantaneous frequency wj(t) was given by: dt td tw jj )( )( θ = (8) c. feature generation: hilbert-huang transform (hht) hht [13, 14, 15] enables the real time signal x(t) into the time frequency domain by merging eemd with the hilbert transform (fig. 3.). the hilbert transform is then implemented for each imf component cj generated with sifting process which is explained in section iii.a. fig. 3. main steps of the feature generation routine with hht iii. experiment set up in this part of the study, pqube was installed to acquire measurements, firstly in basic electricity laboratory for one phase (l1-n) records. secondly it was utilized in computer laboratory for three phases (l1-n, l2-n, l3-n) records which its loads are computers. hht is used in signal processing part of the study for generation of instantaneous amplitude (ia) and instantaneous frequency (if) features. they respectively generated for real time values from pqube one phase in basic electricity laboratory, three phases in computer laboratory for computers. a. real time basic electricity laboratory measurements one phase (l1-n) voltage sag (fig. 4.) event history which recorded by pqube is shown in table i. table i event history (one phase) event_type voltage sag event_magnitude 60.44% event_duration_in_seconds 0.110 trigger_date 2015/11/08 trigger_time t 03:05:51.687 trigger_channel l1-n trigger_threshold 90.0% of nominal trigger_sample_number 257 samples_per_cycle 128 microseconds_per_sample 156.398 fig. 4. power quality monitor with pqube voltage sag condition of signal (2015/11/08) fig. 5. illustrates that first component imf1 the noise (lowest magnitude and highest frequency signal) on the line (l1-n). lower order of imfs means high frequency and oscillation higher order otherwise. fig. 5. imfs for a voltage sag signal processed with eemd -50 0 50 100 150 200 250 300 -500 0 500 time scale signal -50 0 50 100 150 200 250 300 -50 0 50 time scale imf1 (noise) -50 0 50 100 150 200 250 300 -1000 0 1000 time scale imf2 -50 0 50 100 150 200 250 300 -500 0 500 time scale imf3 -50 0 50 100 150 200 250 300 -200 0 200 time scale imf4 -50 0 50 100 150 200 250 300 -200 0 200 time scale residue fig. 6. stockwell transform (st) contours the stockwell transform (st) is developed method related with the gabor transform (gt) and wavelet transform (wt). several works have used st for the analysis of pq disturbance because it allows location in time, real and imaginary components of the spectrum [16, 17, 18, 19]. fig. 6. shows that st can produce proper features for detecting voltage sag. table ii shows the main advantages and disadvantages of two signal processing methods (hht, st). table ii comparison of two feature extraction method for pq disturbances [25, 28] hht st advantages appropriate for feature extraction of nonlinear non-stationary signal, generates perpendicular imfs whereby instantaneous amplitude and phase can be easily assessed maintain time and frequency representation. good time-frequency resolution disadvantages for narrow band conditions is limited, end effects does not accomplish real-time requirement based on block processing, false harmonics measurement owing to dependency of frequency window width. b. real time computer laboratory measurements for 3 phases (l1-n, l2-n, l3-n) real time processing the first intrinsic mode function is removed with the addition (superposition) of remain components to reconstruct the analyzed signal (equ. 1.). respectively, fig. 7. and fig. 8. show that after removing noise component normal and voltage sag cases. fig. 7. after removing first imf normal condition of signal fig. 8. after removing first imf voltage sag condition of signal fig 8. illustrates real time computer laboratory measures that after reconstructing the voltage sag (rate: 86.60% duration: 0.063 sec) signal, namely the first imf removing from the noisy component. in addition, voltage sag occurred on phase b-c as a result of this case different load types and number of computers on line. this is main vision of this scientific work to identify the fault on active different load types. 0 50 100 150 200 250 -300 -200 -100 0 100 200 300 time scale am pl itu de pqube computer lab. (normal conditions) phase a(l1-n) phase b(l2-n) phase c(l3-n) 0 50 100 150 200 250 -300 -200 -100 0 100 200 300 time scale a m p li tu d e pqube computer lab. measurements ("voltage sag","86.60%","0.063 sec") phase a(l1-n) phase b(l2-n) phase c(l3-n) fig. 9. ia corresponding to remove first imf normal condition of signal fig. 10. if corresponding to remove first imf normal condition of signal fig. 11. ia corresponding to remove first imf voltage sag condition of signal the results show explicitly different pattern in fig. 9.normal condition as for fig. 11. – voltage sag condition. also it is clearly shown in fig. 11. voltage sag on two phases (l2n, l3-n). this information will be used for evaluating active loads types and risk management of the grid. fig. 12. if corresponding to remove first imfs voltage sag condition of signal if signal can use for separation for two cases but there is end effect problem that have to be solved. this is another future work of the study. when cubic spline fitting is computationally demanding, generates distortions near the end points. this is a technical problem that causes data 0 50 100 150 200 250 305 310 315 320 time scale in st an ta n eo u s a m p li tu d e (i a ) pqube computer lab. measurements-normal condition after removing first imf(noise) ia-phase a ia-phase b ia-phase c 0 50 100 150 200 250 14 16 18 20 22 24 26 28 time scale in st an ta n eo u s f re q u en cy ( if ) normal condition after removing first imf(noise) if-phase a if-phase b if-phase c end effect 0 50 100 150 200 250 280 285 290 295 300 305 310 315 320 325 330 time scale in st an ta ne ou s a m pl itu de (i a ) instantaneous amplitude after removing first imf (noise)voltage sag condition phase a (l1-n) phase b (l2-n) phase c (l3-n) 0 50 100 150 200 250 13 14 15 16 17 18 19 time scale in st an ta ne ou s fr eq ue nc y (if ) instantaneous frequency after removing first imf (noise)voltage sag condition phase a (l1-n) phase b (l2-n) phase c (l3-n) end effect failures and peaks at the beginning and at the end of the signal. this fault will be investigated on hht (fig. 10. 11.). c. feature selection for diagnosis of disturbances, extracted features are produced from firstly eemd method so as to classify the voltage sags in grid. after reconstruction signal without noisy part, first imf pre-processing stage. second stage ia and if) are generated by means of hht. the statistical analysis and classification for identification power quality disturbances. the following features were extracted: mean, standard deviation, skewness of ia and if. selecting appropriate features of voltage sag events are highly crucial for diagnosis of the disturbance. the primary schematic model consists of four steps as shown in fig. 13. fig. 13. schematic model of identification of pq disturbance iv. pq disturbances classification techniques a. support vector machine support vector machine (svm) methods, which are developed by vapnik, whereby statistical learning technique being the basis contributes a novel machine learning method. svms are linked supervised learning methods used for classification and regression [20, 21, 25, 28]. b. decision trees decision trees are methods that utilize divide-andconquer approaches as structure learning by induction [22, 23]. the c4.5 algorithm was developed by qinlan, contains the generation of a tree whereby a training set, finding the information gain criterion to find the finest attribute/feature to be used at each node. furthermore, the algorithm applies the post pruning approach to diminish the size of the tree and prohibit over fitting. c4.5 is a technique for approximating discrete-valued functions that is powerful tool to noisy data and suitable for learning distinctive statements [23, 24, 25, 26, 27, 28]. v. performances of classification algorithms and discussions to figure out the performance of the proposed power quality classification algorithm, a total number of 30 pqube analyzer real time disturbances data were used. the pq signals are divided into two categories; 20 of them were used for training and 10 of them were used for testing the proposed algorithm with shuffling the data. in the light of table iii., it is concluded that for sigmoid kernel degree 0.01 with dimensionality reduction with singular value decomposition (svd) described in [29] is better result in terms of cpu time (3.56 sec), and for polynomial kernel d=3, is also better result cpu time (3.58) in non linear classification svm. decision tree algorithm has the precision of 100% and cpu time of 4.10 sec. eventually, c4.5 decision tree based method is the best and gives more proper outcomes than the svm technique without svd. (note: the most proper and robust classifiers for each data set are showed by red font in table iii). table iii performances of disturbance diagnose algorithms classifier precision time (sec) svm-linear 50% 2.35 svm-poly d=2 100% 3.75 svm-poly d=2 preprocessing svd (r=2) 100% 3.59 svm-poly d=3 90% 1.91 svm-poly d=3 preprocessing svd (r=3) 100% 3.58 svm-poly d=3 preprocessing svd (r=2) 60% 1.06 svm-rbf sigma =0.01 50% 0.36 svm-rbf sigma =0.01 preprocessing svd (r=3) 100% 3.56 svm-rbf sigma =1 50% 0.40 svm-rbf sigma =1 preprocessing svd (r=3) 50% 0.36 decisiontree c4.5 100% 4.10 decisiontree c4.5 preprocessing svd (r=3) 100% 4.0606 vi. conclusion in this real time analysis, eemd-hht signal processing system was used for generation features of different characteristics ia and if for normal condition and voltage sag http://www.sciencedirect.com/science/article/pii/s0957417407005982#ref_bib7 http://www.sciencedirect.com/science/article/pii/s0957417407005982#ref_bib12 http://www.sciencedirect.com/science/article/pii/s0957417407005982#ref_bib12 cases. the technique reported in this study clearly accomplishes generation of features different for normal – voltage sag cases aiming that identification of smart grid faults. simulations results have illustrated the capability and validity of the hht. this study shows that the proposed approach can be easily used for detecting electromagnetic interference on non-stationary signals. results of the experiments will be conduct for relation on three phases between computer numbers and voltage disturbances for future studies. in pq diagnosis part of the study, svm and decision tree (c4.5) were operated and their results were match for precision and cpu time. in consequence of precision and timing criteria, without dimensionality reduction with svd, svm-rbf (sigma =0.01) algorithm presented the best solution. results from the simulations clarify that the proposed method is effective in detecting nonstationary pq signal. for analyzing the real time power quality disturbance signals and classifying them, matlab ™ toolboxes are used for simulations. acknowledgement this scientific study is supported by tubitak. (project number: 114e919) references [1] j. momoh, “smart grid: fundamentals of design and analysis”, first edition, institute of electrical and electronics engineers, 2012. [2] r. smolenski, “conducted electromagnetic interference (emi) in smart grids”, springer -verlag london, 2012. [3] s. borlase “smart grids: infrastructure, technology, and solutions” , crc press, 2012. [4] report to nist on the smart grids interoperability standarts roadmap, epri, 2009. [5] m. h. j. bollen and i. y. h. guo, “signal processing of power quality disturbances”, new york: wiley, 2006. [6] p. f. ribeiro, c.a. duque, p. m. silveria, and a. s. cerqueira, “ power systems signal processing for smart grids”. chichester, uk: john wiley & sons, inc., 2013. [7] a. mceachern, “practical power quality: an update, large customer conference”, power standards lab, november 25, 2015. [8] k.-m. chang,“arrhythmia ecg noise reduction by ensemble empirical mode decomposition”, sensors, 10, 2010, pp. 6063 6080. [9] n.e. huang, z. shen., s.r. long, m.l. wu, h.h. shih, q. zheng, n.c. yen, c.c. tung, h.h. liu, “the empirical mode decomposition and hilbert spectrum for nonlinear and non-stationary time series analysis”, proc. roy. soc. london a, vol. 454, 1998, pp. 903–995. [10]z. wu, n.e. huang, “a study of the characteristics of white noise using the empirical mode decomposition method,” proc. roy. soc. london a, 2002. [11] s. baykut, t. akgül, s. ergintav, “emd – based analysis and denoising of gps data”, ieee 17th signal processing and communications applications conference, antalya, 2009. [12] t.yalcin, o.ozgonenel, “feature vector extraction by using empirical mode decomposition from power quality disturbances”, ieee siu, fethiye, mugla, 2012. [13] o.ozgonenel, t. yalcin, i. guney, u. kurt, “a new classification for power quality events in distribution system”, electric power system research (epsr), 95, 2013, pp. 192-199. [14] z.wu, n.e. huang, “ensemble empirical mode decomposition: a noiseassisted data analysis method”, adv. adapt. data. anal., 1, 2009, pp.1–41. [15] z. wang, q. zhu, j. kiely, r. luxton, “hilbert huang transform impedance measurement data for cellular toxicity monitoring” international conference on networking, sensing and control, 2009, pp. 767-772. [16] m. uyar, s.yildirim, m.t. gencoglu, “an effective wavelet-based feature extraction method for classification of power quality disturbance signals”, electr. power syst. res, 78, (10), 2008, pp. 1747–1755. [17] t. nguyen, y. liao, “power quality disturbance classification utilizing s-transform and binary feature matrix method”, electr. power syst. res., 79, (4), 2009, pp. 569–575 [18] c.n. bhende, s. mishra, b.k. panigrahi, “detection and classification of power quality disturbances using s-transform and modular neural network”, electr. power syst. res., 78, (1), 2008, pp. 122 –128. [19] s. suja, , j. jerome, “pattern recognition of power signal disturbances using s transform and tt transform”, int. j. power energy syst., 32, (1), 2010, pp. 37–53. [20] b. biswal, pk. dash, s. mishra, “a hybrid ant colony optimization technique for power signal pattern classification”, expert syst appl,38, 2011, pp. 6368–75. [21] k. k. hoong, s. p. lam, c. y. chung. “an output regulation based unified power quality conditioner with kalman filters.”, ieee trans ind electron, 59 (november (11)), 2012, pp. 4248–62. [22] i. h.witten, e. frank, “data mining: pratical machine learning tools and techniques.”, san mateo, ca, usa: morgan kaufmann, 2005. [23] m. t. hagan, m. b. menhaj, “training feedforward networks with the marquardt algorithm,” ieee trans. neural netw., , nov., vol. 5, no. 6, 1994, pp. 989– 993. [24] r. j. quinlan, “c4.5: programs for machine learning”, san mateo, ca, usa, morgan kaufmann, vol. 1, 1993. [25] s.mishra, t. nagwani, “a review on detection and classification methods for power quality disturbances”, international journal of engineering science and computing, volume 6, issue no. 3, 2016. [26] f. a. s. borges, r. a. s. fernandes, i. n. silva, c. b. s. silva, “feature extraction and power quality disturbances classification using smart meters signals”, ieee transactions on industrial informatics, vol. 12, no. 2, 2016. [27] ozgonenel o., thomas d. w. p., yalcin t., “superiority of decision tree classifier on complicated cases for power system protection” , 11th international conference on developments in power systems protection, birmingham, uk, 2012, pp. 134–134. [28] mahela o. p. , shaik a. g. , gupta n., “a critical review of detection and classification of power quality events”, renewable and sustainable energy reviews, volume 41, 2015, pp. 495–505. [29] b. schölkopf, a. smola, k. müller, “nonlinear component analysis as a kernel eigenvalue problem”, neural computation, 10, 1998, pp. 1299– 1319. http://ieeexplore.ieee.org/search/searchresult.jsp?searchwithin=%22authors%22:.qt.zhiyan%20wang.qt.&newsearch=true http://ieeexplore.ieee.org/search/searchresult.jsp?searchwithin=%22authors%22:.qt.quan%20zhu.qt.&newsearch=true http://ieeexplore.ieee.org/search/searchresult.jsp?searchwithin=%22authors%22:.qt.j.%20kiely.qt.&newsearch=true http://ieeexplore.ieee.org/search/searchresult.jsp?searchwithin=%22authors%22:.qt.r.%20luxton.qt.&newsearch=true http://www.sciencedirect.com/science/article/pii/s1364032114007564 http://www.sciencedirect.com/science/article/pii/s1364032114007564 http://www.sciencedirect.com/science/article/pii/s1364032114007564 http://www.sciencedirect.com/science/journal/13640321 http://www.sciencedirect.com/science/journal/13640321 http://www.sciencedirect.com/science/journal/13640321/41/supp/c turgay yalçın received the b.sc. erciyes university (eu), kayseri, turkey in 2006 and m.sc. degrees in electrical engineering from ondokuz mayıs university (omu), samsun, turkey, in 2010 and he is currently pursuing the ph.d. degree in electrical engineering from ondokuz mayıs university (omu). currently, he is an research assistant with the department of electrical and electronics engineering, ondokuz mayıs university (omu), samsun, turkey from 2007. his areas of interest are identification of power quality disturbances, signal processing methods and machine learning algorithms. muammer özdemir received the b.sc. and m.sc. degrees in electrical engineering from black sea technical university (ktü), trabzon, turkey, in 1988 and 1991, respectively, and the ph.d. degree in electrical engineering from the university of texas at austin (ut), austin, tx, usa, in 2002. currently, he is an assistant professor with the department of electrical and electronics engineering, ondokuz mayıs university (omu), samsun, turkey. his areas of interest are power systems harmonics, power quality, and power system analysis. i. introduction ii. feature extraction a. emprical mode decomposition (emd) b. ensemble emprical mode decomposition (eemd) iii. experiment set up a. real time basic electricity laboratory measurements table i event history (one phase) comparison of two feature extraction method for pq disturbances [25, 28] b. real time computer laboratory measurements c. feature selection iv. pq disturbances classification techniques v. performances of classification algorithms and discussions table iii performances of disturbance diagnose algorithms vi. conclusion references [12] t.yalcin, o.ozgonenel, “feature vector extraction by using empirical mode decomposition from power quality disturbances”, ieee siu, fethiye, mugla, 2012. [13] o.ozgonenel, t. yalcin, i. guney, u. kurt, “a new classification for power quality events in distribution system”, electric power system research (epsr), 95, 2013, pp. 192-199. [14] z.wu, n.e. huang, “ensemble empirical mode decomposition: a noise-assisted data analysis method”, adv. adapt. data. anal., 1, 2009, pp.1–41.  short communication transactions on environment and electrical engineering issn 2450-5730 vol 1, no 1 (2015) ©jacek rezmer & zbigniew leonowicz  abstract— this paper presents a novel method for data analysis and visualization, including real-time visual monitoring and proposal for combined area pq indices on the example of the developed and operational comprehensive system of registration, archiving and data processing for the wide-area monitoring of power quality in a separated part of real power grid with distributed renewable generation. real case studies related to power quality disturbances are presented. index terms—power quality, power distribution faults, power distribution reliability, power system restoration, power system transients, relational databases, distributed generation, location of disturbances, voltage dips, gis, data visualization and analysis. i. introduction ew ideas related to power quality data visualization and analysis are presented in this paper. one of the most important problems in assessment of power quality problems is the huge amount of different indices, numbers and information, spread in time and related to different geographical points. aggregation and averaging leads to loss of important information. one of the objectives of received funding was building a distributed power quality observation system. such a system, composed of pq monitoring units is found at the nodes of the particular distribution network, permits the study of phenomena occurring within the system containing distributed generation units. the observation system is provided with stationary time-synchronized power quality recorders with gps synchronization. the software system supports data transmission between analyzers and database. it permits to conduct effective analysis of phenomena occurring in a distribution system [2]. the main requirement was observation of events in networks and power facilities regardless of their location [3]. in addition, there was a requirement for synchronous data capture for quality assessment throughout pq multi-measurements. the analysis this work was supported by the national science centre (poland) under grant dec-2011/01/b/st8/02515. manuscript published 14 dec 2015. of the assumptions and limitations within the context of technological and economic possibilities, led to gsm telecommunications systems, permitting the best coverage of sparsely industrialized and urbanized areas. basic diagram of developed and realized management and measuring system is shown in figure 1 and the system schematic is shown in figure 2. fig. 1. functional diagram of the wide-area pq monitoring system [4, 5]. first contribution of the paper is the idea of plotting one power quality index for all nodes of the wide-area measuring system in form of radar plot, as shown in figure 3. the second contribution is the idea of visualization of pq data (similar to j. rezmer is with wroclaw university of technology, poland (e-mail: jacek.rezmer@pwr.edu.pl). z. leonowicz is with wroclaw university of technology, poland (e-mail: zbigniew.leonowicz@pwr.edu.pl). novel power quality data analysis and reporting framework for wide-area system of registration and processing of power quality data jacek rezmer, zbigniew leonowicz n mailto:jacek.rezmer@pwr.edu.pl mailto:zbigniew.leonowicz@pwr.edu.pl 2 [1]) in form of special-purpose gis (geographical information system) , as shown in figure 4. fig. 2. schematic of the supervised pq monitoring system. fig. 3. radar plot of combined pq indices. ii. data visualization radar plot (figure 3) shows the normalized values of chosen pq index for all nodes and phases of the system for chosen period of time (10 days in this example). area delimited by red line shows the best pq values for the observation period (in this example, the amplitude of 5th harmonic as the percentage of the fundamental), scaled to the range from zero to one, where one corresponds to best power quality. the area delimited by the polygon (with red, green or blue contour) corresponds to the whole-area pq index. the red line corresponds to the best values, blue line to the worst values and green line to the mean values within the time period and within the area of power system. in that way one can combine multiple pq indices into one meaningful result. fig. 4. gis visualization of real-time pq data. the figure shows one frame of the video file. the map of pq distribution over a geographical area is composed of the superposition of the quasi-geographical localization of the nodes of pq monitoring system (as a map overlay) and a layer of 3-d gaussian functions corresponding to the value of chosen index, as follows: 𝜙𝑘(𝑡; 𝜇𝑘, 𝜎𝑘 2) = 𝐴exp⁡(− ||𝑡−𝜇𝑘|| 2 2𝜎𝑘 2 ), 𝑘 = 1, … , 𝐾 (1) where 𝐴 – amplitude is scaled by the chosen pq index localized over the place of recording 𝜇𝑘 setting the location and 𝜎𝑘 2 setting the width of the function. in the example shown in figure 4 the 5th harmonic level spatial distribution is shown and the width of the function corresponds to expected distance of propagation of pq disturbance, being wider on hv level and narrower in mv and lv level. iii. conclusions in this paper two new efficient methods of power quality data for a wide area monitoring system are proposed. radar plot allows monitoring combined pq indices averaged over a period of time and showing the combined pq index as the polygon area for chosen section of the power system. the gis pq visualization shows spatial and temporal distribution of pq indices, also in real-time, allowing analysis of propagation of disturbances between nodes, mutual influence on pq disturbances between nodes, etc. appendix matlab files are available on request from the authors, contact zbigniew.leonowicz@pwr.edu.pl. video file 65,3 mb can be downloaded from: http://teee.eu/public/film3.avi. 6.3 kv 6.3 kv 1250 kva l-2 (20 km) s = 10 kva 1250 kva 20 kv 20 kv 110 kv 110 kv s-1 (35 km) 6 mva 16 mva s1 = 1150 kva s2 = 1150 kva 10 kv 10 kv l-1 (5 km)4 mva 20 kv 20 kv l-3 (8 km) 0.4 kv 350 kva s1 = 850 kva s2 = 850 kva s1 = 720 kva s2 = 720 kva pqa 2 pqa1pqa4 pqa5 pqa3 mailto:zbigniew.leonowicz@pwr.edu.pl http://teee.eu/public/film3.avi 3 references [1] t. cooke. condensing pq data and visualization analytics. presented at 2015 ieee power & energy society general meeting. [online]. available: http://www.ieeepes.org/presentations/gm2015/pesgm2015p-001598.pdf [2] p. p. baker, t.a. short, and c.m. burns. "power quality monitoring of a distribution system" ieee trans. on power delivery, vol. 9, no. 2, , pp. 1136-1142, april 1994. [3] a. gubanski, p. janik, p. kostyla, j. rezmer, t. sikorski t., j. szymanda, z. waclawek: “comparative analysis of the functionality of distributed power quality monitoring systems (in polish)”. energetyka, no 10, pp. 589-594, 2012. [4] z. leonowicz, j. rezmer, t. sikorski, j szymanda, p. kostyla. widearea system of registration and processing of power quality data in power grid with distributed generation: part ii. localization and tracking of the sources of disturbances. 2014 14th international conference on environment and electrical engineering, eeeic 2014 conference proceedings, art. no. 6835904, pp. 414-417. available: http://ieeexplore.ieee.org/xpl/articledetails.jsp?arnumber=6835904 [5] z. leonowicz, j. rezmer, t. sikorski, j szymanda, p. kostyla. widearea system of registration and processing of power quality data in power grid with distributed generation: part i. system description, functional tests and synchronous recordings. 2014 14th international conference on environment and electrical engineering, eeeic 2014 conference proceedings, art. no. 6835904, pp. 175-181. available: http://ieeexplore.ieee.org/xpl/articledetails.jsp?arnumber=6835859 [6] r. essomba. smoothing techniques using basis functions: gaussian basis. [online] available: http://datascienceplus.com/smoothingtechniques-using-basis-functions-gaussian-basis/ jacek rezmer received the m.sc., ph.d. and habilitate doctorate (dr sc.) degrees, all in electrical engineering, from the wroclaw university of technology, poland in 1987, 1995, and 2013, respectively. he has been with the department of electrical engineering, wroclaw university of technology, since 1987. his current research interests are in the areas of transients in power systems, control and protection, and especially application of signal processing methods in power systems. zbigniew leonowicz (ieee m’03– sm’12) became a member (m) of ieee in 2003 and a senior member (sm) in 2012. he received the m.sc., ph.d. and habilitate doctorate (dr sc.) degrees, all in electrical engineering, in 1997, 2001, and 2012, respectively. he has been with the department of electrical engineering, wroclaw university of technology, since 1997. his current research interests are in the areas of power quality, control and protection of power systems, renewables, industrial ecology and applications of signal processing methods in power systems. http://www.ieee-pes.org/presentations/gm2015/pesgm2015p-001598.pdf http://www.ieee-pes.org/presentations/gm2015/pesgm2015p-001598.pdf http://ieeexplore.ieee.org/xpl/articledetails.jsp?arnumber=6835904 http://ieeexplore.ieee.org/xpl/articledetails.jsp?arnumber=6835859 http://datascienceplus.com/smoothing-techniques-using-basis-functions-gaussian-basis/ http://datascienceplus.com/smoothing-techniques-using-basis-functions-gaussian-basis/ transactions on environment and electrical engineering issn 2450-5730 vol 1, no 3 (2016) © luca d’acierno, marilisa botte, claudia di salvo, chiara caropreso, and bruno montella abstract—a rail system may be considered a useful tool for reducing vehicular flows on a road system (i.e. cars and trucks), especially in high-density contexts such as urban and metropolitan areas where greenhouse gas emissions need to be abated. in particular, since travellers maximise their own utility, variations in mobility choices can be induced only by significantly improving the level-of-service of public transport. our specific proposal is to identify the economic and environmental effects of implementing an innovative signalling system (which would reduce passenger waiting times) by performing a cost-benefit analysis based on a feasibility threshold approach. hence, it is necessary to calculate long-term benefits and compare them with intervention costs. in this context, a key factor to be considered is travel demand estimation in current and future conditions. this approach was tested on a regional rail line in southern italy to show the feasibility and utility of the proposed methodology. index terms—microscopic rail system simulation, operational cost definition, public transport management, signalling system, travel demand estimation. i. introduction ccording to the european commission [1], 23.2% of greenhouse gas emissions in 2014 were produced by the transport sector, of which road transport accounted for 72.8%. the development of actions to promote sustainable transportation systems and to reduce vehicular flows on the road system (i.e. cars and trucks) could therefore significantly abate the sector’s emission contributions. in this context, adoption of a public transport system based on the use of a rail technology which makes railways the highperforming mobility backbone represents a sound choice: besides being environmental friendly, rail systems are highperforming (high travel speeds and low headways), competitive (lower unit costs per seat-km or carried passenger-km) and they are able to ensure a high degree of safety thanks to the presence of signalling, control and train protection systems. obviously, the attractiveness of public transport can be this paper was partly supported under research project fersat grant no. pon03pe_00159_4 (italian ministry of education, universities and research). l. d’acierno (corresponding author), m. botte, c. caropreso and b. montella are with the department of civil, architectural and environmental engineering, federico ii university of naples, via claudio 21, naples, 80125 enhanced only by improving service quality and minimising user discomfort. indeed, since according to the assumptions of rational decision-maker each user tends to choose the alternative of maximum utility (i.e. minimum disutility), the goal is to minimise user generalised costs, which represent the weighted sum of times and monetary costs spent by passengers during their trips. such costs may be split into: access and egress times, waiting times, travel times, transfer times and ticket costs. the measures for reducing user generalised cost may be classified according to three main categories: infrastructural measures (new lines or modification of existing lines), fleet improvement (partial or complete replacement of rolling stock) and signalling system modification (replacement or upgrade of trackside and on-board equipment). obviously, each kind of intervention affects a specific component of the user generalised cost. indeed, under the assumption that timetables and all public transport services are integrated and optimised: • access and egress times depend on the location of stops and stations; • waiting times depend on the headway between two successive convoys allowed by the travel speed and signalling system adopted; • travel times depend on rolling stock performance and infrastructure characteristics; • transfer times depend on the layout of stations, platforms and rolling stock; • ticket costs depend on pricing policies adopted by administrations. clearly, an infrastructural intervention requires high funding availability and may be unfeasible in densely populated contexts. however, in certain cases, it could be essential. likewise, the adoption of policies based on replacing existing fleets or reducing fare levels entails increases in national or regional subsidies, which would be difficult to italy (email: luca.dacierno@unina.it; marilisa.botte@unina.it; chiara.caropreso@yahoo.it; bruno.montella@unina.it). c. di salvo is with ge oil&gas, via cassano 77, casavatore (naples), 80020 italy (email: claudia.disalvo@ge.com). a methodology for long-term analysis of innovative signalling systems on regional rail lines luca d’acierno, marilisa botte, claudia di salvo, chiara caropreso, and bruno montella a achieve in the current economic climate. hence the interventions on which to focus concern implementation of innovative signalling systems whose effect is an increase in service frequencies of a rail system and a consequent reduction in passenger waiting times. moreover, such measures have been made very topical by recent european union policy whose aim is to create a single european standard for rail networks. the necessity of the presence of a signalling system lies in the fact that, since the friction between a train wheel (made of steel) and a rail track (also made of steel) imposes stopping distances of several hundred metres (sometimes kilometres), a system solely based on driver visibility is impossible to create. in particular, the safety of a rail system is based on two main aspects: spacing between convoys and train integrity. the former consists in technologically imposing, by means of a signalling system, a minimum distance between two successive trains so that, in the case of the first train slowing or stopping, the following train is able to react and stop safely. the latter consists in verifying the completeness of the train composition while it is in operation. there are several types of signalling systems currently in use. however, as mentioned above, in order to make rail networks interoperable, the european union has promoted the development of the european rail traffic management system (ertms) [2]. in particular, european train control systems (etcs), which represent the signalling, control and train protection systems designed to harmonise all european safety systems, can be implemented on four levels: from level 0 (when an etcs-compliant rolling stock interacts with a line that is non-etcs compliant) to level 3 (when the infrastructure loses any safety and verification function). the higher the implementation level, the higher is the network performance in terms of maximum speed and minimum headway between two successive convoys. in terms of real applications, only level 2 has been applied in actual railways because on-board train integrity verification is still under research and development (see, for instance, [3]). in this context, the italian ministry of education, universities and research (miur) has funded the research project fersat whose aim is to develop a rail signalling system based on satellite technologies in order to apply etcs level 3. in this regard, we analysed the effects of different rail signalling systems in the case of a regional rail line by performing a costbenefit analysis. however, since the proposed signalling system is based on the etcs level 3 paradigm, the achieved results may be easily exported in the case of conventional rail lines, also in the presence of complex nodes. a key factor to be taken into account for carrying out such an analysis is estimation of travel demand in terms of potential or expected passengers with related characteristics (i.e. starting and arrival stations, adopted time slot, trip duration, etc.). indeed, such information is essential for any kind of assessment related to transportation systems. this has generated an extensive literature on the estimation and forecasting methodologies of travel demand, which is summarised below. in general, estimation of current and future demand can be performed by [4]: direct estimation, disaggregated estimation and aggregated estimation. the first approach, indicated in the literature as direct estimation (see, for instance, [5]–[7]), can be adopted to determine only ‘present’ travel demand. it is based on the application of sampling theory in the case of mobility choices. the main limits of this methodology consist in the vast amount of information to be collected and the inability to predict future developments due to transportation network or socio-economic variations. the second approach, known as disaggregated estimation (see, for instance, [8]–[11]), consists in specifying (i.e. providing the functional form and related variables), calibrating (i.e. determining numerical values of model parameters) and validating (i.e. verifying the ability of the model to reproduce original data) a model by means of appropriate data. these data express disaggregate information related to a sample of individuals, where the size and sample characteristics generally differ from those used in the first approach. this methodology allows mobility choices to be simulated in current conditions (based on the ability to reproduce sampling data) and in the case of future conditions (based on the ability to simulate user reactions to transportation network or socio-economic variations). the above disaggregated approach is referred to in the literature as the revealed preference (rp) approach [4] since it is based on the use of data related to real behaviour of travellers. in the last decades (see, for instance, [12] and [13]), the stated preference (sp) approach has been developed, based on the statements of travellers about their appropriately described and designed preferences in hypothetical scenarios. with the use of this second approach the prediction abilities of the calibrated demand models can be improved. finally, the last approach, known as aggregated estimation (see, for instance, [14]–[16]), is based on modifying demand model results after correcting them by means of traffic counts (i.e. vehicular or passenger flows). the aim of this approach is to identify an origin-destination (od) matrix which is closest to its estimation by model and, once it is assigned to the network, generates flows closest to the counting data. therefore we propose a methodology based on the use of different data sources (censuses, historical data, forecasts, counts, etc.) to estimate travel demand in a wide time period (several decades), so as to perform a cost benefit analysis based on a feasibility threshold approach. specifically, this assessment concerns the economic and environmental effects of implementing an innovative signalling system, even combined with infrastructural measures which, as we will see, in some cases become imperative in order to make any kind of further intervention effective. the paper is organised as follows: section 2 describes the main features of the proposed methodology by focusing on travel demand estimation and investigated performance indexes; section 3 verifies the usefulness of the proposed approach by applying it in the case of a real regional line; finally, conclusions and research prospects are summarised in section 4. ii. the proposed methodology a cost-benefit analysis to estimate economic and environmental utility in modifying the current signalling system on a regional rail line requires the simulation of effects of interactions between all components of a rail system, namely: infrastructure, signalling system, rolling stock, timetable and travel demand. as shown by ([17]–[18]), it is possible to simulate in detail the main aspects of a rail system by resorting to a combination of three kinds of models: a service model for simulating train movements depending on infrastructure, signalling system, rolling stock, planned timetable and travel demand ([19]–[24]); a supply model for simulating performance of all transportation systems in the area depending on passenger flows ([4]); a travel demand model for simulating user choices in terms of mobility selections (departure time, modal choice, starting and arrival stations) and platform behaviour (choices of runs, coaches and entering doors). details on demand estimation and related interactions with supply models can be found in [25]–[27]. in particular, the whole simulation of the rail system can be performed by using commercial microsimulation software (opentrack® software [21]) appropriately integrated with adhoc tools. as already shown, the main contribution of a new signalling system is reduced passenger waiting times: reduction in headways allows an increase in the number of convoys per hour. moreover, in certain cases, a different signalling system may also increase travel speeds if they are not limited by infrastructural conditions or close distance between stations. in light of the above considerations and since our purpose is to implement the methodology by means of a feasibility threshold approach, in the following we do not provide any technological detail concerning the new signalling system, but we characterise it only in terms of maximum achievable performance and maximum level of costs provided. in particular, in order to evaluate and compare different intervention scenarios within the cost-benefit analysis, it is necessary to simulate effects on travel demand explicitly (by taking into account its variability in a long time horizon) and establish certain evaluation criteria. the proposed procedure for estimating passenger flows in current and future conditions and the data sources used are set out below, together with the performance indexes adopted. a. travel demand estimation defining travel demand may be considered of primary importance for evaluating effects of any intervention on transportation systems. however, whatever the methodology adopted, the following requirements have to be met: • accurate reproduction of the current situation; • prediction of future conditions arising at least from demographic changes and/or different performance of transportation systems; • travel demand must be considered a random variable and hence not only average values but also their distribution must be analysed. this implies that the model has to be elastic at least at the level of modal choice (in the case of transportation system variations) and trip generation (in the case of demographic changes). hence, in order to meet these conditions, we propose a methodology based on the use of different italian data sources, even if generalisations to different contexts may easily be obtained. in particular, the suggested procedure can be divided into seven steps. the first phase consists in using data from the national census ([28]) which provide revealed information (i.e. related to behaviour actually occurring in the days prior to the survey) concerning mobility choices in terms of origin, destination, daily time period and transport mode. it is worth noting that census data concern systematic trips (i.e. for work or school purposes) during the average working day and origins and destinations are expressed in terms of municipalities. likewise, daily times are indicated as the morning peak hour (7.30-9.29) and the rest of the day. moreover, although trips are generally bidirectional (i.e. from home to the workplace and return), these data provide only outward trips. in order to satisfy the third requirement (i.e. a wide distribution of considered values) and increase our dataset, we propose to analyse data from at least two decades (i.e. data from the 2001 and 2011 italian censuses). the second phase consists in extending information by means of data from mobility observatories (such as [29]). indeed, information such as total daily trips, rates of trips during morning peak hours, rates of trip chains (i.e. trips with intermediate destinations) and regional modal split needs to be collected. indeed, by combining such data, we may generate non-systematic trips during the average working day classified by origin and destination municipality, time period and transport mode used. in the third phase, by using historical data from the resident population ([30]), previous data may be extended from the census period to a successive period by considering the trip generation model as elastic and adopting a variation rate equal to population variation (i.e. a variation in α% of population in municipality a provides a variation in α% of all trips generated in a). the following phase consists in generating travel demand matrices related to all-day trips where the origin and destination are the stations of the rail line in question. this means that two sub-phases may be identified: the first for obtaining round trips from outward trips in the case of all-day trips; the second for transforming trips expressed in terms of origin and destination municipalities into origin and destination stations. obviously, the second sub-phase requires the definition of a regional network model in order to implement a minimum path approach for associating each municipality to each station with suitable assumptions if there are two or more stations in a municipality. the fifth phase consists in correcting origin-destination matrices associated to rail mode (r) by using turnstile counts, as widely shown in the literature by [14]–[16]. the sixth phase consists in the temporal extension to one or more analysis periods of corrected matrices. in particular, the new matrices may be obtained by considering (real or estimated) demographic variations as in the case of the third phase. in order to make demand elastic at least at modal choice level, it is possible to specify, calibrate and validate a suitable choice model by adopting traditional methodology proposed in the literature (see, for instance, [4]). in particular, it is necessary to: • specify a utility formulation and a probability choice model; • calibrate the values of parameters by solving an optimisation problem; • validate results by means of suitable statistical tests. the following phase consists in determining hourly matrices consistent with the corrected matrices and data on daily variation in travel demand. it is worth noting that the above-mentioned procedure makes use of all methodologies previously described for estimating and forecasting travel demand by properly integrating them with each other in a comprehensive theoretical framework. indeed, recourse to data from national census represents direct estimation of travel demand. in addition, considering three different levels of demographic variation allows us to meet the requirement of stochasticity. the fifth phase makes use of data from turnstile counts in order to correct the initial origin-destination matrices so as to reproduce surveyed flows. hence, this step addresses the issue of aggregate estimation of travel demand. finally, the sixth phase includes both forecasting techniques of travel demand (by means of the temporal extension to future analysis periods of corrected matrices) and its disaggregated estimation (by means of the specification, calibration and validation of a suitable modal choice model). b. performance indexes regarding performance indexes, in order to analyse effects of each intervention scenario, we propose the adoption of an objective function which jointly considers the costs of public administration, passengers and society: ecpgcnocofv ++= (1) the first term is represented by the net operational cost (noc) which is equal to the part of operational costs not covered by ticket revenues. it can be expressed as follows: trtocnoc −= (2) where toc is the total operational cost of the rail system and tr is the ticket revenues. national and regional governments are often inclined to finance public transport in order to improve the mass-transit level-of-service and reduce related fares (i.e. increase user utility). obviously, there are some regulations for funding public transport. in particular, in italy, there is a contractual rate (indicated as standard cost) at which the government pays the service company according to transport supply, and a constraint on service effectiveness expressed in terms of the ratio between ticket revenues and operational costs. hence, by adopting a standard cost approach, the term toc can be expressed as: km-trainctoc km-train ⋅= (3) with: tt,ii ti tlkm-train ∆∆ ∆ ϕ ⋅⋅= ∑∑ (4) =∑ t t t∆ ∆ 8,760 hours = 1 year (5) where ctrain-km is the standard cost (expressed in euros per trainkm); train-km is the unit of measurement adopted to quantify the supply service; li is the length (expressed in kilometres) of line i; ϕi,∆t is the service frequency (expressed in trains per hour) of line i during time interval ∆t; t∆t is duration (expressed in hours) of time interval ∆t. ticket revenues (tr), which depend on fare policies and user choices, can be expressed as follows [31]: ( ) t,lj,lj tlj fntctr ∆ ∆ ⋅= ∑∑∑ (6) where tcj is the revenue associated to ticket type j; nl,j is the number of trips made by user category l by using ticket j; fl,∆t is the passenger flow of category l during time interval ∆t. the second term is the passenger generalised cost (pgc), which can be expressed as follows: rcmtpcrpgpgc ++= (7) where rpg is passenger cost on the analysed rail system, mtpc is passenger cost on mass-transit systems except the analysed rail system and rc is user cost on the road system. in particular: mtobwae cttttrpg ++++= (8) mtobwae cttttmtpc ++++= (9) mwob cttrc ++= (10) where tae is the access and egress time, tw is the waiting time, tob is the on-board time, tt is the transfer time and cm is the monetary cost. in the case of the road system, the on-board and waiting times represent, respectively, times spent travelling along road links and waiting at intersections or searching for parking. however, details on the formulation of the above times can be found in [4] and [32]. finally, the third term is the environmental cost (ec) associated to the whole transportation system. it can be formulated as proposed by [32], that is: at,a at km lfcecec ⋅⋅= ∑∑ ∆ ∆ (11) where eckm is the environmental cost (expressed in euros per kilometre) associated to each vehicle in the road system (i.e. car or truck), fca,∆t is the traffic flow associated to road link a during time interval ∆t, and la is the length (expressed in kilometres) of road link a. in particular, the implementation of different signalling systems allows different headways to be adopted between two successive trains, which results in an increase in service frequencies. hence, it is possible to have an increase in total operational costs (toc), a reduction in passenger waiting times (tw), an increase in passenger flows on the rail system (which allows an increase in ticket revenues tr) and a reduction in traffic flows on the road system (which allows a reduction in environmental costs). obviously, it is necessary to verify quantitatively any compensation between increases and reductions. iii. application in the case of a regional rail line in order to verify the feasibility and the utility of the proposed approach, we applied it to the naples–sorrento regional rail line serving the metropolitan area of naples in southern italy (see fig. 1). the line connects the regional capital (i.e. naples) with the sorrento peninsula, where the city of sorrento represents the line terminus. the line can be decomposed into a first part, 24.5 km long, between naples and moregine, based on a double-track framework and a second part, 17.0 km long, between moregine and sorrento, based on a single-track framework. moreover, in barra and torre annunziata there are the junctions respectively for sarno and poggiomarino. hence, between naples and torre annunziata there is the overlap among the different lines. since the average distance between successive stations is about 1.2 km and the maximum acceleration and deceleration is fixed by comfort conditions (i.e. higher values may cause standing passengers to fall over), increases in the maximum speed of lines do not provide significant reductions in travel times. hence, improvements due to signalling systems are mainly related to reductions in headways between two successive rail convoys which mean reductions in passenger waiting times. however, it should be pointed out that in the current framework of the line, the existence of a single-track section represents the real bottleneck of the line operation for any possible improvement. hence, although in a highly populated area such as the analysed contexts (where the average density is 2,631 inhabitants/km2) any infrastructural intervention may require considerable funds, doubling the line, which would cost about € 300-800m, represents a major intervention for optimising the benefits of a new signalling system. fig. 1. general framework of the naples–sorrento regional rail line. table i scenarios analysed scenario description 1 current infrastructure; current signalling system; current timetable. 2 current infrastructure; current signalling system; current timetable for overlapping lines; maximising frequency for naples–sorrento line. 3 current infrastructure; current signalling system; maximising frequency for naples–sorrento line, considering it a priority over other overlapping lines. 4 current signalling system; doubling of moregine–sorrento section; current timetable for overlapping lines; maximising frequency for naples– sorrento line. 5 current signalling system; doubling of moregine–sorrento section; maximising frequency for naples–sorrento line, considering it a priority over other overlapping lines. 6 doubling of moregine–sorrento section; innovative signalling system which allows a 4 minute headway to be achieved between two successive rail convoys; maximising frequency for naples–sorrento line, considering it a priority over other overlapping lines. 7 doubling of moregine–sorrento section; innovative signalling system which allows a 3 minute headway to be achieved between two successive rail convoys; maximising frequency for naples–sorrento line, considering it a priority over other overlapping lines. 8 doubling of moregine–sorrento section; innovative signalling system which allows a 2 minute headway to be achieved between two successive rail convoys; maximising frequency for naples–sorrento line, considering it a priority over other overlapping lines. in the above context, we considered the current situation of the line (scenario 1) and seven additional scenarios of increasing complexity in terms of technological and monetary vesuvius barra naples torre annunziata moregine sorrento city of amalfi city of pompei gulf of naples gulf of salerno branch to sarno branch to poggiomarino double-track section train station single-track section effort. details of the scenarios analysed are summarised in table i. table ii objective function values (ofvs) – year 2016 scenario objective function value minimum average maximum 1 21,612,206 26,001,375 30,390,544 2 21,562,321 25,966,660 30,371,000 3 21,556,056 25,962,951 30,369,845 4 21,350,117 25,822,162 30,294,208 5 21,181,881 25,694,677 30,207,474 6 21,005,063 25,557,593 30,110,123 7 20,363,535 25,030,696 29,697,858 8 18,136,387 22,935,558 27,734,729 table iii objective function values (ofvs) – year 2026 scenario objective function value minimum average maximum 1 20,902,096 25,470,409 30,150,743 2 20,857,177 25,438,469 30,132,151 3 20,851,346 25,434,976 30,131,060 4 20,659,078 25,301,345 30,057,672 5 20,499,187 25,178,209 29,972,295 6 20,330,619 25,045,435 29,876,295 7 19,716,293 24,533,087 29,468,742 8 17,574,679 22,488,373 27,524,045 table iv objective function values (ofvs) – year 2036 scenario objective function value minimum average maximum 1 19,878,911 24,640,361 29,653,055 2 19,841,150 24,612,758 29,636,438 3 19,835,943 24,609,605 29,635,479 4 19,663,374 24,487,163 29,566,759 5 19,515,507 24,370,825 29,484,199 6 19,358,826 24,244,789 29,391,002 7 18,783,695 23,755,185 28,993,229 8 16,765,324 21,789,297 27,086,785 table v objective function values (ofvs) – year 2046 scenario objective function value minimum average maximum 1 18,535,015 23,484,308 28,845,225 2 18,506,654 23,462,747 28,831,814 3 18,502,268 23,460,065 28,831,068 4 18,355,572 23,353,207 28,769,925 5 18,223,498 23,246,337 28,691,938 6 18,082,430 23,129,687 28,603,290 7 17,558,779 22,671,758 28,221,393 8 15,702,281 20,815,657 26,377,041 table vi objective function values (ofvs) – year 2056 scenario objective function value minimum average maximum 1 16,869,305 21,984,231 27,695,675 2 16,852,596 21,970,508 27,686,826 3 16,849,226 21,968,439 27,686,385 4 16,734,599 21,881,802 27,636,024 5 16,622,100 21,787,218 27,564,543 6 16,500,384 21,682,746 27,482,370 7 16,040,541 21,265,919 27,123,063 8 14,384,679 19,552,276 25,367,067 the simulation outcome in terms of objective function values in the analysed time period, detailed for minimum, average and maximum levels of demographic variation, is set out below (tables ii–vi). table vii objective function variations – year 2016 scenario objective function variation minimum average maximum 1 – – – 2 -0.06% -0.14% -0.23% 3 -0.07% -0.16% -0.26% 4 -0.32% -0.74% -1.21% 5 -0.60% -1.26% -1.99% 6 -0.92% -1.81% -2.81% 7 -2.28% -3.93% -5.78% 8 -8.74% -12.20% -16.08% table viii objective function variations – year 2026 scenario objective function variation minimum average maximum 1 – – – 2 -0.06% -0.13% -0.21% 3 -0.07% -0.15% -0.24% 4 -0.31% -0.71% -1.16% 5 -0.59% -1.22% -1.93% 6 -0.91% -1.77% -2.73% 7 -2.26% -3.87% -5.67% 8 -8.71% -12.11% -15.92% table ix objective function variations – year 2036 scenario objective function variation minimum average maximum 1 – – – 2 -0.06% -0.12% -0.19% 3 -0.06% -0.13% -0.22% 4 -0.29% -0.67% -1.08% 5 -0.57% -1.16% -1.83% 6 -0.88% -1.70% -2.62% 7 -2.23% -3.78% -5.51% 8 -8.65% -11.96% -15.66% table x objective function variations – year 2046 scenario objective function variation minimum average maximum 1 – – – 2 -0.05% -0.10% -0.15% 3 -0.05% -0.11% -0.18% 4 -0.26% -0.60% -0.97% 5 -0.53% -1.08% -1.68% 6 -0.84% -1.60% -2.44% 7 -2.16% -3.63% -5.27% 8 -8.56% -11.73% -15.28% table xi objective function variations – year 2056 scenario objective function variation minimum average maximum 1 – – – 2 -0.03% -0.06% -0.10% 3 -0.03% -0.07% -0.12% 4 -0.22% -0.49% -0.80% 5 -0.47% -0.95% -1.47% 6 -0.77% -1.44% -2.19% 7 -2.07% -3.42% -4.91% 8 -8.41% -11.40% -14.73% furthermore, variations in the objective function value with respect to the non-intervention scenario (i.e. scenario 1) are reported in tables vii–xi. our numerical results point to a common conclusion: it is indispensable to double the line in order to fully exploit the advantages provided by the innovative signalling system. indeed, as can be seen, in the current infrastructural configuration of the line (i.e. with a section with a single-track framework) timetable optimisation (i.e. scenarios 2 and 3) provides improvements which are at most equal to 0.26% in 2016. then they drop to 0.12% in 2056. moreover, also complete replacement of the signalling system provides results similar to those of scenario 3, since the major limitation is related to the single-track section. hence, although doubling the line (scenarios 4 and 5) provides maximum improvements lower than 2.0% over the whole examined period, it represents an intervention required to reduce the current minimum headway between two successive rail convoys. indeed, a new configuration of the line, based on a fully double-track framework, confers benefits from an innovative signalling system (scenarios 6, 7 and 8) in terms of a reduction in minimum headways, providing maximum improvements between 16.08% and 14.73%. the trend of objective function value variations, during the tested period, in the case of an average rate of demographic change, is shown in fig. 2. leaving aside the slightly decreasing pattern, simply due to a reduction in demographic terms, the graph shows, once again, the importance of the doubling intervention in order to take full advantage of implementing the innovative signalling system. indeed, although the gap between scenarios 2 and 3 (represented respectively by the black and green line) and scenarios 4 and 5 (represented respectively by the red and blue line) appears limited, without the infrastructural intervention of the doubling of the line it would be impracticable to obtain the benefits provided by scenarios 6, 7 and 8 (represented respectively by the orange, brown and dark green line) which show a far higher gap with respect to the other scenarios. fig. 2. variation of objective function value in average conditions during the analysed time period (2016-2056). fig. 3. simulation results in terms of headway for each scenario analysed. table xii number of convoys required scenario convoys required additional convoys 1 10 0 2 19 9 3 20 10 4 40 30 5 49 39 6 56 46 7 74 64 8 110 100 fig. 3 shows the effects of each intervention scenario in terms of the headway between two successive convoys. as can be seen, thanks to timetable optimisation, the headway can be reduced from 29 to 12 minutes, with a reduction of more than 50%; whereas by doubling the line, we can regain only around 7 minutes. however, this infrastructural intervention is essential in order to reduce headways between two successive convoys to as low as 2 minutes. obviously, in order to ensure such low headways, it is necessary to put in place an appropriate fleet in terms of number of available convoys per rail service: the lower the headway, the higher the number of trains needed (table xii). hence, besides the above-mentioned costs, additional resources are required to acquire a suitable number of vehicles. iv. conclusions and research prospects the paper proposed a methodology for evaluating economic and environmental effects related to implementing an innovative signalling system by performing a cost-benefit analysis based on a feasibility threshold approach. the application in the case of a real regional rail line shows the usefulness of the proposed procedure and points out that, in the considered context, the main limitation to network improvements is represented by the single-track section. hence, the replacement of the existing signalling system may be successfully implemented only if combined with doubling of the line. the costs of infrastructure improvements are clearly high (€ 300-800m). yet it is worth noting that they have the same order of magnitude as benefits achievable after just one year. -14.00% -12.00% -10.00% -8.00% -6.00% -4.00% -2.00% 0.00% year 2016 year 2026 year 2036 year 2046 year 2056 objective function variation [%] scenario 1 scenario 2 scenario 3 scenario 4 scenario 5 scenario 6 scenario 7 scenario 8 29.09 12.47 12.00 5.61 4.62 4.00 3.00 2.00 0 5 10 15 20 25 30 scenario 1 scenario 2 scenario 3 scenario 4 scenario 5 scenario 6 scenario 7 scenario 8 headway [min.] timetable optimisation doubling intervention innovative signalling system a key role in the proposed procedure is represented by the estimation and forecasting techniques for travel demand which is a fundamental factor to consider for evaluating the effects of any intervention in the case of transportation systems. in particular, the suggested methodology makes use of data from italian sources. hence, in terms of future research, we propose to apply the described approach in other contexts both on other italian railways (in order to verify the correctness of the procedure and, in particular, the reliability of the adopted data sets in different network configurations) and other non-italian railways (in order to test the methodology in the case of different data sources). finally, the main limitation of the proposed approach in the case of more complex rail networks is related to the excessively high number of solutions to be analysed and related computation times. however, recently [33] and [34] have proposed some methodologies based on the use of heuristic and/or meta-heuristic algorithms in order to solve these kinds of problems by reducing computational efforts. references [1] european commission, climate action: reducing emissions from transport. available: http://ec.europa.eu/clima/policies/transport/index_en.htm (last access: september 2016). [2] j. pachl, railway operation and control. vtd rail publishing, mountlake terrace (wa), usa, 2009. [3] shift2rail joint undertaking, multi-annual action plan. european union: brussels, belgium, 2015. [4] e. cascetta, transportation systems analysis: models and applications. springer, new york (ny), usa, 2009. [5] m. j. smith, “the existence, uniqueness and stability of traffic equilibria,” transportation research part b, vol. 13, no. 4, pp. 295–304, 1979. [6] w. brog and e. ampt, “state of the art in the collection of travel behaviour data,” in travel behaviour for the 1980’s, special report 201, national research council, washington (dc), usa, 1982. [7] j. de d. ortuzar and l. g. willumsen, modelling transport, 4th ed., john wiley and sons ltd., chichester, united kingdom, 2011. [8] t. a. domencich and d. mcfadden, urban travel demand: a behavioural analysis. american elsevier, new york (ny), usa, 1975. [9] j. horowitz, “identification and diagnosis of specification errors in the multinominal logit model,” transportation research part b, vol. 15, no. 5, pp. 345–360, 1981. [10] c. f. manski and d. mcfadden, structural analysis of discrete data with econometric applications. the mit press, cambridge (ma), usa, 1981. [11] m. ben-akiva and s. r. lerman, discrete choice analysis: theory and application to travel demand. the mit press, cambridge (ma), usa, 1985. [12] m. ben-akiva and t. morikawa, “estimation of switching models from revealed preference and stated intention,” transportation research part a, vol. 24, no. 6, pp. 485–495, 1990. [13] j. de d. ortuzar, “stated preference in travel demand modelling,” in proc. wctr 1992, lyon, france, 1992. [14] h. -p. lo and c. -p. chan, “simultaneous estimation of an origin-destination matrix and link choice proportions using traffic counts,” transportation research part a, vol. 37, no. 9, pp. 771–788, 2003. [15] e. cascetta, a. papola, v. marzano, f. simonelli and i. vitiello, “quasi-dynamic estimation of o-d flows from traffic counts: formulation, statistical validation and performance analysis on real data,” transportation research part b, vol. 55, pp. 171–187, 2013. [16] c. -c. lu, x. zhou and k. zhang, “dynamic origin-destination demand flow estimation under congested traffic conditions,” transportation research part c, vol. 34, pp. 16–37, 2013. [17] l. d’acierno, m. gallo, b. montella and a. placido, “the definition of a model framework for managing rail systems in the case of breakdowns,” in proc. ieee itsc 2013, the hague, the netherlands, pp. 1059-1064, 2013. [18] l. d’acierno, a. placido, m. botte and b. montella, “a methodological approach for managing rail disruptions with different perspectives,” international journal of mathematical models and methods in applied sciences, vol. 10, pp. 80–86, 2016. [19] r. prinz, b. sewcyk and m. kettner, “nemo: network evaluation model for the austrian railroad (öbb),” eisenbahntechnische rundschau, vol. 50, no. 3, pp. 117–121, 2001. [20] m. kettner and b. sewcyk, “a model for transportation planning and railway network evaluation,” in proc. wcits 2002, chicago (il), usa, 2002. [21] a. nash and d. huerlimann, “railroad simulation using opentrack,” computers in railways, vol. 9, pp. 45–54, 2004. [22] t. siefer and a. radtke, “railway simulation: key for better operation and optimal use of infrastructure,” in proc. 1st international seminar on railway operations modelling and analysis, delft, the netherlands, 2005. [23] m. marinov and j. viegas, “a mesoscopic simulation modelling methodology for analyzing and evaluating freight train operations in a rail network,” simulation modelling practice and theory, vol. 19, no. 1, pp. 516–539, 2011. [24] e. quaglietta, “a microscopic simulation model for supporting the design of railway systems: development and applications,” ph.d. dissertation, federico ii university of naples, naples, italy, 2011. [25] g.e. cantarella, “a general fixed-point approach to multimodal multi-user equilibrium assignment with elastic demand,” transportation science, vol. 31, no. 2, pp. 107–128, 1997. [26] s. nguyen, s. pallottino and m. gendreau, “implicit enumeration of hyperpaths in a logit model for transit networks,” transportation science, vol. 32, no. 1, pp. 54–64, 1998. [27] m. ercolani, a. placido, l. d’acierno and b. montella, “the use of microsimulation models for the planning and management of metro systems,” wit transactions on the built environment, vol. 135, pp. 509–521, 2014. [28] istituto nazionale di statistica – istat (italian national institute of statistics). population and housing census. available: http://www.istat.it/it/censimento-popolazione (last access: september 2016). [29] osservatorio sui comportamenti di mobilità degli italiani – audimob (observatory on the italian mobility behaviour). regional mobility statistics. available: http://www.isfort.it/sito/statistiche/audimob.htm (last access: september 2016). [30] istituto nazionale di statistica – istat (italian national institute of statistics). resident population. available: http://www.istat.it/it/popolazione (last access: september 2016). [31] b. montella, l. d’acierno and m. gallo, “a multimodal approach for determining optimal public transport fares,” journal of applied sciences, vol. 14, no. 21, pp. 2767–2781, 2014. [32] m. gallo, b. montella and l. d’acierno, “the transit network design problem with elastic demand and internalisation of external costs: an application to rail frequency optimisation,” transportation research part c, vol. 19, no. 6, pp. 1276–1305, 2011. [33] l. d’acierno, m. gallo and b. montella, “application of metaheuristics to large-scale transportation problems,” lecture notes in computer science, vol. 8353, pp. 215–222, 2014. [34] m. botte, c. di salvo, a. placido, b. montella and l. d’acierno, “a neighbourhood search algorithm for determining optimal intervention strategies in the case of metro system failures,” international journal of transport development and integration, vol. 1, no. 1, pp. 63–73, 2017. luca d’acierno is associate professor at federico ii university of naples, italy. he holds an msc degree in civil engineering (2000) and a phd in road infrastructures and transportation systems (2003), both from federico ii university of naples, italy. his research interests include public transport planning and design, rail system analysis and management, multimodal transportation network design, transportation network assignment, pricing policy analysis, and probe vehicle use. he has authored more than 130 papers in peer-reviewed journals and conference proceedings. marilisa botte is a phd student in civil system engineering at federico ii university of naples, italy, having completed her msc in hydraulics and transportation systems engineering (2014) at the same university. her research interests include rail system analysis and management, and travel demand estimation. she has authored 15 papers in peer-reviewed journals and conference proceedings. claudia di salvo is materials planner at ge oil&gas (italy) and research collaborator at federico ii university of naples. she holds an msc degree in civil and environmental engineering (2012), and has completed postgraduate training for experts in digital pattern techniques (2014), both at federico ii university of naples, italy. her research interests include signalling system analysis, longterm travel demand estimation and rail system simulation. moreover, she has authored five papers in peer-reviewed journals and conference proceedings. chiara caropreso is a scholarship holder at the department of civil, architectural and environmental engineering, federico ii university of naples where she was awarded an msc degree in hydraulics and transportation systems engineering in 2016. her research interests include rail system simulation and long-term travel demand estimation. moreover, she has authored two papers in conference proceedings. bruno montella is full professor at federico ii university of naples, italy. he holds an msc degree in transportation engineering (1973) from the same university. his research interests include transit system analysis and management, multimodal transportation network design and optimisation, and public transport quality. he has authored more than 170 papers in peerreviewed journals and conference proceedings. i. introduction ii. the proposed methodology a. travel demand estimation b. performance indexes iii. application in the case of a regional rail line iv. conclusions and research prospects references  transactions on environment and electrical engineering issn 2450-5730 vol 2, no 2 (2017) © calebe a. matias, girodani pacífico medeiros, pedro h. f. moraes, bruno de a. fernandes, aylton j. alves, wesley p. calixto, geovanne p. furriel abstract— the purpose of the present study is to simulate and analyze an isolated full-bridge dc/dc boost converter, for photovoltaic panels, running a modified perturb and observe maximum power point tracking method. the zero voltage switching technique was used in order to minimize the losses of the converter for a wide range of solar operation. the efficiency of the power transfer is higher than 90% for large solar operating points. the panel enhancement due to the maximum power point tracking algorithm is 5.06%. index terms—energy efficiency, geometric brownian motion, monte carlo simulation, performance measurement and verification, solar water heating. i. introduction he concern to produce clean energy is relevant in view of the global warming and pollution. the search for new technologies to improve power conversion of renewable energy sources is the focus of studies and discussions in many academics centers and industries across the globe [1]. photovoltaic (pv) panels are made of photosensitive semiconductors. their semiconductor cells are hit by solar radiation and produce a difference of potential. the problem is that panels cannot deliver the maximum power by their own considering the impedance matching principle. that is the main reason for using a power converter running a maximum power point tracking (mppt) algorithm [2]. dc/dc converters are used to change the impedance seen by any source, due to control of the trigger circuit of these switches [3]. a dc/dc converter is needed when speaking at tracking the maximum power of an energy source, as a photovoltaic panel [4]. the analysis of the behavior and power transfer ratios of a converter can determine its efficacy on deliver the maximum power to the load. a computer simulation provides lots of information about these characteristics. this can be used to help the development of a real converter [5]. this paper aims to analyze the power transfer ratios and the modified perturb and observe (p&o) mppt performance of an isolated full-bridge dc/dc boost converter considering the power loss on its components in order to verify the feasibility of the development of a real device [6], [7]. in order to minimize the losses on the switches of the converter the zero voltage switching technique was applied to trigger the mosfets. to perform the analysis of the results and approach of the electrical model of a photovoltaic panel was made for generating its characteristic curve. later, there were made an analysis of the operation of the dc-dc converter and observations on the mppt technique. simulation and analysis of an isolated fullbridge dc/dc boost converter operating with a modified perturb and observe maximum power point tracking algorithm t wesley p. calixto geovanne p. furriel school of electrical, mechanical and computer engineering federal university of goiás (ufg), goiânia, brazil email: w.p.calixto@ieee.org calebe a. matias giordani pacífico medeiros pedro h. f. moraes bruno de a. fernandes aylton j. alves experimental and technological research and study group (next) federal institute of goias (ifg), goiânia, brazil email: calebeabrenhosa@gmail.com ii. methodology a simulation was performed aiming to analyze the power transfer ratio for a wide range of solar operation and the mppt method. a. materials the selected pv panel was kc200gt from kyocera, with 54 cells. its main electrical performance under standard test conditions (irradiance 1000w/m2, am 1.5 spectrum and module temperature at 25oc) data are shown in table i. the simulation was performed with spice software (simulation program with integrated circuits emphasis). b. electrical model of a photovoltaic panel there are several electrical models that describe the behavior of a photovoltaic panel, among them stands out the model with one diode, one series resistance and one resistor in parallel [8], [9], as shown in fig. 1. applying kirchhoff’s law on the circuit, the equation of the load current is obtained as in (1). 𝐼 = 𝐼𝑝ℎ − 𝐼𝑑 − 𝐼𝑟 (1) where 𝐼𝑝ℎ is the current generated by the photovoltaic effect, 𝐼𝑑 is the current in the diode and 𝐼𝑟 is the current in 𝑅𝑠ℎ. the 𝐼𝑝ℎ current is dependent on the solar radiation and temperature as (2). 𝐼𝑝ℎ = [𝐼𝑝ℎ,𝑠𝑡𝑐 + 𝐾𝑖(𝑇 − 𝑇𝑠𝑡𝑐 )] 𝐺 𝐺𝑠𝑡𝑐 (2) where 𝐼𝑝ℎ,𝑐 is the current generated by the photovoltaic effect under standard conditions, 𝐾𝑖 is the temperature coefficient of the short circuit current, 𝑇𝑠𝑡𝑐 is the temperature at standard conditions (25∘c), g𝑠𝑡𝑐 is the radiation at standard conditions (1000𝑊/𝑚 2 ). the current in the diode (𝐼𝑑) has a non-linear characteristic and is dependent on such factors as the saturation current (𝐼0), the boltzmann constant (𝑘), the electron charge (𝑞), the ideality factor (𝑎1) and the number of cells in series (𝑛𝑠) as (3). 𝐼𝐷 = 𝐼0 {𝑒𝑥𝑝 [ 𝑞 × (𝑉 + 𝐼 × 𝑅𝑠) 𝑛𝑠 × 𝑘 × 𝑇 × 𝑎1 ] − 1} (3) the calculation of the saturation current considers the temperature coefficient of open circuit voltage (𝐾𝑣), temperature coefficient of short circuit current (𝐾𝑖), the short circuit current 𝐼𝑠𝑐 under standard conditions (𝐼𝑆,𝑠𝑡𝑐) and the open circuit voltage under standard conditions (𝑉𝑂𝐶,𝑠𝑡𝑐) as (4). 𝐼0 = 𝐼𝑆,𝑠𝑡𝑐 + 𝑘𝑖(𝑇 − 𝑇𝑠𝑡𝑐 ) 𝑒𝑥𝑝 [ 𝑞 (𝑉𝑂𝐶,𝑠𝑡𝑐 + 𝑘𝑣(𝑇 − 𝑇𝑠𝑡𝑐 )) (𝑛𝑠 × 𝐾 × 𝑇) ] − 1 (4) the current through the resistor in parallel is as (5). 𝐼𝑅 = 𝑉 + 𝐼 × 𝑅𝑠 𝑅𝑠ℎ (5) there are two other important parameters needed to be calculated: the value of 𝑅𝑠ℎ and 𝑅𝑠. these values lead the calculated maximum power point match the experimental maximum power point (𝑉𝑚𝑝 × 𝐼𝑚𝑝). an iteration algorithm, under pylab environment, that increases the value of 𝑅𝑠 to estimate the 𝐼𝑝ℎ,𝑐, 𝐼𝑝ℎ and 𝑅𝑠ℎ values as (2), (6) and (7). 𝐼𝑝ℎ,𝑠𝑡𝑐 = 𝑅𝑠ℎ + 𝑅𝑠 𝑅𝑠ℎ 𝐼𝑠𝑐,𝑠𝑡𝑐 (6) 𝑅𝑠ℎ = 𝑉𝑚𝑝(𝑉𝑚𝑝 + 𝐼𝑚𝑝 × 𝑅𝑠) [𝑉𝑚𝑝 + 𝐼𝑝ℎ − 𝐼𝑑 − 𝑃𝑚𝑎𝑥,𝐸 ] (7) table i electrical performance under standard test conditions maximum power 200w (+10% / -5%) maximum power voltage 26.3v maximum power current 7.61a open circuit voltage 32.9v short circuit current 8.21a max system voltage 600v temperature coefficiente of 𝑉𝑂𝐶 −1.23𝑥10−1𝑉/°𝐶 temperature coefficient of 𝐼𝑆𝐶 3.18𝑥10−3𝐴/°𝐶 area 1.41𝑚2 fig. 1. single-diode model of the photovoltaic module. the characteristic curve of the photovoltaic panel was obtained using a computational algorithm. the following values of parallel (𝑅𝑠ℎ) and series (𝑅𝑠) resistances used on simulation were: 𝑅𝑠ℎ = 158.66ω 𝑅𝑠 = 0.0053ω c. modified perturb and observe mppt method this method was proposed by [7]. the classical p&o mppt algorithm considers that the pv power variation is caused by only the pv voltage perturbation. in fact the pv power is influenced by both the converter and the environmental conditions, such as irradiance and temperature. during rapidly irradiance changing period, on conventional p&o method, there is a wrong control signal due to simple observation of the pv power and voltage reference. fig. 2 represents the pv power curve at irradiance 𝐼1 and irradiance 𝐼2. considering that the conventional p&o algorithm is running from point a to point b, trying to reach point d, at irradiance 𝐼1, so the control signal must increase the voltage from 𝑉1 to a short period of time (δ𝑡) and soon back to irradiance 𝐼1, the converter might read the power at point c and the next control signal must decrease power, for with the increasing voltage and the power reduction the algorithm will try to reach the maximum power point running the opposite way of the real condition, point d. so the conventional perturb and observe method fails to track the maximum power point [7]. to fix this issue a modified p&o mppt algorithm is needed. this method consists in distinguish the power variation caused by the mppt control and the irradiance. this can be done by adding a pv power measurement between of control period. the diagram in fig. 3 illustrates the process. where 𝑑𝑃0.5, shown in (8), is the power difference between the middle-point (𝑑𝑃𝑘−0.5) and the starting point (𝑑𝑃𝑘−1), which contains the power of both solar radiation and mppt control; 𝑑𝑃1, shown in (9), contains the power caused by only the irradiance variation and 𝑑𝑃 , shown in (10), is the power caused by only the mppt control [10] [7]. 𝑑𝑃0.5 = 𝑃(𝑘 − 0.5) − 𝑃(𝑘 − 1) (8) 𝑑𝑃1 = 𝑃(𝑘) − 𝑃(𝑘 − 0.5) (9) 𝑑𝑃 = 𝑑𝑃0.5 − 𝑑𝑃1 (10) so the algorithm uses the power variation caused by only the mppt control signal to track the maximum power point. this fixes the problem of the conventional p&o method. iii. simulation and results using spice software a simulation was performed to determine the voltage and current values of each component. fig. 2: control signal analysis of the conventional p&o. fig. 3: modified p&o algorithm method. the schematic in fig. 4 represents the circuit used in the simulation. an algorithm running the mathematic model of the solar panel provides the voltage 𝑉𝑃𝑉 to dcdc converter. a. operating mode of full-bridge dc/dc boost converter the isolated full-bridge dc/dc converter, depicted in fig. 5, working as a step-up or boost converter. 𝑉𝑃𝑉 is the operating voltage of the photovoltaic panel, which can be varied until the open circuit voltage limit (𝑉𝑜𝑐). the use of 𝐶𝑃𝑉 decoupling capacitor is recommended to prevent the effects of high frequency current ripple generated by the converter, in the photovoltaic panel. the mosfets are used to generate an alternating waveform in the primary of the transformer with a duty cycle frequency equals 110 khz. fig. 6 shows the actuation cycle of the mosfets and the waveform generated in the primary of the transformer (v1). a charge/discharge time on the mosfets generate losses for there is voltage and current over them at the same time. to minimize these losses a zero voltage switching (zvs) technique is used. therefore a time period in which the voltage at the primary of the transformer (v1) remains zero due to the use of this technique, so the switching time of the mosfets are different, depicted in fig. 6. the transformer amplifies the voltage at a ratio of 𝑛. the next stage is rectifying this voltage and then filters the current e voltage ripple through the inductor l and the capacitor c_link. the dc link voltage is: 𝑉𝑙𝑖𝑛𝑘 = 𝑛 × 𝑉𝑃𝑉 × 2𝑑′ (11) the duty cycle of the mosfets switching is equal d, where d = 2.d’. the voltage blanking time (voltage equals zero) in the primary of the transformer (v1) is changed by the duty cycle of effective work d’ = d/2, as depicted in fig 6. the simulation was performed under standard conditions, so the commercial electronic components are specified for its limits values. a load resistance was attached at the dc-dc converter, so measurements can be made. the voltage and current values were observed on each component; therefore the list of commercial components is determined, seen in table ii. the performance analysis of the dc/dc converter is performed using the specification of the components of the table ii. a 1:18 transformer ratio was used to elevate the voltage to the desired level. after running a new simulation, varying the solar radiation, the values on the table iii was obtained. this information shows that the converter works properly under wide range of solar operation points and it keeps the efficiency higher than 90% for operation points above 5%. when the solar panel is operating under low solar radiation the converter loses the efficiency and the mppt algorithm does not work effectively. fig. 4: schematic of the circuit used in simulation. fig. 5: dc-dc isolated full bridge power converter circuit. fig. 6: mosfets waveform of the converter. fig. 7 represents the output power of the panel due to suddenly irradiance variation. when the irradiance gets from 1000𝑊𝑚 2 to 800𝑊/𝑚 2 the output power decreases. the implemented modified p&o mppt has a batter dynamic response than the classic. b. modified p&o mppt analysis the modified p&o mppt method was proposed to correct the mppt control signal during rapidly solar radiation changing. during steady state both methods work properly, but when the irradiance changes suddenly the recovery time on the modified algorithm is faster than the classic. the efficiency of the panel, 𝜂, can be evaluated as: η = ∫ 𝑝(𝑡)𝑑𝑡 𝑇 0 𝐴𝑐 ∫ 𝐺(𝑡)𝑑𝑡 𝑇 0 (12) where 𝑝(𝑡) is the power output of the panel, g(𝑡) is the solar radiation and 𝐴𝑐 is the area of the panel. a comparison between the panel efficiency under the conditions depicted in fig. 7 using both algorithms gives that the conventional and the modified p&o efficiency is respectively 12.24% and 12,86%. so the enhancement on panel efficiency, using the modified p&o method, was 5.06%, compared to classical method. the efficiency of the panel considers the integration on time interval [0, 𝑇], which means as more irradiance variation over the panel, during the time, higher is its efficiency gain, for the time recovering of the power is lower than conventional p&o. iv. conclusion the performed simulations with the commercial specified components shows that the isolated full-bridge dc/dc boost converter is efficient because it keeps its efficiency above 90% almost in the entire operation range. the modified p&o mppt improved the power conversion compared to the classic method. the panel efficiency enhancement was 5.06%. the results could be better for a real condition of solar radiation, in view of the quick irradiance variation on time due to shadows caused by clouds. table ii components used in the simulation component specification panel kc200gt 200w 𝐶𝑖𝑛 330uf mosfets 𝑅𝑠𝑜𝑛 = 0.07ω, 16 a, 60v diodes 1000v, 2a l 12mh, 5ω transformer 1: 18 c_link 30𝑢𝐹, 450𝑉 𝐶𝑜𝑢𝑡 1000𝑛𝐹, 275 𝑉𝑎𝑐 fig. 7: modified p&o mppt response. table iii simulation results of the dc-dc converter. pv operation point 5% (10w) 25% (50w) 50% (100w) 100% (200w) pv voltage: medium voltage [v] 24.55 27.35 26.11 26.15 pv current: average current [a] 0.4304 1.73 3.83 7.65 pv power [w] 10.56 47.42 100.04 200 duty cycle [%] 8.17 15.96 24.81 35.48 dc link voltage: medium voltage [v] 67.73 146.72 212.66 298.79 dc link current: average current [a] 0.1354 0.29345 0.424533 0.59759 power dc link [v] 9.17 43.05 90.45 179.33 equivalent load resistance [ω] 500.22 499.98 499.99 499.99 efficiency [%] 87 91 90 90 the result shows that the physical implementation of the device is feasible. acknowledgment the authors would like to thank coordination for the improvement of higher education personnel (capes), the national counsel of technological and science development (cnpq) and research support foundation of goias state (fapeg) for financial support research and scholarships. references [1] m. i. hofferta et al., “advanced technology paths to global climate stability: energy for a greenhouse planet,” science, vol. 298, pp. 981– 987, 01 nov 2002. [2] b. parida, s. iniyanb, and r. goic, “a review of solar photovoltaic technologies,” renewable and sustainable energy reviews, vol. 15, pp. 1625–1636, 2011. [3] m. h. rashid, power electronics handbook, third edition ed. butterworth-heinemann, 2010. [4] n. femia, g. petrone, g. spagnuolo, and m. vitelli, power electronics and control techniques for maximum energy harvesting in photovoltaic systems. crc press, 2013. [5] m. g. villalva, m. f. espindola, t. g. de siqueira, and e. ruppert, “modeling and control of a three-phase isolated grid-connected converter for photovoltaic applications,” revista controle & automação, vol. 22, no. 3, pp. 215– 228, 2011. [6] s. b. kjær, “design and control of an inverter for photovoltaic applications,” ph.d. dissertation, fac. of eng. and science, aalborg university, denmark, aalborg, jan. 2005. [7] b. yu, “an improved dynamic maximum power point tracking method for pv application,” ieice electronics express, vol. 11, no. 2, pp. 1–10, 2014. [8] m. g. villalva, j. r. gazoli, and e. r. filho, “comprehensive approach to modeling and simulation of photovoltaic arrays,” ieee transactions on power eletronics, vol. 24, no. 5, pp. 1198–1208, 2009. [9] g.p. medeiros, “analysis and simulation of the p&o mppt algorithm using a linearized pv array model,” 10th brazilian power electronics conference (cobep), 2009. [10] l. piegari, r. rizzo, i. spina, and p. tricoli, “optimized adaptive perturb and observe maximum power point tracking control for photovoltaic generation,” energies 2015, vol. 8, pp. 3418–3436, 2015. i. introduction ii. methodology a. materials b. electrical model of a photovoltaic panel c. modified perturb and observe mppt method iii. simulation and results a. operating mode of full-bridge dc/dc boost converter b. modified p&o mppt analysis iv. conclusion acknowledgment references  transactions on environment and electrical engineering issn 2450-5730 vol 2, no 2 (2017) © alireza rezaee  abstract—this paper proposes a model predictive controller (mpc) for control of a p2at mobile robot. mpc refers to a group of controllers that employ a distinctly identical model of process to predict its future behavior over an extended prediction horizon. the design of a mpc is formulated as an optimal control problem. then this problem is considered as linear quadratic equation (lqr) and is solved by making use of ricatti equation. to show the effectiveness of the proposed method this controller is implemented on a real robot. the comparison between a pid controller, adaptive controller, and the mpc illustrates advantage of the designed controller and its ability for exact control of the robot on a specified guide path. index terms—robot, control, model, prediction. i. introduction model predictive controllers (mpc) are widely adopted in industry as effective tools for dealing with large scale multivariable and multi-constrained control problems (guang et al,2005; camacho and bordons, 1999; nagy et al, 2005). the main idea of mpc lies in online construction of the system model, predicting its future states and generating the required control actions by repetitive solution of an optimal control problem. issues may arise for guaranteeing closed-loop stability, model uncertainty handling and reducing the on-line computations. there are three kinds of mpc controller schemes that use different methods for system modeling but are similar to each other in control process (likar et al, 2007) : 1mac: uses impulse response for system modeling, 2gpc : uses transfer function for system modeling, 3dmc: uses step response for system modeling. these controllers optimize a cost function that depends on the control law (hauge et al, 2002). although dmc is primarily developed for control of chemical processes (camacho and bordons, 1999; garcia et al., 1989; limon et al, 2005), it has been extended assistant professor of department of system and mechatronics engineering, faculty of new sciences and technologies, university of tehran, tehran, iran, arrezaee@ut.ac.ir. successfully to other applications such as motorway traffic systems (bellemans et al, 2006), switching max-plus-linear discrete event systems and simplified model of biped walking models(van et al, 2006; azevedo et al, 2002) . however implementation of this control scheme in robotics has been less reported and seems to be in still in its infancy (limon et al, 2005; kouvaritakis et al, 2006). in this work we concentrate on application of mpc/dmc controllers in position control of robotic systems. the rest of this paper is organized as follows: the next section presents the idea of mpc controller, section 3 describes p2at robot, section 4 discusses the effect of variation of dmc parameters on speed error, section 5 discusses results of the experimental implementation of mpc/dmc on a real robot, and finally the last section contains the conclusions. ii. model predictive controller the main strategy of a model predictive controller is illustrated in figure 1. in a typical mpc algorithm the system outputs are predicted for a certain interval of time (prediction horizon) by making use of a proper system model which is constructed based on the information (inputs and outputs) gathered from the system past as well as future control signals that have to be determined properly. as shown in the figure the control signal is a sequence of step functions with variable amplitude. amplitudes of these inputs are obtained by solving an optimization problem that tries to keep the system output close to the reference set point. objective function of this problem is usually a quadratic function of the difference between the predicted output signals and the reference trajectory. all the mpc algorithms using a linear model have similar behavior. here we demonstrate how dmc works. 0 yuay  (1) where a includes the step response, y is the predicted output, y0 is past output, and u is the control law (azevedo et al, 2002, shridhar and cooper, 1997). model predictive for mobile robot control alireza rezaee fig.1. methodology of mpc. due to uncertainities of the model it is very hard to achieve the exact value of a to satisfy the desired bahavior. to compansate for this problem an error term is added to the system output: n p n yycorrection  (2) 1 1 0 0 ... p f p n n y y y a u y a u y e y y                  (3) where the correction term represents the difference between the current plant actual output and the output extracted from the model. the error vector over prediction horizon is then written as 0 0 e r y r a u y e         (4) using the above expression a quadratic cost function, j, can be defined which is minimized to obtain the optimal controller uwuewej tt u   21 min (5) where w1 and w2 are constants. the modified control law is obtained as: '' 1 1 21 )( ekewawawau c tt   (6) fig. 2 shows the structure of a model predictive controller. in this configuration the block labeled as "model" contains the model of the robot that predicts the behavior of the robot over a certain time horizon. the future inputs (u(t+k|t)) are calculated under constrains and by optimizing a cost function. this process continues until the end of the trajectory. fig. 2. the structure of a mpc. a. the algorithm of a dmc controller the algorithm of a dmc controller is as follows: 1obtain model of the robot to be controlled. 2use the model to predict behavior of the robot over a certain time horizon. 3 calculated the e from equation (4). 4determine the control action by optimizing a performance index, which typically is the error between the outputs predicted from the model and the desired output over the time horizon. 5apply the optimal control actions and then measure robot outputs over the time horizon. the measured values at the final stage will be used as initial conditions of the model in the next iteration. 6repeat steps 2 to 5 until the end of the trajectory. iii. robot control for robot control with mpc controller we need to have the model equation of the robot (fig. 3). the robot under consideration in this study is a four wheel p2at robot in which wheels of the robot are controlled independently. fig. 3. p2at robot. in order to obtain the system model and design the proper controller for it, a sound appreciation of dynamic behavior of the system is needed. to do that a simple sketch of the robot is shown if fig. 4, it is assumed that the distance between each wheel is constant and four wheels have the same radius. kinematic model of the robot is described by x(k + 1) = [ x(k + 1) y(k + 1) θ(k + 1) ] =[ x(k) + ϑ(k)∆tcos(θ(k))cos⁡(α(k)) y(k) + ϑ(k)∆tsin(θ(k))cos⁡(α(k)) θ(k) + +ϑ(k)∆tsin(α(k))/l ]=f[z(k)]=f[ xm(k), u(k)] . . .. . . .. 2cos )cos( 2cos )cos(       w lax s w lax s o i       (7) where u(k) = [ϑ(k), α(k)]t is the control vector for motion tracking,∆t is the sample period and is and o s denote speed of the left and right wheels respectively and the distance between wheels is shown by l and w. moreover, a is the distance between reference point of the robot and the wheels. additionally, the position of the robot in global reference frame is specified by coordinates x and y. the angular difference between the global and local reference frames is given by θ. such a non-linear system is open loop controllable, which can be linearised in order to use traditional linear feedback control to regulate the robot. but if the robot operates over a large range in its state space, especially when the robot turns around corners, the linearization of the kinematics will lead to the loss of controllability. since the mpc’s models are based on linear regressions. fig. 4. schematic of the p2at robot. iv. the effect of dmc parameters on controller performance due to the simple nature of the linear mathematical models (mayne et al, 2000; axehill, 2004; axehila, 2004) most of the mpcs including impulse and step response models and the transfer function model are based on this type of model description (dougherty and cooper, 2004; gilbert and tan., 1991). thus, the first step in controller design is to linearize the model equations and then calculate the control laws. figure 5 shows the schematic of the mpc controller that is connected to the system (robot). rest of this paper is devoted to study effect of different parameters of the model on the controller performance. to do that a series of experiments were conducted on a simple straight path and speed of robot was measured for different instances (figs. 6-13). mpc controller is to optimize a cost index j(x(k),u(k))) under the constraints of formula 7: minu(k)j(x(k), u(k)) (8) the current control vector is chosen to minimise thee state errors and control energy over several steps in future so that the path tracking of the robot is smooth aand stable. therfore, the cost index can be expressed as j = 1 2 ∑ ‖xr(k + i) − x(k + i)‖ 2 + 1 2 n2 i=n1 ∑ i‖∆u(k + i − nμ i=1 d)‖2 (9) fig. 5. connection between mpc controller and the system (robot). a. effect of m (control horizon) figures 6, 7 show the effect of control horizon parameter, m, on the control output and control law respectively. it is observed that by increasing value of m the settling time is decreased and the control effort is increased. this also increases the computational complexity. according to the experimental results the optimal value of m selected as 2. m=1, n=20, p=5, α=0.5, w2=0. (a) m=2, n=20, p=5, α=0.5, w2=0. (b) m=4, n=20, p=5, α =0.5, w2=0. (c) fig .6. output. m=1, n=20, p=5, α =0.5, w2=0. (a) m=2, n=20, p=5, α =0.5, w2=0. (b) m=4, n=20, p=5, α =0.5, w2=0. (c) fig.7.control law b. effect of p (prediction horizon) figures 8, 9 show the system output and control law for two different values of parameter p respectively. it can be seen that by increasing value of p the settling time is decreased and the control effort is increased and the computational complexity is increased simultaneously. . m=2, n=20, p=3, α =0.5, w2=0. (a) m=2, n=20, p=10, α =0.5, w2=0. (b) fig. 8. output m=2, n=20, p=3, α =0.5, w2=0. (a) m=2, n=20, p=10, α =0.5, w2=0. (b) fig. 9. control law. c. effect of w2 figures 10, 11 show the effect of changing weight factor w2 (see eq. 5) on the system output and the control law. it is seen that by increasing the value of w2 increases the settling time while the control effort is decreased and computational complexity is not changed. m=2, n=20, p=5, α =0.5, w2=0.1. (a) m=2, n=20, p=5, α =0.5, w2=1. (b) fig.10. output. m=2, n=20, p=5, α =0.5, w2=0.1. (a) m=2, n=20, p=5, α =0.5, w2=1. (b) fig.11. control law. d. effect of α figures 12 and 13 show the effect of variation of  used in the input signal filter band and applied on the control output and control law respectively. according to these figure increasing the value of  , increases the settling time and decreases the control effort. however the computational complexity is not changed. m=2, n=20, p=5, α =0.7, w2=0. (a) m=2, n=20, p=5, α =0.9, w2=0. (b) fig. 12. output. m=2, n=20, p=5, α =0.7, w2=0. (a) m=2, n=20, p=5, α =0.9, w2=0. (b) fig.13control law. v. comparison of mpc with other control models to show the effectiveness of the mpc controller three different controllers (mpc, pid and adaptive) are implemented on p2at mobile robot and the system is tested in an elliptical path (fig. 14). pid control tuning is described at (gu et al,1997) fig.14.elliptical path. figures 15-17 show the error and its first derivative for different controllers. moreover the real path moved by the robot is given in the subplots. as figure 17 shows, the mpc controller has a lower error compared to the other control methods and can track the path more precisely. fig.15.robot path with pid controller. fig. 16. robot path with adaptive controller. fig.17. robot path with mpc controller. vi. conclusion in this paper implementation of mpc controller on p2at robot was explained. the conducted experiments show effectiveness of the proposed method on control of the mobile robot. furthermore the effects of the model parameters such as control horizon, prediction horizon, weighting factor and signal filter band on the controller performance were studied. finally, a comparison between the designed mpc controller and pid and adaptive controllers was presented demonstrating superior performance of the model predictive controllers. references [1] guang, li., lennox, b., zhengtao, ding., 2005, infinite horizon model predictive control for tracking problems, [2] control and automation,. icca '05. international conference on, pages 516 – 521. [3] camacho, e.f., bordons, c., 1999. model predictive control , springerverlag 2 edition. [4] nagy, z. , franke, r., mahn, b., allg , ower, f., 2005, real-time implementation of nonlinear model predictive control of inite time processes in an industrial,framework. in international workshop on assessment and future directions of nonlinear model predictive control, germany, pages 483-490. [5] likar, b., kocijan, j., 2007, predictive control of a gas-liquid separation plant based on a gaussian process model. computers & chemical engineering, 31(1):142–152. [6] hauge, t.a., slora, r. and lie, b., model predictive control of a norske skog , preliminary study, in proceedings of control systems,2002, june 3-5, stockholm, sweden, pages 75-79. [7] garcia, c.e., prett, d.m, morari, m., 1989. model predictive control: theory and practice, a survey. automatica, 25(3):335-348. [8] limon, d., álamo, t., camacho, e.f., 2005. enlarging the domain of attraction of mpc controllers, automatica. [9] kouvaritakis, b., cannon, m., couchman, p., 2006, mpc as a tool for sustainable development integrated policy assessment. ieee transactions on automatic control, 51(145–149). [10] bellemans, b., schutter, d., de moor, b., 2006, model predictive control for ramp metering of motorway traffic: a case study, control engineering practice, 14(7): 757-767. [11] van den boom, t.j.j., de schutter, b., 2006. mpc of implicit switching max-plus-linear discrete event systems timing aspects, proceedings of the 8th international workshop on discrete event systems (wodes'06), ann arbor, michigan, pages 457-462. [12] azevedo, c., poignet, p., espiau, b., 2002. moving horizon control for biped robots without reference trajectory. in ieee international conference on robotics and automation, pages 2762–2767. [13] shridhar, r., cooper, j., 1997. a tuning strategy for unconstrained siso model predictive control. england: chem.. [14] mayne, d.q., rawlings, j.b, rao, c.v., scokaert, p., 2000. constrained model predictive control: stability and optimality. automatica, 36:789814. [15] axehill d., 2004. a preprocessing algorithm with applications to mpc, 43th ieee conference on decision and control. [16] axehila, d., 2004. a preprocessing algorithm with applications to mpc , fifth conference on computer science and systems engineering, norrköping, sweden, october 21. [17] dougherty d., cooper, d., 2004. a practical multiple model adaptive strategy for single loop mpc. [18] gilbert, e. g., tan., 1991. k., linear system with state and control constraints: the theory and application of maximal output admissible sets. ieee transactions on automatic control, 36:1008-1020. [18]d. gu, h. hu, m. brady, f. li,1997, navigation system for autonomous mobile robots at oxford, proceedings of international workshop on recent advances in mobile robots, leicester, u.k., , 1-2: 24-33. alireza rezaee received his b.sc. degree in control engineering from sharif university of technology, iran 2002 and m.sc and ph.d degree in electrical engineering from amirkabir university of technology, iran (2005 and 2011 respectively). from 2013 till now he is an assistance professor in department of system and mechatronics engineering in new sciences and technology faculty at university of tehran. his research interest are in intelligent robotics, mobile robot, navigation, machine learning, bayesian networks, cognitive science http://ieeexplore.ieee.org/xpl/mostrecentissue.jsp?punumber=10234 i. introduction ii. model predictive controller a. the algorithm of a dmc controller iii. robot control iv. the effect of dmc parameters on controller performance a. effect of m (control horizon) b. effect of p (prediction horizon) c. effect of w2 d. effect of α v. comparison of mpc with other control models vi. conclusion references paper title (use style: paper title) transactions on environment and electrical engineering issn 2450-5730 vol 1, no 4 (2016) © heiko thimm towards a high reliable enforcement of safety regulations a workflow meta data model and probabilistic failure management approach heiko thimm abstract—today’s companies are able to automate the enforcement of environmental, health and safety (eh&s) duties through the use of workflow management technology. this approach requires to specify activities that are combined to workflow models for eh&s enforcement duties. in order to meet given safety regulations these activities are to be completed correctly and within given deadlines. otherwise, activity failures emerge which may lead to breaches against safety regulations. a novel domain-specific workflow meta data model is proposed. the model enables a system to detect and predict activity failures through the use of data about the company, failure statistics, and activity proxies. since the detection and prediction methods are based on the evaluation of constraints specified on eh&s regulations, a system approach is proposed that builds on the integration of a workflow management system (wms) with an eh&s compliance information system. main principles of the failure detection and prediction are described. for eh&s managers the system shall provide insights into the current failure situation. this can help to prevent and mitigate critical situations such as safety enforcement measures that are behind their deadlines. as a result a more reliable enforcement of safety regulations can be achieved. keywords— environmental health and safety, workflow management, workflows, failure detection, failure prediction; i. introduction multiple legal authorities with different responsibility levels obligate companies to follow environmental, health, and safety (eh&s) regulations [1] [2]. due to the enormous size of this ever growing and frequently revised set of eh&s regulations companies are required to establish an efficient and an effective practice for the enforcement of new regulations and of revisions of existing regulations [3]. ideally, this enforcement duty is performed trough carefully selected measures such as employee instruction and training, the use of additional safety devices and facilities, and even product revisions in order to reduce the potential of harm [3]. although there exists a great awareness about the need for a reliable and effective eh&s enforcement practice, often in companies deficiencies can be found in this area [4] [5]. the organizational deficiencies and inappropriate use of information and communication technology (ict) can create substantial eh&s risks and losses in efficiency and effectiveness. that the use of a workflow management system (wms) [6] will lead to a reliable, effective, and efficient eh&s enforcement in companies seems to be a promising approach. traditional wms are designed to enact and manage the execution of workflow instances according to given workflow models. typically, the system notifies participants about assigned activities and provides access to information artifacts. however, traditional wms are not designed to cope with problems that can occur in the context of the enforcement of eh&s regulations. this can lead to an unreliable enforcement of these regulations (i.e. non-compliance) because of activity failures within the execution of eh&s enforcement workflows. as a result critical situations can happen in which the company is threatened by financial losses, by health risks for humans, and by risks for damages to the environment. activity failures can emerge due to a variety of different reasons. first of all, activity failures can be men made. for example, individuals who are expected to complete eh&s activities can be over-challenged by what is demanded from them. they can also be insufficiently experienced/qualified or suffer from human factors. activity failures can also be caused by problems inherent to group work such as a bad group atmosphere and group thinks effects. malfunctioning and defects of components of the corporate technical infrastructure also have to be considered as potential sources for activity failures. the reliability of eh&s workflow completions can significantly be improved through the use of a wms that is capable to detect in realtime activity failures that occurred already or that are likely to occur (i.e. prediction) in the near future. such an enhanced system approach can help companies to prevent and/or mitigate the potential harm resulting from the failures. this consideration presents the overall research objective of the project described in this article. the research is part of a broader project that investigates the integrated use of both wms and eh&s compliance information systems in order to improve reliability of eh&s regulation enforcement. the focus of the article is on the foundation of this integrated approach which is a domain-specific workflow management meta data model. the modelling concepts of this model are specialized to the detection and prediction of activity failures. the modeling concepts are directed at the company specific organizational context, failure statistics, and proxy templates for real world activities. in the article the modelling concepts are exemplified through a concrete workflow example. furthermore, an overview of prototypical implementation is presented. the remainder is organized as follows. an overview of related work is contained in section 2. the domain-specific meta data model is described in section 3. examples of some major modeling concepts of the model are presented in section 4. 978-1-4799-7993-6/15/$31.00 ©2015 ieee section 5 describes the main principles of the detection and prediction of activity failures and section 6 gives an overview of a prototypical implementation. concluding remarks are contained section 7. ii. related work in the literature several projects are described that target the monitoring of workflows in order to detect compliance violations [7] [8]. an overview of the work in this field is given in [9]. it seems however that the core issue investigated in our research which is the prediction of workflow activity failures that may lead to compliance violations has not been addressed before. domain specific modelling has gained considerable attention in the research community. recommendations about when and how to develop domain-specific languages are given by mernik et al. [10]. an example of such a language is the extended compliance rule graph (ecrg) language [11]. an overview of domain-specific extensions of the popular bpmn modelling notation is contained in [12]. several research teams proposed domain-specific extensions for process modelling including extensions for the modelling of clinical pathways in the healthcare domain [13] and extensions to capture specific process requirements of the maintenance management domain [14]. conformance validation through traditional database technology has been the subject of several research teams. snodgrass et al. proposed to store additional information in the database in order to enable a separate audit log validator [15]. another approach is the use of event-condition-action rules. experience with this approach for support of clinical protocols is reported in [16]. various rule-based approaches addressing process monitoring and failure detection have been proposed. the realm approach developed by ibm research [17] is especially directed at compliance automation. regulations are first expressed based on logical models and then automatically mapped into processible rules. a comprehensive survey of online failure prediction methods and a proposal of a respective taxonomy is given in an article of a research group from the humboldt university in berlin [4]. in general, the failure prediction method of our work belongs to the so-called ‘classifiers’ that are one of several specializations of the so-called ‘symptom monitoring approaches’. the classifier approaches evaluate values of system variables directly in order to classify whether the current situation is failure-prone or not. for our system approach a more refined classification scheme has been devised with the categories ‘non-failure-prone’, ‘failure-prone’, and ‘highly failure-prone’. iii. domain specific workflow meta data model for the detection of activity failures companies are often advised to address eh&s regulations by establishing a corresponding management system according to the international norm iso 14001 [18]. a set of clearly defined processes that are oriented at a set of goals such as compliance to eh&s regulations serves as foundation of many management systems. the focus of the research reported in this article is on ict-supported processes to enforce eh&s obligations. in particular the research is focused on three eh&s obligations that require the existence of an eh&s regulation management database referred in the following by rm-db [3]. the three obligations are: 1. the obligation to systematically establish and maintain a central registry of relevant regulations as a part of the rm-db, 2. the obligation to carefully complete routine regulation management activities according to defined procedures (e.g. business processes and workflows, respectively), and 3. the obligation to record regulation management specific information in the rm-db. this documentation obligation includes the recording of context information and status information about workflow activities as well as results of completed activities. a main reason for this documentation task is that through logging of activities valuable persistent data is established. this data is of high importance when internal and external eh&s audits are performed. the above mentioned three central eh&s obligations require from companies to frequently perform eh&s regulation enforcement activities. a correct and careful completion of these activities requires to observe context-specific aspects such as specific organizational characteristics of the company (e.g. number of organizational units and decision boards). another context-specific aspect concerns the set of relevant regulation areas (e.g. occupational safety, waste, fire, air pollution, chemical, transportation safety). only in an ideal world never will required activities be missed and never will they fail the required outcome. for the non-ideal real world, however, one has to consider the possibility that actually required activities will not take place and that executing activities will not lead to the required outcome. we generally refer to such situations by activity failures which may tamper an organization’s efforts to enforce safety regulations with utmost reliability. a workflow management meta data model is proposed that is specialized on the above eh&s obligations. the meta data model is directed especially at activity failures. the model considers activity failures that may occur when eh&s workflows are performed. the model is intended to serve as a foundation of an approach to detect already occurred activity failures and to predict activity failures that are likely to occur. a concept diagram of the meta data model given in the popular martin notation [19] is shown in figure 1. the boxes denote real world phenomena of the universe of discourse that possess an identity of their own. the semantic relations between the modelled phenomena are denoted by labelled edges. the concepts at the top of figure 1 address the company specific eh&s context. the concepts at the middle layer are oriented at template data defined by modelers at workflow modelling time. the concepts at the bottom are directed at monitoring data about executing activities and also failure tracking data. one can envision that the concrete instances of the concept of activity are created (i.e. instantiated) from the corresponding activity templates (i.e. activity type). the activity instances serve as proxies for real world activities that are controlled and monitored for example on the basis of a wms. concepts for company-specific context data. the top part of the model in figure 1 models the eh&s specific company context. the concepts regulation area, organizational unit, 978-1-4799-7993-6/15/$31.00 ©2015 ieee and decision board are addressed. that companies are usually structured into different organizational units for which different sets of the regulation areas are relevant is the first intention addressed by these concepts. the second intention is to reflect that for each relevant regulation area of an organizational unit one may assign an individual set of decision boards that are responsible for decisions in the given area. for example, the decision boards might be responsible for the selection of measures to enforce eh&s regulations [20]. one of the motivation for data modelling of the eh&s specific company context is that the data can be used to obtain indicators about the complexity of activities. it is possible to obtain from these indicators lead times of activities that are useful for failure management. concepts for template data defined by workflow modelers. similar to other workflow data models the proposed model considers the concepts workflow type and activity type that serve as templates for concrete workflows and activities. that in companies with a good eh&s practice a set of pre-specified types of workflows and a set of corresponding types of activities are specified is the intention of these modelling constructs. several modelling concepts are considered in order to model specific details of activity types as concepts of their own. the construct occurrence pattern reflects the occurrence characteristics of activities such as if the activity is repeatedly executed during a given time period or if the activity is triggered by a specific events. the concept execution characteristics reflects the execution characteristics of activities such as if the activity is completed iteratively in several steps or in an all in one approach. the concept outcome constraint refers to the set of conditions by which the completeness and correctness of the activity result can be validated. for most of the activities these conditions specify the set of data values to be contained in the rm-db. the lead time concept is oriented at lead time specifications (i.e. minimum, average, and maximum lead time) per type of activity. note that every lead time specification refers exactly to one particular regulation area and one particular organizational unit. through this specification it is possible to check if given activity deadlines are met. the specification of the lead time of an eh&s activity in the form of an educated guess requires to consider three activityspecific aspects: 1. the type of regulation area that the activity deals with, 2. the characteristics of the business activity of the referred organizational unit (e.g. is hazardous material involved in manufacturing processes), and 3. the number of involved decision boards. consequently, per activity type a set of variants with individual lead times is considered. every variant is associated with an individual combination of regulation areas and organizational units. concepts for monitoring data and failure tracking data. the concept activity stands for activities that are performed according to the referring activity type. failures that already occurred during the activity completion and failures that are likely to occur are addressed through the following three modelling concepts. the concept of missed activity is oriented at activity failures that emerge when an activity that is required according to its occurrence pattern has not been considered until the given deadline. that is, not even a corresponding activity instance has been created. the concept of overdue activity refers to individual activity instances that have been initiated according to their occurrence pattern but that missed their deadline (already). in addition, activity instances are also treated as overdue activities when these activities are likely to miss their deadline. the concept of imperfect activity refers to initiated activity instances that fail to meet the set of outcome constraints at the given deadline. the statistic log record models a comprehensive event log about detected and predicted activity failures. the event log also contains accumulated statistical data such as the failure frequencies for the various different activity types. iv. exemplification of the meta model workflows models in general correspond to formal or semiformal specifications of a set of activities that serve the goal to partially or completely automate business processes [6]. to this end workflow specifications result from a refinement of business processes in terms of concrete activities and of the dependencies between activities such as temporal dependencies and input/output dependencies. the work flow specifications of our research are extended by domain-specific data, i.e. data specific to the domain of eh&s enforcement management. it is the target of this extension to establish a data foundation for the detection and prediction of missed activities, overdue activities, and imperfect activities that constitute activity failures as described above. the acquisition of the domain-specific data, for fig. 1. proposed workflow meta data model. 978-1-4799-7993-6/15/$31.00 ©2015 ieee example, can be performed by a corresponding extension of the tracking and logging system component of a wms. on the basis of the proposed meta data model, it is possible to complement information about completed and still executing workflows and activities by further data. it is possible to use the resulting rich data in order to detect occurred activity failures and to predict activity failures. by executing proper counteracting measures to mitigate and compensate the activity failures, it is possible to enforce eh&s regulations with a high reliability. for the now following description of sample data for activity failure detection and prediction an essential regulation enforcement process is considered. in accordance with its main objective this considered process is sometimes referred by new regulation management (nrm) process [3]. the main tasks of the nrm process are: 1. to ensure that new regulations which are potentially relevant for the company are recognized, 2. to evaluate new regulations in terms of the company specific relevance, and 3. to accordingly enforce new regulations through a careful selection and implementation of proper measures. like for all processes in the field of eh&s regulation management it is also a major task of the nrm process to comprehensively document the actions and the progress. from industry partners we learnt that workflow modelers are advised to establish an nrm workflow that is composed of six activities [3]. these subsequent activities are: a1: monitor, filter, and capture new regulation. the relevant information channels (e.g. eh&s information services) of the eh&s rule setters are monitored. the announcements that pass a first rough relevance check are registered in the rm-db. a2: judge the regulation relevance for the company. an evaluator judges the relevance of a new regulation for the company by assigning a relevance category to the regulation. a3: specify decision schedule for enforcement measures. when a new relevant regulation that requires enforcement measures is observed then a plan is defined for (collaborative) decisions about the set of required measures. among others, one needs to specify who is in charge of the decision and when is the deadline of the decision. a4: organize and complete measure decision(s). a decision manager organizes and controls the completion of the decision plan. a5: implement set of measures. an implementation manager organizes and controls the implementation of the set of measures according to the given implementation plan. a6: evaluate effectiveness of measures. a reviewer checks the effectiveness of the implemented set of measures. when the review result meets given success criteria then a confirmation entry is made in the rm-db. otherwise, another workflow is initiated in order to deal with the problem of the failing measures. when no new announcement of a new regulation was detected for a certain monitoring period – for example a calendar month – then only a very short version of the workflow is executed. the short version consists of the monitoring action of activity a1, the “closing” of the monitoring period, and the documentation that no new announcement was detected during the closed monitoring period. through this approach a coherent and traceable activity documentation for all monitoring periods is established. recall from earlier that the proposed meta data model contains specific concepts to model company-specific context data. the modelled context data can be used to determine the complexity of workflow activities. based on this complexity data and further data about executing workflows one can predict if activity failures are likely to occur. the sample data used to demonstrate company-specific context data correspond to the specific characteristics of a real company referred by the fictive name cexperts [3]. for competitive reasons the real name of the company behind cexperts is not disclosed in this article. the company is a globally acting german mid-sized manufacturer of industrial alcohol, chemicals, and polymer with two different production sites in germany. the eh&s workflows of cexperts are directed at 10 regulation areas that include water, occupational safety, waste, fire, radiation, and chemical. in the end of year 2015 the total body of regulations stored in cexperts’ corporate rm-db comprised roughly 2000 regulations in these 10 areas from several different rule setters at all different levels (world, world region, country, state, community). because, cexperts develops among others special chemical substances the potential enforcement measures include a) product revisions, b) infrastructure and compound revisions, c) manufacturing process revisions, d) workforce trainings and education, e) workforce instructions, f) workforce information. the eh&s organization of cexperts needs to deal with three different corporate organizational units. every unit is assigned to each of its relevant regulation areas a set of four decision boards. the people of these four decision boards possess complementary expertise in the fields of product management, logistics and transportation, occupational safety, and quality management. of the above described activities of the nrm process for three activities sample template specifications are given and are explained in the following. according to our meta data model these templates result from a business process modelling and workflow specification activity. a modelling environment such as the open source environment camunda bpm [21] which is able to derive processible workflow specifications from graphical process models can ease this activity. since the sample templates are intended to exemplify the meta data model in the following the specifications are stated in verbal form. the data values in these verbal specifications reflect the specific characteristics of the sample company cexperts. obviously, in a system implementation the verbal explanations are replaced by respective predefined and thus machine processible codes. table 1 contains the specification data for activity a1 (i.e. monitor, filter and capture new regulations) of the nrm process. the specification data of the occurrence pattern describes the conditions that trigger the execution of activity a1. as given by the sample data activity a1 is triggered when a new regulation announcement is recognized. the execution characteristics state that the activity is typically performed in a single step that requires only little time. that upon completion of activity a1 the rm-db has to contain a description of the new regulation is 978-1-4799-7993-6/15/$31.00 ©2015 ieee specified by the outcome constraint. a general rule for the deadline of activity a1 is specified by the deadline component. the lead times specification data correspond to the minimum, the average, and the maximum lead time of activity a1. table i. details of activity a1monitor, filter, and capture new regulation. concept specification data occurrence pattern activity is started by an individual nrm workflow that is triggered by a new regulation. execution characteristics execution is typically performed in a single step that only requires a negligible duration. outcome constraint rm-db contains a description of the new regulation including the deadline for the relevance evaluation performed in activity a2. deadline activity completion is required within one day. lead times all regulation areas: 1/1/1 in table 2 and table 3 the template data for the activity a3 and a4 are given, respectively. the interpretation of the specification data is strait forward and thus not explicitly described in this article. table ii. details of activity a3specify decision schedule concerning enforcement measures. concept specification data occurrence pattern activity is triggered by a preceding activity a2 when measures are required to enforce a new relevant regulation. execution characteristics execution is typically performed in several steps that require non-negligible durations. the larger the number of organizational units the more complex the decision schedule to be defined and the larger the time demand. outcome constraint the decision schedule that can be composed of a set of sub-decisions is fully described in the rm-db. deadline completion is required within 3 days after the new regulation has been registered lead times all regulation areas: 1/3/5 table iii. details of activity a4 – organize and complete measure decision(s). concept specification data occurrence pattern activity is triggered by a preceding activity a3. execution characteristics execution is typically performed in several steps that require a substantial duration. the more complex the decision schedule the more time is needed to complete the activity. outcome constraint the decision results (i.e. measures) are fully described in the rm-db. deadline completion deadline is given by the decision schedule. lead times water: 4/8/15; safety: 6/10/18; chemical: 10/21/34 v. probabilistic failure management approach today’s workflow management systems (wms) usually perform many tasks in order to execute workflows according to specifications given in the form of workflow models. this includes that for new workflow instances to be executed in the physical world, internal workflow proxy objects are created and maintained [6]. a specific wms component referred to by “workflow engine” usually performs these runtime proxy management tasks. the corresponding workflow models are specified through the use of a workflow modelling environment that is often an integrated component of wms. every proxy object represents and mirrors a referring real world workflow. similarly, workflow engines instantiate and maintain activity proxy objects for the constituent real world activities of workflows. in order to enforce that the execution of workflows and activities conforms to the referring workflow models, wms track workflows and activities in realtime and maintain a corresponding data log. in the following it is described how the proposed workflow meta data model can be used for a new approach to predict and to detect failures of executing safety enforcement workflows. this approach draws on an extension of the traditional workflow management data log by failure management specific data items as considered in the workflow meta data model. based on the extended logging data, a system instance is able to obtain the current failure status of ongoing workflows. the following section a gives an overview of the major principles of the proposed approach. the major considerations for the detection and prediction of activity failures are described in section b. section c contains a brief scientific evaluation of the proposed approach. a. major principles in general, the capabilities of wms to support workflow modeling and runtime execution management of proxy objects are based on a corresponding meta data workflow model. the majority of the existing workflow meta data models do not address domain specific concepts because wms are primarily developed as general purpose systems. as opposed to that, the proposed failure management approach builds on a wms that uses the workflow meta data model described above. that is, the concepts of the meta data model serve as basis for the specification of workflow models by the users and also for the runtime management of proxy objects (workflow instances and activity instances) by the workflow engine. the major principles of the failure detection and prediction approach are illustrated in figure 2. for every activity proxy object that refers to an enforcement activity, there is a comprehensive data set for failure detection and prediction supplied by the corresponding activity template. additionally, eh&s context data specified by the workflow modeler is considered for failure detection and prediction. also used are the statistic failure data and further logging data that are continuously maintained by the wms. the three types of activity failures addressed in the meta data workflow model (i.e. missed activities, overdue activities, and imperfect activities) are 978-1-4799-7993-6/15/$31.00 ©2015 ieee the target of the processing steps shown in figure 2. in the next section these steps are described in more detail. the failures that are identified in the steps are indicated in the form of activity failure objects. b. detection and prediction of activity failures the three processing steps that are shown in figure 2 are oriented at activity failures. step one targets the detection of occurred (i.e. evident) failures that are imperfect activities and overdue activities. the failures are detected by evaluating the constraints that are specified in the activity templates. obviously, the individual data available at runtime for every activity proxy are used for a corresponding constraint check. to give a concrete example, consider the above specification of the nrm process’s activity a1 (a1: monitor, filter, and capture new regulation). the outcome constraint requires that a description of the new regulation has to be available when the activity is finished. when this constraint is not met, an activity failure of type imperfect activity needs to be handled. similarly, by a comparison of corresponding proxy data with the individual activity deadline, it is possible to obtain activity failures that are overdue activities. note that the individual activity deadlines are computed from the generally specified deadline constraint of the referring activity template. it is the goal of step two to make use of available modelling data and runtime data in order to predict activity failures. in particular, the prediction step targets activity failures that did not yet occur but that are expected to happen in the future when no attention is paid to the potential failure cause. figure 3 gives a high-level overview of step two using activity a1 of the workflow described in section 4 as an example. at first a corresponding proxy object for activity a1 is instantiated. next, the set of failure probability indicators for the proxy object is computed based on data generally defined in the indicator formulas using the corresponding current data values. this data includes modeling data of the relevant activity template such as the deadline, the complexity, and the lead times. also logging data such as the completion status of the activity and data about new eh&s regulations are processed in order to obtain the failure probability indicators. the use of the complexity indicators reflects the general fact that usually there exists a direct positive correlation between an activity’s complexity and the likelihood of activity failures. moreover, also taken into account by the failure prediction methods are actual failure statistic data and data about the current status of the activities. the obtained failure probability indicators serve as basis for a decision step that follows next. in general, in this next step it is focused on two questions: 1. how likely is it that the activity will be completed until the given deadline? 2. how likely is it that the results expected from the activity will be achieved? in the decision step, based on the failure probability indicators, a qualitative prediction measure is obtained. this measure determines whether an activity failure for the investigated activity instance has to be considered or not. the activity failures that are predicted by the methods may either correspond to an overdue activity or an imperfect activity. an overdue activity is predicted when the given deadline will most likely be failed by the activity. when it is likely that the activity will not fulfill the specified outcome constraints, then an imperfect activity will be predicted. also in the third step a prediction of activity failures is performed targeting failures that are missed activities. the prediction method for missed activities makes use of the occurrence pattern and execution characteristics that are supplied by the activity templates. it is checked if the specified pattern and execution characteristics imply the existence of an activity. recall that these activities correspond to real world activities that are expected to be performed. in turn, it is checked if a corresponding activity proxy object exists, indeed. when two conditions hold true, 1. no proxy object is found and 2. the fig. 2. principles of failure detection and prediction approach. fig. 3. prediction of overdue activities and imperfect activities. a1: monitor, filter, and capture new regulation. a2: judge the regulation relevance for the company a3: specify decision schedule for enforcement measures. a4: organize and complete measure decision(s). a5: implement set of measures. a6: evaluate effectiveness of measures. system-level data processing for failure detection compute failure probability indicators new a1 proxy object is instantiated evaluate failure probability indicators failure probability for activity a1 is negligible? false activity failure predicted for activity instance a1 true no failure predicted for instance a1 real world completion of nrm process data about new regulation eh&s management database including central regulation repository e.g. constraints, lead times, context complexity e.g. completion status of a1 proxy instance workflow log activity templates domain-specific and general modeling data logging data statistic log eh&s context data step 1: failure detection through evaluation of activity constraints. step 2: failure prediction through evaluation of activity complexity and remaining time. step 3: failure prediction through evaluation of occurence patterns of activities workflow proxy objects activity proxy objects overdue acivities imperfect acivities missed acivities eh&s context data statistic log workflow log activity templates domain-specific and general modeling data logging data failure data proxy data 978-1-4799-7993-6/15/$31.00 ©2015 ieee temporal conditions of the activity and the workflow require the activity to be started, then a missed activity is predicted. c. an initial evaluation to our knowledge the approach to use a meta data workflow model that is specialized on failure detection and failure prediction in the domain of eh&s workflows is a new approach. in the next section it is described how this approach can be implemented leading to an integrated information system solution. the implementation of a corresponding research prototype is an ongoing project. it is planned to use the resulting prototype for comprehensive evaluation studies with the real world data of cexperts described in this work. the lab studies will provide validation data for our approach and insights about possibilities for improvement. looking at the “bigger picture” of our approach, one can already at this early research phase state, that the proposed solution bears promising potential to improve the reliability of safety regulation enforcing workflows. the enforcement workflows are actively monitored with respect to their temporal constraints and outcome constraints. also, the proposed solution provides the capability to detect missing workflows. when occurred failures are detected and failures that are likely to occur are predicted, corresponding failure data is made available for further failure handling actions. one can consider especially the failure handling action to actively provide users with alerts, failure data, and background data to effectively cope with the situation. it can be expected that this kind of active assistance being offered to safety managers for coping with failures in eh&s management tasks will promote reliability of regulation enforcement tasks. vi. prototype implementation a research prototype to demonstrate and evaluate the above probabilistic failure management approach has been devised. the prototype builds on an integrated information system that combines an extended wms with a specialized regulation management information system (rm-is). the extension of the wms concerns the support of workflow models that are augmented with detailed failure management data such as outcome constraints, occurrence patterns, and execution characteristics of activities. the rm-is is specialized towards the processing and storage of failure monitoring and tracking data such as data about missed activities and overdue activities. figure 4 illustrates the high-level architecture of the prototype. the database stores among other data the regulation registry with all regulations and corresponding relevance information, the eh&s workflow models (i.e. templates for workflow instances), data about ongoing and completed workflow instances and activity proxies, and data for the detection of activity failures that already occurred or that are likely to occur in the near future. failure detection query processing against the database is performed on request by interactive users who perform ad hoc queries. additionally, this query processing is also triggered by scheduled query batch jobs such as the generation of failure reports. in order to further clarify the notion of “extended wms” consider that today’s wms usually maintain an online log in order to track the states of ongoing workflows [19] in realtime. based on the status information, the wms determines and manages actions such as requests for completion of activities that are issued to workflow actors. for our approach, an extended wms is envisioned that performs a very fine grained logging of both, workflows and workflow activities, as specified in the workflow meta data model. especially, it is assumed that the begin time and completion time of every activity is logged through corresponding time stamps. the architectural model defines four components that periodically update database objects for failure management purposes. the proxy initializer assigns to each new created activity proxy the corresponding set of initial values such as the individual activity deadline, the appropriate lead time values, and failure probabilities. note that these initial values are copied from the respective activity type. the constraint checker checks the set of outcome constraints of activity proxies and reports the result in the respective activity property (oc_passed). the query set of a proxy that specifies the outcome constraints is executed until one of the following two termination conditions is reached. 1. when all outcome constraints are satisfied (i.e. all queries result to true), then the checking task is completed. 2. when the activity is completed and the query set was executed one more time after the activity completion, then the checking task is finished, too. the statistics updater component maintains data about failures stored in a failure occurrence log. the component also computes from this log statistical failure data in order to keep the corresponding attribute values of activity types up-to-date. that is, the statistics updater periodically updates the database with the latest statistical data about failures. this updating (“learning”) mechanism contributes to a proper degree of precision of predicted activity failures. at the level of activity instances, failure prediction is performed based on a symptom monitoring approach [4]. a set of activitytype specific indicators is periodically evaluated in order to classify whether the current activity execution status is ‘nonfailure-prone’, ‘failure-prone’, or ‘highly failure-prone’. the indicators include failure statistical data stored at the respective activity type (e.g. failure history) and relevant facts about the activity instance such as the complexity of the activity and the tightness of timing constraints. the resulting classification fig. 4. high-level architecture of prototype. statistics updater proxy initializer constraint checkerregulation registry failure detection data eh&s workflow data eh&s workflow models background query processing failure detection query processing specialized regulation management information system extended wms wms log extended failure predictor 978-1-4799-7993-6/15/$31.00 ©2015 ieee decision is reported in the respective activity instance’s attributes (p_iact, p_oact, and p_mact) with values ‘unlikely’, ‘likely’, or ‘highly likely’. the processing according to these principles is performed by the component failure predictor. note when a new activity instance is created the failure probability failures are copied from the respective activity type. these values are periodically updated in every processing cycle of the failure predictor in order to reflect the evolving individual execution situation of the activity instance. in a first implementation step the rm-is has been implemented. the interfacing of the developed rm-is with the extended wms is simulated through corresponding data files. so far, the use of traditional relational database technology for the prototype implementation did not lead to any “dead ends” or extraordinary “workaround approaches”. the latest versions of the popular sql standard [20] supports language extensions by user-defined concepts such as user-defined data types and userdefined functions. for example, for several properties of eh&s activities user-defined datatypes have been developed according to the workflow meta data model. it is expected that with the further advancement of the demonstrator this sql extensibility feature will even be more exploited. the database tables activity and failure-logentry that store data about activities and failure monitoring data are described in table iv and v, respectively. data about activity failures are stored in the tables fail-overdue-act, fail-imperfect-act, and fail-missed-act. the outcome constraints of activities are encoded into sql queries which evaluate if the database contains all of the data items that are required by the outcome constraints. in order to give an overview of the proposed approach for the detection and prediction of activity failures, several sample queries are described next. for the graduation of activity failures with respect to the indication of failure occurrence four categories are used. the category “occurred” refers to evident failures that already occurred. whereas, failure prediction results are classified into the three categories “highly likely”, “likely”, and “unlikely”. table iv. database table ‘activity’ field type description aid int unique identifier of activity act_type int type of activity (foreign key) deadline date absolute deadline of activity lt_min int minimum lead time lt_avg int average lead time lt_max int maximum lead time ts_start datetime time stamp of start of activity ts_end datetime time stamp of end of activity oc_passed bool result of outcome constraint check performed by the constraint checker ts_oc_chec datetime time stamp of constraint check p_iact char probability of imperfect activity with possible values ‘unlikely’, ‘likely’, ‘highly likely’ p_oact char probability of imperfect activity with possible values ‘unlikely’, ‘likely’, ‘highly likely’ p_mact char probability of imperfect activity with possible values ‘unlikely’, ‘likely’, ‘highly likely’ table v. database table ‘failure-log-entry’ field type description lid int unique identifier of log entry ts_entered datetime time stamp of log entry activity int activity concerned (foreign key) fail_type char type of failure with possible values overdue, imperfect, missed fail_occat char occurrence category of failure with possible values ‘occurred’, ‘highly likely’, ‘likely’, ‘unlikely’ fc_before int no. of failures before the failure fc_after int no. of failures after the failure fr_before_ int failure rate before the failure fr_after int failure rate after the failure the following two sample queries are directed at the detection of failures that are overdue activities and imperfect activities, respectively: select aid, “occurred” into fail-overdue-act from activity where not isnull(ts_end) and ts_end > deadline; select aid, “occurred” into fail-imperfect-act from activity where not isnull(ts_end) and not oc_passed; the next two sample queries predict (i.e. search for) activity failures that are highly likely: select aid, “highly likely” into fail-overdue-act from activity where isnull(ts_end) and lt_min > (deadline – now()); select aid, “highly likely” into fail-imperfect-act from activity where isnull(ts_end) and not oc_passed and now() < deadline and p_iact = “highly likely” the third query checks for overdue activities by selecting activities for which the remaining time is smaller than the minimum lead time. the fourth query predicts highly likely failures that are imperfect activities through respective conditions in the where-clause. of the set of not yet terminated activities that did not exceed the deadline, those activities are selected, that did not yet pass the outcome constraint check. recall that the outcome constraints of all currently executing activities are frequently evaluated by the constraint checker in parallel to the other query processing activities. when all defined constraints are fulfilled, then the boolean value ‘true’ is assigned to the property ‘oc_passed’ of the respective activity instance. the clause ‘p_iact = “high likely”’ is directed at restricting the selection to activities for which a high failure likelihood was assessed by the failure predictor. 978-1-4799-7993-6/15/$31.00 ©2015 ieee vii. conclusion the research being described in this article aims on a new domain-specific approach for the real time detection and prediction of failures of workflow activities. the context of the failures are activities to enforce eh&s regulations. it seems to be possible to achieve a more reliable enforcement of safety regulations through a timely detection and prediction of activity failures including missing activities. the proposed failure detection methods and failure prediction methods make use of a diverse data set. this data set includes complexity indicators of activities, failure statistic data, and status data about activity proxy objects. the use of company specific organizational context data in order to obtain an indication of the complexity of activities is one of the novel ideas of the proposed approach. a standalone prototype version of a probabilistic failure detection system is under development that follows the above described approach. the prototypical implementation builds on experience gained with the ccpro system. ccpro is a research prototype of an environmental compliance management information system [20]. traditional relational database technology is used for the prototype in order to make sure that the proposed failure detection and prediction approach can easily be adopted by existing eh&s management systems. it appears that through user-defined functions and active capabilities such as triggers, relational database technology offers sufficient support for the implementation of the intended failure detection and failure prediction methods. in the next phase the prototype will be integrated with a wms that supports the definition of domain-specific modelling concepts such as the yawl system [22]. references [1] aberdeengroup, "compliance management in environment, health and safety, white paper 6991," aberdeengroup, boston, ma, 2011. [2] n. gunningham, "enforcing environmental regulation," journal of environmental law, vol. 23, no. 2, pp. 169-201, 2011. [3] h. thimm, „it-supported assurance of environmental law compliance in small and medium sized enterprises,“ int. journal of computer and information technology, pp. 297-305, 2015. [4] j. petts, "small and medium sized enterprises and environmental compliance: attitudes among management and non-management," in small and medium sized enterprises and the environment, r. hillary, ed., sheffield, greenleaf publishing, 2000, pp. 49-60. [5] b. walker, j. redmond, l. sheridan, c. wang and u. goeft, "small and medium enterprises and the environment: barriers, drivers, innovation and best practice. a review of the literature," edith cowan university, australia, 2008. [6] w. van der aalst and k. van hee, workflow management: models, methods, and systems, cambridge, ma: mit press, 2004. [7] m. weidlich, h. ziekow, j. mendling, o. guenther, m. weske and n. desai, "event-based monitoring of process execution violations," proc. business process management (bpm 2011), clermont-ferrand, france, vol. lncs 6896, pp. 182-198, 2011. [8] a. rozinat and w. van der aalst, "conformance checking of processes based on monitoring real behavior," inf. syst., vol. 1, pp. 64-95, 2008. [9] m. kharbili, a. de medeiros, s. stein and w. van der aalst, "business process compliance checking: current state and future challenges," proc. mobis'08, saarbrücken, germany, pp. 107-113, 2008. [10] m. mernik, j. heering and a. sloane, "when and how to develop domain-specific languages," acm computing surveys, vol. 37, no. 4, pp. 316-344, 2005. [11] d. knuplesch, m. reichert and a. kumar, "visually monitoring multiple perspectives of business process compliance," proc. 13th int. conference on business process management (bpm 2015), innsbruck, austria, vol. lncs 9253, pp. 263-279, 2015. [12] r. braun and w. esswein, "classification of domain-specific bpmn extensions," proc. 7th ifip wg8.1 working conference, poem 2014, manchester, uk, vol. lnbip197, pp. 42-57, 2014. [13] r. braun, h. schlieter, m. burwitz and w. esswein, "extending a business process modeling language for domain-specific adaptation in heathcare," proc. 12th int. conference wirtschaftsinformatik (wi2015), osnabrück, germany, pp. 468-481, 2015. [14] m. lopez-campos and a. c. marquez, "modelling a maintenance management framework based on pas 55 standard," quality and reliability engineering international, vol. 27, no. 6, pp. 805-820, october 2011. [15] r. snodgrass, s. shilong and c. collberg, "tamper detection in audit logs," proc. 30th vldb conference, toronto, canada, pp. 504-515, 2004. [16] k. dube, b. wu and j. grimson, "using eca rules in database systems to support clinical protocols," proc. database and expert systems applications (dexa2002), aix-en-provence, france, vol. lncs 2453, pp. 226-235, 2002. [17] c. giblin, s. müller und b. pfitzmann, „from regulatory policies to event monitoring rules: towards model-driven compliance automation,“ ibm rüschlikon, switzerland, 2006. [18] international standards organization (iso), iso 14001:2015 environmental management system, 2015. [19] j. martin, information engineering: planning & analysis, book ii, englewood cliffs, nj: prentice-hall, 1990. [20] h. thimm, "ict support for collaborative environmental compliance management in smes the ccpro approach," ieee int. conf. collaboration technologies and systems, pp. 295-301, 2015. [21] a. fernandez, "camunda bpm platform loan assessment process lab," 2013. [online]. available: http://fundamentals-of-bpm.org/wpcontent/uploads/2013/11/camunda-bpm-loan-assessment-processlab-v1.0.pdf. [accessed 22 january 2016]. [22] w. van der aalst, l. alfred, m. dumas and a. h. t. hofstede, "design and implementation of the yawl system," in advanced information systems engineering, vol. 3084, springer, 2004, pp. 142-159. heiko thimm has been a full professor for information technology and quantitative methods since 2008 at the school of engineering of pforzheim university in germany. in 2004 he joined kiel university of applied sciences as a full professor for business information systems after working 6 years for sun microsystems and sap, respectively, as it solution architect and system analyst. from 1991-1998 he was a researcher at the german national research centre for information technology (gmd) where he was involved in various pioneering internet and database research projects. he earned a phd degree from the technical university in darmstadt and a msc degree from new jersey institute of technology both in computer science. in his current research he investigates the use of recent it technology advancements such as sensor networks, machine learning, and big data approaches for the development of next 978-1-4799-7993-6/15/$31.00 ©2015 ieee generation environmental compliance management and sustainability management information systems. he is especially focusing on supporting corporate environmental compliance and eh&s managers with active and smart assistance systems that are able to provide, among others, risk management information. prof. thimm is also involved in various industry 4.0 / cpps projects with partners from the german car/automotive manufacturing industry.  transactions on environment and electrical engineering issn 2450-5730 vol 2, no 2 (2017) © thiago martins pereira, rodrigo alves lima, viviane margarida gomes, wesley pacheco calixto, aylton josé alves  abstract— this work study the influence of concrete, plaster, clay and others buried structures in grounding systems. comparison of soil characteristics between dry and rainy seasons on different grounding systems. the study includes comparison of six different grounding system on dry season and wet season. simulations in finite element method was performed for tree layer stratified soil and the electrostatic equipotential surfaces were mapped into the region of interest. index terms— grounding systems, finite elements method. i. introduction erbert g. ufer was in charge of the facilities from davis montana military base, one of the tasks was to protect the bombs warehouse from atmospheric discharge [1] he utilized structural system to reinforce the grounding efficiency from the traditional grounding rod system [2]. subsequent inspections of the installations showed that combined grounding systems presented lower resistance and greater consistency in high electrical resistance soils than grounding systems without concrete structures[3]. concrete is a hygroscopic substance and therefore, absorbs water more easily than lost. for that reason, concrete presence in soil helps to keep soil humidity levels and grants a lower resistance to soil [1]. grounding systems are used for many different functions, from noise reduction for better functioning of electronic equipment to security applications, power systems and substations grounding is important to maintain stable and secure systems for equipment and users [4]. this work starts from the hypothesis that the electrical behavior of soil is altered by climate seasonality [5]. the grounding grid efficiency was studied during the rainy season and during dry season in combined and non-combined systems. concrete used in construction basically consists of a mixture of cement, water and crushed stone. buried concrete block has equivalent behavior to a semiconductor element with resistivity between 30ω and 90 ω as ieee indicates their standardizations [3] [6]. grounding probes were build using different types of material.in monitoring the current on dry soil and moist soil studying behavior of materials on different humidity. concrete hygroscopic feature helps both concrete and soil to remain moist, lowering soil resistance [2]. this moisture is present between the solid particles of the soil, so it consists basically water, organic minerals and dissolved inorganic [7]. utilization of the structural columns was also mentioned to reinforce the grid in its function [8]. in the 70’s it was indicated the use the enclosed electrodes in concrete, with a view to improve grounding grid performance. the lack of standardization in telecommunications wiring was a concern among operators and only in 1991 there was standardization and regulation, facilitating the use of ground [11]. among the main functions of grounding because the grounding system importance, are highlighted four most common applications. protection systems and security function to living beings and protection to equipment [11]. thus, grounding installation purposes includes personal safety in the handling and maintenance of equipment, avoiding dangerous tensions. grounding systems are also capable of providing overvoltage protection, limiting noise and crosstalk in transmission systems and serve as return path for dc circuits. additionally, they serve as protection in case of lightning [12]. grounding probes were built using different types of composition from clay to concrete with salt. the different hygroscopy from materials makes different electric current drained. analysis and monitoring of electrical grounding grid encapsulated with concrete: case study using simulation in finite element method thiago martins pereira, rodrigo alves lima, viviane margarida gomes, wesley pacheco calixto, aylton josé alves h ii. methodology a. data acquisition to verify the influence of concrete efficiency of grounding systems built two-ground grid. these structures were installed on same topographical area and have same number of rods, differentiated only by the presence of one of concrete in mesh. data acquisition made at two-week intervals to monitor the effect of climate seasonality in meshes checking the humidity, temperature and grounding resistance. the wenner method is a method for measuring resistivity of homogeneous soil which four rods are inserted into equally spaced ground like figure 1. the central terminals are used to determine the voltage side terminals are for power insertion into the ground. current flowing between the rods produce a potential in voltage measuring rods, with voltage and current values, wenner show a correlation between soil resistivity and measured resistance. increasingly the rods distance, more current will penetrate into soil and a deeper soil resistivity will be measured. another six grounds grid were built and installed on another place on same topographical area. this six was inserted on soil has different constitution using the same 50mm conductor. the difference on constitution display how de grounding resistance changes in different soil. the figure 2 displays how was disposed and measured the grounding resistance. the first on left has concrete, gravel and salt, second concrete and gravel, third only concrete, fourth plaster, fifth clay and last on right has only conductor buried on soil. the mold used to build grounding system guarantees size of grounding systems. grounding systems has 100cm length, 10cm depth and 10cm of width. the mold used is described on figure 3. the fall-of-potential method was used to obtain curve that represents the locations of grounding resistance. [17] this method consists the equipment called megohmmeter that generates a know current, between earth electrode and the outer stake, while the drop voltage potential is measured between the earth stake and the outer current stake. [18] the distance used between earth stake and current stake was 40m like nbr 15749 says. the potential stake is moved every 5 meters from earth stake to current stake. there is seven different values measured on these grounding systems. according the measured values, is drawn a graphic that shows grounding resistance. the expected graphic on figure 4 shows resistance with distance and the influence of earth stake and current stake on measure. the linearity region distance between earth stake and current stake changes according different soil types. regularly the linearity region is 37,5% to 62,5% between the roads. in case of the curve don’t present the linearity level some mistake may have occurred on rods or cable connections. b. simulation software finite element method (fem) allows to simulate and validate data obtained from soil stratifications [13]. performance of ground and influence of neighboring structures to ground grid fig. 2. simplified disposition in soil of six different grounding grids fig.3. grounding system mold fig. 1. simplified disposition in soil of six different grounding grids are observed for solution of potential surfaces in plane and on edge of layers. simulation allows observing dissipation of electric current and influence of aggregate structures to ground grid [14]. finite element method consists of a mathematical analysis based on discretization of a continuous environment into small elements while maintaining the same characteristics of the original environment. all elements are described in differential equations and then they are solved using mathematical models. the accuracy and performance of method depends on number of elements and nodes. smaller elements and consequently greater amount thereof and greater number of nodes in mesh greater will be precision of resolution of problem. even when dealing with an approximation method, increasing amount of elements size tends to zero and so the amount of us tend to infinity. when this occurs the problem solution tends to an exact solution, i.e., the smaller the size of the largest elements accuracy of the analysis results. modeling involves the reproduction of main geometric and electrical aspects of ground grid [15]. figure 5 presents the soil modeling of details in detail. it illustrated three interconnected rods without concrete and three interconnected concreted rods. soil characteristics, resistivity and depth are calculated and used in simulator in stratified three-tier model and forty meter radius. mesh construction detail is shown in figure 5. rods are made of copper with 5/8 inch in diameter and 2.4 meters long, on right its shown rods combined with concrete, being enclosure is 30 cm radius and 2.40 meters long. distance between rods is 4,5 meters, distance between ground mesh is 3 meters. simulations were done by inserting 200v and 2000v in each grounding system. considering the distance on fall-off-potential 40m between grounding system and current road was used this size to build the model like figure 6. iii. results a. data acquisition the table 1 shows characteristics of soil collected in the field during the rainy season and dry season. these data allow the comparison between soil resistance and response by the absorption and retention of water [16]. second column corresponds to soil characteristics in dry season and third column corresponds to rainy season. figure 7 shows the resistance curves depending on the positioning of electrodes for same periods of year. the red line represents values obtained using three points in mesh without concrete and blue curve is the response of ground with concrete. fig. 6. building of the soil in fem simulation software on the grid and their properties. fig. 5. construction in fem software of ground grid. fig. 4. the fall of potential method figure 8 shows the ground due to the spacing of the rods resistance curve showing the increase of efficiency of the mesh after the concrete insertion along the rods. how much moisture in the soil after rain considerably alters response of soil resistivity, improving ground resistance. blue line is no concrete ground resistance before the rain, and red line is the grounding strength concrete after rain. wenner method used to measure soil resistivity, this method related to the amount of soil resistivity and the resistance measured. wenner's method used in the 4 rods equally spaced ground straight. two side rods used for the insertion of electrical current in the ground, since central rods used to determine tension. wenner method considers the homogeneous ground and if it departs from rods is considered value of resistivity of same depth as distance between the rods [12]. the table 2 shows acquired values were sampled only 4 values starting with 1 meter to 6 meters, because the space was short and the distance of 6 meters between rods already requires an area of length of 18 meters, and the length available for measurement was 20 meters. table 3 shows resistivity of soil layers over a longer period presenting lower humidity 3/10/2013 and higher humidity 04/20/2014, according to values shown in table 3. was built figure 9 with depths of layers having the thicknesses changing according to the moisture retained by the soil. the six grounding probes was tested in different soil moisture, on figure 10 has shown de worst and better grounding resistance curve. the first probe made by concrete, gravel and salt, has the betters results on dry season and wet season. table iv shows de difference of grounding resistance between dry and raining season measured in ohms. the first ground shows that concrete is better than any another grounding on wet season or dry season. the biggest difference fig. 7. chart show grounding resistance at end of dry season in relation to distance from stems using werner method. 0,0 200,0 400,0 600,0 800,0 1000,0 5 10 15 20 25 30 with concrete without concrete probes distance (m) r e si st a n ce ( ω ) table i data collection before the rainy season. feature drought rainy pattern ii soil moisture soil temperature humidity ambient temperature resistance grounding concrete grounding resistance standard precipitation 25.0 % 32.2 ºc 27.0 % 31.0 ºc 275.7 ω 537.0 ω 0,0 mm 70.0 % 26.3 ºc 59.0 % 28.4 ºc 165.3 ω 361.7 ω 45.0 mm 99.925% fig. 8. comparing ground not concretes rain and concreted after rain. 0,0 200,0 400,0 600,0 800,0 1000,0 5 10 15 20 25 30 before rain without concrete after rain with concrete probes distance (m) r e si st a n ce ( ω ) table ii wenner method distance drought rainy pattern ii 1 m 2 m 4 m 6 m 125.0 ω 160.0 ω 110.0 ω 160.0 ω 815.3 ω 2030.2 ω 2771.4 ω 2113.5 ω 99.925% table iii soil resistivity layer 10/03/2013 10/17/2013 12/04/2013 04/20/2014 pattern ii 1 2 3 352.93 ω 1411.61 ω 2795.98 ω 1033.88 ω 10062.9 ω 623.3 ω 749.92 ω 3591.83 ω 404.78 ω 469.44 ω 1663.14 ω 1071.28 ω 99.925% fig. 9. chart the depths of soil layers on dry season was 337% and on dry season 521%. b. computational method figure 11 shows the equipotential lines form insert 200 v on rods of ground grid. interaction with rods concreted ground grid. the outer loop ensures that there is a higher voltage drop, maximizing the absorption of electrical current through the ground grid. simulation of charge distribution in ground by inserting a 2000 v voltage can see in figure 12. iv. conclusions as first measurements and studies in experimental ground on the dates of september 7 and october 3, 2013 has considerable difference in resistance of grounds with and without concrete after a long period of drought and after rain during dry season grounding with the rods wrapped in concrete already had a degree of improved efficiency that grounding without concrete. the concreted loop resistance was 275.7 ω and without concrete showed resistance 537.0 ω a difference of almost 200%. difference is accentuated by concrete characteristics such as moisture, its resistivity is smaller than ground contact area with ground is increased, which helps the current distribution in soil. a study can be done is influence of meshes creating interaction between them, when there injection voltage in a mesh, there is the scattering voltage at this soil. due to the proximity of ground grid, note change in the lines of equipotential caused the second ground grid. figure 5 illustrates the difference in grounding resistance level in mesh without concrete before the rain and the mesh concreted after rain. use of fem to determine equipotential surfaces due to scattering of current through soil and its mesh with next ground with these images it is possible to study and see what electrical interaction between layers and substances buried in ground make to ground. differences in responses are already expected between two meshes, we can put data in a computer simulator for verification of equipotential surfaces and consequently ground to ground response. several aspects to be analyzed, since strength of concrete in ground, moisture retention of concrete in relation to retention of moisture from soil. fig. 11. shows equipotential surfaces in soil and interaction between two grounds studied with inclusion of 200 v. fig. 10. grounding resistance curves 0 200 400 600 800 1000 1200 1400 1600 1800 2000 5 10 15 20 25 30 35 grounding 1 wet grounding 1 dry grounding 2 wet grounding 2 dry grounding 3 wet grounding 3 dry grounding 4 wet grounding 4 dry grounding 5 wet grounding 5 dry grounding 6 wet grounding 6 dry r e si st a n ce ( ω ) probes distance (m) fig. 12. simulation of equipotential lines in plane parallel to surface of soil. table iv grounding resistance gnd 1 gnd 2 gnd 3 gnd4 gnd 5 gnd 6 pattern ii dry wet 524 213 798 367 730 477 643 451 708 603 1769 1111 99.925% references [1] o. vicente, “study on the electrical behavior of concrete used in structural grounding systems,” 2010. [2] a. nadler and h. frenkel, “determination of soil solution electrical conductivity from bulk soil electrical conductivity measurements by the four-electrode method,” soil science society of america journal, vol. 44, no. 6, pp. 1216–1221, 1980. [3] p. w. rowland, “industrial system grounding for power, static, lightning and instrumentation, practical applications,” in textile, fiber and film industry technical conference, 1995., ieee 1995 annual. ieee, 1995, pp. 1–6. [4] l. cong-li and p. minfang, “a new approach for monitoring grounding grid of electrical power system,” in electronic measurement and instruments, 2007. icemi’07. 8th international conference on. ieee, 2007, pp. 4–419. [5] w. p. calixto, “mathematical and computational methods applied to geoelectric prospecting with threedimensional stratification,” in portuguese ph.d. dissertation, 2012. [6] j. b. j. pereira, “modeling of uncertainties in electrical grounding systems” in portuguese 2008. [7] w. p. calixto, l. m. neto, m. wu, h. j. kliemann, s. s. de castro, and k. yamanaka, “calculation of soil electrical conductivity using a genetic algorithm,” computers and electronics in agriculture, vol. 71, no. 1, pp. 1–6, 2010. [8] e. j. fagan and r. h. lee, “the use of concreteenclosed reinforcing rods as grounding electrodes,” industry and general applications, ieee transactions on, no. 4, pp. 337–348, 1970. [9] j. preminger, “evaluation of concrete-encased electrodes,” industry applications, ieee transactions on, no. 6, pp. 664–668, 1975. [10] t. r. brinner, j. d. atkins, and m. o. durham, “electric submersible pump grounding,” industry applications, ieee transactions on, vol. 40, no. 5, pp. 1418– 1426, 2004. [11] m. santos, w. cunha, and a. nascimento, “infrastructure in systems of communication technology”, vol. 1, no. 1, 2010. [12] g. kindermann and j. m. campagnolo, “grounding systems” in portuguese. sagradc luzzatto, 1992. [13] j. paknahad, k. sheshyekani, f. rachidi, and m. paolone, “lightning electromagnetic fields and their induced currents on buried cables. part ii: the effect of a horizontally stratified ground,” ieee transactions on electromagnetic compatibility, vol. 56, no. 5, pp. 1146–1154, oct 2014. [14] y. wang, r. chen, s. lin, j. tian, j. li, and m. chen, “calculation and analysis of the current carrying capability of electric cable based on finite element method,” in electrical insulation and dielectric phenomena, 2009. ceidp ’09. ieee conference on, oct 2009, pp. 307–310. [15] j. a. guemes and f. e. hernando, “method for calculating the ground resistance of grounding grids using fem,” ieee transactions on power delivery, vol. 19, no. 2, pp. 595–600, april 2004. [16] a. habjanic and m. trlep, “the simulation of the soil ionization phenomenon around the grounding system by the finite element method,” ieee transactions on magnetics, vol. 42, no. 4, pp. 867–870, april [17] ma, jinxi; dawalibi, farid p. extended analysis of ground impedance measurement using the fall-of-potential method. ieee transactions on power delivery, v. 17, n. 4, p. 881-885, 2002. [18] ma, jinxi; dawalibi, farid p. influence of inductive coupling between leads on ground impedance measurements using the fall-of-potential method. ieee transactions on power delivery, v. 16, n. 4, p. 739-743, 2001. i. introduction ii. methodology a. data acquisition b. simulation software iii. results a. data acquisition b. computational method iv. conclusions references microsoft word transactions_teee-1 transactions on environment and electrical engineering issn 2450-5730 vol 2, no 1 (2017) © calebe a. matias, licínio m. santos, aylton j. alves, wesley p. calixto abstract— this paper presents the development of a cooling apparatus using water in a commercial photovoltaic panel in order to analyze the increased efficiency through decreased operating temperature. the system enables the application of reuse water flow, at ambient temperature, over the front surface of photovoltaic (pv) panel and is composed of an inclined plane support, a perforated aluminum profile and a water gutter. a luminaire was specially developed to simulate the solar radiation over the module under test in a closed room, free from the influence of external climatic conditions, to carry out the repetition of the experiment in controlled situations. the first case study was published at eeeic2016 conference where the panel was submitted to different rates of water flow, from 1 l/min to 4 l/min. in the test conditions without cooling apparatus, the panel reached about 70°c and produced approximately 63wh. with the cooling apparatus with water flow rate of 2 l/min, the module reached about 50°c and produced approximately 77wh. however, it has been observed that this water flow was overestimated. a second case study was carried out in order to perform the threshold between the flow and the energy produced. the best ratio was flow of 0.6 l/min and net energy of 77.41wh. gain of 22.69% compared to the panel without the cooling system. index terms— efficiency increase, increasing of the electrical performance, photovoltaic panel, water cooling. i. introduction lthough abundant on earth, solar energy for electricity production is still little used. this problem stems from factors such as low efficiency in energy conversion, which can be influenced by external agents as solar radiation and temperature, and the high cost of pv panels [1]. the electrical characteristics of photovoltaic cells can be changed due to intrinsic and extrinsic factors. some factors are resulting from the manufacturing process and the material the authors would like to thank coordination for the improvement of higher education personnel (capes), the national counsel of technological and science development (cnpq) call cnpq-setec/mec no. 17/2014, project 468544/2014-3 and research support foundation of goias state (fapeg) for financial support research and scholarships. c. a. matias {1}, calebeabrenhosa@gmail.com l. m. santos {1}, liciniomoraes@gmail.com a. j. alves{1}, aylton.alves@ifg.edu.br w. p. calixto{1,2}, wpcalixto@ieee.org 1 experimental and technological research and study group (next), federal institute of goias (ifg), goiânia, brazil. 2 school of electrical, mechanical and computer engineering, federal university of goias (ufg), goiânia, brasil. used, and others are environmental factors such as solar radiation and ambient temperature [2]. increasing the operating temperature of the pv panel leads, in particular, to the reduction in open circuit voltage (���) and a slight increase in short circuit current (���) which does not compensate for the loss caused by the reduction of voltage, therefore there is a clear loss of power and loss of power conversion efficiency (�) [3]. the fig. 1 shows the power versus voltage (� � �) curves for different operating temperatures, demonstrating the maximum power point drop and the voltage drop due to the temperature increase in the panel [4]. hybrid systems, photovoltaic/thermal (pv/t), can be used to circumvent the problem of high operating temperature of the panel, for there is a gain in the electrical and thermal efficiency compared to individual systems [5]. this model uses a fluid, usually air, as references [5]-[7], or water, as references [8]-[10], in order to reduce the operating temperature of the panel, transferring heat to the used fluid [11]. the increased electrical efficiency in pv/t systems can reduce the payback period of the investment and also increase the life of the system, which contributes to the growth of economic feasibility for implementation of this type of system and encourages the generation of electricity by this source [1], [12]. this work aims to develop a cooling apparatus of a photovoltaic panel applying reuse water at ambient temperature and analyze the efficiency gains in electricity conversion when the panel is under different rates of water increasing photovoltaic panel power through water cooling technique calebe a. matias1, licínio m. santos1, aylton j. alves1, wesley p. calixto1,2 a fig. 1. dependence of the photovoltaic power output on panel operating temperature. flow and was published at eeeic2016 conference [13]. this method consists of applying water on the front surface of the panel in order to remove the maximum amount of heat, as references [8]-[10]. preliminary studies will be performed a quantitative comparison of the increase in power generated, using the apparatus, when the panel is under different rates of water flow (1 l/min, 2 l/min, 3 l/min and 4 l/min), as reference [13]. an original experiment, second case study, was carried out varying the flow rate from 1 l/min to 0.125 l/min, since the flow observed in case study 1 was overestimated. the power conversion efficiency of photovoltaic systems is defined as the ratio of the generated electric power by the total power incident over the panel: ag iv mm × × =η (1) where � and � are respectively the maximum voltage and maximum output current, is the incident solar radiation on the photovoltaic device and � is the area of the pv panel [14]. the percentage of improved efficiency from the pv panel is given by: 100× × = η ηη α m (2) where � and η are, respectively, the efficiency of the modified module and the efficiency of the original module [15]. ii. methodology a photovoltaic panel is cooled using water at ambient temperature to enhance the voltage and thereby increase the power generated by the panel. a. materials the set of devices used to cool the pv panel consist of a luminaire capable of simulating solar radiation conditions; a water tank, which will supply water through gravitational action; a 140w photovoltaic panel, where will be verified the effect of the temperature in power generation; temperature sensors to measure the ambient and the panel temperature; a flow sensor for monitoring the amount of water delivered to the pv panel; a valve to control the water flow; a maximum power point tracker (mppt) device to extract the maximum power supplied by the panel integrated with a communication device, which will receive all the data and will provide it to a computer from serial link rs-232; a computer for data analysis. the diagram depicted in fig. 2 represents the used system, where � is the luminaire, ��) is the water tank, ��� is the water valve, �� is the water flow sensor, � is the water input on the photovoltaic panel, �� is the water output, ��� is the mppt device and ���� is the computer. the experiment was conducted in the city of goiânia, brazil, in a room, protected from external climatic conditions, where is installed. a luminaire capable of simulating solar radiation was used aiming at providing stable and constant irradiance throughout the experimental procedure to ensure that the same conditions are repeatable. this step was adopted in order to standardize the irradiance conditions. lm35 sensors was used to measure the ambient temperature and the pv panel temperature. ds18b20 sensors measured the inlet and outlet water temperature. the photovoltaic panel was installed on an inclined supporting structure, which allows water runoff through gravitational force. the fig. 3 shows the apparatus developed, where: 1 is the built luminaire capable of simulating solar radiation, 2 is the water inlet, 3 is the perforated metal profile for distribution of water over the module, 4 is the pv module and 5 is the water outlet gutter. b. data acquisition the simulated solar radiation was fixed at 800w/m² for all the tests. fig. 2. operating diagram of the system. fig. 3. developed laboratory bench. the photovoltaic panel was subjected to four different rates of water flow (1 l/min, 2 l/min, 3 l/min and 4 l/min) in order to establish which provide the highest net increase of power. the second case study was carried out varying the flow rate from 1 l/min to 0.125 l/min. the first case study was conducted at an interval of 30 minutes in which the values of voltage, current, power, ambient temperature and photovoltaic panel operating temperature were stored in the data logger. iii. results a. case study 1 fig. 4 indicates that the water flow rate on the photovoltaic panel influences the generated power. fig. 5 presents a wide variation in panel temperature during the experiment. the temperature sensor is positioned in the center of the panel, where it was observed higher values. it is observed in fig. 4 and fig. 5 as the instantaneous power without cooling system decrease when the panel temperature increase. the initial panel temperature is equal the ambient temperature, approximately 25°c, and it gets about 70°c without cooling system. when applied the water flow on the photovoltaic panel its temperature drops and the instantaneous power rise. it was expected that higher water flow rates leads to enhancement of the power production, but it is not observed in fig. 4. the data shows that the water flow rate of 2 l/min generates more power than others. fig.6 depicts the randomness of the water flow on the panel surface. the distribution of temperature is not the same in all points of the panel, which may vary suddenly the path taken by the water, this may explain why the water flow rate of 4 l/min, for example, in a few moments, it was less effective for improving panel efficiency than lower flow rates. table i fig. 4. effect of water flow rate in the generation of the maximum power. fig. 6. thermal image of the photovoltaic panel when the water flow is 2 l/min. fig. 5. effect of water flow rate in the panel temperature. table i temperatures indicated by the thermal imager. markers temperature (°c) p1 58.0 p2 41.8 p3 49.6 p4 68.2 fig. 7. effect of water flow rate in the voltage produced. indicates the temperature provided by a thermal imager and a huge difference in temperature at nearby points. panel work operating temperature declines the voltage and therefore induces the decrease of power. fig. 7 shows the influence of temperature on the voltage produced by the photovoltaic panel. the power, depicted in fig. 4, follows the voltage change, for the current keeps almost constant, approximately 4.7 a, during the experiment, observed in fig. 8. the total energy produced using different water flow rates can be seen in table ii. the water flow corresponding to 2 l/min showed the highest gain efficiency, 22.90%, compared to the test done without the cooling apparatus. b. case study 2 in order to overcome the problem of heating, as shown in fig. 6, a solenoid valve has been inserted into the system to obstruct the passage of water for a certain period of time with the purpose of ensuring that the water flows through the entire surface of the panel. it is observed in case study 1 that the volume of water applied in the panel is exaggerated, since the difference between the generated potencies is not large. a new experiment was performed by setting the water flow rate to 1 l/min and activating the solenoid valve at different time intervals. the tests were performed under the same conditions as in the case study 1. the panel was exposed to the luminaire for 30 minutes without the cooling system. after this period the solenoid valve was switched on for 10 minutes at a constant flow rate of 1 l/min. in order to vary the flow rate of the water, the valve was acted in different time intervals (ton = on and toff = off). the valve opening and closing time is given in seconds and is shown in table iii. 1 l/min is the reference and other values are expressed as a percentage. the interval �� � �� shown in fig. 9 is 30 minutes and the interval �� � �� is 10 minutes. figure 9 shows the data obtained from this experiment, where the first graph shows the power generated by the panel, the second shows the panel temperature, the third shows the fig. 8. effect of water flow rate in the current produced. table ii efficiency gain of the pv panel with the cooling sy stem.. water flow (l/min) energy (wh) efficiency gain (%) without 63.09 1 74.81 18.57 2 77.54 22.90 3 75.62 19.86 4 75.32 19.38 fig. 9. data from the case study 2. water outlet temperature and the fourth shows the average water flow rate applied to the panel. fig. 10 (a) shows the thermal image of the pv panel. as can be seen, the control through the solenoid valve, identified in fig. 10 (b) by the number 1, forces the water to always take a different path cooling most of the photovoltaic panel. table iv shows the temperature markers of fig. 10 (a). table iv temperatures indicated by the thermal imager. markers temperature (°c) p1 42.3 p2 44.6 p3 51.8 p4 59.5 considering the steady state of the power and temperature in the interval between �� � ��, that is, in the last 10 minutes, a comparison is made between the energy generated in each time interval and the gain of the efficiency with respect to the energy produced by the panel without cooling during the stability period. the energy was calculated in relation to the average produced during the stability period. the water flow is in relation to 1 l/min. the efficiency gain is shown in table v. c. data management it was observed that the amount of water applied on the panel was overestimated, as higher flows did not generate more energy. another problem observed was the stagnation of the water path. some regions remained at higher temperatures. in case study 2 the water flow rate varies from 1 l/min to 0.125 l/min. the flow control was performed through the opening and closing period of a solenoid valve. the flow that provided the highest net energy, table v, was 0.6 l/min, showing an efficiency gain of 24.8%. this flow had a level of efficiency gain similar to case 1, since the path traveled by the water varied over the whole area of the photovoltaic panel. table v efficien cy gain of the pv panel with the cooling sy stem.. water flow (%) energy (wh) efficiency gain (%) without 63.09 100 71.12 12.72 87.5 75.65 19.90 80 76.67 21.52 66 76.68 21.54 60 78.74 24.80 50 77.77 23.27 40 76.85 21.81 33 76.23 20.82 20 75.20 19.19 12.5 75.05 18.96 the best case observed was 60% of water flow rate: 0.6 l/min. the reservoir of water is located in the room where the experiment was performed. the tank is at a height of 1.5 meters from the water inlet of the panel. in order to quantify the energy spent in a real system, was proposed a hypothetical situation in which the reservoir would be 6 meters height distant from the capture and treatment system of reuse water installed in facilities of the institution. an analysis was performed to determine the amount of energy needed by a pumping system to elevate the water to a height of approximately 6 meters. it is desirable that the used pump has the lowest power possible, since the flow rate values and head height satisfies the proposed pumping system in order to achieve a more efficient system, with higher yields and that does not compromise achieved power gain using the cooling apparatus. the suggested pump is the etanorm 050-032-125.1220v, single-phase with 139mm rotor, and a 370w motor. this set was obtained by analysis of the technical datasheet using a flow rate of 10 ��/� and head height of 8 mca. table vi shows the amount of water required to cool the panel for 1 hour when subjected to several different flow rates. it was also shown the power required to transport this amount of water using the same pump, but with different operating a) thermal image of the photovoltaic panel when the water flow is 0.6 l/min. (b) position of the solenoid valve. fig. 10. bench structure. time. the net energy gain, calculated by the difference between the energy generated with the cooling technique and the energy expended in the water transport, is shown in the last column. table vi shows that the best water flow rate to cool the photovoltaic panel is 0.6 l/min, because it provided the greatest net energy. there was an increase of 22.69% efficiency, when the temperature was 70ºc on the panel, which may result in reducing the payback period of investment. another effect of the implementation of this apparatus is to produce the same amount of energy in a smaller space than a conventional photovoltaic plant. table vi net efficien cy gain of the pv panel with cooling system. water flow generated energy (wh) pump energy (wh) net energy (wh) efficiency gain (%) without 63.09 4 75.32 8.88 66.44 5.30 3 75.62 6.66 68.96 9.30 2 77.54 4.44 73.1 15.86 1 74.81 2.22 72.59 15.05 0.875 75.65 1.94 73.71 16.83 0.8 76.67 1.77 74.9 18.71 0.66 76.68 1.46 75.22 19.22 0.6 78.74 1.33 77.41 22.69 0.5 77.77 1.11 76.66 21.50 0.4 76.85 0.888 75.96 20.40 0.33 76.23 0.73 75.5 19.67 0.2 75.20 0.44 74.76 18.49 0.125 75.05 0.27 74.78 18.52 the cost of the water used is virtually null, given that the industries should dispose of wastewater effluent properly in rivers, lakes or oceans, according to local regulations, after the treatment of solid waste in water treatment plants. this apparatus therefore can be implemented in industries where there is this kind of water supply, encouraging the generation of photovoltaic electricity. iv. conclusion this work shows that decreasing the panel operating temperature, when subjected to cooling apparatus, is the factor responsible for the increase of the voltage and consequently the increase of the amount of energy produced. in initial conditions, without cooling apparatus, the photovoltaic panel produced 63.09 wh, for 70ºc on the panel, and after using a water flow rate of 0.6 l/min it produced around 78.74wh, for about 50ºc on the panel, a gain of 24.80%. a hypothetical case of water transport was implemented to simulate the amount of energy needed to cool the panel for 1 hour. then the comparative analysis of the increase in efficiency using the apparatus reveals that the water flow of 0.6 l/min on the front surface of the panel provides the highest net power increase, 22.69%. the apparatus, consequently can be implemented in industrial facilities, where there is reuse water potential, in order to increase the amount of energy produced and/or reduce the payback period of the investment. the results show that the water distribution system, under the pv panel can be improved to optimize the efficiency of the water flow used. references [1] h. a. hussien, a. h. numan, and a. r. abdulmunem, “improving of the photovoltaic/thermal system performance using water cooling technique,” iop conf. series: materials science and engineering, vol. 78, no. 012020, 2015. [2] m. sharma, k. bansal, and d. buddhi, “real time data acquisition system for performance analysis of modified pv module and derivation of cooling coefficients of electrical parameters,” procedia computer science, vol. 48, pp. 582–588, 2015. [3] s. dubey, j. n. sarvaiya, and b. seshadri, “temperature dependent photovoltaic (pv) efficiency and its effect on pv production in the world a review,” energy procedia, vol. 33, pp. 311–321, 2013. [4] e. skoplaki and j. palyvos, “on the temperature dependence of photovoltaic module electrical performance: a review of efficiency/power correlations,” solar energy, vol. 83, pp. 614–624, 2009. [5] s. c. solanki, s. dubey, and a. tiwari, “indoor simulation and testing of photovoltaic thermal (pv/t) air collectors,” applied energy, vol. 86, pp. 2421–2428, 2009. [6] s. dubey, g. s. sandhu, and g. n. tiwari, “analytical expression for electrical efficiency of pv/t hybrid air collector,” applied energy, vol. 86, pp. 697–705, 2009. [7] s. dubey, s. c. solanki, and a. tiwari, “energy and exergy analysis of pv/t air collectors connected in series,” energy and buildings, vol. 41, pp. 863–870, 2009. [8] y. m. irwana et al., “indoor test performance of pv panel through water cooling method,” energy procedia, vol. 79, pp. 604–611, 2015. [9] v. eveloya, p. rodgers, and s. bojanampati, “enhancement of photovoltaic solar module performance for power generation in the middle east,” 28th ieee semi-therm symposium, pp. 87–97, 2012. [10] d. kim, d. h. kim, s. bhattarai, and j.-h. oh, “simulation and model validation of the surface cooling system for improving the power of a photovoltaic module,” journal of solar energy engineering, vol. 133, no. 041012, 2011. [11] h. zondag, “flat-plate pv-thermal collectors and systems: a review,” renewable and sustainable energy reviews, vol. 12, pp. 891–959, 2008. [12] j. ji, j. lu, t. chow, w. he, and g. pei, “a sensitivity study of a hybrid photovoltaic/thermal water-heating system with natural circulation,” applied energy, vol. 84, pp. 222–237, 2007. [13] matias, c. a.; santos, l. m.; alves, a. j. and calixto, w. p. electrical performance evaluation of pv panel through water cooling technique. (eeeic) ieee 16th international conference on environment and electrical engineering, 2016. doi: 10.1109/eeeic.2016.7555643. [14] m. chandrasekar and t. senthilkumar, “experimental demonstration of enhanced solar energy utilization in flat pv (photovoltaic) modules cooled by heat spreaders in conjunction with cotton wick structures,” energy, vol. 90, pp. 1401–1410, 2015. [15] t. chow, “a review on photovoltaic/thermal hybrid solar technology,” applied energy, vol. 87, pp. 365–379, 2009. nazwa pliku: transactions_teee-1 katalog: c:\users\admin\documents\dokumenty_msi\dydaktyka\konferencja\e eeic2017 szablon: c:\users\admin\appdata\roaming\microsoft\szablony\normal.dotm tytuł:  temat: ieee transactions on magnetics autor: słowa kluczowe: komentarze: data utworzenia: 2017-02-20 13:00:00 numer edycji: 3 ostatnio zapisany: 2017-02-20 13:01:00 ostatnio zapisany przez: zbigniew leonowicz całkowity czas edycji: 11 minut ostatnio drukowany: 2017-02-20 13:08:00 po ostatnim całkowitym wydruku liczba stron: 6 liczba wyrazów: 2 849 (około) liczba znaków: 17 095 (około) transactions on environment and electrical engineering issn 2450-5730 vol 2, no 2 (2017) © antonio marcelino silva filho, carlos leandro borges silva, marco antonio assfalk oliveira, thyago gumeratto pires, aylton josé alves, wesley pacheco calixto, marcelo gonçalves narciso  abstract—this paper presents the study of the relationship between electrical properties and physical characteristics of the soil. measures of apparent electrical resistivity of the soil were made for different types of soil, varying moisture content gradually while maintaining a constant compaction, and then varying the compaction and relating it to a constant humidity. development of a correlation surface is proposed in order to identify granulometry of the soil from moisture and compaction measurements. for the study of spatial variability, two areas were chosen to allow the change of moisture content and compaction in order to verify the measurement capacity of apparent electrical resistivity of the soil as methodology to identify change in soil dynamics. results obtained show correlations among apparent electrical resistivity of the soil, moisture, soil compaction and clay content. index terms— geoelectric prospecting, apparent electrical resistivity, soil compaction, moisture content. i. introduction oil can be considered as an electrical conductor having a tortuous path or a large number of conduction paths with variable lengths and cross sections. soil property to conduct electric current is called apparent electrical conductivity of the soil σa, which can be calculated from measurements of the voltage v collected in field, after applying a current i to the soil. methods used in precision agriculture require fast and accurate answers to map the potential productivity of a given area [1]. one of the principles of precision agriculture is based on the property of soil to vary the apparent electrical conductivity according to variation of its physical and chemical properties such as texture, moisture, compaction, hydraulic potential, organic matter content, etc. for these reasons, apparent electrical conductivity of the soil has attracted the attention of researchers and investors for being a fast, noninvasive and inexpensive method [2]. in addition, investigations related to σa can aid in the measurement of clay and water content [10]. on a plot, areas with homogeneous physical and chemical characteristics can be identified with the mapping of σa. with these values mapped geographically within land, it is possible to divide the regions in management areas and then proceed to collect some samples for analysis and, depending on their physical and chemical properties, make decisions on how to intervene with the inputs, pesticides and irrigation. the mapping of σa is becoming an effective tool in the investigation of physical and chemical behavior and spatial variability of the soil, allowing to identify areas with similar properties and easily define management zones [3]. this paper strives to present the development of methodology to correlate soil electrical conductivity, soil moisture content (w), clay content (δ) and soil compaction (c). the development of theoretical and experimental mechanism allowing abacus production for identification of soil granulometry from moisture and compaction measurements also comprises a goal of this paper. ii. methodology electric conductivity is an intrinsic property of the whole electric current conductive material. in geoelectric prospecting, the conductor is the soil in which the electric current flows through the presence of free salts in the soil solution (liquid phase) and also due to the exchangeable ions in the particle surface [6]. however, unlike what happens with a wire conductor of electricity, the electric current in the ground can cycle through several ways, as illustrated in fig. 1. in moist soils, the current conduction occurs mainly through the salt content in the soil water which occupies the largest pores, region 1 in fig. 1. however, there are also solid phase contribution to the electrical conductivity in moist soils mainly by exchangeable cations associated with the clay mineral, region 2 in fig. 1. a third path for electric current in the soil exists through particles in direct and continuous contact with each other, region 3 in fig. 1. these three paths of current flow contribute to the overall electrical conductivity of the soil known as apparent electrical conductivity of the soil σa [5]. geoelectric method applied in correlation between physical characteristics and electrical properties of the soil antonio marcelino silva filho, carlos leandro borges silva, marco antonio assfalk oliveira, thyago gumeratto pires, aylton josé alves, wesley pacheco calixto, marcelo gonçalves narciso federal university of goias, brazil s apparent electrical resistivity of the soil ρa (which is the inverse of σa) is obtained through field measurements using geoelectric prospecting methods, the wenner method [7]. the calculation of soil electrical resistivity is performed by an instrument which measures soil electrical resistance rm [8]. this instrument has four terminals, t1, t2, t3, t4, connected to four rods spiked at depth p at the points q1, q2, q3, q4, aligned and separated by the same distance a. electric current i is injected into the terminal t1 (q1) and collected in the terminal t4 (q4), which produces electric potential v at points q2 and q3. with i and potential difference v between q2 and q3, resistance rm can be calculated. thus, apparent electrical resistivity of the soil ρa [ωm] is given by [4]:   2222 )2()2( 2 )2( 2 1 4 pa a pa a ar a m a        (1) this value ρa varies with distance a, since the soils are heterogeneous. thus, the magnitude happens to be called apparent electrical resistivity ρa(a) where the resistivity values are a function of a [11]. fig. 1. path of electric current in the soil. a. lateral profiling lateral profiling method is the geoelectric investigation technique used in the horizontal mapping, which consists of the relocation of electrodes in the following points at each measurement, keeping a fixed distance among the rods [9]. wenner array is used for the application of this technique, where apparent electrical conductivity of the soil σa is determined at midpoint of the arrangement between potential electrodes. entire area can be mapped by identifying these midpoints, delimiting different conductivity regions. fig. 2 illustrates the technique for research. in this case, the method applies in column 1 at a1, b1, c1 and d1. in this example the wenner method is used to apply the lateral profiling. the horizontal mapping of apparent electrical conductivity is determined by performing electrical pathway of columns 2, 3 and 4. fig. 2. lateral profiling method. fig. 3. pvc cell. b. experimental procedure laboratory study for the application of the proposed methodology, a pvc container (cell) is built in order to contain the soil samples. constructed cell has r = 37,5mm internal radius, l = 80mm height, wall thickness of e = 4mm and 4 holes for insertion of current and potential electrodes for the application of wenner arrangement, as illustrated in fig. 3. 1) sample preparation sample preparation consists in doping the same with known amounts of clay. from the sandy soil, samples are prepared with different proportions of clay, varying the texture of each sample. each portion (sample) must contain the same mass. in each portion, withdraws certain amount of soil and inserts the same amount of clay, in order to obtain samples with a known clay content. for example, to prepare the sample with clay content of 20%, it is weighed initially the full portion of collected soil and withdraws the equivalent of 20% of its weight, inserting in place the same amount of clay, in order to obtain the sample doped with 20% clay. this methodology is adopted in order to obtain samples with a gradual variation in granulometric classification of the soils. after this doping step, samples were sent to laboratory for analyze granulometric classification of the same. the tab i provides the classification results. 2) relation among electrical resistance (rm), soil moisture (w) and clay content (δ) relationship among electrical resistance of the soil rm, soil moisture w and the clay content δ consists of performing rm measures in soils with different clay content, gradually varying the moisture content, keeping the compaction c constant. the rm values are obtained from geoelectrical prospection methods which detect effects produced by the flow of electric current in the soil. method of geoelectric prospection utilizes wenner arrangement for data collection, where electric current is injected and captured in the soil by current electrodes, and voltage is measured at two points of surrounding region by potential electrodes. soil samples are first taken to the oven (heater) in order to reduce moisture levels to values close to 0%. after passing through the desiccator cooling process, it is gradually added portions of water equivalent to 5% of the mass of the dry sample. at each increment of water portion to the sample, the corresponding value of electrical resistance rm is measured by the electrical resistivity meter as illustrated in fig. 4(a). for each sample, the compaction load is constant. 3) relation among electrical resistance (rm), soil compaction (c) and clay content (δ) the practical experiment to relate rm, c and δ consists of performing rm measurements in soils with different clay content, gradually varying the compaction c keeping moisture content w constant. soil samples are taken to the oven (heater) and cooled in desiccator. after passing through the cooling process, all samples are gradually moistened until they have constant moisture, w ≈ 20%. at each increase in compaction values, the corresponding value rm is measured by the electrical resistivity meter. apparatus of fig. 4(b) is designed in order to apply pressure on the sample that is within the pvc cell. under the pvc cell, it places the scale to measure the force exerted on the sample. rm is measured at each increment of 1 kg in compaction. fig. 4. pvc cell connected to electrical resistivity meter: (a) rm x w x δ and (b) rm x c x δ. c. experimental procedure field study two areas are chosen to enable the change of moisture content and compaction, in order to verify the measurement capability of ρa to identify changes in soil dynamics. wenner method is used along with lateral profiling method for determining ρa. the entire area is mapped in the state it is in. one sub-region within the area should be chosen, where water should be added (sub-region with different color in fig. 5), modifying the moisture. again, the survey of ρa is performed in the same region. this same methodology can be applied to the mapping of the soil penetration resistance by simply, instead of adding water to modify the dynamics of the system, change the soil compaction. iii. results the soil for laboratory analysis was classified as gleysol and collected in location s1346’14.5” w5048’16.1”. a. relation among electrical resistance (rm), soil moisture (w) and clay content (δ) fig. 6 illustrates equipment and connections used to analyze the correlation rm x w x δ, where red cables are electrical current injection electrodes and black cables are electrical potential electrodes. initially, the humidity of each sample is measured by moisture meter, in order to verify whether soil samples are with humidity values near 0%. then, the mass of table i granulometric classification of the samples clay [%] granulometric classification 0 sand 20 sandy loam 40 sandy clay loam 60 sandy clay 80 sandy clay 100 clay each dry sample is measured in order to calculate the mass of water corresponding to be inserted gradually in the sample. fig. 5. delimited area for the application of lateral profiling method. fig. 6. equipment for the analysis of correlation rm x w x δ. at each insert 5% water portion, resistance rm is measured with electrical resistivity meter, for each one of the 6 samples. in any sample could be obtained value of rm for dry soil, therefore, for this situation the resistance value is greater than the set full scale. values of rm were obtained for moisture contents above w = 10%, except for gleysol sample. this type of soil allows to obtain resistance values at lower moisture levels by their sandy granulometry, which, due to its porosity, facilitates the circulation of electric current through the liquid phase located in the larger pores. fig. 7(a) shows the curves of 6 soil types under study in order to evaluate the effect of moisture w and clay content in rm values. resistance rm varies considerably with the change of soil moisture content w and clay content δ (fig. 7(a)). maintaining constant compaction, electrical resistance rm decreases with increasing moisture and clay content. due to the grain size, the sandy sample (0% clay) has a higher pore area than the other samples. with increased moisture content, these spaces occupied by air become occupied by water allowing smaller values of rm with increasing humidity, resulting in high percentage value in the reduction of electrical resistance rm. b. relation among electrical resistance (rm), soil compaction (c) and clay content (δ) apparatus of fig.8 was developed to apply pressure on the samples for the experiment of correlation rm x c x δ. a scale placed below the cell performed measurements of the force applied on the sample. it changed gradually the pressure, in order to perform measurements rm at each 1 kg of applied load. pressure applied to the soil sample is calculated through the value read on the scale. red cables in fig.8 are electrical current injection electrodes and black cables are electrical potential electrodes. fig. 7. correlation curves: (a) rm x w x δ and (b) rm x c x δ fig. 7(b) shows the curves of the 6 soil types under study in order to verify the effect of compaction c and clay content δ in rm values. each 1 kg load applied, the resistance rm is measured by earth meter for each of the 6 samples. for all of them, humidity is approximately constant w ≈ 20%. resistance decreases with increasing clay content and the increase in the levels of soil compaction, particularly on soils with greater presence of clay fractions in its constitution. for samples of sandy soil and sandy loam soil, where there is a predominance of sand fractions, the influence of compaction on rm values was lower when compared to samples with higher clay content, indicating rm behaves differently for sandy and clay soils. fig. 8. apparatus developed for the analysis of correlation rm x c x δ. c. correlation surface with the data of fig. 7, correlations can be established among resistance rm, moisture, soil compaction and clay content in order to facilitate the identification of soil granulometry from moisture and compaction measurements. fig. 9 shows the correlation surface rm x c x w x δ, where the plane x,y represents the relationships rm x w and rm x c and the z axis represents the clay content. on this surface, the data used were the saturation, or the average moisture of w = 28,99% and average compaction of c = 37,78kpa. the color grading in the figure corresponds to granulometric variation of soil. fig. 9. correlation surface. from the saturation data of rm in fig. 7, clay content can be determined through the correlation surface. for example, in the case of moisture saturation was obtained rm = 1252ω (fig. 7(a)), while for the case of compaction saturation was obtained rm = 2930ω (fig. 7(b)). with these two points in the x,y plane, clay content in z axis can be obtained from correlation surface (fig. 9). note that the value found is δ = 0%, or sandy soil. thus, for the case of an argisoil with moisture content w ≈ 28% and compaction c ≈ 37kpa, granulometry of the soil can be identified on the correlation surface. for this, the use of tab i is required, which defines the intervals: 0% ≈ 20% sand soil; 20% ≈ 40% sandy loam soil; 40% ≈ 60% sandy clay loam soil; 60% ≈ 100% sandy clay soil and equal to 100% clay soil. d. identification of soil moisture stains through lateral profiling method. the region defined for the study has 5 meters long and 5 meters wide. lateral profiling was used with spacing a = 1m and p = 20cm depth. with data of rm and through equation 1, the apparent resistivity matrices before and after the insertion of water are produced, fig. 10(a) and fig. 10(b) respectively. fig. 10. moisture stains mapping through lateral profiling method (ρa). the colors of fig. 10(a) and fig. 10(b) represent areas with homogeneous physical and chemical characteristics. each of these 5 color shades are associated to soil portion with same properties. to map the soil of this region, for example, would be required to collect 5 soil samples to be analyzed in laboratory, perform soil mapping and, based on their physical and chemical properties, make decisions about management of inputs, pesticides and irrigation. fig. 11(a) demonstrates the dynamics occurred in the system, obtained by subtracting of matrices from fig. 10(a) and fig. 10(b). formation of isolines are observed in region where water was added, showing that proposed method has sensitivity to detect differences in apparent electrical resistivity of the soil, mapping its dynamics. fig. 11. moisture stains mapping: (a) proposed method and (b) moisture meter tdr. in this same area and at the same time it was used a time-domain reflectometer (tdr) to measure the soil moisture content in order to compare the results found in the proposed method. fig. 11(b) shows the result obtained by tdr, after subtraction of data measured before and after insertion of the water. the proposed method, fig. 11(a), obtains more details than tdr, fig. 11(b), indicating the proposed method is more sensitive than tdr. e. identification of soil compaction stains through lateral profiling method. in another region, soil compaction is changed, using the same proposal, producing matrices before and after compaction. fig. 12(a) represents apparent electrical resistivity data collected for original soil conditions and fig. 12(b) represents measured values for soil after compaction. as in section iii-d, each of 5 colors shades are associated to soil portion with same properties allowing, analogously, the mapping of spatial variability of the soil. fig. 12. compaction stains mapping through lateral profiling method (ρa). fig. 13(a) shows the result of the difference between the matrices before (fig. 12(a)) and after (fig. 12(b)) compaction in subregion of study area. hot colors of fig. 13(a) represents the differences occurred in dynamics of the area where soil was compacted, showing that stain compaction was detected by measuring of apparent electrical resistivity of the soil. in this same area and at the same time a digital penetrometer was used to measure soil compaction in order to compare the results found by the proposed method. thus, using the digital penetrometer, soil compaction was measured before and after the change of the dynamic, generating two matrices. fig. 13(b) shows result of the subtraction of these two matrices. again, it is observed that proposed method has isolines with greater precision details, see relationship between the curves of fig. 13(a) and (b). fig. 13. compaction stains mapping: (a) proposed method and (b) penetrometer. iv. conclusions the main purpose of this work was to develop methodology for correlating water content in soil, clay content and compaction with electrical properties of the soil, measured by geoelectrical prospecting methods. resistance rm varies considerably with the change in moisture content of the soil w. rm also decreases with increasing w and the measured resistance decreases with increasing clay content. clay soils conduct better than sandy soils. the compaction c influences the values of rm, especially for samples with higher clay content, but at a lower degree than the moisture w. through the correlation surface (fig. 9) it is possible to correlate electrical resistance of the soil, moisture, soil compaction and clay content in order to identify the granulometry of the soil. field experiments indicate that lateral profiling method is appropriate for measuring regions with same characteristics. these experiments were able to accurately measure the change in soil dynamic produced intentionally. the proposed method is more accurate both in detecting moisture as in detection of soil compaction. the lateral profiling method is more sensitive to plot curves and, from correlation abacus, humidity values and compaction can be inferred directly on readings of apparent electrical resistivity of the soil. another observation subject of argumentation is related to different color shades of fig. 10 and fig. 12. each tonality is associated to specific soil area with different physical and chemical characteristics. in possession of this georeferenced surface it is possible to proceed data collect on the regions of different shades and utilizing the consideration that the same tonalities has the same physical and chemical characteristics, considerably reduce the amount of sample collection and the number of laboratory analysis. thus, lateral profiling is a noninvasive, fast and low cost method for mapping spatial variability of the soil, supporting decision making on how and when to intervene with inputs, pesticides and irrigation. acknowledgment the authors would like to thank coordination for the improvement of higher education personnel (capes), the national counsel of technological and scientific development (cnpq) and research support foundation of goias state (fapeg) for financial support research and scholarships. references [1] e.d. lund, p.e. colin, d. christy, and p.e. drummond. applying soil electrical conductivity technology to precision agriculture. saint paul: proceedings asa/cssa/sssa, 1998. [2] a. tabbagh, m. dabas, a. hesse, and c. panissod. soil resistivity: a non-invasive tool to map soil structure horizontal. geoderma, vol.97, p. 393-404, 2000. [3] e.l. eisenreich. electrical conductivity mapping for precision agriculture. proceedings. montpellier, ecole national superiure agronomique: european conference on precision agriculture, v. 3, 2001. [4] a.m. silva filho, g.p. furriel, w.p. calixto, a.j. alves, f.a. profeta, j.l. domingos, e.g. domingues, and m.g. narciso. methodology to correlate the humidity, compaction and soil apparent electrical conductivity. ieee congreso chileno de ingenieria electrica, electronica, tecnologias de la informacion y comunicaciones (ieee chilecon 2015), 2015, santiago. actas de ieee chilecon 2015, universidad central de chile, p. 1-7, 2015. [5] d.l. corwin and s.m. lesch. application of soil electrical conductivity to precision agriculture:theory, principles, and guidelines. agronomy journal, v.95, n.3, 2003. [6] s. p. friedman. soil properties influencing apparent electrical conductivity: a review. computers and electronics in agriculture, 2005. [7] w.m. telford, l.p. geldart, and r.e. sheriff. applied geophysics, cambridge university press, 1990. [8] j.d. rhoades, and r.d. ingvalson. determining salinity in soils with soil resistance measurements. soil science of american proceedings, vol.35, 1971. [9] o. banton, m.k. seguin, and m.a. cimon. mapping field-scale physical properties of soil with electrical resistivity. soil science societies american journal, vol.61, 1a. ed. sssaj, 1997. [10] b.g. williams, and d. hoey. the use of electromagnetic induction to detect the spatial variability of the salt and clay content of soil. australian journal of research, v.25, n.1, p.21-7, melbourne, 1987. [11] w.p. calixto, a.p. coimbra, b. alvarenga, j.p. molin, a. cardoso, and l. martins neto. 3-d soil stratification methodology for geoelectrical prospection. ieee transactions on power delivery, v. 27, p. 1636-1643, 2012. i. introduction ii. methodology a. lateral profiling b. experimental procedure laboratory study 1) sample preparation sample preparation consists in doping the same with known amounts of clay. from the sandy soil, samples are prepared with different proportions of clay, varying the texture of each sample. 2) relation among electrical resistance (rm), soil moisture (w) and clay content (δ) relationship among electrical resistance of the soil rm, soil moisture w and the clay content δ consists of performing rm measures in soils with different clay content, gradually varying the moisture content, keeping the compaction c constant. the rm values are obtained from geoelectrical prospection methods which detect effects produced by the flow of electric current in the soil. method of geoelectric prospection utilizes wenner arrangement for data collection, where electric current is i... soil samples are first taken to the oven (heater) in order to reduce moisture levels to values close to 0%. after passing through the desiccator cooling process, it is gradually added portions of water equivalent to 5% of the mass of the dry sample. a... 3) relation among electrical resistance (rm), soil compaction (c) and clay content (δ) the practical experiment to relate rm, c and δ consists of performing rm measurements in soils with different clay content, gradually varying the compaction c keeping moisture content w constant. soil samples are taken to the oven (heater) and cooled in desiccator. after passing through the cooling process, all samples are gradually moistened until they have constant moisture, w ≈ 20%. at each increase in compaction values, the corresponding value rm is measured by the electrical resistivity meter. apparatus of fig. 4(b) is designed in order to apply pressure on the sample that is within the pvc cell. under the pvc cell, it places t... c. experimental procedure field study iii. results a. relation among electrical resistance (rm), soil moisture (w) and clay content (δ) b. relation among electrical resistance (rm), soil compaction (c) and clay content (δ) c. correlation surface d. identification of soil moisture stains through lateral profiling method. e. identification of soil compaction stains through lateral profiling method. iv. conclusions the main purpose of this work was to develop methodology for correlating water content in soil, clay content and compaction with electrical properties of the soil, measured by geoelectrical prospecting methods. resistance rm varies considerably with t... field experiments indicate that lateral profiling method is appropriate for measuring regions with same characteristics. these experiments were able to accurately measure the change in soil dynamic produced intentionally. the proposed method is more a... another observation subject of argumentation is related to different color shades of fig. 10 and fig. 12. each tonality is associated to specific soil area with different physical and chemical characteristics. in possession of this georeferenced surfa... acknowledgment references  transactions on environment and electrical engineering issn 2450-5730 vol 1, no 4 (2016) © jiao jiao and r. m. nelms  abstract— explored in this paper is the grid impedance effect on the stability of a single-phase grid connected inverter with an lc filter based on an analysis of the inverter output impedance. for a single-phase grid connected inverter, a pi controller is often used to regulate the current injected into the grid. however, the control performance can be influenced when the inverter is connected to a weak grid. also, the utility grid has background harmonic noise, which can make the injected current distorted. therefore, analysis of the output impedance of a single-phase grid connected inverter is important for the robustness and stability of the system. by modeling the output impedance of inverter, it can be determined that the proportional gain and integral gain of the controller have an effect on the output impedance. analytical results show that by adjusting the pi controller parameters, the ability for harmonic reduction and stability of the system can be improved. simulation and experiments using a 1 kw single-phase grid connected inverter verify the effectiveness of the theoretical analysis. index terms— grid impedance, inverter output impedance, lc filter, pi controller, single-phase inverter i. introduction istributed generation technologies such as solar panels and wind turbines are being investigated because they are environmentally friendly. the voltage source inverter, which is the connection interface between distributed generation and the utility grid, plays an important role and has received more and more attention. an lc filter or an lcl filter is commonly used to reduce the pulse width modulation (pwm) switching harmonics. generally, the lcl filter has better performance in attenuating higher order harmonics and smaller component size and weight compared to an lc filter. however, the lcl filter is third order, which can introduce a resonant peak into the system that will cause an oscillation. for a small power inverter (a few kw), an lc filter is a better choice for the harmonic attenuation. in this paper, we employ an lc filter for the single-phase grid connected inverter. in the case of long distribution lines and lower power submitted on: 27 sep. 2016. accepted on: 30 oct. 2016. jiao jiao is with the electrical and computer engineering department, auburn university, auburn, al 36849 usa. (e-mail: jzj0034@auburn.edu). r. m. nelms is with the electrical and computer engineering department, auburn university, auburn, al 36849 usa. (e-mail: nelmsrm@auburn.edu). transformers in a distribution system, the grid can have a large impedance, which is a typical weak grid. however, the controller design for a grid-connected inverter usually doesn’t take the grid impedance into consideration. when a voltage source inverter is connected into a weak grid, the inverter control performance can be influenced [1]. the variation of the grid impedance may decrease the current controller bandwidth. in order to find the effect of grid impedance on the stability of the system, the external characteristics of the inverter need to be explored. generally, it can be modeled by the output impedance of the inverter. the relationship between the output impedance of the inverter and the grid impedance has been studied by many researchers [2-4]. for stability analysis, the impedance-based stability criterion can be used to examine the ratio of the inverter output impedance to the grid impedance [2]. however, it doesn’t mention a way to improve the stability of the system. in order to reduce the effect of grid impedance, the output impedance can be regulated accordingly. the output impedance depends on the design of the lc filter and the control structure and parameters. to improve the stability of the system and the ability for harmonic rejection, the output impedance can be changed by adjusting the controller parameters. therefore, the output impedance based method is explored under a weak and distorted grid. in this paper, a 1 kw single-phase grid connected inverter is used to investigate the output impedance of the system. by analyzing the output impedance in the frequency domain, the controller parameters can be adjusted to regulate the output impedance. by increasing the proportional gain of the pi controller, the magnitude of the output impedance can be increased to improve the ability for harmonic rejection. by increasing the integral gain of the pi controller, the phase of the output impedance can be increased to improve the stability of system. experimental results are in agreement with simulation analysis. ii. stability analysis of a single-phase inverter considering grid impedance variation a. modeling of the single-phase inverter the structure of a single-phase grid connected inverter with an lc filter is presented in fig. 1. vdc is the input dc link voltage, vinv is the output voltage of the h-bridge inverter and adjusting output impedance using a pi controller to improve the stability of a single-phase inverter under weak grid d jiao jiao and r. m. nelms vg is the utility grid voltage. the lc filter consists of inverter side inductances l1 and l2 and a filter capacitance c. the inductances l1 and l2 are two equal inductances used for attenuating the common mode noise current in the circuit. in addition, the electromagnetic interference (emi) effects are alleviated. lg is the grid impedance. the inductor current il, is used for current control to regulate the injected current with lower harmonics and unity power factor. the utility grid is sensed using a phase loop lock (pll) to make the single-phase inverter synchronize with the grid voltage. pll pispwm  dc v 1 l 2 l c   ref i ref i l i g v inv v g l   l i g i 1 r 2 r   c v   sin fig. 1. single-phase grid connected inverter with lc filter according to kirchhoff’s voltage law, we can get the circuit equation shown as follows:     cinvl l vvirr dt di ll  2121 (1) where r1 and r2 are the parasitic resistances of the inductors. by transforming to the frequency domain, the transfer function between inductor current and the difference between the inverter output voltage and the capacitor voltage can be derived.         rlssvsv si sg cinv l p     1 (2) where 1 2l l l  , 1 2r r r  . b. output impedance of a single-phase inverter many power converters are digitally controlled which will introduce sampling and computational delays into the system. also, the pwm process will introduce delay due to a zeroorder hold [5]. therefore, the total delay can be approximated by one and a half sampling period, which can be expressed by gd (s) shown below.   ts esg ts d 5.11 15.1    (3) the variable t is the sampling period. a pi controller is applied for current control in this inverter. the transfer function of a pi controller is shown as follows:           s ksg i pc  1 1 (4) here kp is the proportional gain and τi is the integral gain. the system control block diagram is shown in fig. 2 [6]. fig. 2(a) can be simplified to fig. 2(b) by control block equivalent transformations. the resulting transfer function of g1(s) and g2(s) can be expressed by:          sgsgkrls sgsgk sg dcpwm dcpwm   1 (5)           122    cssgsgkrcslcs sgsgkrsl sg dcpwm dcpwm (6) the output impedance of the inverter can be defined by:        sg si sv sz refi g c o 2 0     (7)    1 ls r pwm k cg s l i 1 sc  1 g sl g v g i  c v  invv  dg s ref i (a)   li 1 sc  1 g sl g v g i  c v   1 c d pwm g g k  1g s ref i  l i  1 g sl g v g i  c v   1g s  2g s ref i (b) fig. 2. control block diagram for a single-phase grid connected inverter (a) and its equivalent transformation (b) therefore, the single-phase grid connected inverter with an lc filter can be modeled by the norton equivalent circuit, as shown in fig. 3. the inverter is represented by a current source and a parallel output impedance, and the utility grid is modeled by a voltage source and a grid impedance [2]. in this effort, only the grid inductance is considered in the analysis as a worst case scenario. in reality, the resistance in the grid will help to stabilize the system.      gz s gi s  gv s oz s si s  ov s fig. 3. norton equivalent circuit an expression for the inverter output current can be derived by:      sv zz z si zz si g og o s og g     1 1 1 1 (8) from (8), it can be found that in order to mitigate the effect caused by grid voltage and grid impedance variation, the output impedance should be designed as high as possible to operate stably [2]. it can be found from (7) that the output impedance depends on the design of the lc filter and the controller structure and parameters. therefore, the inverter output impedance can be shaped by adjusting the controller parameters to improve the system stability. c. relationship between inverter output impedance and grid impedance a current controller is designed by assuming that the inverter is connected into an ideal grid (lg = 0). the bode plot of the uncompensated system (without a current controller) and the bode plot of the compensated system (with current controller) are shown in fig. 4. when the inverter is connected into a utility grid, the grid impedance can influence the inverter control performance. fig. 5 shows the relationship between inverter output impedance and grid impedances in the frequency domain. according to [3-4], the stability of the inverter depends on the inverter output impedance’s phase at the intersection point of zo and zg. with an increase in the grid impedance, the phase of the output impedance at the intersection frequency point is decreasing, which implies the system is less stable. in order to enhance the stability of the system under a wide range of grid impedance, the phase of the inverter output impedance at the intersection point should be increased by shaping the current controller parameters. the magnitude of inverter output impedance also needs to be designed higher to achieve better harmonic rejection ability. fig. 4. bode plot of the compensated and uncompensated system fig. 5. bode plot of the inverter output impedance and different grid impedances d. shaping pi controller parameters to improve the stability of system under weak grid by analyzing (7), it can be determined that the output impedance of the inverter can be changed by adjusting the parameters of the pi controller. as seen in fig. 6, the lower frequency part of the bode diagram shows that the magnitude of the output impedance increases with increasing proportional gain kp, which improves the ability for harmonic reduction. however, the phase of the inverter output impedance is decreasing, which means the system might be becoming less stable. as for increasing the integral gain of the pi controller, the phase of the inverter output impedance at the intersection frequency is increased, but the magnitude of the output impedance is decreased in the low frequency range. in the high frequency range, the magnitude and phase of inverter output impedance does not change much. it can be concluded that increasing the integral gain has the opposite effect on the output impedance as compared to increasing the proportional gain of the controller. therefore, it is possible to increase the magnitude and phase of the output impedance by adjusting the pi controller parameters without affecting the performance of the inverter. fig. 6. bode plot of inverter output impedance and grid impedance (a) proportional gain changes (top) (b) integral gain changes (bottom) e. output impedance sensitivity to circuit parameters from (7), it can be found that the inverter output impedance depends on the circuit parameters and controller parameters. in reality, the circuit parameters can change due to the temperature variation of the surrounding environment. therefore, it is necessary to explore the effect of circuit parameter variations on the inverter output impedance. since the parasitic resistance of the filter inductors has little effect on the inverter output impedance, the resistance variation is not investigated here. fig. 7. the inverter output impedance sensitivity to inductance variation fig. 8. the inverter output impedance sensitivity to capacitance variation from fig. 7, it can be seen that the frequency response of the inverter output impedance shifts to the left slightly as the inductance increases. it mainly affects the magnitude and phase of inverter output impedance around the peak. as shown in fig. 8, with the increase of filter capacitance, the magnitude of the inverter output impedance remains unchanged in low frequency range, but it decreases in the high frequency range. the phase margin of the inverter output impedance is also decreased. iii. simulation results the system shown in fig. 1 is simulated by matlab to validate the theoretical analysis. the system parameters are given in table i. for the simulation, the utility grid voltage (vg in fig. 1) is modeled by 7% third harmonics and 5% fifth harmonics and 3% seventh harmonics with phase 30°, 90° and 270°, respectively. fig. 9 to fig. 10 show the simulation results when the proportional gain was changed from kp = 2 to kp = 3; the grid impedance was 19.5 mh. the grid current total harmonic distortion (thd) is 4.30% and 3.76%, respectively. the harmonic reduction ability is improved, which validates the analysis in section ii. fig. 11 to fig. 12 show the simulation results when the integral gain was changed from τi = 0.0005 to τi = 0.01; the grid impedance was 19.5 mh. the grid current total harmonic distortion (thd) is 4.51% and 4.83%, respectively. table ii and table iii show the harmonic analysis of the inverter output current. fig. 9. output voltage and output current when lg = 19.5 mh (kp = 2) fig. 10. output voltage and output current when lg = 19.5 mh (kp = 3) fig. 11. output voltage and output current when lg = 19.5 mh (τi = 0.0005) fig. 12. output voltage and output current when lg = 19.5 mh (τi = 0.01) iv. experimental results a 1 kw texas instruments single-phase grid connected inverter with an lc filter was utilized for the experimental table i circuit parameters circuit parameter symbol value dc-link voltage vdc 380 v utility grid voltage vg 120 v fundamental frequency f0 60 hz inverter inductance l 7 mh inductance parasite resistance r 0.4 ω filter capacitance c 1 μf switching frequency fs 19.2 khz table ii output current harmonic analysis harmonic order lg = 6.5 mh lg = 19.5 mh kp = 2 kp = 3 kp = 2 kp = 3 3rd 0.71% 0.73% 1.12% 0.88% 5th 0.34% 0.42% 0.53% 0.17% 7th 0.45% 0.33% 0.53% 0.32% 9th 0.16% 0.17% 0.24% 0.25% 11th 0.13% 0.15% 0.19% 0.18% 13th 0.16% 0.10% 0.12% 0.18% thd 3.76% 3.51% 4.30% 3.76% table iii output current harmonic analysis harmonic order lg = 6.5 mh lg = 19.5 mh τi = 0.0005 τi = 0.01 τi = 0.0005 τi = 0.01 3rd 0.52% 0.74% 0.74% 0.98% 5th 0.34% 0.29% 0.14% 0.49% 7th 0.37% 0.46% 0.28% 0.45% 9th 0.14% 0.25% 0.28% 0.40% 11th 0.05% 0.01% 0.20% 0.12% 13th 0.15% 0.10% 0.11% 0.04% thd 3.51% 3.77% 4.51% 4.83% investigation. an ideal and a distorted grid voltage were simulated using a programmable ac source. in order to examine the effect of the grid impedance, an adjustable impedance is inserted between the inverter and the ac source. the measured single-phase inverter output voltage and output current are presented in fig.13fig. 22. the inverter output current harmonics measured by tektronix oscilloscope are given in table iv and table vii. under the ideal grid, the thd of inverter output current was reduced from 4.94% to 4.65% by increasing the proportional gain of current controller when the grid impedance is 19.5 mh. the output voltage and output current without pi controller is also measured for comparison, which demonstrate the current distortion can be improved by shaping pi controller parameters. the thd of the inverter output current was decreased from 5.09% to 4.73% by decreasing the integral gain of the current controller when lg is 19.5 mh. under the distorted grid, the thd of inverter output current was reduced from 5.21% to 4.69% by increasing the proportional gain of current controller when the grid impedance is 19.5 mh. the thd of the inverter output current was decreased from 5.28% to 4.74% by decreasing the integral gain of the current controller. as can be seen from these results, the thd of the inverter output current can be reduced by adjusting the pi controller gains. because the single-phase inverter is connected to the ac source through a transformer and relay, which makes the grid side inductance larger than the grid impedance in the simulation, the thd of experimental results are higher than the thd of simulation results. but it’s still in agreement with the theoretical analysis. fig. 13. output voltage and output current when lg = 19.5 mh (without pi) fig. 14. output voltage and output current when lg = 19.5 mh (kp = 2) fig. 15. output voltage and output current when lg = 19.5 mh (kp = 3) fig. 16. output voltage and output current when lg = 19.5 mh (τi = 0.0005) fig. 17. output voltage and output current when lg = 19.5 mh (τi = 0.01) table iv output current harmonic analysis harmonic order lg = 6.5 mh lg = 19.5 mh no pi kp = 2 kp = 3 no pi kp = 2 kp = 3 3rd 6.87% 2.34% 1.54% 7.29% 2.70% 1.55% 5th 5.53% 1.52% 0.91% 5.71% 1.63% 0.72% 7th 2.83% 1.77% 1.37% 2.64% 1.91% 1.67% 9th 1.70% 1.97% 1.61% 1.61% 2.20% 1.61% 11th 1.09% 1.00% 0.73% 1.09% 1.18% 0.86% 13th 1.03% 0.69% 0.77% 0.89% 0.91% 0.68% thd 6.83% 4.85% 4.64% 6.61% 4.94% 4.65% current (2a/div) voltage (50v/div) voltage (50v/div) voltage (50v/div) voltage (50v/div) current (2a/div) voltage (50v/div) current (2a/div) current (2a/div) current (2a/div) fig. 18. output voltage and output current when lg = 19.5 mh (without pi) fig. 19. output voltage and output current when lg = 19.5 mh (kp = 2) fig. 20. output voltage and output current when lg = 19.5 mh (kp = 3) fig. 21. output voltage and output current when lg = 19.5 mh (τi = 0.0005) fig. 22. output voltage and output current when lg = 19.5 mh (τi = 0.01) table v output current harmonic analysis harmonic order lg = 6.5 mh lg = 19.5 mh τi = 0.0005 τi = 0.01 τi = 0.0005 τi = 0.01 3rd 1.82% 2.29% 2.33% 2.05% 5th 0.98% 1.80% 0.92% 2.01% 7th 1.57% 2.02% 1.54% 1.77% 9th 2.13% 1.96% 1.58% 1.93% 11th 1.31% 0.72% 0.77% 0.63% 13th 0.94% 0.54% 0.54% 0.40% thd 4.78% 5.28% 4.73% 5.09% table vi output current harmonic analysis harmonic order lg = 6.5 mh lg = 19.5 mh no pi kp = 2 kp = 3 no pi kp = 2 kp = 3 3rd 7.62% 3.80% 2.71% 8.10% 3.47% 2.39% 5th 3.56% 1.68% 1.52% 4.43% 1.92% 0.76% 7th 4.51% 1.90% 1.28% 3.90% 2.02% 1.24% 9th 1.64% 1.57% 1.14% 1.61% 1.95% 1.45% 11th 1.13% 0.85% 0.55% 1.09% 1.22% 0.94% 13th 1.05% 0.92% 0.52% 1.08% 0.74% 0.74% thd 6.91% 5.09% 4.77% 6.86% 5.21% 4.69% table vii output current harmonic analysis harmonic order lg = 6.5 mh lg = 19.5 mh τi = 0.0005 τi = 0.01 τi = 0.0005 τi = 0.01 3rd 2.78% 3.65% 2.75% 2.89% 5th 1.59% 1.51% 1.62% 1.06% 7th 1.36% 2.64% 1.59% 2.32% 9th 1.59% 1.67% 1.89% 1.71% 11th 1.06% 0.86% 0.85% 0.60% 13th 0.95% 0.89% 0.78% 0.58% thd 4.85% 5.51% 4.74% 5.28% voltage (50v/div) voltage (50v/div) voltage (50v/div) voltage (50v/div) current (2a/div) current (2a/div) current (2a/div) current (2a/div) voltage (50v/div) current (2a/div) v. conclusion introduced in this paper is a method to increase the output impedance of a single-phase grid connected inverter with an lc filter to improve the stability and harmonic reduction ability of the system when the inverter is connected to a weak distorted grid. by modeling the output impedance of the inverter, the relationship between output impedance and grid impedance can be investigated. the grid impedance can degrade the control performance of the inverter and make the system less stable. in order to mitigate this effect, pi controller parameters are adjusted to increase the output impedance, which can improve the ability for the harmonic reduction and the stability of the system. finally, simulation and experiment results for a 1 kw single-phase grid connected inverter with an lc filter verify the effectiveness of the proposed method. references [1] liserre, m., teodorescu, r., blaabjerg, f., "stability of photovoltaic and wind turbine grid-connected inverters for a large set of grid impedance values," power electronics, ieee transactions on , vol.21, no.1, pp.263,272, jan. 2006. [2] jian sun, "impedance-based stability criterion for grid-connected inverters," in power electronics, ieee transactions on , vol.26, no.11, pp.3075-3078, nov. 2011. [3] m. céspedes and j. sun, "impedance shaping of three-phase gridparallel voltage-source converters," 2012 twenty-seventh annual ieee applied power electronics conference and exposition (apec), orlando, fl, 2012, pp. 754-760. [4] d. yang, x. ruan and h. wu, "impedance shaping of the gridconnected inverter with lcl filter to improve its adaptability to the weak grid condition," in ieee transactions on power electronics, vol. 29, no. 11, pp. 5795-5805, nov. 2014. [5] xiao-qiang li, xiao-jie wu, yi-wen geng, and qi zhang; “stability analysis of grid-connected inverters with an lcl filter considering grid impedance”, journal of power electronics, vol. 13, no. 5, september 2013. [6] v. blasko and v. kaura, "a novel control to actively damp resonance in input lc filter of a three phase voltage source converter," applied power electronics conference and exposition, 1996. apec '96. conference proceedings 1996., eleventh annual, san jose, ca, 1996, pp. 545-551 vol.2. [7] f. wang, j. l. duarte, m. a. m. hendrix and p. f. ribeiro, "modeling and analysis of grid harmonic distortion impact of aggregated dg inverters," in ieee transactions on power electronics, vol. 26, no. 3, pp. 786-797, march 2011. [8] y. tao, q. liu, y. deng, x. liu and x. he, "analysis and mitigation of inverter output impedance impacts for distributed energy resource interface," in ieee transactions on power electronics, vol. 30, no. 7, pp. 3563-3576, july 2015. jiao jiao received her b.s. and m.s. degree in electrical engineering from china agricultural university, beijing, china, in 2011 and 2013, respectively. she is currently pursuing her ph.d. degree in electrical engineering in the electrical and computer engineering department at auburn university. her current research interests include grid-connected inverter modeling, simulation and control. r. m. nelms (f’04) received the b.e.e. and m.s. degrees in electrical engineering from auburn university, auburn, al, usa, in 1980 and 1982, respectively, and the ph.d. degree in electrical engineering from virginia polytechnic institute and state university, blacksburg, va, usa, in 1987. he is currently professor and chair of the department of electrical and computer engineering, auburn university. his research interests are in power electronics, power systems, and electric machinery. dr. nelms is a registered professional engineer in alabama. in 2004, he was named an ieee fellow “for technical leadership and contributions to applied power electronics.”  transactions on environment and electrical engineering issn 2450-5730 vol 2, no 1 (2017) © márcio r. c. reis, wanderson r. h. araújo and wesley p. calixto  abstract — this article introduces the switched reluctance machine operating as a generator. this kind of electrical machine delivers cc power at the output and the energy generated can be controlled through several variables. in this work, the switching angles of the machine's power converter are optimized using deterministic and heuristic techniques so that the output power is kept constant via pi controller while guaranteeing maximum efficiency value, even for different excitation values and mechanical power on the shaft. index terms — switched reluctance generator, optimization, control, genetic algorithm. i. introduction he constant search for clean energy at present creates the need for studies and tests related to energy sources such as solar and wind. moreover, the climate change is a contemporary issue that requires a decrease in gas emissions. according to [1], existing research shows that in the last decades wind generation has become one of the topics discussed worldwide. this type of alternative source of energy has several advantages, such as: i) low environmental impact for installation of wind turbines. in this way, a small area is required for installed kwh compared to conventional power sources, ii) negligible greenhouse effect in comparison with other power sources, iii) low installation and maintenance costs, and iv) as far as the electronic advances, there are several ways to control the electric machine for better performance. ii. switched reluctance generator switched reluctance machines have a rotor with no winding. single phase windings are concentrated in the stator. both rotor and stator are made of ferromagnetic material and have projecting poles. therefore, the machine is suitable for operating in a wide speed range. there are several configurations for switched reluctance machines. fig. 1 shows the laminated package for reluctance machine with 6x4 1experimental & technological research and study group (next), federal institute of goias (ifg), goiania, goias state, brazil (e-mails: marciorcreis@gmail.com, wandersonrainer@gmail.com and wpcalixto@gmail.com). 2school of electrical engineering, mechanical and computer (emc), federal university of goias (ufg). configuration [4]. fig. 1. stator and rotor of a 6x4 srm. the phase windings are concentrated only in the stator. so for the srg 6x4, there are three phases. this winding must be energized individually, and the energy application in each phase winding is provided by a power converter, as shown in fig. 2 [4]. fig. 2. partial scheme for the half bridge converter. the production of torque occurs due to the tendency of alignment of the stator and rotor poles, when the respective phase is energized, and the position of reluctance is minimal for the established magnetic circuit and maximum inductance. on the other hand, if the mechanical energy is applied to the axis of the machine, the energized phase provides restorative torque resulting in additive counter-electromotive force, generating electric energy. the position of the rotor is required instantaneously, as the power converter energizes the winding phase at the correct time. the inductance profile for a phase of the srg is shown in fig. 3. efficiency improvement of switched reluctance generator using optimization techniques márcio r. c. reis1,2, wanderson r. h. araújo1,2 and wesley p. calixto1,2 t fig. 3. inductance profile for a srg phase winding. the machine shaft has coupled sensors that indicate the position of the rotor instantaneously so that the machine can be operated as a motor or generator. considering the switched reluctance engine operating as a motor, the torque produced is given by (1) [5]. t= 1 2 ×i 2 × dl dq (1) in (1), i is the current applied to the machine phase winding, l is the phase inductance and  is the position of the rotor. therefore, (1) indicates that the production of positive torque occurs when the phase winding is energized when inductance is increasing. if the primary motor supplies torque to the machine shaft, applying current to the winding of a phase during the decreasing inductance, it causes the restoring torque to be converted into electrical energy, which can be routed to the load. this condition makes the machine work as a generator. the srg delivers cc power at the output, eliminating the need for one of two converters used in induction machines, ac-dc and dc-ac, to connect to the power grid. iii. optimization techniques optimization is the mathematical tool to be used in solving problems in which the most efficient solution among all the existing possibilities must be found. such are called optimal or optimized solutions to the problem considered. the optimization of the system by minimizing the function f(x) is performed considering x î â n . a basic optimization algorithm consists of determining the direction from each point obtained to take the next step. since the goal is to minimize f(x), it is reasonable for the function to decrease in the chosen direction [6] [7]. optimization processes may be deterministic or heuristic in nature. in this work, both techniques are applied to the drive parameters of the switched reluctance generator in order to increase the efficiency of the machine, allowing the generation of energy, consuming less excitation power or requiring lower values of mechanical power in the generator axis. quasi-newton methods are based on newton's method to find the stationary point of a function, where the gradient is zero. newton's method is deterministic and assumes that the function can be locally approximated as a quadratic in the region around the optimum, and uses the first and second derivatives to find the stationary point. in higher dimensions, newton's method uses the gradient and the hessian matrix of second derivatives of the function to be minimized. genetic algorithms are heuristic methods of random search inspired by evolutionary biology, such as: i) heredity, ii) mutation, iii) natural selection, and iv) recombination. they are implemented through a computer where the population of abstract solution representations is selected for improvements. another method used in this work is the hybrid method, in which the quasi-newton method is used in conjunction with the genetic algorithm. in this case, the genetic algorithm provides the quasi-newton method with its best individual. from it, the method will find the new individual, that is, the new, better adapted set of solutions [8][9]. normally, hybridization is used to: i) improve the performance of existing techniques, ii) seek better solutions, and iii) divide complex problems by decomposing them into sub-problems where each algorithm / technique solves part of it. iv. methodology the use of the switched reluctance machine as a generator requires the application of current on the phase winding during the decreasing inductance. in addition, the initial instant of application of the current on and the instant when the phase winding is disconnected from the power source off act directly on the output voltage generated and, as a result, on the output power. other factors that interfere with the output power are the vexc excitation voltage and the rotor speed . firstly, the mathematical model for the srg 6x4 associated with power elements will be implemented in order to observe the operation of the machine as a generator. fig. 4 illustrates the model to be used in simulation. fig. 4. scheme for matlab/simulink simulation. the firing angle of the semiconductor switches of the power converter, on, was established at the maximum inductance (alignment position) for all three phases, and the off angle (off) was set at 30° after on. therefore, the triggering of the power converter switches has been set with a 30° window for all the phases of the machine. this driving window is suitable for 6x4 machines as indicated in [6]. for all simulations, the load was set at 269 , 2.8 mf, and the machine parameters are shown in table 1. table i srg parameters symbol quantity value ns stator poles 6 nr rotor poles 4 lmin inductance for unaligned position 35 mh lmax inductance for aligned position 135 mh r stator resistance 3.25  b friction 0.006 kg.m2/s j inertia 0,04806 kg.m2 the activation of the srg with excitation voltage of 120 v, driving angle of 30° and speed of 1000 rpm causes the machine to generate power of 1056.9 w with a yield of 0.8, as shown in fig. 5. fig. 5. srg power and efficiency (open loop driving). with these preliminary results, the objective of this study is to develop three methodologies: i) to apply the pi controller to the vexc excitation voltage in order to guarantee fixed output power, 1000 w was chosen, ii) to apply the heuristic optimization method, genetic algorithm (ga), to find the best on and off firing angles of each of the three phases, seeking to maximize the efficiency of the generator and iii) after establishing the optimized values of on and off, apply them to the drive of the half-bridge converter and use the hybrid optimization method (deterministic and heuristic) to optimize machine axis speed and power pi controller gains. the objective is to find the best performance of the machine for a certain speed of rotation of the shaft, seeking lower values of excitation or mechanical power, in order to guarantee the generation of 1000 w of electric power with higher efficiency. v. results as part of the evaluation function, the optimization algorithms use the srg yield given by (2). fx=1-h (2) where  is the srg yield, given by (3). h= p o p i (3) where po is the electric power in the load and pi is the sum of the excitation and mechanical powers of the srg, both in watts. the genetic algorithm was simulated with an initial population of 20 individuals, containing as seed the angles cited before with the 30º window, which are: i) on_c=30º and off_c=60º, ii) on_b=60º and off_b =90º and iii) on_a=90º and off_a=120º. the mutation rate was 1% in the initial generation and 80% in the final generation. the crossing rate was 80% in the initial generation and 15% in the final generation. the selection method used was the tournament [7]. the maximum generation number (gmax) was defined as 100 generations. the non-uniform mutation operator and the dispersed crossover operator [7] were used. the pi controller was constructed in order to maintain fixed generation power through srg excitation. the setpoint of the control system was 1000 w. fig. 6 shows the evolution of the genetic algorithm, with an initial evaluation function of approximately fx = 16 and after the convergence of the algorithm fx = 13.15. fig. 6. fitness evolution of the genetic algorithm. the genetic algorithm converged in 40 generations and obtained an improvement of approximately 18% in the optimization process. the values of the angles found by the ga were: i) on_c = 29.22º and off_c = 54.46º, ii) on_b = 58.01º and off_b = 83.50º and iii) on_a = 86.00º and off_a = 113.62º. fig. 7 shows the graph of srg output power (po) with the conventional angles and the optimized angles, both under the influence of the pi controller on the excitation. fig. 7. srg output power. fig. 7 shows the generated power, which reaches the system set point at approximately 1000 w, using the pi controller. there was a slight change in the transient between the two tests, because the pi controller, as it is a linear system, behaves differently when the switching angles change because of the non-linearity inserted into the system. fig. 8 illustrates the power supplied for srg excitation. fig. 8. srg excitation power. the power supplied for the srg excitation is highest (590.49 w) when the angles found by the ga are used whereas the test with the standard angles indicates an excitation power of 561.49 w. there was an increase in the electrical power supplied to srg excitation of approximately 5% in the test with angles found through the ga. this would imply some worsening with respect to srg yield, but some results will still be analyzed. fig. 9 shows the mechanical power applied to the srg axis for said tests. fig. 9. srg input mechanical power. with the test using the activation angles of the phases found by the ga, the mechanical power supplied to the srg is reduced by approximately 16.5%. the mechanical power produced through the standard test was 675.21 w and through the ga, it was 563.05 w. accordingly, it can be emphasized that there was a small increase in the electric power provided by the voltage source responsible for srg excitation and a greater decrease in the mechanical power supplied to the srg. in this way, the process is validated, since the srg produces the same electric power, with greater difference in the mechanical supply on the axis. this occurs because there is a pi controller in the excitation and the generation happens with an increase in the mechanical power, reducing the excitation power, as desired. the srg yield for the standardized and optimized test is shown in fig. 10. fig. 10. srg efficiency. the yield of the srg provided by the activation angles of the phases produced by the ga increases by approximately 9.5%. the test performed with the standard angles and window fixed at 30º, has an output of 0.809, and with the ga, 0.868. it is a satisfactory improvement, considering that the srg generates power of 1000 w with reduced mechanical power, with the only inconvenience that the power converter undergoes a slight overload. this causes the srg to achieve a considerable reduction in energy consumption or to produce more power for the same supply. the switching angles obtained by the optimization with the genetic algorithm were used as initial points for another optimization process, but now focusing on obtaining the angular velocity value in the generator axis that provides the best performance. however, with the speed change in the generator axis, it is necessary to obtain new parameters for the pi controller due to the fact that it is a linear control and does not allow the same performance for different speeds of rotation. the optimization of the switching angles on and off previously presented reduced the evaluation function fx = 16 to fx = 13.15. as the new optimization proposal takes into account the angular velocity and pi controller parameters, the fx evaluation function was reformulated as described in (4). fx=[0,9×(1-h)]+[0,1×e p ] (4) this new evaluation function considers the yield of the srg and the variable ep, corresponding to the power error generated. this variable was considered for the adequate adjustment of the parameters kp and ki of the pi controller, but with less influence on fx than the machine performance. thus, the six values obtained for the switching angles on and off of the three machine phases using ga were used as starting points for the new evaluation function. fig. 11 illustrates the evolution of this optimization starting from fx=13.15 to find the angular velocity value that maximizes the srg efficiency associated with the controller which allows, at this new speed, the power of 1000 w generated in the load to be kept. fig. 11. ga fitness evolution (ga for , kp and ki). it can be observed in fig. 11 that fx was minimized to 11.2681, so as to maximize yield and minimize error ep of the pi controller. in order to improve the results obtained, this genetic algorithm was hybridized with the quasi-newton method. the values obtained in the optimization, illustrated by fig. 11, were hybridized and fx had its value reduced to 11.0683, as shown in fig. 12. fig. 12 shows that the initial value for fx obtained in the previous optimization (fx = 11.2681), after some iterations, was applied to the hybrid ga causing a further reduction in the evaluation function fx = 11.0683. fig. 12. hybrid ga fitness evolution. this reduction in fx provided changes in the electric and mechanical variables of the srg and guaranteed an improvement in the yield for the generation of the amount of 1000 w of electric power in the load. fig. 13 shows the power generated for the new speed value ( = 133.66 rad/s) from the hybrid ga. fig. 13. srg output power after hybrid ga optimization. it is also possible to observe in fig. 13 an electric output power of approximately 995 w, causing a 0.5% error for the power pi controller. fig. 14 illustrates the excitation power of the srg for generating the 995 w, a result also obtained through the hybrid ga. fig. 14. srg excitation power after hybrid ga optimization. among all the methods used in this article, the activation of the srg with the values obtained b through y the hybrid ga is the one that required the least input electrical power (excitation) to generate the necessary output power. as shown in fig. 14, a power of 547.73 w was required, whereas the activation with the conventional angles of 30º of driving window, for an angular velocity of 100 rad/s, required a power of 561.66 w. such values of excitation power allow variable e, which characterizes the electric efficiency of the generator, to be obtained, that is, the relation between the electric power generated at the output and the electric power of excitation at the input. thus, e represents the electric power balance in the machine, indicating how the device can generate power for a certain input power. therefore, e can be calculated as shown in (5). he = p o p exc (5) where pexc represents the excitation power of the generator. fig. 15 shows the behavior of the mechanical power in the srg axis for the power generation indicated in fig. 13. fig. 15. srg mechanical power after hybrid ga optimization. just as occurred with the excitation power, the mechanical input power was reduced from 675.22 w to 571.18 w using the hybrid ga. both powers contribute positively to the efficiency of the generator if they are minimized, as indicated in (4). fig. 16 shows the efficiency of the generator with the drive performed with optimized parameters via the hybrid ga. fig. 16. srg efficiency after hybrid ga optimization. as expected, the reduction in mechanical and excitation powers caused an increase in the srg yield, raising this variable from 0.81 to about 0.89. the results indicate that the ga, in spite of providing improved performance with the reduction of excitation power [9], if integrated with the quasinewton method, becoming a hybrid method, contributes positively with variables already optimized by the ga [10]. table 2 and table 3 show the results obtained quantitatively. table 2. simulation results angles (on e off) excitation power (pexc) mechanical power (pmec) generated power (po) default 561.66 w 675.20 w 1001.75 w ga 590.50 w 563.05 w 1001.78 w hybrid ga 547.73 w 571.18 w 995.00 w table 3. simulation results angles (on e off) srg efficiency () electrical efficiency (e) rotor speed default 0.809 1.783 100 rad/s ga 0.868 1.696 100 rad/s hybrid ga 0.889 1.816 133.66 rad/s analyzing the results of table 2 and table 3, it is possible to verify that, although the mechanical power required for 1000 w generation with the srg is higher in the hybrid ga than in the simple ga, the electric efficiency (e) is higher for all the methodologies adopted. that is, it requires less excitation power to generate the same power when the appropriate speed is used for the generator and the optimized angles are used in the power converter. vi. conclusions this article presents the results of simulation for three methodologies of optimization of parameters of switched reluctance generators. the generator is driven with the help of a pi controller for the excitation of the machine in order to maintain 1000 w of power generated over resistive load. the generator is activated with the help of a pi controller for the excitation of the machine so it keeps a generated power of 1000 w over resistive load. the switching angles of the power converter were optimized using the genetic algorithm and the results obtained from this optimization were used in a new process, in which the variables to be optimized were the rotor angular velocity and the parameters of the excitation pi controller. therefore, a speed value was obtained at which the machine provides best performance. the hybrid genetic algorithm, coupled with the quasi-newton method, despite having provided results that indicate an increase of the mechanical power needed to generate 1000 w at the output of the generator, reduced the machine's excitation power the most, resulting in gain in the overall yield, approximately 10%. these studies show how the use of srg for power generation can be made more effective, especially in applications with a wide range of speed variation, considering the constructive aspects of the machine. references [1] z. qixue. a small single-phase switched reluctance generator for wind power generation. icems 2001 proceeding of the fifth international conference electrical machines and systems. 2001. vol 2. aug 2001 pages 1003-1006. [2] m. nassereddine, j. rizk, and m. nagrial. switched reluctance generator for wind power applications. world academy of science, engineering and technology international journal of mechanical, aerospace, industrial, mechatronic and manufacturing engineering vol:2, no:5, 2008. [3] k. bouchoucha, h. yahia, m. n. mansouri. particle swarm optimization of switched reluctance generator based distributed wind generation. journal of multidisciplinary engineering science and technology (jmest) issn: 3159-0040 vol. 1 issue 4, november 2014. [4] juha pyrhonen. electrical drives. lut (lappeenranta university of technology), department of electrical engineering. finland. [5] araujo, wandeson r. h.; ganzaroli, cleber a. ; calixto, wesley. p. ; alves, aylton j. ; viajante, ghunter p. ; reis, marcio r. c. ; silveira, augusto f. v. firing angles optimization for switched reluctance generator using genetic algorithms. in: 2013 13th international conference on environment and electrical engineering (eeeic), 2013, wroclaw. 2013 13th international conference on environment and electrical engineering (eeeic), 2013. p. 217. [6] t. j. e. miller. switched reluctance motors and their control. oxford, u.k.: clarendon, 1993. [7] michalewicz, zbigniew. genetic algorithms + data structures= evolution programs. springer, (1996). [8] calixto, wesley pacheco, et al. "parameters estimation of a horizontal multilayer soil using genetic algorithm". power delivery, ieee transactions on 25.3 (2010): 1250-1257. [9] reis, m. r. c. ; calixto, w. p. ; araújo, wanderson rainer hilário de; alves, a. j. ; domingos, j. l. . switched reluctance generator efficiency improvement for wind energy applications. in: 16th ieee international conference on environment and electrical engineering, 2016, florença. 16th ieee international conference on environment and electrical engineering, 2016. [10] reis. m. r. c. ; márcio rodrigues da cunha reis. análise comparativa de métodos de otimização aplicados à sintonia do controlador pi. márcio rodrigues da cunha reis has graduation at engenharia elétrica by pontifícia universidade católica de goiás (2011) and master's at engenharia elétrica e de computação by universidade federal de goiás (2014) . currently is of instituto federal de educação, ciência e tecnologia de goiás e of companhia energética de goiás. has experience in the area of electric engineering , with emphasis on sistemas microprocessados. wanderson rainer hilário de araújo is graduate at engenharia elétrica from universidade católica de goiás (2003), graduate at redes de comunicação from centro federal de educação tecnológica de goiás (2002) and master's at electric engineering from universidade federal de goiás (2006). has experience in electric engineering, focusing on transmission of the electric energy, distribution of the electric energy, acting on the following subjects: engenharia elétrica, acionamento, relutância and motor. wesley pacheco calixto is professor in physics , m.sc. in electrical and computer engineering from the federal university of goias. completed his phd in electrical engineering from the federal university of uberlandia with part of this time at the university of coimbra, portugal. currently he is professor/researcher at the experimental & technological research and study group (next/ifg).. i. introduction ii. switched reluctance generator iii. optimization techniques iv. methodology v. results vi. conclusions this article presents the results of simulation for three methodologies of optimization of parameters of switched reluctance generators. the generator is driven with the help of a pi controller for the excitation of the machine in order to maintain 10... references  transactions on environment and electrical engineering issn 2450-5730 vol 2, no 1 (2017) © a. ales, m.a. cheurfi belhadj, j-l. schanen, d. moussaoui, j. roudet  abstract—in this paper we proposes a synthesis of different mathematical models of power electronic converters based on thevenin/norton equivalent circuits. those models, composed by impedances and harmonic noise sources, are helpful to predict the conducted electromagnetic interferences (emi) generated by converters connected to the electrical network. moreover, the extracted impedances are determining for sizing emc filters. the proposed analytical model is tested with pspice simulations and validated by experimental measurements, from dc frequency until 30mhz. index terms—emc modeling, differential mode, common mode, emc filter, lisn, emi, conducted noise, noise source. i. introduction owadays, modern networks -such as the car’s network supplying many electrical actuatorsinclude many converters to manage efficiently the power transfer, which creates new issues and therefore the emc study becomes more and more complex. generally, in modern embedded networks, integrated actuators are supplied by dc power, the reason that dcconverters are strongly needed, to manage the corresponding energy. for example, the input power at aeroplane turbine sides may be 10 times greater than at actuator sides. the velocity and position control need also the control of the input power. in order to comply with electromagnetic (emc) standards, an accurate prediction of a common mode (cm) and differential mode (dm) conducted noise is necessary. furthermore, since electromagnetic interference (emi) filters are coupled to converters, the optimisation is necessary reached with the knowledge of converter’s impedances [1– 3].several papers treat the converter’s modelling with different techniques: terminated/unterminated models [4–6], lumped circuits [7–10]...etc. however, the proposed models deal only with one kind of converter, this cannot be generalised to other conditions, due to the parameter’s dependency. this paper, presents a noise source analytical models of two converter’s topologies, in order to highlight the impact of some parameters in the spectrum model profile, and highlights factors which actually make difference between such or that models. this paper is organised in five sections. section ii gives the analytical computing process of the model. section iii is the application of the model on two converter structures. in the section iv we expose the impact of some parameters on the impedance. the last section concludes this paper. ii. general purpose dc-converters under some operation conditions are linear and time invariant systems (lti), as mentioned in [2]. based on this affirmation, they may be modelled by an equivalent norton/thevenin circuit. moreover, converters under study are checked for this characteristic as demonstrated on the fig.1. for frequencies higher than some tens of khz, the input impedance is the same whatever the switches states. fig. 1. input dm impedance of a buck converter for two extreme case of switch state [2]. a. converter impedances computing the main purpose of this work is to make a study about dm converters’ impedances taking into account its global behavior, such as the switching operation and pcb imperfections, and then extract a general law about a propagating path of dm conducted noise, by making a comparison point. power converters, especially dc-converters are supplied by two wires “l” for “plus” and “n” for “minus”, but also connected to a third wire “g” which is the safety conductor commonly called “ground” conductor, since the converter is the main source of conducted interferences, due to its operation principle. thus, dc-converters may be considered as three ports systems, as drawn in the fig.2 [11].where: zlg, analytical models synthesis of power electronic converters a. ales, m.a. cheurfi belhadj, j-l. schanen, d. moussaoui, j. roudet n zng, and zln are respectively impedances supposed between two conductors l-g, n-g and l-n. zlaod l n g cbus (a) g l n ylg yng yln (b) fig. 2. converter circuit topology: (a) converter circuit designed as an active multi-port circuit, (b) the equivalent π-quadripole circuit in addition, by definition both cm impedance zcm and dm impedance zdm are computed according to the equation (1) [1] and [12]. where idm is the current circulating between the two conductors “l” and “n”, vdm is the voltage between the same conductors. icm is the current circulating between the two previous wires including the safety conductor “g”, vcm is half the sum of the l-g and n-g voltages (fig.2.a).          cm cm cm dm dm dm i v z i v z (1) according to the equation (1) and the circuit’s topology of the fig.2.b, the dm admittance ydm becomes as expressed in (2). note that, this impedance depends, not only on the impedance “zln” defined between the two wires l-n (which defines the dm concept), but also on impedances connected to the ground wire “g”, zlg and zng which are commonly supporting the cm current. ) 4 ( nglg lndm yy yy   (2) b. the context of the study the main idea of this paper is to extract a reliable analytical model to compute the dm impedance of dc-converters topology, connected to an electrical network (fig.3), on a frequency range from very low frequencies up to few tens of mhz. the purpose in computing this impedance is multiple:  to compute network impedance,  to manage to make an emc analysis of a whole network,  to contribute on the emi filter optimization,  to enhance the converter design. filter 2 converter 2 filter 1 converter 1 s o u r c e equipement under study zl dc-network converter p x filter fig. 3. synoptic diagram of a dc-network iii. analytical modelling a. the switching function any input signal sin(t) crossing throw any switching cell, as drawn in the fig.4, is chopped according to the control law sw(t) of the switching cell. the system outputs may be expressed according to (3).   )()( tstswts inout  (3) sw(t) sin(t) sout(t) fig. 4. input and output signals crossing the system including the switching function sw(t) the switching operator sw(t), which is a tsw-periodical function, actually establishes the switching law of the converter’s devices and depends on the control strategy, which in this case concerns the buck and the boost converters’ control, considering the duty cycle“α”. since the function sw(t) corresponds to the switching signals, we can imagine that its shape in the temporal domain, is a “one” and “zero” sequence level as drawn in the fig.5. according to the signal processing theory, the switching operator may be expressed as in (4), by the convolution between two conventional functions, [1-3] and [11]:   t swt 2   is the window function,   t swt  is the dirac comb,  where tsw=fsw -1is the switching period.  αduty cycle,                    t t ttsw swsw t sw t 22   (4) in the frequency domain, and by developing the convolution product, the output signal sout(f) (fig.4) can be expressed as in (5), where fsw is the switching frequency. in fact, the sout(f) signal may be the converter current or voltage chopped at fsw frequency whose importance will be highlighted when developing the expression.           k swin k j out kffsekfs 2 2 sinc    (5) t(s) gsw(t) 1 αtsw tsw (n+α)tswntsw(1+α)tsw fig. 5. the switching function. b. the modelling process as claimed previously, any three ports’ converter (fig.2.a) may be represented by an electrical circuit as depicted in the fig.6 [11]. zlisn is the impedance of the line impedance stabilisation network “lisn”. vlg g zlgzlisn zlisn zng zlnvin vng l n il in fig. 6. a three ports converter model. the input voltage vin between wires “l” and “n”, is in fact the dm voltage as expressed in (6). the dm current idm is expressed in (7). since dc-converters, especially under study, are “lti” systems as checked in the previous section, any multiport system may be formatted as an equivalent “norton” (or thevenin) circuit, regarded between two considered wires, such drawn in the fig.7 [11] and [13]. for a consequence, the dm current related to the dm voltage may be expressed in (8). where ih regroup the noise current source, generated when the converter’s operation. ymd is the converter’s dm admittance [3]. ydm ihvin iin fig. 7. converter’s models.     hmdmd nl md ifvy ii fi    2 (8) iv. converters under study a. converters’ topologies the dc-dc converters under study are schematised on fig.8 and fig.9. all essential coupling paths are introduced in order to be more accurate as in experiment, such as parasitical capacitances to the ground (clg, cng, and cm for both the buck and the boost converter), parasitical line inductances lp, parasitical elements of the boost inductor and imperfections of both capacitors cf and co. [10]. cf zl clg cng cm m ~vin il im ing il g b esr esl lp lp zlisn+ zlisnin vng vlg ilg imc a fig. 8. the buck converterunder study. vin iin m s zboost zcf zco zlcm lp clg cng lp vng vlg fig. 9. the boost converter under study. b. systems identification the identification, of different parameters of converters’ topologies, drawn in fig.8 and fig.9, is performed by the impedance analyser “agilent 4294a”. the measurement setup is illustrated in fig.10. fig. 10. the identification of converters’ parameters.   nglgdmin vvvv  5.0 (6)   2 nl md ii fi   (7) 1) capacitors’ model identification the decoupling capacitor’s impedance zcx (cx for cf or co), is modelled by three serial elements: the capacitance “cx”, the serial inductive element “esl”, and the serial resistive element “esr” as expressed in (9). those elements are identified by the impedance analyser 4294a. fig.11 shows the comparison between the model and the measured impedance.   esrpesl pc pz x cx    1 (9) fig. 11. the input capacitor’s impedance: (red) the serial element model,(blue) the measured impedance. 2) parasitical elements converters’ parasitical elements are identified by the impedance analyser 4294a. the parasitical elements such as, the line inductances lp, the capacitances to the ground clg, cng and cm, are deduced from measured impedances of the fig.12, depicted between the “l” (or “n”) and the “g” wire. the cm capacitance, between the heatsink and the ground, is estimated according to (10), which is the plane capacitor equation. e s c rm 0 (10)  ε0 and εr are respectively the permittivity of the air and the relative permittivity of the insulator mica (εr=5)  zlg: impedance between plus-ground wires,  zng: impedance between minus-ground wires, (a) (b) fig. 12. impedances of the quadripolar model of the fig.6: (a) zlg impedance between plus-ground wires, (b) zngimpedance between minus-ground wires. 3) boost inductor the boost inductor is the essential element in the boost converter structure, since it is one which controls the variation of the current (state variable). hence, its impedance is not without effect on the global dm impedance of the converter. moreover, the equivalent hf model given in the fig.13 is composed by a parallel resistive element “epr” and a parallel capacitive element “ecp”. epr ecp lboost fig. 13. the equivalent model of the boost inductor zboost fig.14 is the boost inductor impedance measured by the impedance analyser agilent 4294a. fig. 14. identification of the boost inductor impedance v. model applying a. equations model applied on the buck converter in this work, we are not interested to the harmonic source ih, expressed in (8) and represented in the fig.7. actually, ylg and yng, expressed in (2) and appearing in the model of (8), are parasitical capacitances’ impedances to the referential conductor “g” [3]; they are expressed, respectively in (11) and (12). the measurement is done by an experimental test using the impedance analyser 4294a. they are not dependent on the switching phenomena [3, 11]. note that, in (11) and (12) impedances are an addition of two terms. the first one, is dependent on the inner capacitance clg (or cng), the second term at (1-α) time, there is an addition capacitances (cm of the middle point “m”) which comes to be added on this inner capacitance. moreover, the capacitance cm is added alternately, sometimes to the capacitance clg sometimes to cng. yln expressed in (13), is the switching impedance, depending on: devices switching, the input capacitor impedance and the load impedance [3]. neither cng nor clg are included in the yln expression. as a result, the dm impedance of the buck converter is depicted in the fig.15, for different operation points, depending on the duty cycle α. the result is compared to the decoupling capacitor impedance “zcf” measured by the impedance analyser 4294a (yellow colour), and its serial element synthesised model (in black colour). note that this impedance matches with the input capacitor impedance in yellow colour all along a large frequency band, until 100mhz. this result is helpful to identify dc-converters especially at the design stage. according to (2), the dm impedance at low frequencies, trend to be as expressed bellow: for a consequence, at low frequencies under 100hz (this limit depends actually on converter’s components), the dm converter’s impedance (red and blue colour), depends on the output load impedance zl, and varies with the variation of the duty cycle α (α=0.3, 0.5, 0.8). in other words, it depends on converter’s operation points. the resonance appeared around 100mhz, is due to the interaction between esl of the input capacitor cf and the parasitical line impedances lp. this will be detailed in the next section. after that frequency, the dm impedance takes the value of parasitical line impedance zlp. fig. 15. the dm impedance of the buck converter. b. model applied on the boost converter the previous model is also applied on one more system which is a boost converter, in order to make a comparison and may be extract a general rule about the emc converter identification. the same analytical process is performed in this case, and is expressed on (15). where:  zboost is the boost inductor impedance,  zcf is the input capacitor impedance,  zl is the load impedance,  zlp is the layout parasitical inductance,  zc0 is the output capacitor impedance,                  ifcc ifl ifc ifl y mlg p lg p lg       2 1 2 1 2 1 2 (11)                  ifcc ifl ifc ifl y mng p ng p ng       2 1 2 2 1 2 1 (12)                                   m swllcf ln mffz m zz y 2 2 2 sinc1    (13)   2  fz z l dm  (14)         nboostcf nboost cfdm kfzfz kfz fzz    (15)              0 2 0 2 1sin//1 n swcswln ncnffznffzk  (16) the dm impedance of the boost converter is depicted in the fig.16 for three duty cycle (α=0.3, 0.5, 0.8) and compared to the input filter impedance. it is important to outline that the dm converter’s impedance matches also here with the input filter impedance zcf, (red colour) along the frequency band of interest (150 khz to 30mhz). some distinctions appear due to resonances with the boost inductor around 1.5 khz. note that, at low frequencies (less than 100hz) the impedance depends also on the duty cycle α, as outlined in the buck converter case. fig. 16. impedance zmd of the the boost converter as showed in the fig.15 and fig.16, both spectrums of dm impedances of the buck or the boost converter give the impression that they are different –more resonances in the boost. however, in a major way it’s necessary to point out that both impedances are matching the impedance of the input decoupling capacitors zcf (red colour), all along a large frequency band. in other words, the differential mode impedance ymd of the buck or the boost converter is the same. as a consequence, the knowledge of the converter impedance, which is connected directly to the emc filter, can significantly enhance its optimisation and therefore reduce the emc security margin. vi. the effect of some intrinsic parameters in this section we will study the effect of some parameters of the buck converter on its dm impedance profile. equations (2) and (11–13), reveal that for the buck converter case, the impedance model depends on some intrinsic converter’s parameters:  the input capacitor value “cf”,  the “esl” of the input capacitor “cf”,  parasitical capacitances to the ground clg, cng and cm,  parasitical capacitances on power devices sides: the mosfet capacitance cds (between the drain and the source), the diode capacitance cdiode.  parasitical line inductances lp, those parameters have an actual effect on the impedance profile evolution in the spectral domain. a. the effect of the input capacitor cf according to (2) and (11–13), the dm impedance depends on the zcf impedance, otherwise on the value of the capacitor cf. two profiles of the dm impedance associated to two input capacitors’ value are represented and compared on the fig.17. note that there is an obvious difference between the two profiles. for the blue case (cf = 10µf), the shape of the dm impedance seen at low frequencies, as expressed in (14) is extended until 300hz, than that recorded for the red case (cf = 100µf). fig. 17. the comparison between two dm impedances profile about two values of the input capacitors cf={10µf, 100µf}. b. the effect of the esl of the capacitor cf for the same reason, the serial inductor element of the input capacitor “esl” has a major impact on the dm impedance profile especially at high frequencies. this is clearly visible on the fig.18, between the blue and the red profiles, around the band of [1mhz, 100mhz]. fig. 18. the comparison between two dm impedances profile about two values of the “esl = {10nh, 50nh}” of the input capacitors cf. c. the effect of parasitical capacitances parasitical capacitances are the essential coupling path, supporting the cm current generated by converters, and throwing to the ground [10]. in this paper, the parasitical capacitances accounted are:  clg the mutual parasitical capacitances seen between the “line” wire and the referential plane,  cng the mutual parasitical capacitances seen between the “neutral” wire and the referential plane,  and cm the mutual parasitical capacitances seen between the “middle point” of the switching cell and the referential plane, the fig.19 shows two dm impedance profiles for two parasitical capacitances values. those values introduced in the model of the dm impedance expressed in (2) are comparatively exaggerated in order to make strongly in evidence the actual effect. note that, there is an impact in the impedance profile, appeared as resonances around 60mhz. fig. 19. the comparison between two dm impedances profile for two values of parasitical capacitances. d. the effect of capacitances of the switching devices switching devices of the buck converter, studied in this paper (fig.8), are diode and mosfet. those devices are coupled to parallel capacitances. two profile cases are considered for two values of the mosfet capacitances (cds=50pf and 100pf). the result is given in fig.20 and fig.21. note that, there is a significant difference between two impedances ymd deduced by two capacitances’ values, around 10mhz. vii. conclusions in this paper we have presented a synthesis of input impedances of two dc-dc converters’ structures. results point out that, the differential mode input impedance are matching along a large frequency band and are identical to the input decoupling capacitor. this result is helpful for filter optimisation. in addition, we have presented some parameters impact on this impedance, such as the parasitical capacitances of the switching devices, the esl of the input capacitor and the parasitical capacitances to the ground. the effect of each parameter appears and impact on the dm impedance profile. fig. 20. ymd impedance for two capacitances value of the mosfet fig. 21. zoom of ymd impedance in the impact frequency zone references [1] a. ales, jl. schanen, d.moussaoui, j. roudet, "experimental validation of a novel analytical approach about a dc-dc converter input impedance", epe'2013 ecce europe, lille, france, septembre 2013. [2] a. ales, jl. schanen, d.moussaoui, j. roudet, "impedances identification of dc/dc converters for network emc analysis", ieee transactions on power electronics, vol.pp , issue: 99, page 1, 11 febrary 2014. [3] a.ales, z. gouichiche, b. karouche, d. moussaoui, j-l. schanen, j. roudet, “the accurate input impedances of a dc-dc converters connected to the network “, 2015 ieee 15th international conference on environment and electrical engineering (eeeic), rome 10-13 june 2015. [4] m. foissac, j.l. schanen, c. vollaire, “black box emc model for power electronics converter”, energy conversion congress and exposition, 2009. ecce 2009. ieee , pp. 3609 – 3615, 20-24 september 2009. [5] weichang cheng, xiaoying he, shen xu, weifeng sun, analysis of common-mode electromagnetic interference noise in a flyback converter using a self-supply power control integrated circuit, power electronics, iet (volume:8 , issue: 9 ) , pp. 1749 1757, 31 août 2015. [6] h. bishnoi, a.c baisden, p. mattavelli, d. boroyevich, "analysis of emi terminal modeling of switched power converters", power electronics, ieee transactions on, pp 3924 – 3933, vol 27, september 2012. [7] l. gerbaud, b.touré, j.-l. schanen, j-p carayon, "modelling process and optimization of emc filters for power electronics applications", the international journal for computation and mathematics in electrical and electronic engineering, vol. 31 iss: 3, pp.747 763, 2012. http://ieeexplore.ieee.org/xpl/mostrecentissue.jsp?punumber=5305884 http://ieeexplore.ieee.org/xpl/mostrecentissue.jsp?punumber=5305884 [8] arnold s. de beer, member, ieee, “noise source model for power converters incorporating a single source with imbalance, ieee international conference on industrial technology (icit) 2013, cape town, 25-28 feb. 2013. [9] jian sun, “input impedance analysis of single-phase pfc converters”, applied power electronics conference and exposition, 2003. apec '03. eighteenth annual ieee, vol. 1, pp. 361 – 367, 9-13 february 2003. [10] n. k. poon, b.m.h. pong, c.p. liu, c.k. tse, “essential-coupling-path models for non contact emi in switching power converters using lumped circuit elements”, power electronics, ieee transactions on, vol.18, pp. 686 – 695, 26 march 2003. [11] a. ales, jl. schanen, d.moussaoui, j. roudet, "a new analytical emc model of power electronics converters based on quadripole system: application to demonstrate the mode decoupling condition", the applied on power electronics conference and exposition, charlote dc, usa, mars 2015. [12] a. ales, g. frantz, jl. schanen, d.moussaoui, j. roudet, “common mode impedance of modern embedded networks with power electronics converters", emc europe 2013, brugge, belgium, sept 2013. [13] ma. cheurfi belhadj, a.ales, a. zaoui, z. chebbat, jl.schanen, j.roudet, “analytical model of dc-dc converters based onswitching impedances and emi sources”, emc europe 2016, wroclow, pologne, sept. 2016 achour ales received the diploma in electrical engineering from ecole militaire polytechnique school, algiers, algeria in 2006. he obtained his master in electrical engineering from ecole militaire polytechnique school , algiers, algeria in 2009. he received his ph.d at grenoble university, grenoble, france since 2015. his current research interest the electromagnetic compatibility of the embedded network and power electronics. jean-luc schanen (m’99–sm’04) was born in 1968. he received the electrical engineering diploma and ph.d. degrees from grenoble institute of technology, grenoble, france, in 1990 and 1994, respectively. he is currently a professor with grenoble institute of technology. he has been with the grenoble electrical engineering laboratory, st. martin d’hères, france, since 1994, working in the field of power electronics. his main activities concern the technological design of power converters. his research team uses (or develops if not available) all kinds of modeling tools in order to improve the performance of power electronics converters, including electromagnetic compatibility and thermal aspects. prof. schanen is a senior member of the ieee power electronics society and ieee industry applications society, and was chairman of the power electronics devices and components committee of the ieee industry applications society between 2006 and 2007. djelloul moussaoui graduated in electrical engineering from ecole militaire polytechnique d'algiers, algeria in 1987. he obtained his ph.d in electrical engineering from grenoble institute of technology, france in 1997. he became professor in 2010, and is currently leading research in emc in power electronics. james roudet was born in 1962. he holds phd (1990) and electrical engineering diploma (1986). he is professor at grenoble university (université joseph fourier, france), within the g2elab (grenoble electrical engineering lab), in the field of power electronics. first research interest concerned resonant converters. afterwards, he promoted emc activities in the field of power electronics, and developed a leading activity in the technological design of power converters. he currently holds the director position of g2elab, after several years of leading the power electronics team. http://ieeexplore.ieee.org/xpl/recentissue.jsp?punumber=63 i. introduction ii. general purpose a. converter impedances computing b. the context of the study iii. analytical modelling a. the switching function b. the modelling process iv. converters under study a. converters’ topologies b. systems identification 1) capacitors’ model identification 2) parasitical elements 3) boost inductor v. model applying a. equations model applied on the buck converter b. model applied on the boost converter vi. the effect of some intrinsic parameters a. the effect of the input capacitor cf b. the effect of the esl of the capacitor cf c. the effect of parasitical capacitances d. the effect of capacitances of the switching devices vii. conclusions references teee-68_er.pdf analysis of loss distribution of conventional boost, z-source and y-source converters for wide power and voltage range brwene gadalla(1), erik schaltz(1), member ieee, yam siwakoti(2), member ieee, frede blaabjerg(1), fellow, ieee department of energy technology, aalborg university(1) aalborg 9220, denmark department of electrical, mechanical and mechatronic systems, university of technology sydney(2) sydney, australia bag@et.aau.dk, esc@et.aau.dk, yam.siwakoti@uts.edu.au, fbl@et.aau.dk abstract— boost converters are needed in many applications which require the output voltage to be higher than the input voltage. recently, boost type converters have been applied for industrial applications, and hence it has become an interesting topic of research. many researchers proposed different impedance source converters with their unique advantages as having a high voltage gain in a small range of duty cycle ratio. however, the thermal behaviour of the semiconductor devices and passive elements in the impedance source converter is an important issue from a reliability point of view and it has not been investigated yet. therefore, this paper presents a comparison between the conventional boost, the z-source, and the y-source converters based on a thermal evaluation of the semiconductors. in addition, the three topologies are also compared with respect to their efficiency. in this study the results show that the boost converter has higher efficiency than the zsource and y-source converter for these specific voltage gain of 2 and 4. the operational principle, mathematical derivations, simulation results and final comparisons are presented in this paper. key wordsboost converter; z-source converter; y-source converter; winding losses; core losses; gain; thermal design; reliability i. introduction b oost type converters are essentially needed for many re-newable energy applications such as photo voltaic (pv), wind turbine (wt) and automotive applications (electric and hybrid vehicles) as these often have lower input voltage than the required load voltage. in conventional boost converters, the demanded voltage gain normally requires higher duty cycle (sometimes close to unity), which leads to high conduction losses, higher voltage and current stresses on the switching devices. however, the aforementioned stressor factors may critically affect the reliability and the lifetime of the power electronic components. according to a review based on condition monitoring for device reliability in power electronic systems presented in [1], semiconductor and soldering failures in device modules are sharing totally 34% of converter system failures in fig. 1. in today’s perspective toward the reliability assessment of power electronic components and systems, three main aspects should be considered as shown in fig. 2 [2]. pcb 26% capacitors 30%solder 13% semiconductors 21% connectors 3% others 7% fig. 1. ranking and failure distrbution of power electronic components in power converters [1]. the design and verification aspect could be related to cover the aforementioned shortcomings in the conventional boost converter shown in fig. 3. both z-source and y-source as shown in fig. 4 and fig.5 converters were proposed by the researchers as impedance source network converters to compromise the high voltage gain with small duty cycle ratio. due to their flexibility for a wide voltage ranges and power conversions (dc-dc, dc-ac, ac-ac, and dc-ac) [8], various types of impedance source networks were reported as a solution to overcome the limitations of the voltage source inverter vsi, current source inverter csi and some of the conventional uni/bi directional converters [9]. moreover, an important advantage of the impedance nettransactions on environment and electrical engineering issn 2450-5730 vol , no (201 ) © z power electronics reliability control and m onitoring intelligent control condition m onitoring d es ig n an d ve rif ic at io n d es ig n fo r re lia bi lit y ro bu stn es s va lid at io n analytical physics physics of failure (pof) component physics fig. 2. aspects in power electronics reliability assessment [2]. works is the small duty cycle ratio, which reduces the losses of the switch [10]. therefore, many new topologies are proposed with each being claimed to have improved performances [8]. at present, a collective investigation of some of the existing boosting converters has not been initiated especially with reference to their thermal and reliability issues. furthermore, an investigation of the y-source converter from the point of view of thermal performance at high power ratings has not been reported as well. the junction temperature is one of the important factors that is affecting the thermal performance of the converter, and also the reliability [11]. in this paper the conventional boost, z-source and y-source converters are compared in terms of their efficiencies and junction temperature with respect to a wide power range, different voltage gains and assuming a constant ambient temperature. moreover, calculations of all the relevant losses (e.g switching and conduction losses) of the power switching devices during the operation is also considered in the comparison [12]. this paper conducts a comprehensive investigation of the mapping of the losses of the power converter. section ii is focusing on the theory of operation and design of each converter. section iii describes the evaluation of the power losses and thermal performance of the three converters. section iv describes the simulation results and discuss the results. finally, the conclusion is given in section v. ii. converter design and theory of operation in this section the theory of operation of the converters and design formulas are presented. a. conventional boost converter a boost converter is a step-up converter converting the voltage from low input voltage to higher voltage requiring a duty cycle (0 <d<1). thus its simple theory of operation as well as component count wise allows it to be a competitor with other boosting converters. the basic structure of the boost converter circuit, the equivalent circuit for on state mode of operation and off state mode of operation are shown in fig. 3. it compromise of one active switch sw, a diode d1, an inductor l1, and a capacitor c1 for introducing a high voltage boost with (d>0.5). the two modes of operation are as following: a) during the on-state: the switch is closed, the current flows through the inductor and store the energy in a magnetic field. b) during the off-state: the switch is open, the current passed will be reduced as the voltage across the inductor is reversed and the magnetic field previously created will decrease to maintain the current flow to the load and the current through the diode will charge the capacitor giving a higher voltage. the input/output voltage relationship is expressed in (1) as: vout = vin 1 − d (1) where vout is the output voltage, vin is the input voltage and d is the duty cycle needed for the required voltage gain [13]. b. z-source converter the z-source converter (zsc) is a very convenient topology in many alternative energy sources and other different applications [3, 4]. the zsc has the capability of ideally giving an output voltage range from zero to infinity regardless of the input voltage. the z-source converter circuit, and its two modes of operation are shown fig.4. it consists of two inductors (l1, l2) and two capacitors (c1, c2) connected in x shape to be coupled to the dc voltage source. the zsc can produce a required dc output voltage regardless of the input dc source voltage. the two modes of operation are as the following: a) in the on-state: the switch is closed and the impedance capacitors (c1, c2) release energy to the inductors (l1, l2) and then the voltage source and the load will disconnect the z-source network due to the turn off of the diodes (d1, d2). the major concern is the large conduction current passing through the switch during the on state, which may decrease the converter efficiency. b) in the off-state: the switch is opened and the input voltage will supply energy to the load through the two inductors as well as add energy to the two capacitors to compensate the energy lost during the on state. the input/output voltage relationship is expressed in (2) as: vout = vin 1 − 2d (2) where vout is the output voltage, vin is the input voltage and d is the duty cycle needed for the required voltage gain [3, 4]. c. y-source converter the y-source converter is a promising topology for higher voltage gain in a small duty ratio and has a very wide range l1 cout d1 rl sw l1 (a) (b) (c) vin cout rlsw l1 cout d1 rl vinvin + vout _ iin il iout fig. 3. a) boost converter circuit topology. b) equivalent circuit for on state. c) equivalent circuit for off state [13]. l1 c2 c1 sw l2 l1 c1 l2 d1 cout rlvin (a) (b) (c) rlcout l1 c2c1 l2 d1 cout rlvin sw d2 d2 c2+ vout _ iin il ic1 iout fig. 4. a) z-source converter circuit topology. b) equivalent circuit for on state. c) equivalent circuit for off state [3, 4]. vin d1 cout n1 n3 d2 n2 c1 sw cout n1 n3 n2 c1 cout n1 n3 n2 c1 rl rl (b)(a) (c) vin sw + vout _ iin iout vinrl in2 lm1 im1 fig. 5. a) y-source converter circuit topology. b) equivalent circuit for on state. c) equivalent circuit for off state [5–7]. of adjusting the voltage gain [5–7]. the range of duty cycle in the y-source is narrower than the z-source and the boost converter. fig. 5 shows the y-source impedance network and its two modes of operation. it is realized by a three-winding coupled inductor (n1, n2, and n3) for introducing the high boost at a small duty ratio for the sw. it has an active switch sw, two diodes (d1, d2), a capacitor c1, and the windings of the coupled inductor are connected directly to sw and d1, to ensure a very small leakage inductance at its winding terminals. the two modes of operation are as the following: a) in the on-state: the switch is closed, d1 and d2 are off causing the capacitor c1 to charge the magnetizing inductor of the coupled transformer and capacitor c2 discharge to power the load. b) in the off-state: the switch is opened, d1 starts to conduct causing the input voltage to recharge the capacitor c1 and the energy from the supply and the transformer will also flow to the load. when d2 starts conducting, it recharges c2 and the load is to be continuously powered. the input/output voltage relationship is expressed in (3) as: vout = vin 1 − kd (3) where vout is the output voltage, vin is the input voltage, d is the duty cycle [5–7] and k is the winding factor. the winding factor k is calculated according to the turns ratio of the three-winding coupled inductor and it is expressed in (4) as: k = n1 + n3 n3 − n2 (4) where (n1 : n2 : n3) are the winding ratios of the coupled inductor. a comparison between the inductors, the capacitors design, voltage and current ripples for the three converters is shown in table i. table i. component design for the boost, z-source and y-source converters components boost z-source y-source current ripple across inductor 2 20%0. outl in li p i v 20%l li i 20%l mi i voltage ripple across capacitor 2% outc out v v 2%c outv v 2% outc outv v inductor equation in l s dv l i f 1 2 c l t v l l l i 1.056 2 3 0.29210 , cn l lik c n lik n a l a n l l l l l capacitor equation out max out s out i d c f v 1 2 % l c c t c v i v c c out out l s out v d c r f v st o 1 2 s o d 1 p c 1 2% 1 f 1 k d vd out out l s out v d c r f v im: magnetizing current, ln: nominal inductance, llik: leakage inductance, to = dt= d/fsw iii. evaluation of power losses and thermal performance in this section, the formulas for calculating the relevant power losses are presented. plecs toolbox is used for the three converter analysis. the parameters selected for each converter are compared according to the passive components counts and their voltage and current ripples are as shown in table i. the same for the switching devices, which are designed according to each converter requirements for the voltage and current ratings for a realistic comparison. a. switching and conduction losses calculations switching losses occur when the device is transitioning from the blocking state to the conducting state and vice-versa. this interval is characterized by a significant voltage across its terminals and a significant current through it. the energy dissipated in each transition needs to be multiplied by the switching frequency to obtain the switching losses; the switching losses psw are expressed in (5) as: psw = (eon + eoff ) × fsw (5) where eon and eoff are the energy losses during turn on and turn off of the switch, fsw is the switching frequency. conduction losses occur when the device is in full conduction mode. these losses are in direct relationship with the duty cycle. the average conduction losses pcond are expressed in (6) as: pavg.cond = 1 t ∫ t 0 [vce(t) × ice(t)] dt (6) where vce is the on state voltage, an ice is the on state current. the time period t is as given in (7): t = 1 fsw (7) where fsw is inversely proportional to the time period t . b. capacitor esr losses calculations the capacitor equivalent series resistance (esr) is the value of the resistance, which is equal to the total effect of a large set of energy loss mechanisms occurring under the operating conditions where it can be a parameter to measure the capacitor losses. the capacitor losses are expressed in (8) as: pcap.loss = i 2 cap. × esr (8) where icap. is the rms current passing through the capacitor, and esr is the equivalent series resistance measuring the effect of the losses dissipated in the capacitor. c. winding and core losses calculations according to the steinmetz’s equation, which is a physics based equation used to calculate the core loss of magnetic materials due to hysteresis. the core losses are expressed in (9): pv = kf αb̂β (9) where b̂ is the peak flux density excitation with frequency f, pv is the time-average power loss per unit volume, and(α, β, k) are the material parameters found by curve fitting. the improved generalized steinmetz’s equation is expressed in (10): pv = 1 t ∫ t 0 ki ∣∣∣∣dbdt ∣∣∣∣ α ( δbβ−α ) dt (10) where δb is the flux density from peak to peak and in (11): ki = k (2π) α−1 ∫ 2π 0 |cosθ|α × 2β−αdθ (11) where θ is the angle of the sinusoidal waveform simulated. the copper losses in the winding describe the energy dissipated by the resistance in the wire used in the coil. it is divided in to 2 types (dc and ac winding loss). the dc winding losses can be calculated in (12) as: pdc = i 2 av × rdc (12) where (pdc ) is the dc copper losses in the winding, iav is the average current passing through the wire, and rdc is the dc resistance of the wire. ac copper losses can be significant for large current ripple and for higher frequency. it can be calculated through the skin effect, where the current density is an exponentially decaying function of the distance into the wire, with the characteristic length δ is known as the skin depth in (13) as: δ = 7.5√ fs (13) where δ is the skin depth in cm, and fs is the switching frequency which in our design is 20 khz. in order to calculate the ac resistance rac , the thickness h of the wire should be known since it is a function of the dc resistance rdc which can be calculated in (14): rac = h δ × rdc (14) where h is the thickness of the wire in cm. the ac winding losses can be calculated as given in (15) as: pac = i 2 ac−rms × rac (15) where pac is the ac winding loss, iac−rms is ac ripple rms current passing through the wire, and rac is the ac winding resistance. d. magnetic core design calculations in this section, the magnetic core design [14] is illustrated through the following steps: 1) in order to select a proper core size, the dc current idc in ampere and the inductance l in mili henry required with dc bias should be known to select the core from the core selector chart according to the calculated value (mh.a2) in (16): li2dc = value (16) a high flux 58337 core [14] was selected for the 3 converters in order to have fair comparison from an efficiency point of view for the voltage gain of 2. 2) inductance, core size and permeability are now known, then calculating the number of turns by determining the minimum inductance factor almin by using the worst case negative tolerance (generally −8%) given in the core data sheet in (17) and (18) almin = al − 0.08al (17) n = √ l × 103 almin (18) where al is the inductance factor found in the core data sheet (nh/t2), almin is the minimum inductance factor (nh/t 2), and l is the inductance in (μh). 3) choosing the suitable wire size according to rated power and calculated number of turns (n), is the last step before calculating the dc resistance according to the wire size with window fill assumed to be 40% in (19) as: ca = wf × wa n (19) where ca is the wire area, wf in the window fill, and n is the no. of turns. 4) the dc resistance can be estimated after knowing the winding factor of the core, wire gauge (awg), and the number of turns. the dc resistance can be calculated in (20) as: rdc = mlt × n × ω/length (20) where mlt is the mean length per turn, and ω/length is the resistance per meter. furthermore, in the voltage gain of 4 metglas power-lite c-core [15] is used and kg-method is applied [16]. iv. simulation results and discussion in this section, different power loadings for the voltage gain equal to 2 and 4 are presented in order to demonstrate a fair comparison between the 3 topologies with respect to the thermal performance and the losses (switching, conduction, capacitor esr losses, core and winding losses) for calculating the efficiency of each converter. thermal and efficiency investigation are presented in a separate subsection. table ii summarizes the specifications and the requirements used in the simulation results. the design specifications for each voltage gain are given separately for each topology as it can be seen from table iii, which summarize the semiconductor devices average current and voltage ratings used in the 3 converters. these ratings are based on the required voltage gain for each converter separately. a. junction temperature investigation of the switch under different power loading for each semiconductor a heat sink has been designed. a maximum junction temperature of 125 ◦c has been used a design constraint. the estimation of the junction temperature table ii. common specifications and simulation parameters for the boost, z-source and y-source converters simulation parameters boost z-source y-source gain 2 duty cycle d 0.5 0.25 0.167 no. of turns 64 55 ( 32:32:64 ) switch rms current 71 a 100 a 120 a gain 4 duty cycle d 0.75 0.375 0.25 no. of turns 27 30 ( 7:7:14 ) switch rms current 171 a 237 a 346 a * input voltage for gain 4 common converter specifications for gain 2 and 4 maximum power rating 20 kw input voltage vin 200 v \ 100 v * output voltage vout 400 v switching frequency fs 20 khz resistive load rl maximum junction temperature tj-max. 125 °c table iii. semiconductor devices selection for the three converters and their different voltage gains. converter semiconductor devices gain 2 gain 4 b oo st igbt (ixxx200n60c3) 600 v and 200 a (mg06600wb-bn4mm) 600 v and 600 a diode (d1) (idw100e60) 600v and 100 a (db2f200n/p6s) 600v and 200 a z -s ou rc e igbt (mg06400d-bn4mm) 600 v and 400 a (mg06600wb-bn4mm) 600v and 600 a diode (d1) (ds1f300n6s ) 600v and 300 a (sd600n/r series) 600v and 600 a diode (d2) (ds1f300n6s ) 600v and 300 a (ds1f300n6s) 600v and 300 a y -s ou rc e igbt (mg06600wb-bn4mm) 600v and 600 a (mg06600wb-bn4mm) 600v and 600 a diode (d1) (vsk.9112 ) 1200v and 100a (skn 501/12 semikron) 1200v and 720 a diode (d2) (ds1f300n6s ) 600v and 300 a (ds1f300n6s) 600v and 300 a of the switches are done according to the thermal model and the mapped losses using the plecs toolbox. the estimation of the junction temperatures are different for the 3 topologies, since the desired thermal resistance of the heat sink is not the exact calculated value found in the manufactured heat sinks. 1 5 10 15 20 0 20 40 60 80 100 120 130 140 ju nc tio n te m pe ra tu re t j ( °c ) load power (kw) boost z-source y-source at ambient temperature ta= 25 °c fig. 6. junction temperature variation of the switch at different power loading and using a voltage gain of 2. in this case, the load power is varying from 1 to 20 kw, and a constant ambient temperature is assumed which is 25 ◦c. the junction temperature variation results of the compared 1 5 10 15 20 0 20 40 60 80 100 120 130 140 ju nc tio n te m pe ra tu re t j ( °c ) load power (kw) boost z-source y-source at ambient temperature ta= 25 °c fig. 7. junction temperature variation of the switch at different power loading and using a voltage gain of 4. topologies are shown in fig. 6 for voltage gain of 2. fig. 7 shows the junction temperature variation at different loading power for voltage gain of 4. b. efficiency investigation under different power loading in this subsection the efficiency is calculated according to the total power losses for each converter as listed in the beginning of section iv using the same conditions listed in table iv. distribution of the different losses for the boost converter at 20 kw load power and two different voltage gain. voltage gain boost converter g ai n 2 total loss: 1.7 % g ai n 4 total loss: 3.9 % switching loss 83 w 26% conduction loss 169 w 53% capacitor esr loss 1.5 w 1% core loss 3.3 w 1% dc winding loss 61 w 19% ac winding loss 1 w 0% switching loss conduction loss capacitor esr loss core loss dc winding loss ac winding loss switching loss 340 w 44% conduction loss 231 w 30% capacitor esr loss 3 w 0% core loss 63 w 8% dc winding loss 135 w 18% ac winding loss 2 w 0% table v. distribution of the different losses for the z-source converter at 20 kw load power and two different voltage gain. voltage gain z-source converter g ai n 2 total loss: 3.3 % g ai n 4 total loss: 5 % switching loss 335 w 52% conduction loss 198 w 31% capacitor esr loss 3.1 w 0% core loss 7.4 w 1% dc winding loss 103 w 16% ac winding loss 1.4 w 0% switching loss conduction loss capacitor esr loss core loss dc winding loss ac winding loss switching loss 321 w 31% conduction loss 324 w 32% capacitor esr loss 23.7 w 2% core loss 145 w 14% dc winding loss 204 w 20% ac winding loss 2 w 0% table vi. distribution of the different losses for the y-source converter at 20 kw load power and two different voltage gain. voltage gain y-source converter g ai n 2 total loss: 4.4 % g ai n 4 total loss: 6.3 % switching loss 474 w 51% conduction loss 228 w 24% capacitor esr loss 12 w 1% core loss 18 w 2% dc winding loss 200 w 22% ac winding loss 1.64 w 0% switching loss conduction loss capacitor esr loss core loss dc winding loss ac winding loss switching loss 682 w 42% conduction loss 218 w 13% capacitor esr loss 42 w 3% core loss 329 w 20% dc winding loss 340 w 21% ac winding loss 19.5 w 1% table ii. the results in fig. 8 show that the boost converter has the highest efficiency of 98% compared with the y-source converter of 96% and the z-source converter of 96.7% at 20 kw loading power. the measured efficiencies from low power loading (1 kw) to higher power loading (20 kw) is also shown in 8. the same analysis is repeated for voltage gain of 4 as shown in 9. 1 5 10 15 20 87 90 92 9596 9798 99100 e ff ie ci en y (% ) load power (kw) boost z-source y-source at ambient temperature ta= 25 °c fig. 8. the efficiency at different loading power and using a volatge gain of 2. c. total losses at 20 kw power loading in this section a better understanding is given for the efficiency and loss mapping. six pie charts are presented in tables iv, v, and vi for the same power loading 20 kw and different voltages gain (2 and 4). the total loss listed in table iv, v, and vi were calculated from the total power loss of each converter by measuring the total efficiency as summarized in 1 5 10 15 20 80 82 85 87 90 92 94 96 98 100 e ff ie ci en y (% ) load power (kw) boost z-source y-source at ambient temperature ta= 25 °c fig. 9. the efficiency at different loading power and using a volatge gain of 4. table viii. the switching and the conduction losses are the total losses generated from the semiconductor devices (switch and diodes). in table viii comparison of the total efficiencies using voltage gains of 2 and 4 for the compared converters at 20 kw load power. in voltage gain 2, the magnetic losses which in the y-source converter is sharing 34% of the total losses is the double percentage of the magnetic losses in the z-source converter and 1.5 times the percentage in the boost converter. the capacitor losses in percentages are almost the same in the 3 converters. the switching and conduction losses are the lowest in the boost converter compared to the zsource and y-source converters. the switching and the conduction losses are varied based on the semiconductor devices ratings, as these devices are designed according to the required voltage gain, converter specifications, and to withstand the maximum ratings of each operated converter. in the voltage gain 4 case, the magnetic losses in the ysource is sharing 42% of the total losses is more than double the percentage of the magnetic losses in the boost converter and 1.2 times the percentage in the z-source converter. table vii. comparison of the total efficiencies using gain 2 and gain 4 for the converters at 20 kw load. efficiency boost z-source y-source gain 2 98.3 % 96.7% 95.6% gain 4 96.1 % 95% 93.7% v. conclusions in this paper a comparison between the y-source, z-source and the conventional boost converter has been performed with respect to their thermal behaviour and efficiency. different loading conditions between 1 kw and 20 kw are considered during the studies of the efficiency and junction temperature of the converters for two different voltages gain (2 and 4). the junction temperature variation in voltage gain of 4 is higher than the junction temperature variation in voltage gain of 2. investigations on both the magnetic and electrical losses are also given. the magnetic losses which in the y-source converter is sharing 34% and 42% of the total losses in voltage gain of 2 and 4 receptivity which is higher than in the boost and z-source converters. in the electrical losses it can be noticed that the total electrical loss for voltage gain of 4 is lower than for voltage gain of 2 which clarify that having higher current ratings devices improve the efficiency. this paper summarizes the comparison between the type of losses at constant loading condition. the thermal performances are quite similar in the 3 converters for both voltage gains. the boost converter has better efficiencies in the two selected voltage gains, but it has also the highest decrease in the efficiency from gain 2 to gain 4 at 20 kw power loading compared with the z-source and y-source converters. references [1] s. yang, d. xiang, a. bryant, p. mawby, l. ran, and p. tavner, “condition monitoring for device reliability in power electronic converters: a review,” ieee trans. on power electronics, vol. 25, no. 11, pp. 2734–2752, nov. 2010. [2] h. chung, h. wang, f. blaabjerg, and m. pecht, reliability of power electronics converter systems. the institution of engineering and technology (iet), dec. 2015. [3] k. ramtek and y. nagpur, “dynamic modelling and controller design for z-source dc-dc converter,” international journal of scientific engineering and technology, vol. 2, no. 4, pp. 272–277, apr. 2013. [4] f. z. peng, m. shen, and z. qian, “maximum boost control of the z-source inverter,” in proc. of ieee 35th annual power electronics specialists conference (pesc), vol. 1, june 2004, pp. 255–260 vol.1. [5] y. siwakoti, p. c. loh, f. blaabjerg, and g. town, “ysource impedance network,” in proc. of 29th annual ieee applied power electronics conference and exposition (apec), march 2014, pp. 3362–3366. [6] y. siwakoti, p. c. loh, f. blaabjerg, s. andreasen, and g. town, “y-source boost dc/dc converter for distributed generation,” ieee trans. on industrial electronics, vol. 62, no. 2, pp. 1059–1069, feb 2015. [7] b. gadalla, e. schaltz, y. siwakoti, and f. blaabjerg, “investigation of efficiency and thermal performance of the y-source converters for a wide voltage range,” journal of renewable energy and sustainable development, vol. 1, no. 2, pp. 300–305, jan. 2016. [8] y. siwakoti, f. z. peng, f. blaabjerg, p. c. loh, and g. town, “impedance-source networks for electric power conversion part i: a topological review,” ieee trans. on power electronics, vol. 30, no. 2, pp. 699–716, feb 2015. [9] f. z. peng, “z-source inverter,” ieee trans. on industry applications, vol. 39, no. 2, pp. 504–510, mar 2003. [10] w. t. franke, m. mohr, and f. w. fuchs, “comparison of a z-source inverter and a voltage-source inverter linked with a dc/dc-boost-converter for wind turbines concerning their efficiency and installed semiconductor power,” in proc. of ieee power electronics specialists conference (pesc), june 2008, pp. 1814–1820. [11] n. baker, m. liserre, l. dupont, and y. avenas, “improved reliability of power modules: a review of online junction temperature measurement methods,” ieee industrial electronics magazine, vol. 8, no. 3, pp. 17–27, sept 2014. [12] b. gadalla, e. schaltz, y. siwakoti, and f. blaabjerg, “thermal performance and efficiency investigation of conventional boost, z-source and y-source converters,” in ieee 16th international conference on environment and electrical engineering (eeeic), june 2016, pp. 1–6. [13] d. j. s. newlin, r. ramalakshmi, and s. rajasekaran, “a performance comparison of interleaved boost converter and conventional boost converter for renewable energy application,” in international conference on green high performance computing (icghpc), march 2013, pp. 1– 6. [14] magnetics. magnetics powder core catalog. [online]. available: http://www.mag-inc.com/company/ news/new-powder-core-catalog [15] metglas. powerlite inductor cores. [online]. available: http://www.elnamagnetics.com/wp-content/ uploads/catalogs/metglas/powerlite.pdf [16] r. w. erickson and d. maksimovic, fundamentals of power electronics, second edition. springer science + bussiness media, llc, may 2001.  transactions on environment and electrical engineering issn 2450-5730 vol 1, no 4 (2016) © g. belli, g. brusco, a. burgio, d. menniti, a. pinnarelli, n. sorrentino and p. vizza  abstract – the increase of renewable non-programmable production and the necessity to locally self-consume the produced energy led to utilize ever more storage systems. to correctly utilize storage systems, an opportune management method has to be utilized. this paper implements a multi-period management method for storage systems, using different management strategies. the method aims to minimize the total absorbed and supplied energy or the peak power exchanged with the grid. the results show the effectiveness of the method in diminishing the energy exchanged with the grid and also the possibility to optimize the performance of the storage systems. keywords — storage management, multi-pediod scheduling, prosumer, optimal energy management i. introduction he common tendency to produce on site the necessary energy to the end user by means of small size plants, generally from non-programmable renewable sources, led to a rapid increase of installed photovoltaic (pv) power. although pv generation offers economic and environmental benefits, its non-programmability can reduce these benefits. to increase the possibility to self-consume the produced energy and increase the profit for the user, storage systems should be utilized. otherwise, it is worth to underline that storage systems have a limited lifetime, related to their charge and discharge cycles. for this reason, it is opportune to manage storage systems with a specific strategy that can ensure them a long lifetime [1-2]. in order to better manage storage systems, realizing an accurate scheduling, generation and load forecasting systems would be useful to help the management of the grid. several methods carrying out storage management, considering different strategies, have been proposed in literature [3-7]. such methods focus on an economical optimization, or a real time management without considering this work was financed by the italian ministry of economic development (mise) and the ministry of education, university and research (miur) through the national operational program for development and competitiveness 2007-2013, project domus pon 03pe_00050_2. g. belli, g. brusco, a. burgio, d. menniti, a. pinnarelli, n. sorrentino, and p. vizza are with department of mechanical, energy and management engineering (dimeg) university of calabria via bucci 42c, arcavacata di rende cs, italy (e-mail: name.surname@unical.it). the optimization of storage lifetime. in [3] the “economical” optimal management of a storage system is carried out in a single period (24 hours ahead); if the optimization is not required 24h ahead, the method utilizes a real time approach. in [4] an appropriate management to reduce losses and increase distribution grid capacity is implemented and to this aim, a distributed storage is utilized; such a method allows to improve the utilization of renewable energy resources minimizing the energy adsorption from the grid. both methods [3,4] are implemented on a 24h period and they consider an economical optimization, moreover a multigeneration/multi-storage scheduling is realized. it would be interesting to observe the benefits for a single user equipped with a pv generation system and a storage system, if the considered period is greater than 24h and a multi-period optimal management method is adopted. in [5] users equipped of energy storage systems are considered; it explains how the interaction of different storage systems can be harmful for the grid, although such storage systems have been introduced to protect it. so a novel management technique for distributed storage systems is implemented. the utilization of this technique led to saving up to 13% of the electricity bill for each consumer with a 4-kwh storage system. in [6] a day ahead optimization algorithm is implemented to provide the optimal storage and/or production scheduling strategy for a single user. the real load profile is considered exactly as the programmed 24h ahead profile. at the user level, that is a prosumer level, it is advantageous to know generation and load profiles forecast to choose the better storage power scheduling. indeed, a goal for the user should be to make “himselfsustainable”, satisfying his total energy demand by means of local energy production, minimizing the exchanges of energy with the grid. this paper presents a multi-period storage management method and considers a prosumer equipped with a storage system. in particular basing on weather forecast, generation and load forecasts are obtained; artificial neural networks are utilized to implement the two forecasting models. starting from these load and generation forecasts, different simulations are performed: the method minimizes the overall energy exchanged with the grid, the power peaks between prosumer a multiperiodal management method at user level for storage systems using artificial neural network forecasts g. belli, g. brusco, a. burgio, d. menniti, member, ieee, a. pinnarelli, member, ieee, n. sorrentino, member, ieee, and p. vizza, student member, ieee t and grid or the energy in a particular time interval; moreover, a limit for the power exchanged with the grid is considered. worth noting that the benefits of the storage systems have been also demonstrated by the same authors in [8]; in particular, they underline the suitability of the li-ion batteries compared with lead-acid batteries. indeed, in this paper the economic aspects concerning the storage systems and in particular the management of the storage system are not examined. in this paper, an energetic and power analysis is carried out. the rest of the paper is structured as follows: in the second section, the implemented forecasting models are illustrated; in the third part, storage management method is described; in the last part, simulation results are shown and the benefits in the use of the method are underlined. i. pv and load forecasting models the multi-period storage management (mpsm) method has for input the renewable production profile and the load profile of a user. to generate these profiles, the mpsm method uses renewable production and load forecasting models. several production and load forecasting models exist in literature; they present different accuracy which depends on several factors, such as utilized input data, utilized methodology, and so on. independently from the used method, the results of the generation and load power forecast can be evaluated in different way. a. accuracy estimation many performance parameters are used according to forecast purpose. in general, all these parameters represent the error between the forecasted and the actual value. the mean absolute error (mae) [9] measures how close forecasts are to the outcomes. it is more sensitive to high value of the discrepancy between the forecasted and the actual power profiles, so it is used to underline the presence of discrepancy peaks (emphasizing the high error). it allows to know the consequent average power imbalance; the mae is given by: mae = 1 n ∑|fi − yi| = n i=1 1 n ∑|ei| n i=1 (1) where fi is the forecast value, yi is the real value. moreover, the evaluation of the maximum absolute error (maxae) is important to know the maximum difference between fi and yi, in particular it is useful to know the maximum power imbalance resulting from the forecast; it is given by: maxaerr = max(|fi − yi|) (2) the mean squared error (mse) [9] measures the mean square discrepancy between fi and yi. the mse is more sensitive than the mae to high error value. the mse square root provides another statistical quantity, that is the root mean squared error (rmse). the mse and the rmse are defined as: 𝑀𝑆𝐸 = 1 𝑁 ∑(fi − yi ) 2 𝑁 𝑖=1 (3) 𝑅𝑀𝑆𝐸 = √𝑀𝑆𝐸 = √ 1 𝑁 ∑(fi − yi ) 2 𝑁 𝑖=1 (4) another important statistical quantity is the mean absolute percentage error (mape) [9]. it measures the forecasting accuracy and expresses this accuracy as a percentage. it is calculated by: 𝑀𝐴𝑃𝐸 = 1 𝑁 ∑ | fi − yi fi | 𝑁 𝑖=1 (5) moreover, the maximum percentage error (maxpe) allows knowing what the maximum percentage discrepancy between the forecasted and the actual value is: maxpe = max (| fi − yi fi |) ⩝ i = 1 … n (6) such statistical parameters allow to estimate the accuracy of the several forecasting models, to find the best suited to the case under estimation. b. pv forecasting model referring to renewable production forecasting model, it is supposed that the power source is a photovoltaic (pv) plant. several pv forecasting models exist in literature: they often present accurate results. however, such models are very sophisticated and require information not always available, as those presented in [10-12]; the pv forecasting model implemented by the authors uses available input data and for this reason it is defined a “practical” forecasting model. such a pv generation forecasting model is implemented by an artificial neural network (ann). for the implementation of the ann, the neural network toolbox of matlab is utilized. the chosen ann typology is a multi-layer perceptron (mlp), with a supervised training algorithm; in particular, the back-propagation algorithm is used. moreover, only a hidden layer is utilized, the activation function of all the neurons is tan-sigmoidal, while the training function is the levenberg-marquardt method [13]. chosen the ann typology, it is necessary to provide to the method historical input data and historical output (target) data (in the case under examination the collect of the pv power production data really generated by the pv plant). the number of neurons of the input layer is chosen considering the input data, while the number of neurons of the hidden layer has been determined empirically. in fact, a sensitivity analysis has been conducted: the input data are kept constant and the number of neurons of the hidden layer are changed; under this condition, several tests to calculate the mape have been made. the optimum number of neurons for the hidden layer, to minimize the mape is 30, as shown in figure1. fig. 1. sensitivity analisys so, summarizing, the implemented ann consists in (fig. 2):  5 input neurons (meteorological condition of the considered hour h, meteorological condition of the hour h+1, meteorological condition of the hour h-1, the hourly irradiance, the considered hour);  30 hidden layer neurons;  1 output neuron (hourly forecasted power production). fig. 2. ann pv forecasting representation meteorological data about sky conditions are transformed to be used as input data for the ann, in a number. a “1” to “5” scale has been used to represent the increasing of cloudiness. the minimum value (“1”) indicates “clear-sky” condition while the maximum value (“5”) indicates “storm” (tab.1). table i. weather conditions coding code weather condition 1 clear sky 2 nearly cloudless, scattered clouds 3 few clouds 4 partly cloudy 5 covered, storm the model has been tested on a specific a pv plant and good results are obtained both for clear and non-clear sky conditions. indeed, the results show that: in clear sky conditions the mape is less than 10%, while in non-clear sky conditions the mape is about 42%. the mae, compared to pv power plant, is equal to 2.6% for clear sky conditions and 6.8% for nonclear sky conditions. figure 3 depicts the results of the pv forecast for the days from 30 august to 2 september; for these days, clear-sky conditions were predicted. the mape is calculated for all the four days: it is equal to 9.8%. whereas, the percentage error is maximum and equal to 45% in the hours close to sunrise and sunset; while the percentage error becomes minimum (less than 1%) for the hours of higher production. fig. 3. clear-sky days results c. load forecasting model referring to the load forecasting model, in [14] the relevance to use load forecast for several purposes is underlined, especially for islanded operation, because it is necessary to guarantee grid stability, in addition to allow generation programmable system to work at a maximum efficient point. in literature, there are a large number of load forecasting models as those presented in [15]. reference [16] highlights that is more difficult to predict the individual load than an aggregate of loads. nevertheless, in this paper a predictive model for an individual load is implemented. the load forecasting model implemented in this paper utilizes accessible data; such a model predicts also individual and aggregate loads. although its simplicity the results demonstrate the good performances of the model. a feed-forward multi-layer perceptron (mlp) ann, supervised by a back-propagation algorithm, has been implemented (fig. 4). for ann training, the collection of consumption data of the considered user is necessary. similarly than pv forecasting model, also for load forecasting model a sensitivity analysis has been carried out, so to detect the number of the hidden layer neurons which lead to a better accuracy of the model. the implemented ann consists in:  7 input neurons (month, day, day type, hour, daily maximum temperature, daily minimum temperature and daily average temperature);  30 hidden layer neurons;  1 output neuron (hourly forecasted power consumption). 9 11 13 15 17 19 10 15 20 25 30 35 40 45 mape[%] number of neurons fig. 4. ann load forecasting representation the input data are so defined: the day, month and hour identify the period which the forecast is required; the day type identify if the considered day is a workday or a holiday, if it is a day before or after a holiday. moreover, the minimum, maximum and average temperature are utilized; these are useful specially if the electric air conditioning is utilized, in particular maximum temperature is useful for cooling, whereas minimum temperature is useful for heating. the load forecasting model has been tested on real data of a typical residential user. in fig. 5, the forecasted and the real load profiles are shown, for three days (thursday, friday and saturday). the obtained mean absolute error (mae), compared to the rated power of the considered user’s contract is less than 6%, that is an acceptable error for the purpose of the method. fig. 5. forecasted load profile and real load profile figure 5 depicts the results of the load forecast model; the mape is calculated and it is less than 20%. whereas, the percentage error is maximum in the hours and day with a low consumption, and the minimum percentage error occurs for the workdays, when the consumption is high, and it is less than 1%. ii. storage management method description the most important variables used in the multi-period storage management (mpsm) method are reported in table 2. table ii. mpsm method variables nomenclature 𝑃𝑔 𝑡 𝑑 grid transferred power; (time t, day d) 𝑃𝐿 𝑡 𝑑 load power; (time t, day d) 𝑃𝑆 𝑡 𝑑 storage transferred power; (time t, day d) 𝐸𝑆 𝑡 𝑑 storage energy level; (time t, day d) 𝐸𝑆 𝑚𝑖𝑛, 𝐸𝑆 𝑚𝑎𝑥 minimum and maximum storage energy level 𝑃𝑆 𝑚𝑖𝑛, 𝑃𝑆 𝑚𝑎𝑥 minimum and maximum storage power 𝐸𝑆 𝐼𝑛𝑖𝑡 initial energy stored the main equations which describe the mpsm method are as follow: 𝑂𝐹: min (∑ 𝑓(𝑃𝑔 𝑡 𝑑 ) 𝐷 𝑇 𝑡=1 𝑑=1 ) (7) s.t. 𝑃𝑔 𝑡 𝑑 = 𝑃𝐿 𝑡 𝑑 − 𝑃𝑃𝑉 𝑡 𝑑 − 𝑃𝑆 𝑡 𝑑 (8) 𝐸𝑆 𝑡+1 𝑑 = 𝐸𝑆 𝑡 𝑑 + 𝑃𝑆 𝑡 𝑑 ∗ 𝑡 (9) 𝐸𝑆 𝑚𝑖𝑛 ≤ 𝐸𝑆 𝑡 𝑑 ≤ 𝐸𝑆 𝑚𝑎𝑥 (10) 𝑃𝑆 𝑚𝑖𝑛 ≤ 𝑃𝑆 𝑡 𝑑 ≤ 𝑃𝑆 𝑚𝑎𝑥 (11) 𝑠𝑖𝑔𝑛(𝑃𝑆 𝑡 𝑑 ) = 𝑠𝑖𝑔𝑛(𝑃𝑃𝑉 𝑡 𝑑 − 𝑃𝐿 𝑡 𝑑 ) (12) 𝐸𝑆 𝑡=1 𝑑=1 = 𝐸𝑆 𝐼𝑛𝑖𝑡 (13) the mpsm method can be utilized for more days (d) and every day is divided in more time intervals (t); such time intervals are the same of that used in the forecasting models. in objective function (of) (7), 𝑓(𝑃𝑔 𝑡 𝑑 ) indicates different goals of energy exchange optimization. indeed, the user can require to: minimize overall the energy exchanges with the grid, minimize the power peaks or minimize the energy exchanged for a particular time period with the grid the mpsm method is subjected to the constraints from (8) to (13). constraint (8) is used to calculate the power exchanged with the grid, 𝑃𝑔 𝑡 𝑑 . the constrains (9), (10), (11), (12) and (13) concern the storage. in (9) the variation of the stored energy (between two time intervals) is calculated, in (10) the stored energy is limited between a minimum and maximum value (depending on the used storage); in (11) the charge and discharge storage power is limited, (12) indicates that the storage can charge only if there is a power surplus (pv power 𝑃𝑃𝑉 𝑡 𝑑 is greater than the load power 𝑃𝐿 𝑡 𝑑 ), vice versa storage can only discharge. in (13) the initial stored energy is defined as 𝐸𝑆 𝐼𝑛𝑖𝑡 . once load and pv production power forecasts for the user are obtained, the difference between the two profiles is calculated. this difference profile represents the input of the mpsm method, which solves the objective function (of), taking into account the constraints. the method returns the storage power exchange profile and the consequent grid power exchange profile. iii. simulation to test the effectiveness of the mpsm method, some simulations are carried out, considering as prosumer a build of university of calabria: this is a business user, equipped with photovoltaic plants. the considered build has a maximum power consumption of 25 kw and the installed pv plants power is 45 kw. the test considers a time period of 7 days, from the 10th to 16th october 2015. in fig. 6, load and pv power forecast profiles of the considered 7 days are reported. after determining load and pv power profiles, it is necessary to sizing storage system for the required function. a. non optimized pv power first of all, the storage capacity is calculated to supply loads and limit the exchange of energy with the grid. storage capacity will be the smallest between the resulting average daily energy purchased and supplied to the grid, which are calculated as the difference between pv production and load profiles. this analysis is carried out for profiles of a typical day. in the present case, the daily purchased energy is almost 170 kwh, whereas the energy supplied to the grid is 80 kwh; so the storage would have a capacity of 80 kwh. after calculating storage capacity, an overestimation to be conservative will be necessary: an increase of 20% will be considered. in addition, in order to safeguard the storage useful life, a residual state of charge (soc) of 40% has to be considered as a further increase of the estimated storage capacity. considering the previous estimated storage capacity (80 kwh) and the increases of 20% and 40%, the obtained storage capacity is about 140 kwh. starting from the calculation of the difference between load and pv power forecasts, it is used for two groups of simulations. the first group of simulation aims to minimize the total exchange of energy with the grid, trying to make the prosumer self-sustainable and to avoid congestions on the grid. instead, the second group of simulations aims to minimize only the peaks of energy during the day, trying to reduce the costs of energy supply and to avoid worthless oversize of the generation plants. for both the groups of simulations, the comparison is made between the condition with storage working in “real time”, that is no storage management is taking into account, and the condition with storage managed by mpsm method. the starting point is the value of the overall energy exchanged between the user and the grid without using a storage system: this value is equal to 1.68 mwh, where 1.13 mwh is the energy adsorbed by the user and 0.55 mwh is the energy left to the grid. for the first groups of simulations, when none management method is utilized, the total energy exchange with the grid decreases until 0.69 mwh, where 0.60 mwh is the purchased energy by the user and 0.09 mwh supplied to the grid. if the storage is managed by mpsm method, the total energy exchanged with the grid is equal to 0.68 mwh. respect the previous case, the difference is very limited. although this difference is only of 0.01 mwh, the positive effect of the mpsm method consists in the possibility to maximize the performances of the storage. indeed, using a “real time” operation strategy, the charge and discharge cycles are not optimized because they are partial cycles, while with mpsm method, the storage executes always full cycles of charge and discharge (fig.7). only in a few hours, a distorted trend is visible in fig. 7, due to the high variability of weather conditions on the 5th day, that involves to have a partial cycle of charge and discharge. for the second group of simulations, minimizing only the peaks of power exchanged with the grid, the power exchanged in “real time” reaches 23 kw, while using mpsm method, it is about 8 kw. in fig. 8 the exchanged energy profile with and without mpsm method are depicted. fig. 6. load and pv profiles fig. 6. stored energy profile with and without mpsm fig. 7. exchanged power profiles with and without mpsm b. optimized pv power in this section, starting from the average daily load profile and the monthly average daily pv production profile, to minimize the exchange of energy with the grid, the pv plant is sized to cover the daily energy demand. considering this, the obtained rated pv power is about 56 kw. similarly to the previous subsection a, the storage system is properly sized and the obtained capacity is about 240 kwh. with such data the method is utilized to carry out the same test of the previous case. first of all, the mpsm method is utilized to minimize the exchange of energy with the grid; the obtained result of the total energy exchanged with the grid is equal to 0.41 mwh, where 0.25 mwh is the purchased energy by the user and 0.16 mwh is the energy supplied to the grid. in this case, as the rated pv power and the storage capacity are optimized, the use of the mpsm method, compared to the real time management, does not contribute to many advantages in the management of the charge/discharge storage cycles. the real advantage would occur in the management of the exchanged energy with the grid for the days with non-clear sky conditions. in fig. 9 the exchanged energy profile with the grid, with and without the mpsm method is depicted. moreover, referring to fig. 9, it is possible to observe that the energy supplied to the grid is greater than the energy purchased from the grid; only the 5th day the purchased energy is greater than the supplied one, because it is not a clear sky day. fig. 8. exchanged grid power profile with and without mpsm worth noting that the pv power is sized to cover the daily energy demand; so if the rated pv power increase, obviously the produced energy increase and as a consequence the total energy exchanged with the grid increases. in fact, for example, if the rated pv power is 60 kw the total exchanged energy is 0.49 mwh, instead of 0.41 mwh. this shows that before to use the mpsm method, optimal pv and storage sizing is necessary. c. grid power restriction in this section, it is supposed that the considered prosumer has a limit for 𝑃𝑔 𝑡 𝑑 . this can be due to different reason, for example if the prosumer has a contract with the energy provider for a reduced power, or if the power line is designed for a limited power. in fact, this kind of optimization allows to reduce the problems due to the congestion problem and any restrictions of the power interface devices. in particular two cases are examined: in the first case the maximum 𝑃𝑔 𝑡 𝑑 (pg_max) is 10 kw, in the second case pg_max is 7 kw. worth noting that the limit for the power is both for the supplied and delivered energy. the mpsm method is completed using the sequent equations: |𝑃𝑔 𝑡 𝑑 | ≤ 𝑃𝑔_𝑚𝑎𝑥 (14) the of is implemented to minimize the entire energy exchanged with the grid, as implemented above. it is worth to underline that this test is different to the previous minimization of the peak power, in fact in the previous case a restriction for 𝑃𝑔 𝑡 𝑑 is not utilized but solely the peaks of 𝑃𝑔 𝑡 𝑑 are reduced. in the first test pg_max is equal to 10 kw and the constraint (12) is relaxed, in this way the storage can be charged also by the grid and can discharge also if there is a surplus of energy, this is limited only through the of. the utilized storage capacity is 240 kwh and the rated pv power is 56 kw; the load profile is reported in fig. 6. in this first case, the obtained result of the total energy exchanged with the grid is equal to 0.41mwh, where 0.26 mwh is the purchased energy by the user and 0.15 mwh is the energy supplied to the grid. in the second case, pg_max is equal to 7 kw, the total energy exchanged with the grid is also equal to 0.41mwh, and the purchased energy is equal to 0.26 mwh whereas the supplied energy to the grid is equal to 0.15 mwh. such results demonstrate that the constrains on 𝑃𝑔 𝑡 𝑑 are almost irrelevant for the of, in fact the quantity of energy exchanged with the grid is the same of the previous case. the only difference is for the profile of 𝑃𝑔 𝑡 𝑑 : in figs. 10 and 11 the profiles of power exchanged with the grid for either cases are reported. fig. 9. exchanged grid power profile with the constrain pg_max=10 kw fig. 10. exchanged grid power profile with the constrain pg_max=7 kw the figures show that the trend of 𝑃𝑔 𝑡 𝑑 is constant in the area where the power peak occur, this means that for those time and day the bonds are achieved; this is particularly observable for pg_max is 7 kw. moreover, worth noting that the maximum power pg_max (7 kw), obtained in this case, is less than the maximum power obtained with the minimization of the peak of power (in subsection a) where pg_max is 8.2 kw. it is important to observe the behaviour of the storage system when pg_max is equal to 7 kw compared to the case when there is not a constrain for 𝑃𝑔 𝑡 𝑑 : in figure 12 this comparison is reported. fig. 11. stored energy profile with and without pg_max constrain it is possible to observe the differences between the two profiles, especially for the fifth day. in fact, to limit the power exchanged with the grid, in particular the power drawn from the grid, the stored energy is maintained as long as it is not used to decrease 𝑃𝑔 𝑡 𝑑 : the battery is not discharged just when there is a deficit of energy but when this energy is utilized to limit the maximum 𝑃𝑔 𝑡 𝑑 . thanks to the mpsm method the prosumer can employ a reduced power contract with consequent less costs for the prosumer. at the same time the distribution system operator (dso) can design the line for a reduced power, with further savings. iv. conclusion the paper shows the importance of an opportune management method for storage devices. in fact, the positive effect resulting from the use of storage systems, particularly in relation to non-programmable resources, can be increased if an appropriate management strategy is utilized. one of the feature of the implemented method is its multi-periodicity. in fact, if the management is made on more days, there are more data input and the storage can be managed in a better way. the presented storage management method implements different management goals. first of all, the method minimizes the total energy exchanged with the grid to make users self-sustainable. this implies the use of opportune pv and load forecast models. secondly the method is also utilized to reduce the peak of power exchanged with the grid, decreasing from 23 kw to about 8 kw. moreover, it is underlined that an accurate sizing of pv and storage systems is necessary before to implement and utilize a management strategy. the results are compared with the real time storage management; such a comparison shows the effectiveness of the method. the results show also the possibility to optimize the performance of the storage device in terms of charge and discharge cycles. at the end, the behaviour of the management method, if a constrain for the maximum 𝑃𝑔 𝑡 𝑑 is utilized, is evaluated. simulations are carried out; they demonstrate that despite a further constrain is utilized, the entire energy exchanged with the grid is minimized. this can be a good result both for the prosumer and for the dso, indeed they can respectively reduce the contract power and reduce the line capacity, with consequent savings. moreover, it would be interesting evaluate the behaviour of the method if a schedulable load is adopted. references [1] d. menniti, a. pinnarelli, n. sorrentino, a. burgio, and g. brusco, “energy management system for an energy district with demand response availability”, smart grid, ieee transactions on, vol. 5(5), 2014, pp. 2385-2393. [2] d. menniti, a. pinnarelli, n. sorrentino, g. belli, a. burgio, “demand response program in an energy district with storage availability”, international review of electrical engineering, in press. [3] a. nottrott, j. kleissl, and b. washom, “energy dispatch schedule optimization and cost benefit analysis for grid-connected, photovoltaicbattery storage systems”, renewable energy, vol. 55, 2013, pp. 230-240. [4] n. jayasekara, p. wolfs, and m.a.s. masoum, “an optimal management strategy for distributed storages in distribution networks with high penetrations of pv”, electric power systems research, vol. 116, 2014, pp. 147-157. [5] p. vytelingum, t.d. voice, s.d. ramchurn, a. rogers, and n.r. jennings, “agent-based micro-storage management for the smart grid”, in proc. of the 9th international conference on autonomous agents and multiagent systems, vol. 1, international foundation for autonomous agents and multiagent systems, pp. 39-46, may 2010. [6] i. atzeni, l.g. ordóñez, g. scutari,d.p. palomar, and j.r. fonollosa, “demand-side management via distributed energy generation and storage optimization”, smart grid, ieee transactions on, vol. 4(2), 2013, pp. 866-876. [7] a. mohamed, and o. mohammed, "real-time energy management scheme for hybrid renewable energy systems in smart grid applications", electric power systems research, vol. 96, 2013, pp. 133-143. [8] d.menniti, a. pinnarelli, n. sorrentino, a. burgio, g. brusco, “the economic viability of a feed-in tariff scheme which solely awards the self-consumption for promoting the use of integrated photovoltaicbattery systems”, applied energy, in press. [9] s. makridakis, and m. hibon, “evaluating accuracy (or error) measures”, working paper 95/18/tm, insead, (1995) france. [10] y. zhang, m. beaudin, raouf taheri, h. zarcipour, and d. wood, “day-ahead power pv power production output forecasting for small scale soar photovoltic electricity generators”, ieee transactions on smart grid, vol. 6, no. 5, september 2015 [11] c. chen, s. duan, t. cai, and b. liu, “online 24-h solar power forecasting based on weather type classification using artificial neural network”, solar energy, vol. 85, no. 11, 2011, pp. 2856-2870. [12] c. w. chow, b. urquhart, j. kleissl, m. lave, a. dominguez, j. shields, and b. washom, “intra‐hour forecasting with a total sky imager at the uc san diego solar energy testbed”, solar energy, vol 85, no. 11, 2011, pp 2881–2893. [13] j. j. moré, “the levenberg-marquardt algorithm: implementation and theory”, in lecture notes in mathematics, no. 630–numerical analysis, springer-verlag, 1978, pp. 105–116. [14] n. hatziarg, microgrids: architectures and control, wiley-ieee press, february 2014. [15] h. s. hippert, c. e. pedreira, and r.c. souza, “neural networks for short-term load forecasting: a review and evaluation”, power systems, ieee transactions on, vol. 16(1), 2011, pp 44-55. [16] h. chitsaz, h. shaker, h. zareipour, d. wood, and n. amjady, “shortterm electricity load forecasting of buildings in microgrids”, energy and buildings, vol. 99, 2015, pp. 50-60. grazia belli (italy, 1985) received her degree in energetic engineering in 2011 and her ph.d in science of complex systems in 2016 from the university of calabria. her current research interests concern renewable energy sources, distributed generation, smart grid technologies and electricity local market. giovanni brusco (italy, 1980) received his degree in electronics engineering from the university of calabria, italy, in 2007 and his ph.d. in computer and system engineering in 2013 at the electronic, computer and systems science department of the university of calabria, italy. his current research interests concern renewable energy sources, distributed generation, harmonic analysis and smart grid technologies. alessandro burgio (italy, 1973) received his degree in management engineering from the university of calabria in 1999 and his ph.d. in computer and system engineering in 2006 at the electronic, computer and systems science department of the university of calabria, italy. his current research interests include electrical power systems, distributed generation, renewable energy, power electronics and harmonics, electronic ballast. daniele menniti (italy 1958) received his degree in electrical engineering from the university of calabria, cosenza, italy and his ph.d. degree in electrical engineering from the university of naples, italy, in 1984 and 1989 respectively. he is an associate professor at the mechanical, energetic and management department of the university of calabria, italy. his current research interests concern electrical power system analysis, real-time control and automation. anna pinnarelli (italy, 1973) received her degree in management engineering from the university of calabria in 1998 and her ph.d. in electrical engineering in 2002 from the electrical engineering department of the university of naples, italy. she is an assistant professor at the mechanical, energetic and management department of university of calabria, italy. her current research interests concern facts technology, harmonic analysis, electrical system automation, decentralized control and smart grid technologies. nicola sorrentino (italy, 1970) received his degree in management engineering in 1994 and a ph.d. in computer and system engineering in 1999 at the electronic, computer and systems science department of the university of calabria, italy. he is a researcher at the mechanical, energetic and management department of the university of calabria, italy. pasquale vizza (italy, 1990) received his degree in energetic engineering in 2014 from the university of calabria; he is currently attending the phd school at the same university. his current research interests include renewable energy sources, smart grid technologies, energy storage economics, generation and load forecasting. i. introduction i. pv and load forecasting models a. accuracy estimation b. pv forecasting model c. load forecasting model ii. storage management method description iii. simulation a. non optimized pv power b. optimized pv power c. grid power restriction iv. conclusion transactions on environment and electrical engineering issn 2450-5730 vol 1, no 4 (2016) © alana s. magalhães, pedro h. f. moraes, alan h. f. silva, pedro h. g. gomes, aylton j. alves, wesley p. calixto  abstract—the purpose of this paper is to compare mathematical modeling and practical bench in order to validate the electrical interactions between an induction generator and a synchronous generator. two generators was connected to a common bus in steady state, subject to non-linear load. the results comparing modeling and bench tests show that the induction generator besides the active power increasing, has a better way for harmonic currents flowing in common bus. it was concluded that the induction generator repowering and attenuates current harmonic components present at the connection point, improving the network voltage profile. index terms—repowering, induction generator, synchronous generator, harmonics. i. introduction epowering hydroelectric power plants has been increasing the power generated. since there is spare capacity of turbine power and that is not being exploited by the generator already installed, it can be repowered. there are three possible ways to repowering: i) replacing the synchronous generator for a bigger one; ii) adding a second synchronous generator through double coupling on the turbine shaft; iii) adding a second generator coupled to the turbine shaft, but in this case an induction generator. the induction generator is a viable technical and economical option to power generation [1]. the induction generator is used in electrical power plants repowering therefore has a low cost, is more robust, has simple construction, lower cost and less maintenance when compared to a synchronous machine. as disadvantage, external resources are required to compensate reactive power. on repowering, smaller induction generator is connected on a common bus to a larger synchronous generator and thus induction generator may have its reactive power compensated a. s. magalhães{1,2}, alanadsm@gmail.com l. c. a. junior{1}, leovir.engmecatronica@gmail.com c. a. matias{1}, calebeabrenhosa@gmail.com a. h. f. silva{1,2}, alanhfs@gmail.com e. g. domingues{1}, eldergd@ifgoias.edu.br a. j. alves{1}, aylton.alves@ifg.edu.br w. p. calixto{1,2}, wpcalixto@ieee.org 1experimental and technological research and study group (next), federal institute of goias (ifg), goiânia, brazil. 2 school of electrical, mechanical and computer engineering, federal university of goias (ufg), goiânia brasil.} by synchronous generator, without power factor losses in the coupling point between them and can be dispensed of the control voltage, as this will be determined by system [2]. the induction generator besides low maintenance, does not require dc excitation and synchronization. in machines parallel operation is necessary to use of motorized thermomagnetic circuit breakers and in the case of induction generators, where the synchronization is not required, it reduces the cost of the circuit breaker [3]. in the distribution system the impact of induction generators connection is studied in [4]. to stabilize reactive field, the induction generator needs to reactive power on system input. the system supplies this reactive power, affecting losses and system voltage drops. the results show the relationship between losses due to change of the voltage profile, and as a solution indicates the power factor correction. the hydro-quebec system in canada, the demand for small generators connection is increased [5]. recently, study [6] shows that rural electrification can be supplied by small hydropower through induction generator and intelligent controllers in more economic schemes and cost-effective options. studies in [7] compares the use of conventional synchronous machines together with the static frequency converter (sfc) in the kadamparai plant with substitution by a variable speed induction generators to utilize the grid load variation effectively. the results show that the plant can be operated by variable speed machines. the parallel operation voltage and frequency control was performed in [8], where induction generator can provide constant power and does not have excitation control. the synchronous generator has variable excitation in different load conditions. the results also show that changing the reactive load consumption can be supplied by the synchronous generator, keeping the voltage constant to 1 pu. the induction generator operates at full rating and does not respond does not respond to load change in the consumer. in the interconnected electric power system ieps there is presence of a large number of synchronous generating units of high power and non-linear loads. the application of rules aiming to limit the harmonic content of tensions on possible values of maintaining acceptable power quality is recommended [9]. in [10] is presented tests that induction generator does not introduce harmonics in power system. [11] shows the repowering system using the common bus two machines of the same power, a synchronous generator and an reconditioning in synchronous operation with one parallel induction generator alana s. magalhães1,2, pedro h. f. moraes1, alan h. f. silva1,2, pedro h. g. gomes1,2, aylton j. alves1 , wesley p. calixto1,2. r induction generator. the results show that the induction generator reduces harmonic content in the common bus. in [13] proposed the development of a simulation model for repowering steam plant, providing dark areas of links aimed at accelerating the power system restoration process services. recently, [14] conduct a study to evaluate the economic and repowering prospects of a plant into disuse in the territory of petralia sottana (sicily). the work shows that the refurbishment of the plant "catarrate" contributes to the energy independence of the local community, with an estimated annual production of renewable energy of approximately 220 mwh and at the same time, the preservation of industrial heritage. this paper aims to repowering the system. furthermore, the objective is that the induction generator insertion in the common bus to a synchronous generator, can improve the sinusoidal profile of voltage and current. thus, it is noted that the induction generator is still a preferential path for harmonic currents becoming protection synchronous generator, which is a more expensive and less robust machine that induction one. ii. mathematical modeling a. three-phase induction generator under non-sinusoidal steady state fig. 1 presents electrical circuit that models the induction machine in non-sinusoidal steady state, where 𝑋𝐸 is stator leakage reactance and 𝐸𝑎ℎ is the ℎ order harmonic component of voltage, induced in phase 𝑎 machine stator, by the magnetic field produced by sinusoidal spatial distribution of rotating magneto-motive force of ℎ order, 𝑓𝑚𝑚𝐸0ℎ [12]. fig. 1. induction machine representative electrical circuit. considering odd values for ℎ index, which are most likely harmonic components produced by non-linear loads, one can write: �̇�ℎ = �̇�ℎ ∙ 𝐼ℎ̇ (1) with such assumptions the equivalent circuit becomes purely inductive, and impedance �̇�ℎof the circuit is expressed by: �̇�ℎ = 𝑗ℎ(𝑋𝐸 + 𝑘𝑅 ∙ 𝑋′𝑅𝐵 ) (2) as 𝑋′𝑅𝐵 has very similar value to 𝑋𝐸 and 𝑘𝑅 tends to one, can be a approach to accept �̇�ℎ to: �̇�ℎ ≅ 𝑗2ℎ𝑋𝐸 (3) therefore, (1), (2) and (3), leads to: �̇�ℎ ≅ 𝑗2ℎ𝑋𝐸 𝐼ℎ̇ (4) b. three-phase synchronous generator under nonsinusoidal steady state for all phases of synchronous machine, and adopting usual nomenclature to represent harmonic reactance proposed in [12], we have (5) where 𝑟𝐸 is per phase stator resistance, 𝑋𝑆 is synchronous reactance at frequency 𝜔 and 𝑋𝑎𝑓 is stator-rotor mutual reactance at frequency 𝜔. �̇�ℎ = [𝑟𝐸 + 𝑗ℎ𝑋𝑆 ] ∙ 𝐼ℎ̇ + 𝑗 ℎ𝑋𝑎𝑓 2 ∙ 𝐼�̇�(ℎ) (5) in practice 𝑟𝐸 ≪ 𝑋𝑆 and representing the last term by (6), leads to (7): �̇�ℎ = 𝑗 ℎ𝑋𝑎𝑓 2 ∙ 𝐼�̇�ℎ (6) �̇�ℎ ≅ 𝑗ℎ𝑋𝑆 𝐼ℎ̇ + �̇�ℎ (7) expression (7) suggests the circuit of fig. 2. fig. 2. cylindrical rotor synchronous machine equivalent circuit. from undertaken mathematical and physical analyzes, it is concluded that power flowing through terminals �̇�ℎ is practically inductive reactive, therefore suggesting, there is only inductive impedance in circuit which relates �̇�ℎ and 𝐼ℎ̇ , which may be represented by ℎ𝑋𝑆. c. association between induction and synchronous generator assuming two machines, one synchronous and other an induction one, of same power, connected to same bus, fig. 3, it is possible to make comparative analysis of harmonic current components in both. fig. 3. parallel machines. expressions (4), induction machine and (7), synchronous machine, can be rewritten as illustration of fig. 3, and (8) and (9), respectively. �̇�ℎ ≅ 𝑗2ℎ𝑋𝐸 𝐼ℎ̇𝐼 (8) �̇�ℎ ≅ 𝑗ℎ(𝑋𝑆 + 𝑋)𝐼ℎ̇𝑆 (9) where 𝑋 is equivalent reactance between terminals �̇�ℎ . substituting (8) in (9) and through algebraic manipulation, it has: 𝐼̇ℎ𝐼 𝐼̇ℎ𝑆 = 𝑋𝑆+𝑋 2∙𝑋𝐸 (10) assuming threshold condition, where 𝑋 is is negligible in comparison to 𝑋𝑆 and 𝑋𝑆 = 10 ∙ 𝑋𝐸 , from (10) we have: 𝐼̇ℎ𝐼 𝐼̇ℎ𝑆 = 5 (11) in expression (11) 𝑋𝑆 represents phase leakage reactance of the synchronous machine, plus armature reaction, while 𝑋𝐸 is stator leakage reactance of induction machine. by boundary condition, it is possible to ensure the inequality: 𝐼ℎ̇𝐼 > 5 ∙ �̇�ℎ𝑆 (12) in (12) it is conclude that in same bus, harmonic components of currents will flow with higher intensity to induction machine. this fact justifies the proposal of this work, of using induction machine as a means to absorb harmonic components of currents, attenuating its flow to synchronous machine. it follows that when the machine is seen only by the fundamental sinusoidal component, the power flowing in the rotor is almost exclusively active, while, when viewed for a single harmonic component, the power flowing in the rotor is almost entirely inductive reactive. it allows to assume the intensities as irrelevant, or even the direction of electromagnetic torque (motor or generator), to simulate the conditions of harmonic mitigation in synchronous machine. iii. methodology the methodology will be developed in following steps: i. modeling the illustrated electrical system in fig. 4 with the characteristics tab. i; ii. conducting testing connected to common bus nonlinear load 𝑁𝐿 ; iii. conducting testing connected to common bus nonlinear load 𝑁𝐿 and synchronous generator 𝑆𝐺 ; iv. conducting testing connected to common bus nonlinear load 𝑁𝐿 , synchronous generator 𝑆𝐺 and an induction generator 𝐼𝐺 . v. conducting testing connected to common bus nonlinear load 𝑁𝐿 and an induction generator 𝐼𝐺 . the power values will be recorded in meter 𝑀1 in order to prove the increase in power output. for more information on harmonics attenuation, the harmonic content will be recorded at the point of measurement 𝑀1, 𝑀2, 𝑀3 and 𝑀4 to have better understanding of harmonic flows in the system. a. connection machine and loads for case study 1 laboratory tests will be carried out, for ieps shown in fig. 4, where 𝑀1, 𝑀2, 𝑀3, 𝑀4 and 𝑀5 are points for quantities measurements. fig. 4. interconnected electrical power system ieps for case study 1. experimental tests of this work were performed in the laboratory with a system composed of two generating units, a synchronous and another induction. both units are in parallel by feeding the first rectifier which constitute the nonlinear load. the 𝑁𝐿 load is a resistive load of 500 watts, fed by a rectifier. to regulate properly the speed of generators, 𝑆𝐺 and 𝐼𝐺 , they used dc motors. the fig. 5 presents the equipment used in the laboratory. fig. 5. equipments utilized in laboratory tests. b. connection machine and loads for case study 2 laboratory tests will be carried out, for ieps shown in fig. 6, where 𝑀1, 𝑀2, 𝑀3 and 𝑀4 are points for quantities measurements. fig. 6. interconnected electrical power system ieps for case study 2. experimental tests of this work will be performed in the laboratory with system composed of two generating units, a synchronous and another induction. both units will be in parallel by feeding nonlinear load 𝑁𝐿 consisting of triac rectifier feeding sets of lamps. two phases with total power of 5 kw and the third phase with 4kw. to regulate generators speed, 𝑆𝐺 and 𝐼𝐺 . were used diesel engine and induction motor with frequency inverter, respectively. since the induction motor will be fed by the common bus, through 𝑆3 key. the fig. 7 presents the equipment used in the laboratory, in which fig. 7(a) the induction generator and fig. 7(b) shows the synchronous generator. (a) induction generator (b) synchronous generator fig. 7. equipments utilized in laboratory tests. iv. results a. case study 1 1) experimental tests components and values of ieps of fig. 4 are reported in the tab. i, along with their values. table i. acronyms and values of the components from ieps. variable components components values of used 𝑆𝐺 synchronous generator (main generator) 2kva, 230v, three-phase, salient, 4poles, 60hz 𝐼𝐺 induction generator 2kva, 220v, three-phase, cage rotor, 4poles, 60hz 𝑁𝐿 nonlinear load 500w three-phase, 380v, 60 hz 𝑇1 transformador 5kw, 380/220 v, δ/y aterrado 𝑆1, 𝑆2, 𝑆3 interrupter the main objective of ieps experimental testes is to obtain the increment of power generated plant at the measurement point 𝑀1 and results of total harmonic distortion of current 𝑇𝐻𝐷𝑖 , measured in points 𝑀1, 𝑀2, 𝑀3 and 𝑀4, maintaining total harmonic distortion of voltage 𝑇𝐻𝐷𝑣 within standard limits. the limit established by standard and presented in ieeestd-519-1992, [9], for voltage harmonic distortions, varies according to the voltage class in the measured point. in this case, as the measurement points has a 380 v voltage level, the limit of total harmonic distortion of voltage 𝑇𝐻𝐷𝑣 should be 5.0% and the limit of the individual distortion should be 3.0%. 2) repowering the tab. ii shows the operating conditions of the synchronous generator 𝑆𝐺 and the induction generator 𝐼𝐺 for experimental testing. the values of active, reactive and total power and power factor of the 𝑆𝐺 and 𝐼𝐺 were obtained in the measurement points and 𝑀4 and 𝑀3, respectively for loads 𝑁𝐿 . the tab. iii present data of active, reactive and total power and power factor to the measuring point 𝑀1 for the various configurations proposed to loads 𝑁𝐿 . table ii. active, reactive and total power and power factor in 𝑆𝐺 e 𝐼𝐺 for 𝑁𝐿. operation p(w) q(var) s(va) fp 𝑆𝐺 -1085 -3595 i 3767 0.291 ig -1007 3783 c 3767 0.256 table iii. active, reactive and total power and power factor in 𝑀1 𝑓𝑜𝑟 𝑁𝐿. configuration p(w) q(var) s(va) fp 𝑁𝐿 500 174.9 i 531.3 0.943 𝑆𝐺 + 𝑁𝐿 -681 -3379 i 3454 0.2 𝑆𝐺 + 𝐼𝐺 + 𝑁𝐿 -1644 818 c 1865 0.92 𝐼𝐺 + 𝑁𝐿 -514 3949 c 3995 0.133 the tab. iii present the data powers in secondary side of the transformer for various configurations with two types of nonlinear load connected to the system. in the configuration where only 𝑁𝐿 is connected, the network is providing active power of 500w. with the synchronous generator connection, configuration 𝑆𝐺 + 𝑁𝐿 , the network is providing active power of 681 w. connecting the induction generator, setting 𝑆𝐺 + 𝐼𝐺 + 𝑁𝐿 , the network is receiving active power of 1644 w. note that with the inclusion of the induction generator is repowering of the system. note also that the power factor in 𝑀1 the configuration 𝑆𝐺 + 𝐼𝐺 + 𝑁𝐿 is 0.92. 3) harmonics the values shown in tab. iv and tab. v illustrate for a total harmonic distortion of voltage and for a total harmonic distortion of current to the measuring points 𝑀1, 𝑀2, 𝑀3 , 𝑀4 and 𝑀5 , respectively. by measuring 𝑀1, presented in tab. iv and tab. v is observed that the value total harmonic distortion of voltage increases of 1.9% for 2.1% in the configuration 𝑆𝐺 + 𝑁𝐿 , and mitigates to 1.7% in 𝐼𝐺 + 𝑁𝐿 . in the setting 𝑆𝐺 + 𝐼𝐺 + 𝑁𝐿 , mitigates the amount to 1.5%. the total harmonic distortion of current generated for the setting 𝑁𝐿 in 𝑀1 is 23.1%. in setting 𝑆𝐺 + 𝑁𝐿 mitigates the value to 4.6% and setting 𝐼𝐺 + 𝑁𝐿 mitigates the value to 5.6%. in the setting 𝑆𝐺 + 𝐼𝐺 + 𝑁𝐿 mitigates the amount to 12.1%. this proves that both the synchronous generator as induction generator mitigates the harmonic distortion in ieps. in setting 𝑆𝐺 + 𝐼𝐺 + 𝑁𝐿 the value of 𝑇𝐻𝐷𝑖 is 2.8% in 𝑀3 and 1.5% in 𝑀4 and the value of 𝑇𝐻𝐷𝑣 is 1.5% in 𝑀3 and 1.4% in 𝑀4, showing that the induction generator behaves as a preferential path for harmonic. table iv. values of thdv (%) in 𝑀1, 𝑀2, 𝑀3 𝑒 𝑀4 with 𝑁𝐿 . tdhv configuração 𝑀5 𝑀1 𝑀3 𝑀4 cnl 1.7 1.9 sg+cnl 1.7 2.1 2.0 sg+ig+cnl 1.6 1.5 1.5 1.4 ig+cnl 1.7 1.7 1.7 table v. values of thdi (%) in 𝑀1, 𝑀2, 𝑀3 𝑒 𝑀4 with 𝑁𝐿 . tdhi configuração 𝑀5 𝑀1 𝑀3 𝑀4 cnl 17.8 23.1 sg+cnl 4.3 4.6 1.8 sg+ig+cnl 17.9 15.4 2.8 1.5 ig+cnl 5.5 5.6 3.3 these results reaffirm the proposed use of induction generators to mitigate the harmonics in the main generators of power plants. b. case study 2 1) experimental tests components and values of ieps of fig. 6 are reported in the tab. vi, along with their values. table vi. acronyms and values of the components from ieps. variable components components values of used 𝑆𝐺 synchronous generator (main generator) 37kva, 380v, three-phase, salient, 4poles, 60hz 𝐼𝐺 induction generator 7.5kva, 380v, three-phase, cage rotor, 4poles, 60hz 𝑁𝐿 nonlinear load 14kw three-phase, 380v, 60 hz 𝑆1, 𝑆2, 𝑆3 interrupter 2) repowering the operating conditions of synchronous generator 𝑆𝐺 and induction generator 𝐼𝐺 for experimental testing are presented in tab. vii. the values of active, reactive and total power and power factor were obtained in measurement points 𝑀4 and 𝑀3, for loads 𝑁𝐿 . the tab. viii present data of active, reactive and total power and power factor to measuring point 𝑀1 for various configurations proposed to loads 𝑁𝐿 . the excitement of synchronous generator was tuned to get the best power factor 𝑁𝐿 on 𝑆𝐺 + 𝐼𝐺 + 𝑁𝐿 configuration. table vii. active, reactive and total power and power factor in 𝑆𝐺 e 𝐼𝐺 for 𝑁𝐿. operation p(w) q(var) s(va) fp 𝑆𝐺 -23003 -7912 24343 0.945 ig -4011 4737 6214 0.645 table viii. active, reactive and total power and power factor in 𝑀1 𝑓𝑜𝑟 𝑁𝐿. configuration p(w) q(var) s(va) fp 𝑁𝐿 1452 2996 5713 0.254 𝑆𝐺 + 𝑁𝐿 -21682 -4751 22699 0.955 𝑆𝐺 + 𝐼𝐺 + 𝑁𝐿 -19847 -553 22011 0.895 𝐼𝐺 + 𝑁𝐿 2830 8985 13332 0.334 the tab. viii presents the data powers in secondary side of the transformer for various configurations with types of nonlinear load connected to system. in the configuration where only 𝑁𝐿 is connected, the network is providing active power of 1452 w. in 𝑆𝐺 + 𝑁𝐿 configuration, with synchronous generator connection, that provides active power of 23003 w, as tab. vii. network starts to receive active power of 21682 w in the case. connecting induction generator, setting 𝑆𝐺 + 𝐼𝐺 + 𝑁𝐿 , the network is receiving active power of 19847 w. in this case, has a load receiving 452 w, synchronous generator providing 23003 w, induction generator providing 4011 w, as tab. vii. the primary machine of induction generator, connected to 𝑆3 key receive 5368 w. note that with induction generator inclusion, the system is repowering. note also that the power factor in 𝑀1 the configuration 𝑆𝐺 + 𝐼𝐺 + 𝑁𝐿 is 0.895, this is due to the power factor of synchronous generator manufacturer that is 0.8. 3) harmonics the values shown in tab. ix and tab. x illustrate for a total harmonic distortion of voltage and for a total harmonic distortion of current to measuring points 𝑀1, 𝑀2, 𝑀3 and 𝑀4. table ix. values of thdv (%) in 𝑀1, 𝑀2, 𝑀3 𝑒 𝑀4 with 𝑁𝐿. tdhv configuration 𝑀1 𝑀2 𝑀3 𝑀4 cnl 1.6 1.6 sg+cnl 1.5 1.5 sg+ig+cnl 1.5 1.5 1.5 1.4 ig+cnl 1.5 1.5 1.5 table x. values of thdi (%) in 𝑀1, 𝑀2, 𝑀3 𝑒 𝑀4 with 𝑁𝐿. tdhi configuration 𝑀1 𝑀2 𝑀3 𝑀4 cnl 137.9 137.9 sg+cnl 21.9 3.4 sg+ig+cnl 41.9 137 3.7 3.4 ig+cnl 81.2 137.1 3.5 by measuring 𝑀1, presented in tab. ix and tab. x is observed that the value total harmonic distortion of voltage is 1.6% in 𝑁𝐿 configuration and mitigates to 1.5% in configurations 𝑆𝐺 + 𝑁𝐿 , 𝐼𝐺 + 𝑁𝐿 , and 𝑆𝐺 + 𝐼𝐺 + 𝑁𝐿 . the total harmonic distortion of current generated for the 𝑁𝐿 setting in 𝑀1 is 137.9%. in setting 𝑆𝐺 + 𝑁𝐿 mitigates the value to 21.9% and setting 𝐼𝐺 + 𝑁𝐿 mitigates the value to 41.9%. in the setting 𝑆𝐺 + 𝐼𝐺 + 𝑁𝐿 mitigates to 81.2%. this proves that both synchronous and induction generator decrease the harmonic distortion in ieps. in setting 𝑆𝐺 + 𝐼𝐺 + 𝑁𝐿 the value of 𝑇𝐻𝐷𝑖 is 3.7% in 𝑀3 and 3.4% in 𝑀4 and the value of 𝑇𝐻𝐷𝑣 is 1.5% in 𝑀3 and 1.4% in 𝑀4, showing that the induction generator behaves as a preferential path for harmonic. these results reaffirm the proposed use of induction generators to repowering and attenuation the harmonics in the main generators of power plants. the fig. 8 shows the current waveform with non-linear load connected in system. the total harmonic distortion of voltage 𝑇𝐻𝐷𝑣 and current 𝑇𝐻𝐷𝑖 with nonlinear load connected to system was 1.6% and 137.9%, respectively. all individual harmonics were significant with values above 18.1%, individual harmonic values are shown in tab. xi. fig. 8. waveform in 𝑀1 with 𝑇𝐻𝐷𝑖𝑁𝐿 connected. table xi. values in 𝑀1 with 𝑇𝐻𝐷𝑖𝑁𝐿 connected. 𝑇𝐻𝐷𝑣 1.6% 𝑇𝐻𝐷𝑖 137.9% harmonic ab bc ca 60 hz (fnd)) 100% 100% 100% 180 hz (h3) 89.2% 89.5% 89.4% 300 hz (h5) 70.0% 70.2% 70.8% 420 hz (h7) 48.1% 48.3% 49.3% 540 hz (h9) 29.9% 30.2% 30.6% 660 hz (h11) 21.4% 21.1% 20.7% 780 hz (h13) 20.7% 19.9% 19.1% 900 hz (h15) 19.5% 18.8% 18.1% the fig. 9 shows the current waveform, after entry of synchronous generator with non-linear load connected to system. the total harmonic distortion of voltage 𝑇𝐻𝐷𝑣 and current 𝑇𝐻𝐷𝑖 after synchronous generator switching with a nonlinear load connected was 1.5% and 21.9%, respectively. the total harmonic distortion of current 𝑇𝐻𝐷𝑖 attenuated from 137.9% to 21.9% and the most significant individual harmonic orders were the third order ℎ3 an attenuation from 89.5% to 14.1% and fifth order ℎ5 with an attenuation from 70.8% to 13.4%, the values of the other harmonics are listed in tab. xii. the reduction is due to the fact that the synchronous generator is overexcited in order to supply reactive induction generator, while maintaining the power factor as close to 0.92 in 𝑀1, when the configuration 𝑆𝐺 + 𝐼𝐺 + 𝑁𝐿 . fig. 9. waveform in 𝑀1 with 𝑇𝐻𝐷𝑖𝑆𝐺 + 𝑁𝐿 connected. table xii. values in 𝑀1 with 𝑇𝐻𝐷𝑖𝑆𝐺 + 𝑁𝐿 connected. 𝑇𝐻𝐷𝑣 1.5% 𝑇𝐻𝐷𝑖 21.9% harmonic ab bc ca 60 hz (fnd)) 100% 100% 100% 180 hz (h3) 13.7% 13.0% 14.1% 300 hz (h5) 13.4% 11.4% 13.0% 420 hz (h7) 8.3% 6.5% 8.3% 540 hz (h9) 5.0% 3.9% 4.7% 660 hz (h11) 3.6% 2.7% 3.4% 780 hz (h13) 3.0% 2.1% 2.7% 900 hz (h15) 3.2% 2.3% 2.9% the fig. 10 shows the current waveform, after induction generator input with synchronous generator and non-linear load connected to system. the total harmonic distortion of voltage 𝑇𝐻𝐷𝑣 and current 𝑇𝐻𝐷𝑖 after induction generator switching with a synchronous generator and a nonlinear load connected was 1.5% and 41.9%, respectively. the total harmonic distortion of current 𝑇𝐻𝐷𝑖 increased from 21.9% to 41.9% and the most significant individual harmonic orders were the third order ℎ3 an increment from 14.1% to 45.6% and fifth order ℎ5 with an increment from 13.4% to 38.3%. the values of the other harmonics are listed in tab. xiii. the purpose of this configuration is to keep the power factor as close to 0.92 in 𝑀1, which means that there is reduction of harmonics in relation to the configuration 𝑁𝐿 , but increase over the 𝑆𝐺 + 𝑁𝐿 . fig. 10. waveform in 𝑀1 with 𝑇𝐻𝐷𝑖𝑆𝐺 + 𝐼𝐺 + 𝑁𝐿 connected. table xiii. values in 𝑀1 with 𝑇𝐻𝐷𝑖𝑆𝐺 + 𝐼𝐺 + 𝑁𝐿 connected. 𝑇𝐻𝐷𝑣 1.5% 𝑇𝐻𝐷𝑖 41.9% harmonic ab bc ca 60 hz (fnd)) 100% 100% 100% 180 hz (h3) 12.2% 35.3% 45.6% 300 hz (h5) 12.1% 38.3% 27.5% 420 hz (h7) 7.4% 0.6% 9.3% 540 hz (h9) 4.3% 8.3% 4.1% 660 hz (h11) 3.1% 3.5% 1.4% 780 hz (h13) 2.7% 2.2% 1.0% 900 hz (h15) 2.8% 5.3% 1.6% the fig. 11 shows the current waveform, after entry of induction generator with non-linear load connected to system. the total harmonic distortion of voltage 𝑇𝐻𝐷𝑣 and current 𝑇𝐻𝐷𝑖 after synchonous generator swiching off, with the induction generator and a nonlinear load connected to system was 1.5% and 81.2%, respectively. the total harmonic distortion of current 𝑇𝐻𝐷𝑖 attenuated from 137.9% to 81.2% and the most significant individual harmonic orders were the third order ℎ3 an attenuation from 89.5% to 83.0% and fifth order ℎ5 with an increment from 70.8% to 76.6%, the values of the other harmonics are listed in tab. xiv. it shows a decrease with respect to 𝑁𝐿 configuration, but it is important to note that with induction generator connection there feeding of primary machine that increases distortion at 𝑀3 measuring point. furthermore, the induction generator is a smaller machine than the synchronous one. fig. 11. waveform in 𝑀1 with 𝑇𝐻𝐷𝑖𝐼𝐺 + 𝑁𝐿 connected. table xiv. values in 𝑀1 with 𝑇𝐻𝐷𝑖𝐼𝐺 + 𝑁𝐿 connected. 𝑇𝐻𝐷𝑣 1.5% 𝑇𝐻𝐷𝑖 81.2% harmonic ab bc ca 60 hz (fnd)) 100% 100% 100% 180 hz (h3) 37.6% 83.0% 61.1% 300 hz (h5) 29.4% 76.6% 30.5% 420 hz (h7) 20.7% 5.8% 10.1% 540 hz (h9) 12.6% 20.6% 3.5% 660 hz (h11) 9.1% 9.1% 1.0% 780 hz (h13) 8.8% 6.8% 1.2% 900 hz (h15) 8.3% 11.8% 2.1% the individual harmonic distortion of current generated at 𝑀1 is higher than 𝑁𝐿 configuration. both for 𝑆𝐺 + 𝑁𝐿 setting as for 𝑆𝐺 + 𝑁𝐿 there is an attenuation in individual distortions, to be more significant in configuration 𝑆𝐺 + 𝑁𝐿 . it is necessary to conduct a detailed analysis 𝑆𝐺 + 𝑁𝐿 and 𝐼𝐺 + 𝑁𝐿 configurations, where it is important to consider that: i) the synchronous generator is configured to supply the reactive induction generator keeping the power factor as 0.92 in 𝑀1, which makes work in the region where shows attenuation characteristic of harmonics; ii) at 𝑀3 measuring point is included the induction generator and the primary machine, increasing total harmonic distortion of this configuration, since the primary machine still has power biphasic and iii) the induction generator is active power machine approximately five times smaller than synchronous one. v. conclusions this work confirmed through the results that induction generator in connection with a synchronous generator and a nonlinear load has the ability, to increase the power generated, available for the electrical system, besides increasing the power generation available for the electrical system, attenuating harmonic distortion current and voltage in commom bus. the induction machine besides showing low cost, robustness, simple construction, lower cost and less maintenance compared with synchronous machine, repowering the system. the results showed that harmonic distortion bus suffers reductions for synchronous generator connection as for induction generator connection. it is noted in results that induction generator provided a preferred path for current harmonic order, even when two machines produce or consume equivalent and proportional reactive power. acknowledgment the authors would like to thank coordination for the improvement of higher education personnel (capes), the national counsel of technological and scientific development (cnpq) and the research support foundation of goias state (fapeg) for financial support research and scholarships. references [1] j. m. chapallaz; j. d. ghali; p. eichenberger and g. fischer. "manual on motors used as generators". mhpg series, vol. 10, friedr. vieweg & sohn verlagsgesellschaft mbh, germany. [2] d. m. medeiros. "the use of pumps operating as turbines and induction generators to generate electricity". thesis in portuguese, federal university of itajubá, itajubá, minas gerais, brazil, 2004. [3] k. d pham. "cogeneration application: interconnection of induction generators with public eletric utility". rural electric power conference repc, 1991. [4] v, pongpornsup. "impacts of non-utility induction generator to distribution network". ieee transmission and distribution conference and exhibition, vol. 2, pages 1352-1356, 2002. [5] r. behome; m. plamondon, h. nakra, d. desrosiers, c. gagnon. "case study on the integration of a non-utility induction generator to the hydro-quebec distribution network". ieee transactions on power delivery, vol. 10, no. 3, 1995. [6] j.a. laghari, h. mokhlis, a. h. a. bakar, m. hasmaini, "a comprehensive overview of new designs in the hydraulic, electrical equipments and controllers of mini hydro power plants making it cost effective technology". renewable and sustainable energy reviews, vol. 20, pages 279-293, 2013. [7] n. sivakumar, das. devadutta, padhy, n.p. padhy. "variable speed operation of reversible pump-turbines at kadamparai pumped storage plant a case study". elsevier energy conversion and management, pages 96 104, 2014. [8] p.j. reddy, s.p. singh, "voltage and frequency control of parallel operated synchronous and induction generators in micro hydro scheme". computation of power, energy, information and communication (iccpeic), vol. 1, pages 124 129, 2014. [9] ieee std 519-1992. "ieee recommended practices and requirements for harmonic control in electrical power systems". new york, 1993. [10] r. l. nailen. "spooks on the power line? induction generators and the public utility". ieee transaction on industry applications, vol. 1a-18, no. 6, 1982. [11] a. s. magalhães; l. c. a. junior; c. a. matias; a. h. f. silva; e. g. domingues; a. j. alves and w. p. calixto. "repowering of a synchronous generation plant by induction generator." ieee congreso chileno de ingeniería eléctrica, electrónica, tecnologías de la información y comunicaciones (ieee chilecon 2015), 2015, santiago. [12] e. delbone. " harmonic attenuation in the synchronous generator due the nonlinear loads using induction generators". thesis in portuguese, federal university of uberlandia, uberlandia, minas gerais, brazil, 2012. [13] r. behome; m. plamondon, h. nakra, d. desrosiers, c. gagnon. " steam power plant re-powering to provide black-start ancillary service and speed up power system restoration". ieee bologna powertech conference, 2003. [14] a. gaglianoa, g. m. tinab, f. noceraa , f. patania. " technical and economic perspective for repowering of micro hydro power plants: a case study of an early xx century power plant". energy procedia, 6th international conference on sustainability in energy and buildings, seb-14, 2014. i. introduction ii. mathematical modeling a. three-phase induction generator under non-sinusoidal steady state b. three-phase synchronous generator under non-sinusoidal steady state c. association between induction and synchronous generator iii. methodology a. connection machine and loads for case study 1 b. connection machine and loads for case study 2 iv. results a. case study 1 1) experimental tests 2) repowering 3) harmonics b. case study 2 1) experimental tests 2) repowering 3) harmonics v. conclusions acknowledgment references  transactions on environment and electrical engineering issn 2450-5730 vol 2, no 1 (2017) © florent becker, ehsan jamshidpour, philippe poure, shahrokh saadate  abstract— in this paper, an open-switch fault diagnosis method for five-level h-bridge neutral point piloted (hb-npp) or t-type converters is proposed. while fault tolerant operation is based on three steps (fault detection, fault localization and system reconfiguration), a fast fault diagnosis, including both fault detection and localization, is mandatory to make a suitable response to an open-circuit fault in one of the switches of the converter. furthermore, fault diagnosis is necessary in embedded and safety critical applications, to prevent further damage and perform continuity of service. in this paper, we present an open-switch fault diagnosis method, based on the switches control orders and the observation of the converter output voltage level. in five-level converters such as hb-npp and t-type topologies, some switches are mostly 'on' at the same time. therefore, the fault localization is quite complicated. the fault diagnosis method we proposed is capable to detect and localize an open-switch fault in all cases. computer simulations are carried out by using matlab simulink and simpowersystem toolbox to validate the proposed approach. index terms— open-switch fault; fault diagnosis; fault detection; fault localization; multilevel converter; neutral point piloted converter; t-type converter; h-bridge. i. introduction n the recent decades, the high-power and medium-voltage (mv) industrial applications have increased significantly. mv grids connection of these applications with respect of device rating limits requires some series-parallel combinations of power semiconductor switches. the multilevel converter (mc) topologies could provide significant advantages for these applications, such as output waveforms improvement and low switching frequency. therefore, in a mc, the output filters and the use of passive components could be reduced. on the other hand, a low switching frequency allows performing high efficiency. among the different mc topologies, the neutral point piloted (npp) converter is one of the widely used in high-power industrial applications, more f. becker and p. poure, institut jean lamour (umr7198) university of lorraine, vandoeuvre les nancy, france (email: florent.becker@univlorraine.fr ; philippe.poure@univ-lorraine.fr ) f. becker and s .saadate, green laboratory university of lorraine, vandoeuvre les nancy, france (email: shahrokh.saadate@univ-lorraine.fr ) e. jamsidpour, icube (umr 7357) université of strasbourg , illkirch, france and ecam strasbourg-europe, schiltigheim, france (email: ehsan.jamshidpour@ecam-strasbourg.eu) particularly in high voltage-direct current transmission as well as in power quality improvement of pv generation and wind energy systems [1]. a combination of the npp and h-bridge topologies (hbnpp) that is depicted in fig. 1 allows generating five voltage levels (vdc, vdc/2, 0, -vdc/2, -vdc). the five voltage levels are summarized in table i with their associated switching states, considering the current i(t) direction. the table i details the passing components associated to each voltage level for i(t)>0. by the same, table ii is dedicated to i(t) <0. the derived hbt-type topology (fig. 2) allows generating the same voltage levels. fig. 1 hb-npp five level topology. fig. 2 hb-t-type five level topology. table i voltage levels and possible corresponding states for the hb-npp and t-type topologies for i(t)>0. voltage level state passing components hb-npp t-type vdc 1 t11, t12, t23, t24 t1,t4 vdc /2 2 tc1+, dc1-, t23, t24 tc1+, dc1-, t4 3 t11, t12, tc2-, dc2+t1, tc2-, dc2+ 0 4 tc1+, dc1-, tc2-, dc2+ tc1+, dc1-, tc2-, dc2+ 5 t11, t12, d22, d21 t1, d3 6 d13, d14, t23, t24 d2, t4 vdc /2 7 d13, d14, tc2-, dc2+ d2, tc2-, dc2+ tc2+ dc2+ dc2tc2t11 d11 t12 d12 t13 d13 t14 d14 t21 d21 t22 d22 t23 d23 t24 d24 tc1+ dc1+ dc1tc1vo vdc 2 vdc 2 n i(t) tc2+ dc2+ dc2tc2t1 d1 t2 d2 tc1+ dc1+ dc1tc1vo vdc 2 vdc 2 n t3 d4 t4 d4 i(t) study by modeling and simulation of openswitch fault diagnosis for five-level converters florent becker, ehsan jamshidpour, philippe poure, and shahrokh saadate i mailto:florent.becker@univ-lorraine.fr mailto:florent.becker@univ-lorraine.fr mailto:philippe.poure@univ-lorraine.fr mailto:shahrokh.saadate@univ-lorraine.fr 8 tc1+, dc1-, d21, d22 tc1+, dc1-, d3 vdc 9 d13, d14, d21, d22 d2, d3 when fault occurrence is considered, one of the most critical elements in power electronic converters are the semiconductor switches. switches or gate drivers faults, resulting in open-circuit fault (ocf) or short-circuit fault (scf), affect the power generation and may lead to its shutdown. more than 30% of malfunctions and breakdowns are reported to be due to power semiconductor failures [2]. table ii voltage levels and possible corresponding states for the hb-npp and t-type topologies for i(t)<0 voltage level state passing components hb-npp t-type vdc 1 d11, d12, d23, d24 d1,d4 vdc /2 2 tc1-, dc1+, d23, d24 tc1+, dc1-, t4 3 d11, d12, tc2+, dc2-d1, tc2+, dc2 0 4 tc1-, dc1+, tc2+, dc2 tc1-, dc1+, tc2+, dc2 5 d11, d12, t21, t22 d1, t3 6 t13, t14, d23, d24 t2, d4 vdc /2 7 t13, t14, tc2+, dc2t2, tc2+, dc2 8 tc1-, dc1+, t21, t22 tc1-, dc1+, t3 vdc 9 t13, t14, t21, t22 t2, t3 typically, industrial gate drivers include scf protection (resulting in ocf) but ocf detection must be diagnosed to accomplish fault tolerance and continuity of service. therefore, a fast and robust ocf diagnosis is required. a few works have studied open-switch fault detection in h-bridge converter and multilevel matrix converters [3-7]. after fault detection, the localization of the faulty switch is mandatory to manage the post-fault operation as proposed in [8-11]. in this paper, a new ocf detection method with faulty switch localization capability is proposed. it can be applied to both hb-npp and t-type topologies as well. the principle of the proposed fault detection method is summarized and detailed in the next section. more, some simulations have been performed to validate the proposed method in matlab-simulink environment, by using simpowersystems toolbox. the simulation results shown in section iii confirm the robustness, the rapidity of the ocf detection and the localization capability. fig. 3 fault diagnosis principle for a t-type converter fig. 4. open switch fault diagnosis in state 2 (i(t)>0). control pwm vo i vref tabl e i i(t)>0 voe vo state fault dete ctionfault detection localization localization algorithm δi δn δi δn sign i(t) vo i vo hb-t-type ti tn tabl e ii i(t)<0 n t1 t2 t3 t4 dc1+ tc1+ tc1dc1dc2+ tc2+ tc2dc2vo i(t) vo=vdc/2 n t1 t2 t3 t4 dc1+ tc1+ tc1dc1dc2+ tc2+ tc2dc2vo i(t) vo=0 => tc1+ faulty n t1 t2 t3 t4 dc1+ tc1+ tc1dc1dc2+ tc2+ tc2dc2vo i(t) vo=-vdc/2 => t4 faulty if tc1+ faulty if t4 faulty healthy conditions (state 2, i(t)>0) 2 dcv 2 dcv 2 dcv 2 dcv d1 d2 d3 d4 d3 d4 d1 d2 d3 d4 d1 d2 d3 d4 fig 5. open switch fault diagnosis in state 3. ii. principle of the fault diagnosis method fig. 3 shows the principle of the proposed fault diagnosis [13]. the converter is controlled by a classical carrier-based pwm [12]. thus, the switching pattern (δi…δn) is generated by the control system. by using the switching pattern and the tables i and ii, the actual state (sx), passing components and estimated output voltage (voe) can be determined. in normal condition, the measured output voltage vo is equal to the estimated voltage (voe). otherwise, in the case of an ocf, these voltage values (vo and voe) are not equal. the major problem in the studied five-level converters is the fault localization. the localization block (fig. 3) identifies the faulty vertical or horizontal switch. based on tables i and ii, several switches could be faulty in each state. to precise the fault location, the measured output voltage level is considered. to clarify the localization principle, let us suppose an ocf vo=vdc/2 vo=0 vo= 0 if t1 is faulty if tc2is faulty healthy conditions (state 3, i(t)>0 n t1 t2 t3 t4 dc1+ tc1+ tc1dc1dc2+ tc2+ tc2dc2vo i(t) n t1 t2 t3 t4 dc1+ tc1+ tc1dc1dc2+ tc2+ tc2dc2vo i(t) n t1 t2 t3 t4 dc1+ tc1+ tc1dc1dc2+ tc2+ tc2dc2vo i(t) tc1+ is switched off by the localization algorithm vo=0 => tc2faulty 2 dc v 2 dc v n t1 t2 t3 t4 dc1+ tc1+ tc1dc1dc2+ tc2+ tc2dc2i(t) vo=-vdc/2 => t1 faulty 2 dcv 2 dcv 2 dcv 2 dcv 2 dcv 2 dcv n t1 t2 t3 t4 dc1+ tc1+ tc1dc1dc2+ tc2+ tc2dc2vo i(t) 2 dcv 2 dcv vo d1 d2 d3 d4 d1 d2 d3 d4 d1 d2 d3 d4 d1 d2 d3 d4 d1 d2 d3 d4 it is not possible to discriminate the faulty switch beetween t1 and tc2 detection when the converter operates in state 2 (i(t)>0) (fig.4). in this case, according to table 1 for the hb-t-type topology, the faulty switch can be t4 (t23 or t24 for the hbnpp topology) or tc1+. in the hb-npp topology case, the discrimination between t23 or t24 is not necessary to perform post-fault operation. by the same, if the declared fault was due to a fault occurrence on dc1-, considering tc1+ as faulty will lead to the same and suitable post fault operation. here, the output voltage level will be equal to 0 if tc1+ is faulty (fig.4). otherwise, it will be equal to (-vdc/2) if t4 (t23 or t24 for the hb-npp topology) is faulty (fig.4). to clarify the previous explanations, the equivalent circuits and the associated current paths are presented in fig.4, in heathly conditions and when tc1+ or t4 is faulty. table iii diagnosis of the faulty switch for i(t)>0 state healthy conditions faulty conditions voltage level faulty switch 1 vo=vdc vo=0 t4 vo=vdc/2 t1 2 vo=vdc/2 vo=0 tc1+ vo=-vdc/2 t4 3 vo=vdc/2 vo=0 tc2tc1+ is switched off vo=0 vo=0 t1 tc1+ is switched off vo=-vdc/2 4 vo=0 vo=0 tc2t1 is switched on vo=0 vo=0 tc1+ t1 is switched on vo=vdc/2 5 vo=0 vo=-vdc/2 t1 6 vo=0 vo=-vdc t4 7 vo=-vdc/2 vo=-vdc tc2 8 vo=-vdc/2 vo=-vdc tc1+ nevertheless in some states, it is not possible to determine the faulty switch by only comparing the estimated and measured output voltages. for example, let us consider the state 3 (i(t)>0). fig. 5 presents the current paths when t1 or tc2is faulty: in both cases, the output voltage vo is equal to 0 and it is not possible to discriminate the faulty switch. in fact, even if the only passing switches in the state 3 are t1 and tc2-, the switching order applied to tc1+ by the pwm control is ‘1’. in heathly conditions, tc1+ is open, even if an order equal to ‘1’ is applied to its driver (tc1+ reverse biased, fig.5). however, when an ocf of in t1 occurs, tc1+ is no more reverse biased and conducts the output load current. notice that in the case of an ocf in tc2, the switch t1 remains reverse biased (fig.5). consequently, after switch fault detection (voe ≠ vo i.e vo≠vdc/2) the localization algorithm (fig. 3) switches off tc1+ which results in a new output voltage value (fig. 5). if the resulting value of vo is equal to –vdc/2, the faulty switch is t1; otherwise, vo remains equal to 0 and the faulty switch is tc2-. if necessary, the same approach can be applied in the suited states to perform ocf diagnosis (for example, sates 3 and 4 when i(t) > 0). table iii summarizes the fault diagnosis for i(t)>0. iii. simulation results to validate the performances of the proposed fault diagnosis method, some simulations are performed for a hb t-type converter feeding a (r, l) load with r=30ω and l=10mh (fig. 6). the vdc is 1200 v. the pwm switching frequency is 2 khz. the simulations are realized in matlab/simulink environment by using sympowersystem library. the following simulation results are presented in two state cases, when an ocf occurs in the state 2 (subsection a) and when an ocf occurs in the state 3 (subsection b), as discussed in section ii. two ocf cases in state 2 are presented in the next subsection: the first case is an ocf of tc1+ and the second one is an ocf of t4. in subsection b, we discuss the ocf diagnosis in state 4: first when t1 is faulty and secondly when tc2is faulty. the same validation by simulation has been performed for all other states. fault detection and localizationvref vo i vdc o 2 vdc 2 vdc hb-t-type vo l r switching patterns fig. 6 hb-t-type connected to a (r, l) load. a. fault in state 2 fig 7. shows the simulation results when an ocf is occurred in tc1+. to generate the ocf, the switching pattern of tc1+ is forced to '0' by using the signal “fault generation” (fig 7). this signal modifies the control order applied to tc1+, by forcing it to '0' when “fault generation” is switched to '1'. by this way, at the time t = 6.1ms, an ocf is generated on tc1+. a zoom around the fault occurrence is provided in fig. 8 in order to give more details on fault diagnosis. when the fault is occurred at t=6.1 ms, tc1+ is off. therefore, until t=6.46 ms, the values of voe and vo are equal because tc1+ is not switched on to generate the output voltage level. thus, the ocf cannot be detected. at t=6.46 ms, the command of tc1+ switches to '1', but the switch remains off because of the generated ocf. as a result, the values of vo and voe become different; then the fault can be declared (signal “fault detection in fig 8). fig. 7 simulation results when tc1+ is faulty (ocf). fig. 8 zoomed simulation results when tc1+ is faulty (ocf). after fault detection, fault localization must be performed. the proposed method not only declares the fault but also discriminates the faulty switch by considering the actual state during fault apparition, by using the switching pattern and table i or table ii. in this case, the converter is in state s2 with i(t) >0. thus the faulty switch can be t4 or tc1+ and the output voltage should be vdc/2 (heathly conditions). as it can be seen in fig. 8, the output voltage is equal to '0'. as mentioned in section ii, the localization algorithm that observes the output voltage value can discriminate the faulty switch, here tc1+. fig. 9 simulation results when t4 is faulty (ocf). fig. 10 zoomed simulation results when t4 is faulty (ocf). fig. 9 shows the simulation results for an ocf in t4. in fig. 10 which is a zoomed view of fig. 9, after the fault occurrence, the ocf in t4 is generated by the signal “fault generation”. nevertheless, this ocf cannot be detected while the command of t4 is '0' (t4 is off). when the command of t4 goes to '1', it remains off and then the output voltage of the h-t-type converter becomes equal to –vdc/2 instead of vdc/2 (value of voe). by using table i, the fault is declared when the converter is in state s2. in healthy conditions, in this state, vo should be equal to vdc/2. therefore, an ocf can be declared. the fault localization is done by the localization algorithm and t4 is declared as the faulty switch. b. fault in state 3 in this subsection, an ocf in state 3 is considered. first, an ocf is artificially applied to t1 by using the suited signal “fault generation” (see fig. 11): consequently, at t = 8.7 ms, an ocf is generated in t1. figure 12 shows a zoom view of fig. 11, for a short duration around the fault occurrence. as 0 1 2 3 4 5 6 7 0 1000 20 00 0 1 2 3 4 5 6 7 0 1000 20 00 0 1 2 3 4 5 6 7 0 1 2 0 1 2 3 4 5 6 7 0 1 2 0 1 2 3 4 5 6 7 0 1 2 0 1 2 3 4 5 6 7 0 1 2 vo (v) voe (v) fa ult generati on fa ult det ection fa ult loca liza tion tc1+ t c1+ swit chi ng patt er n time (ms) 6 6.1 6.2 6.3 6.4 6.5 6.6 0 1000 2000 6 6.1 6.2 6.3 6.4 6.5 6.6 0 1000 2000 6 6.1 6.2 6.3 6.4 6.5 6.6 -3 0 1 2 6 6.1 6.2 6.3 6.4 6.5 6.6 0 1 2 6 6.1 6.2 6.3 6.4 6.5 6.6 0 1 2 6 6.1 6.2 6.3 6.4 6.5 6.6 0 1 2 vo (v) voe (v) fa ult generati on fa ult det ection fa ult loca liza tion tc1+ tc1+ time (ms) 5 5.5 6 6.5 7 7.5 8 8.5 9 9.5 -500 0 500 10 00 5 5.5 6 6.5 7 7.5 8 8.5 9 -500 0 500 10 00 5 5.5 6 6.5 7 7.5 8 8.5 9 9.5 0 1 2 5 5.5 6 6.5 7 7.5 8 8.5 9 9.5 0 1 2 5 5.5 6 6.5 7 7.5 8 8. 5 9 9.5 0 1 2 5 5.5 6 6.5 7 7.5 8 8.5 9 9.5 0 1 2 9.5 vo (v) voe (v) fa ult generati on fa ult det ection fa ult loca liza tion t4 t4 swit chi ng patt er n time (ms) 8.7 8.75 8.8 8.85 8.9 8.95 9 -500 0 50 0 1000 8.7 8.75 8.8 8.85 8.9 8.95 9 -500 0 50 0 1000 8.7 8.75 8.8 8.85 8.9 8.95 9 0 1 2 8.7 8.75 8.8 8.85 8.9 8.95 9 0 1 2 8.7 8.75 8.8 8.85 8.9 8.95 9 0 1 2 8.7 8.75 8.8 8.85 8.9 8.95 9 0 1 2 vo (v) voe (v) fa ult generati on fa ult det ection fa ult loca liza tion t4 t4 swit chi ng patt er n time (ms) one can see in fig. 12, when the ocf is artificially generated, the switching pattern of t1 is zero (time  in fig. 12), thus the hb-npp converter operates correctly. consequently, no fault occurrence must be detected because t1 is normally open (area  in fig. 12). at the time  in fig. 12, the switching pattern of t1 becomes one and the output voltage v0 remains equal to zero instead of switching from zero to vdc/2. at this time, an ocf is declared because v0 is different from v0e (see  in the signal “fault detection” in fig. 12) but the faulty switch cannot be determined, as explained in fig. 5. to discriminate the faulty switch between t1 and tc2-, tc1+ is switched to zero (see  in the signal “fault detection” fig. 12): thus, the output voltage v0 switches from zero to (-vdc/2) and the fault localization can be performed: t1 is declared as the faulty switch (see  in fig. 12). 0 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 0.01 -2000 0 2000 0 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 0.01 -2000 0 2000 0 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 0.01 0 0.5 1 0 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 0.01 0 0.5 1 0 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 0.01 0 0.5 1 0 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 0.01 0 0.5 1 temps (s) vo(v) voe(v) fault generation fault detection fault localization t1 tc1+ switching pattern fig.11: simulation results when t1 is faulty (ocf in state 3). 8.6 8.7 8.8 8.9 9 9.1 9.2 x 10 -3 -1000 0 1000 8.6 8.7 8.8 8.9 9 9.1 9.2 x 10 -3 -1000 0 1000 8.6 8.7 8.8 8.9 9 9.1 9.2 x 10 -3 0 0.5 1 8.6 8.7 8.8 8.9 9 9.1 9.2 x 10 -3 0 0.5 1 8.6 8.7 8.8 8.9 9 9.1 9.2 x 10 -3 0 0.5 1 8.6 8.7 8.8 8.9 9 9.1 9.2 x 10 -3 0 0.5 1 8.6 8.7 8.8 8.9 9 9.1 9.2 0 0.5 1 temps (ms) vo(v) voe(v) fault generation fault detection fault localization t1 t1 switching pattern tc1+ switching pattern 1 2 4 5 6 fig.12: zoomed simulation results when t1 is faulty. in the same spirit as the simulation previously discussed, an ocf is artificially applied to tc2(see signal “fault generation in fig. 13): consequently, at t = 8.96 ms, an ocf is generated in tc2-. figure 14 shows a zoom view of fig. 13, for a short duration around the fault occurrence. as one can see in fig. 14, when the ocf is artificially generated (time  in fig. 14), the switching pattern of tc2is one, thus the hbnpp converter do not operate correctly after the fault occurrence. the fault occurrence is quickly detected because the output voltage v0 becomes equal to zero instead of vdc/2. (time  in fig. 14). at this time , an ocf is declared because v0 is different from v0e but the faulty switch cannot be determined, as explained in fig. 5. to discriminate the faulty switch between t1 and tc2-, tc1+ is switched to zero (see  in the signal “fault detection” in fig. 14): thus, the output voltage v0 remains equal to zero and the fault localization can be performed: tc2is declared as the faulty switch (see  in the signal “fault detection” in fig. 14). 0 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 0.01 -2000 0 2000 0 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 0.01 -2000 0 2000 0 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 0.01 0 0.5 1 0 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 0.01 0 0.5 1 0 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 0.01 0 0.5 1 0 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 0.01 0 0.5 1 temps (s) vo(v) voe(v) fault generation fault detection fault localization tc2tc1+ swtiching pattern fig.13 : simulation results when tc2is faulty (ocf in state 3). 8.8 8.9 9 9.1 9.2 x 10 -3 -2000 0 2000 8.8 8.9 9 9.1 9.2 x 10 -3 -2000 0 2000 8.8 8.9 9 9.1 9.2 x 10 -3 0 0.5 1 8.8 8.9 9 9.1 9.2 x 10 -3 0 0.5 1 8.8 8.9 9 9.1 9.2 x 10 -3 0 0.5 1 8.8 8.9 9 9.1 9.2 x 10 -3 0 0.5 1 8.8 8.9 9 9.1 9.2 0 0.5 1 temps (ms) vo(v) voe(v) fault generation fault detection fault localization tc2m tc1+ switching pattern tc2switching pattern 1 2 3 4 fig.14 : zoomed simulation results when tc2is faulty. iv. conclusion h-bridge neutral point piloted or t-type converters are increasingly being used in industrial applications. generally, they are used in applications where continuity of service is mandatory. furthermore, switch faults are the most common faults in power electronics converters. after a fault occurrence, in order to avoid its propagation in the whole system, a fast and robust fault diagnosis method must be implemented to perform the reconfiguration of the converter. even if short-circuit faults are handled by the switches' drivers, this is not the case for open-switch faults. an openswitch fault diagnosis method for hb-npp or t-type converters is proposed in this paper. this method is based on the switches states and on the output voltage level observation. the validity of the diagnosis is illustrated by simulations in some ocf cases. references [1] b. li, s. shi, b. wang, g. wang, w. wang, d, xu,"fault diagnosis and tolerant control of single igbt open-circuit failure in modular multilevel converters", ieee transaction on power electronics, vol. 31, no. 4, pp. 3165-3176, april 2016. [2] e. jamshidpour, p. poure and s. saadate, "photovoltaic systems reliability improvement by real-time fpga-based switch failure diagnosis and fault-tolerant dc–dc converter", in ieee transactions on industrial electronics, vol. 62, no. 11, pp. 7247-7255, nov. 2015. [3] m. aleenejad, h. mahmoudi, p. moamaei, r. ahmadi, " a new faulttolerant strategy based on a modified selective harmonic technique for three-phase multilevel converters with a single faulty cell", ieee transaction on power electronics, vol. 31, no. 4, pp.3141-3150, april 2016. [4] j. xu, p, zhao, c. zhao," reliability analysis and redundancy configuration of mmc with hybrid submodule topologies", ieee transaction on power electronics, vol. 31, no. 4, pp. 2720-2729, april 2016. [5] shuai shao, p. w. wheeler, j. c. clare and a. j. watson, "fault detection for modular multilevel converters based on sliding mode observer", in ieee transactions on power electronics, vol. 28, no. 11, pp. 4867-4872, nov. 2013. [6] h. w. sim, j. s. lee and k. b. lee, "a detection method for an openswitch fault in cascaded h-bridge multilevel inverters ", 2014 ieee energy conversion congress and exposition (ecce), pittsburgh, pa, 2014, pp. 2101-2106. [7] m. shahbazi and m. zolghadri, "fast detection of open-switch fault in cascaded h-bridge multilevel converter", power electronics, drives systems & technologies conference (pedstc), 2015 6th, tehran, 2015, pp. 538-543. [8] binbin li, shaolei shi, bo wang, gaolin wang and dianguo xu, "fault diagnosis and tolerant control of single igbt open-circuit failure in modular multilevel converters", in ieee transactions on power electronics, vol. 31, no. 4, pp. 3165-3176, april 2016. [9] f. deng, z. chen, m. r. khan and r. zhu, "fault detection and localization method for modular multilevel converters", in ieee transactions on power electronics, vol. 30, no. 5, pp. 2721-2732, may 2015. [10] qichen yang, jiangchao qin and m. saeedifard, "analysis, detection, and location of open-switch submodule failures in a modular multilevel converter", in ieee transactions on power delivery, vol. 31, no. 1, pp. 155-164, feb. 2016. [11] t. wang, h. xu, j. han, e. elbouchikhi and m. e. h. benbouzid, "cascaded h-bridge multilevel inverter system fault diagnosis using a pca and multiclass relevance vector machine approach", in ieee transactions on power electronics, vol. 30, no. 12, pp. 7006-7018, dec. 2015. [12] h. shin, k. lee, j. choi, s. seo and j. lee, "power loss comparison with different pwm methods for 3l-npc inverter and 3l-t type inverter", 2014 international power electronics and application conference and exposition, shanghai, 2014, pp. 1322-1327. [13] f. becker, p. poure, e. jamshidpour, s. saadate, “open-switch fault diagnosis for five-level h-bridge neutral point piloted or t-type converters”, 2016 ieee 16th international conference on environment and electrical engineering (eeeic), 7-10 june 2016, florence, italy. florent becker was born in france, in 1987. he received the m.s degree from université de nancy, nancy, france, in 2011. he is currently working toward the ph.d. degree in electrical engineering in the faculty of sciences and technologies of université de lorraine, nancy, france. his research interests are high voltage direct current (hvdc), modular multi-level converter (mmc), renewable energy, and fault tolerant converters. ehsan jamshidpour was born in kermanshah, iran, in 1975. he received the b.s. degree from university of tabriz, tabriz, iran, in 1999, the m.s. degree from sharif university of technology, tehran, iran, in 2001 and ph.d. in electrical engineering from university of lorraine, nancy, france, in 2014. from 2003 to 2010; he was an assistant professor in electrical engineering with the institute for energy and hydro technology, kermanshah, iran. he is currently with laboratory icube (umr 7357) université of strasbourg , illkirch, france. his research interests are renewable energy, power electronic converters and fault tolerant converters. philippe poure was born in 1968. he received the engineer degree and ph.d. degree in electrical engineering from inpl-ensem-green, france, in 1991 and 1995 respectively. from 1995 to 2004, he was an associate professor and worked at the university louis pasteur of strasbourg, france, in the field of mixed-signal system-on-chip for control and measurement in electrical engineering. since september 2004, he joined the université de lorraine, nancy france and works on fault tolerant power systems, fpga based real time applications and energy harvesting. shahrokh saadate received the diplôme d’ingenieur (1982), diplôme d’études approfondies (1982), ph.d. degree in electrical engineering (1986) and habilitation à diriger les recherches scientifiques (1995) from ensem-inpl-green, nancy, france. since 1996 he is full professor in university of lorraine in nancy. he has been the head of green laboratory from 2009 to 2012. his main research domain is power systems reliability, power quality and renewable energies. i. introduction when fault occurrence is considered, one of the most critical elements in power electronic converters are the semiconductor switches. switches or gate drivers faults, resulting in open-circuit fault (ocf) or short-circuit fault (scf), affect the power... ii. principle of the fault diagnosis method iii. simulation results a. fault in state 2 b. fault in state 3 iv. conclusion references  transactions on environment and electrical engineering issn 2450-5730 vol 2, no 1 (2017) © joão r. b. paiva, viviane m. gomes, bernardo a. rodrigues, lais f. a. silva, bruno c. m. aniceto, geovanne p. furriel, wesley p. calixto  abstract—this paper proposes a methodology based on system connections to calculate its complexity. two study cases are proposed: the dining chinese philosophers’ problem and the distribution center. both studies are modeled using the theory of discrete event systems and simulations in different contexts were performed in order to measure their complexities. the obtained results present i) the static complexity as a limiting factor for the dynamic complexity, ii) the lowest cost in terms of complexity for each unit of measure of the system performance and iii) the output sensitivity to the input parameters. the associated complexity and performance measures aggregate knowledge about the system. index terms—complexity, connections, discrete events systems, modeling, simulation. i. introduction ystems are studied in different areas by observing their parts and their behavior. a system consists of elements, arranged in a natural or controlled manner to fulfill a certain goal. to rechtin and maier [1], a system is a collection of things or elements, that by working together produce results that are impossible to be obtained by the elements themselves individually. each part interacts directly or indirectly with each other and performs functions on behalf of the whole. the systems’ behavior in time can be investigated to verify patterns, relationships, hierarchy and other features. many systems have a high variability in their parameters. to per bak [2], such characteristic defines them as complex. this type of systems are non-linear, hierarchical, emerging and selforganizing [3], i.e., they have variables that may emerge at any time or assume different levels of importance according to the system dynamics, producing a behavior that is difficult to predict. in complex systems, the whole is considered as more than just the sum of the parts, referring to the fact that the set of properties is not easily inferred from the properties of the parts the authors would like to thank the national council for scientific and technological development (cnpq), the foundation for research in the state of goiás (fapeg) and the higher education personnel improvement coordination (capes) for financial assistance to this research. joão r. b. paiva, viviane m. gomes, bernardo a. rodrigues, bruno c. m. aniceto, geovanne p. furriel and wesley p. calixto are with the experimental and technological research and study group (next) from federal institute of goias (ifg), goiânia, go 74055-110 brazil (e-mail: and the laws of their interactions [4]. in a wide range of scales, the variability may express itself, leading to the identification of the separate parts. this fact can be observed in the relationship between humans, who are able to recognize other humans because they are all different. in this context, the brain can be considered as the most complex system of all, because it can form a representation of the complex external world [2]. faced with the need to quantify the complexity of systems, several authors have proposed different metrics. according to lloyd [5], these metrics are developed to respond questions about the system with respect to: i) difficulty of description, ii) difficulty in creating or iii) degree of organization. the system complexity can be evaluated for its dynamic operation or for its a static configuration. the static complexity measure is independent of the simulation and is considered as a reference. in this work, the maximum number of active connections between elements of the system is adopted to calculate the static complexity. the dynamic complexity measure considers the evolution of the system over time, in which the number of active connections varies at every state change [6]. events contribute to the systems’ dynamics, provoking changes of state [7]. during the process of simulating discrete events systems, the input parameters are changed and the output is observed. the simulation provides data to the analysis of parameter sensitivity, and the goal is to verify each input variable’s contribution to some output. according to saltelli [8], the sensitivity analysis is the study of how uncertainty in the model can be divided to different sources of doubt in the inputs. the sensitivity analysis is relevant to the study of complexity because certain variables can eventually emerge and cause significant impact to the system. to holland [3], this emergence characterizes complex systems and help distinguish such systems from others. the purpose of this paper is to present a methodology for measuring the static and dynamic complexities of systems, jricardobpaiva@yahoo.com.br) and school of electrical, mechanical and computer engineering (emc) from federal university of goiás (ufg), goiânia, go 74605-010 brazil (e-mail: wpcalixto@gmail.com). lais f. a. silva is with experimental and technological research and study group (next) from federal institute of goias (ifg), goiânia, go 74055-110 brazil. metric for calculation of system complexity based on its connections joão r. b. paiva, viviane m. gomes, bernardo a. rodrigues, lais f. a. silva, bruno c. m. aniceto, geovanne p. furriel, wesley p. calixto s using as case studies two systems modeled by discrete events. section ii presents the problems that underlie these systems modeling and section iii brings the complexity metrics. section iv presents the methodology used for the computer simulation and the complexity measure of the proposed systems. in section v, the results are presented. ii. problems a. dining chinese philosophers the dining chinese philosophers problem, proposed by the dutch computer scientist edsger dijkstra in 1965, became a classic representation of synchronization in concurrent programming, where there is competition for limited resources [9]. the problem consists of a situation in which there are five philosophers sitting at a circular table, there is a plate of pasta for each philosopher and a hashi between each pair of plates, as illustrated in fig 1. fig. 1. dining chinese philosophers. each philosopher needs two hashis to eat. thus, when a philosopher wants to eat, he tries to catch the hashis that are adjacent to his plate, one at a time. in case both are free, he starts to eat for a predetermined time. when he finishes eating, he releases the hashis on the table so other philosophers may use them. if only one hashi is available, he holds it and keeps waiting for the release of the missing one. this problem has features of discrete event systems (des), as it presents events and states that can be represented by discrete sets. the states of philosophers in this system are: i) philosophizing; ii) waiting and iii) eating. based on the states, there are the following events: a) begin philosophizing; b) end philosophizing; c) begin waiting; d) end waiting; e) start eating; f) end eating. b. the distribution center the problem of the distribution center consists of a delivery logistics for requests according to the demand, which can form a queue. the freight for each order comes to the dock and is loaded into the truck by a team of four people [10]. the distribution process works accordingly to the following logical sequence: 1) requests arriving at the distribution center in time intervals t1; 2) requests waiting in queue while the resources are not available; 3) each truck staying for a time t2 on the dock while the loading process is performed; 4) truck leaving to make a delivery, while dock and boots are released for new loading; 5) request being transported to the destination during a period of time t3; 6) truck returning to the distribution center in a time interval t4 after the delivery. after step 6, the empty truck is available for new deliveries. in the model, values t1, t2, t3, and t4 are random values given by probability distributions that best represent each time. the distribution center is a typical discrete event system, which features entities, queues, and resources. the entities are the requests, which are waiting in line for the availability of the resources: docks, trucks, and group of loaders. the set of discrete states concerning requests are: i) waiting in the queue; ii) being loaded; iii) being transported. based on the states, it is determined how many resources are being used at every instant of time t. the events in this system are: a) receiving a new order; b) allocating a group of loaders to perform loading; c) starting to use the dock; d) allocating truck; e) starting to load truck; f) finishing to load truck; g) finishing to use dock; h) deallocating group of loaders; i) transporting request to the recipient and j) deallocating truck. iii. complexity metric several metrics to calculate the complexity have been developed based on the size of the system, entropy, information, cost, hierarchy, organization and other criteria [5]. in many cases, the complexity measure is dimensionless, so it only makes sense if compared to another measured value in the system itself or in a separate system, as long as the nature of the analyzed systems allows comparison [11]. before applying the chosen complexity metric, the focus of analysis must be defined and a system model that objectively represents the various interactions of its parts must be created [10]. based on shannon’s studies [12] about entropy in information exchange, some complexity metrics were developed [11] [13] [14] [15] [16]. lemes [13] adapted shannon modeling to measure the complexity of the system connections using (1):      2, , , 1 γ log s i j i j i j s p x p x    (1) where: γ(𝑠) is the complexity of the system connections equivalent to the entropy in information exchange, s is the set of connections between elements of the system, |s| is the total of connections between the elements of the system, 𝑝(𝑥), ∀x ∈ s, is the frequency in which the connections between elements i and j occur. iv. methodology a. proposed metric the proposed method is based on lemes’ modeling [13] in order to measure the complexity of real systems, but it ignores the exchange of information among its members. the proposed metric maps the active connections between entities, resources and queues at any given time t, expressed through the relationship matrix m. thus, one can measure the static and dynamic complexity of any real system using (2) and (3).      2 1 . log i i i s p c p c     (2)     1 . i r q p c e n n   (3) where: 𝛾(𝑠) is the system complexity, e is the number of entities that each resource can attend, nr is the number of resources, nq is the number of queues, p(c)i is the probability that the connection i occurs and 𝜌 is the number of active connections at the moment t, expressed by (4). 1 . k cj ej j n n   (4) where: k is the number of entity states, 𝑛𝑐𝑗 is the number of active connections per entity at the jth state and 𝑛𝑒𝑗 is the number of entities at the j state. b. model for the dining chinese philosophers problem the problem of the dining philosophers is simulated from a computational routine that implements the three proposed states in [17] for the philosophers: i) ap philosophizing; ii) aw waiting and iii) ae eating. to perform the simulation, it is necessary to define 1) the number of philosophers n; 2) the philosopher pi that starts the simulation, with i = 1, 2, 3, ..., n; 3) the orientation of the simulation (clockwise or counterclockwise); 4) the amount m of periods of time t that the philosophers can stay in state ae; 5) the vector tj consisted of periods of time t to the state ae with j = 1, 2, 3, …, m, which can be stochastically generated. the philosopher who is in the state ae uses two hashis. if the philosopher has just one hashi, he is waiting to eat, therefore he is in state aw. in case the philosopher has no hashi with him, meaning that he is philosophizing, he in state ap. considering clockwise orientation, the philosopher pi takes the hashi released from philosopher pi-1, from his right, and the other hashi released from philosopher pi+1, from his left, as explained in (5). 1 1 1 1 0 1 1 i n i p p if i p p if i n             (5) each philosopher pi stays in state ae during the period of time given by tj, in its j-th position, which corresponds to the j-th time that a philosopher assumes the state ae. when the philosopher is finished eating, he mandatorily switches to state ap and must wait, at least, a whole round through the table to come back to state ae. the simulation ends when the vector tj is totally gone through, with j = m, and when all philosophers come out of the state ae. by the end of the simulation, there are the times tae, taw and tap during which the entity pi kept in the states ae, aw and ap, respectively. the total time of simulation t(pi) is the sum of times tae, taw and tap for the entity pi. two approaches are considered in the model of the dining chinese philosophers: 1) the philosophers always want to eat and 2) the philosophers do not always want to eat. in the first one, the resources are required at all times, unlike the second approach, in which philosophers may decide not to eat even if the resources (hashis) are available. the simulation clock varies in a time unit, [ut], one by one, corresponding to one round. at the beginning of each round, the philosophers can change their status according to the rules of the simulation. in order to apply the complexity metric to the dining chinese philosophers, it is considered that the number of active connections 𝜌 in this system in a given moment is provided by (6), where: nap, naw and nae are the number of philosophers in ap, ae and aw, respectively. when the philosopher is eating, two connections are added to the system, (one for hashi used); when the philosopher is waiting, one connection is added (with hashi that was already allocated); and when the philosopher is just philosophizing, they do not add any connection to the system. hence, (4) assumes the following configuration: 2 . 1 . 0 . ae aw ap n n n    (6) the probability of occurrence for a connection in this system is obtained by applying (3), in which the number of entities e that each resource can attend is equal to two, as each hashi attends up to two different philosophers, and nr corresponds to the number of resources (hashis). since this is a closed system, the number of queues nq is not taken into consideration. the calculation of complexity is performed based on the active connection at each instant of simulation. in the dining chinese philosophers problem, it is possible to have the configuration represented by the relationship matrix m at some instant t expressed in (7), in which the columns represent the philosophers from p1 to p5 and the lines represent the hashis from h1 to h5. in this matrix, the philosophers p1 and p4 are philosophizing, p3 is waiting to eat and p2 and p5 are eating. 0 1 0 0 0 0 1 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 0 0 1 m                (7) c. model for the distribution center problem different performance measures can be chosen for the problem of the distribution center, e.g. the average waiting time in queue, the average time for transportation, or the percentage of use of the system resources. it is important to note that unlike the dining chinese philosophers problem, the distribution center is an open system because new entities can be integrated into the system during operation. thus, the number of entities (requests) and therefore, the demand for resources (docks, trucks and loaders) can vary over time. for this case, the delivery time td of requests is considered as a performance measure, which is the time taken from the arrival of the request at the distribution center to the moment it is delivered to the recipient. this time is composed of the sum of the times i) waiting in queue until it gets carried away trq, ii) loading the request into the truck trl and iii) transporting the request from the distribution center to the recipient trt, as in (8): d rq rl rt t t t t   (8) the measure for complexity is calculated by (2). the active connections 𝜌 are mapped based in the states pq, pl and pt, which correspond to the queued requests, in loading process and being transported, respectively. the number of active connections is determined by (4), being that each request in state pq adds one connection to the system (with the request ahead of it), each request in state pl, adds three connections (one with the dock, one with the truck and another with the loaders group) and finally a request in state pt contributes with one connection to the system (with the truck). thus, (4) results in (9). 1 . 3 . 1 . rq rl rt n n n    (9) where: nrq, nrl and nrt are the number of requests in queue, being loaded and being transported, respectively. the probability of occurrence for a connection in the distribution center problem in given by (3), in which the number of entities e that each resource can attend is equal to the total number of requests in the system at the instant t, considering that each entity (request) can be attended by any system resource. therefore, e is the sum of nrq, nrl and nrt. in this model, nr is equal to the sum of the number of docks, trucks and loaders groups. the value of nq corresponds to the number of queues adopted in the system modeling, which is in this case, one queue. considering a configuration of 6 requests with 2 docks, 3 trucks and 2 groups of loaders, it is possible at some instant t of this system, to set up the following relationship matrix m expressed in (10), in which the columns represent the requests, from r1 to r6, and the lines represent the queue (q), docks (d1 and d2), trucks (from t1 to t3) and loaders group (g1 and g2), in this order. 0 0 0 1 1 1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 1 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 m                          (10) in (10), it is verified that the requests r1 and r2 are being transported in trucks t2 and t3, respectively, the request r3 is being carried by the loaders group g1 into the truck t1 parked in the dock d2 and the requests r4, r5 and r6 are waiting in line. it is noteworthy that the number of active connections can be computed from the matrix m v. results a. study case 1: dining philosophers the static and dynamic complexities were calculated from the model presented in section iv-b. for calculating the static complexity, the maximum number of active connections was considered, which is only possible in closed systems. this is the case of the dining philosophers, where new entities cannot be added during its operation. in this model, the maximum number of active connections 𝜌 is equal to the number of resources. in order to obtain the dynamic complexity 𝛾𝑑 (𝑝) during the simulation of the system, (2) was used at each event occurrence, which alters the state of the system. therefore, the configuration of the system is considered at each instant of time t. for this problem, two policies were used to express the behavior of philosophers concerning their desire to eat. in the first one, philosophers want to eat whenever possible. the second approach considers that each philosopher may or may not want to eat at each round. the policy in which the philosophers always wants to eat was presented in [6], where it was observed that at every instant of time t, the dynamic complexity was maximum, equal to the static complexity 𝛾𝑠 (𝑝). thus, (11) is obtained.     : d s jp p t t    (11) in this approach, it is noted that the philosophers (entities) always use all hashis (resources) at every round. however, in most real systems, the resources are not fully utilized at all times, varying the number of active connections in time as the system evolves. for the case in which philosophers can decide whether or not to eat, a new simulation was performed because the number of connections varied. all philosophers started in the state ap, philosophizing, so all the hashis were free. the simulation clock forwarded one unit ut at a time, which corresponds to a round when each philosopher has its state assessed. the simulation was performed during 200 ut for the scenario with five philosophers and five hashis. at each change of state of the system, (2) was used to assess the dynamic complexity, observing the number of active connections. in fig. 2, it is noted the behavior of the system relative to the static complexity 𝛾𝑠 (𝑝), dynamic 𝛾𝑑 (𝑝) and average 𝛾𝑎 (𝑝). fig. 2. static, dynamic and average complexities in the dining chinese philosophers problem. in fig. 2, static and average complexities were approximately 1.66 and 1.42, respectively. the dynamic complexity varied during the simulation process, reaching the maximum value in few instants. as each philosopher could choose to eat or not, the number of active connections in the system changed every round. it was also noted also that dynamic complexity has a ceiling limit defined by the value of the static complexity. another observation can be made regarding the mode values, being approximately: 0.38; 0.68; 1 and 1.35. the repeated values are repeated are an indicative of the number of resources used. b. study case 2: distribution center the model of the distribution center, presented in section iv-c, was simulated based on the following operation dynamics: 1) the arrival of the requests happens in interval of time t1 with mean average of 120 minutes, according to an exponential distribution; 2) the loading process of each truck lasts a certain time interval t2, following a normal distribution with mean value of 100 minutes and standard deviation of 30 minutes; 3) when a truck leaves to delivery, the dock and the loaders group are released to a new loading process; 4) the transportation of each request to the recipient is performed in a period of time t3, evenly distributed from 120 to 240 minutes; 5) after delivery, the empty truck takes an amount of time t4 to return to the distribution center, and t4 follows the same probabilistic distribution of t3. the number of docks used in the simulation ranges from 1 to 10, the amount of trucks, ranges 1 to 15, and the number of loaders groups ranges 1 to 10. therefore, there are 1,500 different scenarios, obtained by combining the amounts of resource values. the simulation was performed for 180 days for each scenario, considering 24 daily hours of operation and that every truck is loaded with only one request at a time. in the 1,500 simulated scenarios, the delivery time td in minutes and the complexity 𝛾(𝑑) of the system were calculated. the normalized values of td and 𝛾(𝑑) for all simulation scenarios are presented in fig. 3. as the simulation is the combination of resources, docks, trucks and loaders groups, in fig. 3 the peaks of td (in blue) correspond to the combinations in which they had only one truck. at every change in the number of docks, a truck was used for 10 scenarios, leading to higher values of td and 𝛾(𝑑) (in red). the oscillations that occur between the values 0.1 and 0.3 of 𝛾(𝑑) are derived from changes in the number of trucks, indicating greater sensitivity in this system parameter. table i has some of the shortest delivery time values td obtained in the simulation. by analyzing fig. 3, it is possible to see that for values of td close to zero, there are several system configurations that result in the same value of td. however, table i presents the five scenarios with the lowest complexity values found for the five smallest times td. table i scenarios with smallest delivery time and correspondent complexity fig. 3. relationship between delivery time and complexity. the smallest complexity that was found in the 1500 scenarios is 𝛾(𝑑) = 0.2697, which corresponds to the third smallest delivery time and to the scenario with the maximum number of available resources, as disposed in table i. the worst delivery time td was 340 times bigger than the smallest time occurred in 100 different scenarios. that is illustrated in the peaks of fig. 3, where each peak contains 10 scenarios. the biggest complexity was 𝛾(𝑑) = 2.7906, which relates to the worse delivery time since the calculation for both the complexity and delivery time considers the permanence of the requests in line. as each request in queue corresponds to one connection, the td 𝛾(𝑑) docks trucks l. groups 276.9357 0.2963 10 15 6 276.9569 0.3038 6 14 10 277.0154 0.2697 10 15 10 277.0368 0.2821 9 14 10 277.0613 0.3041 5 15 10 longer the queue, the bigger will be the time td and the complexity 𝛾(𝑑). it was noted that the value 𝛾(𝑑) is intrinsically related to the configuration of the system. by putting the results from the 1500 scenarios in descendant order of 𝛾(𝑑), among the five highest values, there is td ≈815. this represents a value close to 3 times the smallest td that was found, as displayed in table ii. it is verified that for td = 815.7777, approximately 84% fewer resources were used when compared to the scenario of smallest td. considering rc in (12) and that 𝛾(𝑑) is calculated from the connections, rc contains the relationship between time and the configuration of the system.   d c d t r   (12) table ii relationship between the delivery time and the complexity table ii displays the values found for rc. it is noted that the lowest value refers to the time of 815.7777, whose scenario presented a reduced amount of resources. even so, the efficiency of the system was 115 times greater than the worst case and only 3 times smaller than in the best case. this shows that the use of resources was optimized, as the ratio value rc indicates the lowest cost in terms of complexity for each minute of permanence of the request in the system. high complexity values may reflect significant sizes of the queue if the delivery time is high. however, high complexity can also be indicative of optimal settings for the system. therefore, the number of active connections observed in the calculation of complexity can express both the queue formation and the use of resources for loading and transportation requests. thus, it becomes necessary to use rc to check if the high value of complexity is indicative of queue or of full operation, as described in (13): d d t queue t operation         (13) c. sensitivity analysis of distribution center this case study aims to confirm the hypothesis that the truck parameter has the largest sensitivity. from the simulated results, the following input parameters were analyzed: dock, truck, and loaders group. delivery time and complexity were considered as output parameters. the following values were considered: 5 docks, 7 trucks and 5 loaders groups. these values were chosen because they correspond to the average points of the following resource ranges: 1 to 10 docks, 1 to 15 trucks and 1 to 10 groups of loaders. in fig. 4 and fig. 5, each parameter was varied from its base value of -100% to 100%, according to a univariate analysis. fig. 4 represents the relation between the amount of resources and delivery time, and fig. 5 represents the relation between resources and complexity. fig. 4. changes in multiple parameters (resources) for a single output variable (time delivery). fig. 5. changes in multiple parameters (resources) for a single output variable (complexity). it can be observed that the truck parameter has the largest sensitivity in the distribution center, especially in the interval between -85.7% to 28.6% of its base value. fig. 5 shows that the curve related to the truck resource has the largest distance from the base (dotted) line, which indicates the largest sensitivity in the analyzed scenarios. this analysis reflects the fact that the truck has a larger demand for resources, because it is utilized during the states where the order is being i) loaded and ii) transported, differently from other resources that are requested only when the order is being loaded. td 𝛾(𝑑) rc docks trucks l. groups 276.9357 0.2963 934.6463 10 15 6 291.6034 0.4083 714.1128 2 11 7 297.8970 0.5892 505.5827 3 7 2 322.1149 0.4690 686.8121 2 6 9 815.7777 2.0263 402.5947 1 3 1 vi. conclusion this work presented a methodology for measuring the static and dynamic complexity of systems. the presented metric can be applied in real systems, provided that they are modeled in terms of events and states that express the existing connections. static complexity was calculated only in the dining philosophers problem because it is a closed system. this measurement indicated the maximum complexity of the system, upwardly limiting the dynamic complexity. in the distribution center problem, it was found that the complexity associated with the performance measure provides knowledge about the system. the lowest value of the relationship between delivery time and complexity was in a system configuration that showed high complexity, although the demand was met with fewer resources. the truck resource presented the largest sensitivity, causing the highest impact on the system when its quantity is limited. the presented complexity metric can support decisionmaking policies related to resource management, optimization processes (as a constraint or goal), security policies, or appreciation of systems. references [1] e. rechtin and m. w. maier, the art of systems architecting. crc press, 2010. [2] t. per bak and k. wiesenfeld, “self-organized criticality: and explanation of 1/f noise,” phys. rev. let, vol. 59, pp. 381–384, 1987. [3] j. h. holland, complexity: a very short introduction. oxford university press, 2014. [4] h. a. simon, “the architecture of complexity,” proceedings of the american philosophical society, vol. 106, no. 6, pp. 467–482, 1962. [5] s. lloyd, “measures of complexity: a nonexhaustive list,” ieee control systems magazine, vol. 21, no. 4, pp. 7–8, 2001. [6] l. santos, c. silva, j. paiva, v. gomes, s. oliveira, a. alves, and w. calixto, “a methodology for calculation of complexity in systems: case study,” in ieee congreso chileno de ingeniería eléctrica, electrônica, tecnologías de la informacíon y comunicaciones, pp. 213–218, 2015. [7] c. g. cassandras, s. lafortune. introduction to discrete event systems. secaucus, nj: springer science & business media, 2008. [8] a. saltelli, et al. global sensitivity analysis. england: john wiley & sons, ltd, 2008. [9] e. w. dijkstra, “hierarchical ordering of sequential processes,” acta informatica, vol. 1, no. 2, pp. 115–138, 1971. [10] l. chwif and a. c. medina, modeling and simulation of discrete events (in portuguese). campus-elsevier, 4 ed., 2014. [11] d. w. repperger, r. g. roberts, and c. g. koepke, “quantitative measurements of system complexity,” us patent 8,244,503 b1, aug. 8, 2012. [12] c. e. shannon, “a mathematical theory of communication” the bell system technical journal, vol. 27, pp. 379–423, 623–656, 1948. [13] m. j. r. lemes, “complexity, coupling and criticality (c2a) as risk factors in system design (in portuguese),” phd thesis, university of são paulo, 2012. [14] c. gershenson and n. fernández, “complexity and information: measuring emergence, self-organization, and homeostasis at multiple scales,” complexity, vol. 18, no. 2, pp. 29–44, 2012. [15] c. gershenson, “harnessing the complexity of education with information technology,” complexity, vol. 20, no. 5, pp. 13–16, 2015. [16] c. e. maldonado, g. cruz, and a. nelson, “biological hypercomputation: a new research problem in complexity theory,” complexity, vol. 20, no. 4, pp. 8–18, 2015. [17] a. s. tanenbaum and h. bos, modern operating systems. prentice hall press, 2014. i. introduction ii. problems a. dining chinese philosophers b. the distribution center iii. complexity metric iv. methodology a. proposed metric b. model for the dining chinese philosophers problem c. model for the distribution center problem v. results a. study case 1: dining philosophers b. study case 2: distribution center vi. conclusion this work presented a methodology for measuring the static and dynamic complexity of systems. the presented metric can be applied in real systems, provided that they are modeled in terms of events and states that express the existing connections. stat... references  transactions on environment and electrical engineering issn 2450-5730 vol 2, no 1 (2017) © rodrigo a. lima, a. c. paulo coimbra, tony almeida, viviane margarida gomes, thiago m. pereira, aylton j. alves and wesley pacheco calixto  abstract—the objective of this work is to investigate the influence of slotted air gap constructive parameters on magnetic flux density of rotating machines. for this purpose, different approaches were used to solve the air gap field diagram using finite element method and the magnetic field distribution uniformity was evaluated by carter's factor calculation on twodimensional and three-dimensional models. sensitivity analysis of slot constructive parameters was performed and results show that slot geometry modifies the magnetic flux on air gap and shifts the air gap magnetic equipotential midline of double slotted machines. finally, minimization of carter’s factor on twodimensional model presents an optimized slot geometry with a near uniform magnetic flux density distribution. index terms—carter's factor, finite elements method, rotating machines. i. introduction agnetic circuits of rotating machines are briefly composed by its ferromagnetic parts (rotor and stator) and air gap [1]. thus, analytic equations and numerical method simulation are both used on magnetic circuit design process to determine parameters such as torque and machine excitation current [2], [3]. magnetic flux distribution on air gap has great influence on machine performance because the most part of magnetic energy distribution is contained in air gap domain [1], [4]. once there are magnetic flux density fluctuations in air gap domain due the presence of slots in ferromagnetic parts, air gap reluctance is dependent of the rotor and stator relative position [5], [6]. this work was supported by the coordination for the improvement of higher education personnel-capes—pdse process number: 99999.003605/2014-00, national council of scientific and technologic development of brazil—cnpq and research support foundation of goiás state-fapeg. r. a. lima (corresponding author), t. m. pereira and w. p. calixto are with electrical and computer engineering school, federal university of goias, ufg, av. universitaria, 1488 qd. 86 bl. a zip 74605-010, goiania, goias, brazil and with experimental & technological research and study group, next of federal insitute of goias, ifg, rua 75, 46, centro, zip: 74055-110, goiania, goias, brazil (email: rodrigo.lima@ifg.edu.br, thiago.pereira@ifg.edu.br, wpcalixto@gmail.com ). a. c. p. coimbra and t. almeida are with institute of systems and robotics, isr of coimbra university, uc, rua silvio lima, zip 3030-194, coimbra, portugal (email: acoimbra@deec.uc.pt , tony@deec.uc.pt ). v.m. gomes and a. j. alves are with experimental & technological research and study group, next – ifg (email: vivianemargarida@gmail.com, aylton.alves@ifg.edu.br ). although analytic description for magnetic flux density in air gap domain is not an easy task, f. w. carter presented an analytic equation to quantify the magnetic flux reduction in slotted air gap by introducing the concept of equivalent air gap [7][9]. carter considered that magnetic flux reduction is equivalent to replace the slotted air gap with length g for an equivalent smooth air gap with length eq c g k g  ( 1 c k  ). the term c k is called carter's factor and its value is influenced by air gap magnetic flux distribution, where 1 c k  is equivalent to a uniform magnetic flux distribution. the purpose of this paper is to present a study of slot geometry influence on magnetic flux density distribution in air gap dominion using carter's factor to evaluate the flux distribution uniformity. the finite element method (fem) was used to solve magnetic flux distribution in air gap domain in two and three dimensional models and different approaches to evaluate carter's factor were compared. additionally, it was performed a sensitivity analysis to study the influence of air gap constructive parameters influence on carter’s factor rodrigo a. lima, a. c. paulo coimbra, tony almeida, viviane margarida gomes, thiago m. pereira, aylton j. alves and wesley pacheco calixto calculation of the influence of slot geometry on the magnetic flux density of the air gap of electrical machines: three-dimensional study m fig. 1. simplified air gap of a machine with slots on stator's surface, where, t is stator tooth pitch, / t s w w tooth/slot width, s h stator slot height and / s r r r stator/rotor yoke height. blue lines represent uniform magnetic flux, n  . mailto:rodrigo.lima@ifg.edu.br mailto:thiago.pereira@ifg.edu.br mailto:wpcalixto@gmail.com mailto:acoimbra@deec.uc.pt mailto:tony@deec.uc.pt mailto:vivianemargarida@gmail.com mailto:aylton.alves@ifg.edu.br considering two different slot patterns. finally, double-slotted air gap geometry effect was considered on a case study where different slot patterns were combined. ii. carter’s factor carter's original study considers a simplified air gap geometry to quantify the magnetic flux reduction by the slot presence. fig.1 presents the constructive relations considered in air gap. to describe the magnetic flux density in air gap domain, carter considered that regions close to teeth have uniform magnetic flux (blue arrows in fig. 1) while regions near of slot opening have null magnetic flux. null flux regions will reduce the mean magnetic flux in air gap domain and the carter's factor is defined by [1], [4], [7]. . c s t k t w   (1) the value of  on (1) depends on air gap constructive parameters and is related to the magnetic flux distribution near the teeth. carter found (2) using schwartz-christoffel's conformal mapping on a simplified slot geometry [7]. other methodologies may be found in literature to determine  . however, this work will consider the empirical expression (3) proposed by langsdorff for comparison of analytic results [10]. 2 2 arctan ln 1 2 2 s s s w wg g w g                          (2) / 5 / s s w g w g    (3) on double slotted air gaps, the carter’s factor is defined by the product of the rotor carter’s factor, cr k , and the stator carter’s factor, cs k . in this case, cr k and cs k are independently calculated using (1) and  is determined by equation (2) or (3) [1], [5], [6]. c cr cs k k k  (4) although many authors consider sufficient in practice, the results obtained by expressions (1), (2), (3) and (4) are not accurate [11][15]. more accurate methods involve air gap magnetic field diagram solution by numerical methods. in this work the finite element method is applied on different approaches for magnetic field diagram solution in air gap domain. the first approach, using the field diagram solution, considers that neither electric charges nor currents exist within the air gap domain. in this case, the problem is totally characterized by setting magnetic scalar potential m v . magnetic induction lines are mapped by solving two dimensional laplace's equation with respective neumann's and dirichlet's boundaries conditions on analogy to electrostatic behavior [1], [6], [13]. fig. 2 illustrates neumann’s boundary conditions (blue and green lines), and red regions represent the known potential of dirichlet condition applied to the rotor and the stator surface. fig. 3 illustrates the magnetic equipotential lines mapping (blue lines) and the magnetic flux density b (red lines). the intermediate equipotential line obtained in diagram has greater length than intermediate equipotential line of a smooth air gap. fig. 2. discretized domain representation of air gap using fem. red boundaries and green boundaries represents dirichlet's known potential and blue boundaries represents neumann boundary condition. fig. 3. magnetic flux diagram determined by fem in air gap domain with length l . green line represents intermediate magnetic equipotential line with length i l numerically determined. (a) (b) fig. 4. (a) domain representation and current density considered. (b) flux diagram determined by fem. the intermediate equipotential line length is given by i l and is determined numerically [13]. the ratio of the actual intermediate equipotential line length i l by theoretical smooth air gap length l calculates carter’s factor [16], [17]. i c l k l   (5) another carter’s factor definition using fem is the ratio between magnetic flux density peak value in air gap, max b , and the magnetic flux density's average along the stator pitch [1], [14]. max c av b k b  (6) where av b is the average value of magnetic flux density along the tooth pitch axis x , given by [1], [14], 2 0 1 ( ) 2 t av b b x dx t   (7) the second approach for the air gap field diagram calculation considers the electric current in the slots windings and maps the induction lines considering both air gap and ferromagnetic parts of the machine. fig. 4a indicates the direction of current density ( f ) of each winding. the result is shown in fig. 4b and since this approach do not considers equipotential lines, carter's factor is determined only by (6) and (7). although results obtained by two-dimensional magnetic field mapping provides more accurate results than analytical method they don’t describes the dynamical behavior of rotating machines. on double slotted air gaps the magnetic flux depends on relative position of rotor and stator slots. the squirrel cage rotor has an angular deviation of one slot step to reduce the magnetomotive force loss in machine. however, the angular deviation of rotor breaks the axial symmetry and two-dimensional approximation of laplace’s equation are not possible because magnetic scalar potential and magnetic flux density has axial dependency. this paper proposes a three-dimensional analysis of magnetic field diagram in airgap domain and the generalization of equations (5), (6) and (7) for carter’s factor calculation. for magnetic scalar potential approach, the equation (6) can be extended to three-dimensional domain taking into consideration magnetic equipotential surfaces in place of magnetic scalar equipotential lines of twodimensional approach. k mef c g a a  (8) where, mef a is the intermediate magnetic equipotential surface numerically determined in airgap domain and g a is the theoretical intermediate magnetic potential of the smooth air gap. the three-dimensional evaluation of carter’s factor by (8) is equivalent to the mean of the results of (5) taken into infinitesimal angular displacements in the one slot step interval. in addition, three-dimensional generalization of excitation current approach leads to the modification of the calculus of the mean magnetic field density in axial coordinates 1 ( , , ) av mean g b b r z d dz a    (9) where, mean r the axial coordinate equivalent to the radius coordinate of minimum axial distance between rotor and stator teeth. the maximum value of magnetic flux density max b is determined on axial coordinate mean r and the carter’s factor is then calculated by (6). iii. methodology a generic slot was created to study the geometry effect in air gap magnetic flux. changing its main constructive parameters, it is possible to observe changes in slot pattern. in fig. 5 are presented some possible changes in the slot geometry. the variation of parameters max h , max w , r , slot w , med w and c s creates new air gaps patterns. thus, the field diagram is solved for magnetic potential approach and magnetizing current approach. carter’s factor (a) (b) (c) (d) fig. 5. (a) generic slot representation and main parameters to be varied 120sc   . (b) resulting geometry for slot med max w w w  and r . (c) max r w and 270sc   (pattern i). (d) 60sc   and max r w (pattern ii). table i slot constructive initial values max h (mm) max w (mm) slot w (mm) t (mm) 12 6 2 10 then quantifies flux’s uniformity [1], [13], [14]. a small induction machine stator inspired the initial values of the slot. the slots constructive parameters are shown in table i. values of max h and max w are fixed as indicated in table i to ensure that the winding will be accommodated. the variation of the parameters r and c s is divided in two slots patterns. first pattern has r and c s relations given by (10a) and (10b) and is illustrated in fig. 5c. max r w (10a) 180 , 2 2 slot c w s arcsin r            (10b) second pattern is illustrated in fig. 5d and the parameters r and c s are given by (11a) and (11b), respectively.  max max , r w h (11a) 2 2 max w sc arcsin r         (11b) both patterns relate slot opening slot w and intermediate opening med w by (12a) and eq.(12b). [ , ] slot min max w w w (12a) 7 3 10 3( ) 3( ) max min max min med slot max min max min w w w w w w w w w w        (12b) where (12b) came from slot geometrical analysis for med slot w w when slot max w w . in (12a) was adopted 1 min w  mm. the viable space is represented by equations (10b), (10a) and (12a). five case studies analyses the effect of air gap variation using both magnetic field mapping approaches. the parameters variation and domain discretization are defined by (11) and (12) and resulted in the sensitivity study of the construction parameters [18]. the sensitivity analysis study is based on the factorial experiment design. for textual comprehension, ‘factor’ in the factorial experiment design is related to the decisions variables considered in sensitivity study. on this paper, carter’s factor is considered the system response or experimental result of the factorial experimental design. the factorial experiment design takes on all possible level combinations of decisions variables and schematically organizes their combinations in a planning diagram. the planning diagram relates the decision variables combined levels with their respective system response. considering two generic variables p and q , with two discrete levels, the experiment is called 2 2 factorial design and has four system response j y . in this case, the planning diagram can be picture by fig. 6. the principal effect of decisions variables is given by, 1 32 4 2 2 p yy y y     (13a) 3 4 1 2 2 2 q y y y y    (13b) the respective sensitivity analysis of p and q considers the mean of different 2 2 factorial design principal effects calculated by (13) and is expressed by its relative value. the respective mean relative percentage value is denoted by p sens and q sens , respectively for p and q . a. case i: the first case study is a sensitivity analysis on r , c s and slot w . in this case, air gap magnetic equipotential lines are mapped applying the scalar magnetic potential formalism m v and carter's factor is determined by (5) [1], [13], [14]. the sensitivity analysis is then performed by taking a discretized domain from equations (10) and (11). for slot pattern i, the principal effect of c s and slot w are evaluated using (13) and the sensitivity of each parameter is calculated as the average of several 2 2 factorial design in distinct two levels domains. analogously, the sensitivity analysis of r and slot w is performed to slot pattern ii. additionally, this case study, magnetic flux density lines are mapped on the air gap domain. the maximum magnetic flux density max b and mean magnetic flux density av b by fem and the carter’s factor is then evaluated using (6). the sensitivity analysis study is then performed. b. case ii the sensitivity analysis on r , c s and slot w for air gap magnetic flux setting the winding current. carter's factor is determined by (6) [1], [13], [14]. althought the winding current approach is also a fem method to determine magnetic flux diagram is not possible determine scalar magnetic equipotential curves and carter’s factor determination from (5) . the sensitivity analysis is then performed only to the excitation current approach by the calculation of the respective principal effects of decision variables in the discretized domain in analogous form of case i. c. case iii: this case study shows that air-gap equipotential midline, fig. 6. system response diagram for a 2 2 factorial experiment design. defined by the magnetic equipotential line with minor influence by the air gap geometry, is not always located in the middle of the airgap. different geometries of double slotted air gaps were considered and the results show that magnetic equipotential middle line position approach to the rotor surface or stator surface [16], [17]. d. case iv on this case study an air gap geometry is proposed by using constrained optimization of carter’s factor by real-coded genetic algorithm [19]. this case study considers the minimization problem . . min c s o x k  (14) where ( , , , ) c slot med x r s w w is the decision variables vector, and  is the viable space defined by equations (10), (11) and (12). on an optimization process, each decision vector x is an individual belonging to a finite population that evolves by crossover and mutation agents on a natural selection scheme inspired by darwin evolution’s theory. the population evolves by successive iterations called generations until an individual satisfies a determined stop criteria. on this case study a population of 50 individuals evolves until the error defined by 1 c k   has a value minor to 3 10  or the number of iteration achieves 100 generations. the carter’s factor is calculated using magnetic field diagram by fem and (6). on the first geometry proposed, the constructive parameters on table i are fixed and the optimization process takes into consideration only the parameters r and c s . the objective of this study is to compare the carter’s factor of an optimized geometry with the original geometry from the real induction machine. on the second geometry proposal, (14) is solved to x . on both geometries, the carter’s factor obtained by fem is compared to analytical calculations by (1), (2) and (3). e. case v the final case study uses a generic three-dimensional model of a double slotted air-gap created in fem. the air gap constructive parameters were parametrized with initial values given in table i. using magnetic scalar equipotential approach the magnetic field is mapped the intermediate equipotential surface are numerically determined. the carter’s factor for the theoretical double slotted air gap is then evaluated by (8) and compared with the results evaluated by the analytical approach. additionally, the effect of the rotor twist angle, rot  , analyzed varying the value of rot  from 5º to 40º. the fig. 7a represents the generic air gap model and the detailed view of finite elements mesh is depicted on fig. 7b. the air gap region was divided on four layers to increase the number of finite elements on the air gap domain. iv. results a. case i sensibility analysis of carter’s factor due to slot geometrical parameters variation was performed calculating the relative deviation of c k by an associated perturbation on parameters viable space. carter's factor was calculated by (5) using magnetic scalar potential mapping by fem. results for slot pattern i are presented in the first and second columns of table ii, where sc sens and wslot sens are the respective mean sensibility of c s and slot w parameters taken by the 2 n factorial combination of perturbations in viable space. additionally, a relative deviation in c s with respect the initial values of table i and their respective relative deviation on c k is depicted in fig. 8. (a) (b) fig. 7 (a) three-dimensional model of double slotted airgap. (b) detailed view of finite elements mesh on airgap domain. table ii case study i – magnetic potential approach pattern i pattern ii sc sens wslot sens r sens wslot sens 2.013% 97.987% 0.075% 99.925% on the other hand, sensitivity analysis of r and slot w parameters to slot pattern ii are presented in third and fourth columns of table ii. it is possible to observe that in both patterns the major influence for c k provided by slot w parameter, although in slot pattern i c s has significant contribution when compared with r parameter in slot pattern ii. after the magnetic scalar potential analysis, the magnetic field density was evaluated in the air gap tooth axis and carter's factor is determined by (6). sensitivity analysis of carter's factor using the magnetic field density approach was carried out analogously to analysis performed for the magnetic scalar potential approach. table iii presents the respective sensitivity of c s and slot w for slot pattern i in the first and second columns. additionally, the sensitivity of r and slot w parameters for slot pattern ii are presented in third and fourth columns of table iii. b. case ii magnetic field density determination uses fem and excitation current approach in the tooth axis of air gap. because the magnetic field is directly determined on this method, carter's factor is determined only by (6). following the same methodology described in case i, sensitivity analysis was performed for both slot patterns considering relative perturbations in their respective viable spaces and calculating the resulting relative deviation in c k . sensitivity of c s and slot w for slot pattern i are presented in the first and second columns of table iv. the sensitivity values of r and slot w are presented in the third and fourth columns of table iv. fig. 8. sensitivity analysis of c s parameter (blue axis) and sensitivity analysis of slot w parameter (black axis) for slot pattern i by magnetic scalar potential calculation of c k by (5). table iii case study i – magnetic field density approach pattern i pattern ii sc sens wslot sens r sens wslot sens 0.3% 99.7% 0.008% 99.992% (a) (b) (c) fig. 9. double slotted air-gaps patterns and their respective equipotential midline (red line). (a) slots a and b, (b) slots a and c, (c) slots a and d. c. case iii this section presents the analysis of double air gap geometry influence in air gap magnetic equipotential midline position. equipotential midline is defined by the equipotential line that have minor influence by slot’s presence in air gap ferromagnetic surfaces. due to space limitations, the four slot patterns illustrated in fig.9 were combined to form only three different types of air gap geometries. table v summarizes the slots constructive parameters. the red line depicted in fig. 9 represents the equipotential midline mapped by fem in air gap domain. table vi presents the respective axial position of midline and the relative deviation from theoretical position. axial position line y is measured from the rotor surface to the equipotential midline’s unperturbed region and is presented in second column of table vi. third column shows the deviation line  from the theoretical position by relative to the total size of the air gap. d. case iv optimized slot geometries are presented in fig. 10. the optimization of slot constructive parameters results in an error of 6 1 10    and is equivalent to a uniform magnetic field flux on air gap. table vii summarizes slot constructive parameters and compares their respective carter’s factor by fem and analytical methods. the results of carter’s factor calculation in different methodologies show that the proposed slot geometry have improved the simulated magnetic field uniformity. however, only the fem methodology can determine accurately the contribution of r and c s parameters. e. case v on this case study the magnetic field mapping of the geometry depicted in fig.7a with parameters indicated by table i result in an intermediate equipotential surface. the total area of the equipotential was numerically determined by fem. table viii presents the comparison of carter factor evaluated to the three-dimensional air gap methodology (3dfem) and the analytical equations presented by carter and langsdorff. the results evaluated by the 3d-fem methodology show little divergence to the analytical methodology. the reason for the divergence on the values of carter’s factor can be explained by the magnetic field mapping in the air gap domain. the twist angle of one slot step allow to evaluate different relative positions between the stator and rotor slots. the resulting intermediate magnetic scalar equipotential surface also represents the dynamical character of the magnetic flux uniformity on this case, once the equipotential surface area is conserved with the angular displacement of the rotor. the effect of twist angle is depicted on the fig. 11, where an 18 slots rotor is represented with twist angles of 5º, 20º and 40º, respectively. usually the value of rot  is equivalent to one slot step in the machine. however, using rot  as a constructive parameter of the three-dimensional simulation the effect of rot  on magnetic flux uniformity can be analyzed. it is possible to observe from fig.12 that carter’s factor decrements asymptotically with twist angle. a possible explanation is that with greater twist angle more stator and table iv case study ii pattern i pattern ii sc sens wslot sens r sens wslot sens 0.26% 99.74% 0.02% 99.98% table v slot pattern pattern slotw (mm) r (mm) cs (deg) a 2 3 180 b 6 c 2 3 270 d 1 9 19.47 table vi midline position and displacement pattern liney (mm) line  (mm) a & b 0.68 18 a & c 0.47 3 a & d 0.30 20 fig. 10. optimized slot geometry by genetic algorithm and carter’s factor calculation by fem. table vii slot constructive parameters and carter’s factor comparison parameter original optimized r 3.00 mm 11.87 mm slot w 2.00 mm 1.00 mm c s 180.00º 29.28º med w 3.89 mm 3.33 mm c k (6) 1.062 1.000 c k (1) and (2) 1.057 1.017 c k (1) and (3) 1.060 1.015 rotor slots encounter in an alignment position and the distortion of the magnetic flux is compensated. v. conclusions this paper presented a study of air gap geometry influence on the magnetic field flux density on rotating machines. finite elements method was successfully used to measure uniformity of the magnetic flux distribution by calculating carter's factor. magnetostatic approach shows that slot constructive parameters type i can contribute in 2% for the sensitivity of the value of the carter factor. thus, even though slot w , max h and max w are the most responsible for the magnetic flux uniformity on the air gap it is possible that search for optimized geometries has to take into consideration parameters intrinsically linked to the slot geometry. furthermore, slot geometry has great influence on magnetic equipotential midlines in doubled slotted air gaps. deviations of 3% and 20% was observed, even on similar patterns of slots on rotor and stator surfaces showing that the equipotential midline is not necessarily located in the middle of the gap length. additionally, genetic algorithm optimization process and carter’s factor calculation presented a slot geometry with an approximately uniform magnetic flux on air gap. although real design machine process demands more restriction conditions on air gap constructive parameters the proposed optimization method is easily adapted including penalization factors to optimization problem and offers an efficient method to air gap optimization process. finally, the three-dimensional magnetic field mapping on the double air gap domain was performed to measure the carter’s factor. the proposed methodology calculates the carter’s factor using the intermediate magnetic scalar potential surface in air gap domain. the effect of the rotor twist angle was analyzed and show that carter’s factor decreases asymptotically with twist angle. although the steady-state study performed in the simulation, the results represents the dynamic character of the magnetic flux uniformity in the machine, once the area of scalar magnetic equipotential surface remains constant over angular displacements of the rotor. table viii carter’s factor comparison evaluated on three-dimensional airgap approach equation ck carter (1), (2) and (4) 1.118 langsdorff (1), (3) and (4) 1.120 3d-fem (8) 1.090 (a) (b) (c) fig.11 different twist angles in a rotor with 18 slots. (a) 5º, (b) 20º and (c) 40º. fig.12 graphic representation of the effect of twist angle rot  on carter’s factor. references [1] v. h. juha pyrhönen, tapani jokinen, design of rotating electrical machines. john wiley & sons, ltd, 2014. [2] d. a. lowther, "the development of industrially-relevant computational electromagnetics based design tools," ieee trans. on magnetics, vol. 49, no. 5, pp. 2375-2380, may 2013. [3] g. kron, "induction motor slot combinations rules to predetermine crawling, vibration, noise and hooks in the speed-torque curve,” transactions of the american institute of electrical engineers, vol. 50, no. 2, pp. 757-767, june 1931. [4] t. a. lipo, introduction to ac machine design. wisconsin power electronics research center, wisconsin, 2004. [5] e. m. freeman, "the calculation of harmonics, due to slotting, in the flux-density waveform of a dynamo-electric machine," in proceedings of the iee part c: monographs, vol. 109, no. 16, pp. 581-588, september 1962. [6] g. liebmann, "the change of air-gap flux in electrical machines due to the displacement of opposed slots," in proceedings of the iee part c: monographs, vol. 104, no. 5, pp. 204-207, march 1957. [7] f. w. carter, "corrigendum: the magnetic field of the dynamo-electric machine," in electrical engineers, journal of the institution of, vol. 65, no. 371, pp. 1025-, november 1927. [8] f. w. carter, "note on air-gap and interpolar induction," in electrical engineers, journal of the institution of, vol. 29, no. 146, pp. 925-933, july 1900. [9] f. w. carter, “air-gap induction”, electrical word and engineer, vol. 38, no. 22, pp. 884-888, november 1901. [10] a. langsdorf, principles of direct current machines. macgraw-hill: new york, 1959. [11] h. vuxuan, d. lahaye, h. polinder and j. a. ferreira, "improved model for design of permanent magnet machines with concentrated windings," 2011 ieee international electric machines & drives conference (iemdc), niagara falls, on, 2011, pp. 948-954. [12] h. vu xuan, d. lahaye, h. polinder and j. a. ferreira, "influence of stator slotting on the performance of permanent-magnet machines with concentrated windings," in ieee transactions on magnetics, vol. 49, no. 2, pp. 929-938, feb. 2013. [13] w. p. calixto, b. alvarenga, a. p. coimbra, a. j. alves, l. martins neto, m. wu, w. g. da silva and e. delbone, “carter’s fator calculation using domain transformations and the finite elemento method,” international journal of numerical modelling: electronic networks, devices and fields, vol. 25, n0. 3, pp. 236-247, 2012. [14] a. c. viorel, i. a. viorel and l. strete, "on the calculation of the carter factor in the slotted electric machines," 2014 international conference and exposition on electrical and power engineering (epe), iasi, 2014, pp. 332-336. [15] z. x. fang, z. q. zhu, l. j. wu and z. p. xia, "simple and accurate analytical estimation of slotting effect on magnet loss in fractional-slot surface-mounted pm machines," 2012 xxth international conference on electrical machines, marseille, 2012, pp. 464-470. [16] w.p. calixto, j. c. da mota and b. p. alvarenga, “methodology for the reduction of parameters in the inverse transformation of schwartz-christoffel applied to electromagnetic devices with axial geometry”, international journal of numerical modelling, vol. 24, 2001. [17] w. p. calixto, e. g. marra, l. da cunha brito and b. p. alvarenga, “a new metthodology to calculate carter fator using genetic algorithms,” international journal of numerical modelling, vol. 24, 2011. [18] j.p.c. kleijen, experimental design for sensitivity analysis, optimization and validation of simulation models, pp. 173-223, johm wiley & sons, inc. 2007, [online] available: http://dx.doi.org/10.1002/ 9780470172445.ch6. [19] k. l. du and m. n. s. swamy, search and optimization by metaheuristics: techniques and algorithms inspired by nature, birkhäuser, 2016 i. introduction ii. carter’s factor iii. methodology a. case i: b. case ii c. case iii: d. case iv e. case v iv. results a. case i b. case ii c. case iii d. case iv e. case v v. conclusions references  transactions on environment and electrical engineering issn 2450-5730 vol 2, no 2 (2017) © carlos l. b. silva, thyago g. pires, wesley p. calixto, diogo n. oliveira, luis a. p. souza and antonio m. silva filho  abstract—this paper deals with the computation of ground resistance, surface voltage, touch voltage and step voltage, to mesh with horizontal wires arranged in different angles. the computer program implemented used in the mathematical modeling is based on the method proposed by heppe, which allows obtaining the grounding parameters for homogeneous soil and soil stratified in two layers. the results obtained with the proposed method will be compared with other methods in literature. also will be presented the results of a grounding grid using wires at various angles. index terms— grounding grids parameters, heppe, soil stratified in two layers. i. introduction he study and analysis of grounding grids brings great concern to engineers, as is the initial step in the process of building a substation. the main purpose of the grounding grid design is to keep the step voltages, touch and electrical resistance to earth within tolerable limits [1]. the classic method of grounding grid design [2] is a method that does not require computing resources and its intended to be easy to use. however, it has some limitations for heterogeneous soil, to the analysis of potential on the ground’s surface and the geometry of the ground grid. it can only be used in cases where the wires are equidistant and in grounding grids with the following shapes: square, rectangular, l-shape and t-shape. the geometry of the grounding grid depends on the area of the substation [3] and several studies prove a greater effectiveness of the unequally spaced grounding grids as regards the trend the touch voltages [4]. the methodology used in this paper to obtain the ground resistance and the potential on the soil surface is based on heppe [5] using the method of images and the average potential method. the examples shown in [5] used only grids containing conductors placed in parallel and perpendicular to each other, deployed on homogeneous soil. however, our method enables the use of meshes in any relative positions with conductors placed in soil stratified in two layers. the computer program was developed to implement the mathematical model and allows the calculation of the grounding potential rise, the potential on the soil surface and the ground resistance. the touch voltages and the step voltages obtained from de surface potential. some results of grounding grids will be presented in standard formats, which are compared with traditional methods. results of a ground grid of unconventional geometry are also presented. ii. methodology the grid conductors are conceptually divided in rectilinear segments in order to discretize the system. the accuracy of the modeling is associated with the number of segments used. the greater the number of segments, the more precise is the modeling. in each segment, it is considered that the distribution of leakage current is constant throughout its length, but distinct from segment to segment. it is assumed that all segments have the same voltage, which is equal to the ground potential rise (gpr). after the division, the leakage current of each segment and gpr are calculated. then, the leakage current is used to calculate the ground resistance and the voltage at the ground surface at any desired point. to find the leakage current (i) in each segment the linear equation shown in (1) must be solved. where m is the number of segments. mmmmmmm mm mm mm viriririr viriririr viriririr viriririr          332211 33333232131 22323222121 11313212111  (1) the above system can be written in matrix form as:   the total current injected into the grid ( gi ) is equal to the sum of leakage current of all segments, as shown in (3).     m k gk ii 1                                                     mmmmmmm m m m v v v v i i i i rrrr rrrr rrrr rrrr       3 2 1 3 2 1 321 3333231 2232221 1131211 calculation of grounding grids parameter at arbitrary geometry carlos l. b. silva, thyago g. pires, wesley p. calixto, diogo n. oliveira, luis a. p. souza and antonio m. silva filho t appending (3) in (2), we have:                                                                   g mmmmmm m m m igpr i i i i rrrr rrrr rrrr rrrr 0 0 0 0 01111 1 1 1 1 3 2 1 321 3333231 2232221 1131211          thus the gpr becomes a system variable, because the total current injected into the grid is usually a project information and not the potential of electrodes. next, computation of mutual and self-resistance of (4), the ground resistance and voltages will be explained. all terms are calculated for each individually segment, without any symmetry of the grid as used in [5]. to calculate the mutual resistance and the voltage at the ground surface the method of images is used. a. mutual resistance the mutual resistance (rjk) is the ratio of the voltage produced on the segment k by leakage current of segment j. the symmetry of mutual resistance allows. the self-resistance (rjj) is the ratio between the voltages produced on the segment by its own leakage current. considering a soil composed of two layers with the upper layer having resistivity ρ1 and depth h, and lower layer having resistivity ρ2 and extending to a great depth. the mutual resistance between a segment j and a segment k, and their images, buried at the same depth (d) in the upper layer of soil is given by (5) and in the bottom layer is given by (6). considering a soil composed of two layers with the upper layer having resistivity ρ1 and depth h, and lower layer having resistivity ρ2 and extending to a great depth. the mutual resistance between a segment j and a segment k, and their images, buried at the same depth (d) in the upper layer of soil is given by (5) and in the bottom layer is given by (6). fig. 1 is the corresponding diagram to the terms of (8) and (9). the images of segment are in different planes. the point c is in the same plane of segment ab and point g is in the same plane of segment ef.     where k is the reflection factor. 12 12     k   the term m is given by (8), for . 0   the term  is the following equation:   in the case of parallel segments, when θ decrease towards zero, the term cg.ω /sin θ approaches be+af-bf-ae. to compute the self-resistance a hypothetical segment parallel and identical to the original segment separated by a distance equal to the radius of the conductor is considered. b. ground resistance the ground resistance (rg) is the ratio between the gpr, computation with (4), and the total current injected into the grid. tg igprr    c. voltage on soil surface once the leakage currents in each segment is found, the voltage at a point on the soil surface due to the contribution of a leakage current of a segment located in a upper layer is                         )22(2 22)2( 4 1 0 1 dhnmhnmk dhnmhnmk ll r n n n n kj jk                       dhnmkk dhmkm ll r n n kj jk 221 220 4 0 2 2     sin '' ' ln ' ' ln ' ' ln ' ' ln)(                                                cg aeae bebe ge afaf bfbf gf eaae faaf ca ebbe fbbf cbcgm                                             ae ge cg ca ae cg af gf cg ca af cg be ge cg cb be cg bf gf cg cb bf cg          sin tan tan sin tan tan sin tan tan sin tan tan 1 1 1 1 c,g a a� b b� f f� e e� lj x y  fig. 1. angled segments. calculated by (11) and of a segment located in a bottom layer is calculated by (12).                                                                 lxdhnylx xdhnyx lxdhnylx xdhnyx k lxdylx xdyx l i v n n 222 222 1 222 222 222 222 1 2)( 2 ln 2)( 2 ln )( ln 2                                                                1 222 222 222 222 1 2 2 ln ln 2 )1( n ppp pppn ppp ppp xdhnyx xldhnylx k xdyx xldylx lπ kiρ v   therefore, the voltage at a point on the soil surface is calculated by superposition, by the sum of the contribution of all segments. d. touch, mesh and step voltages with the surface voltages, the other voltages can be determined. the touch voltages is the potential difference between the gpr of a ground grid and the surface potential at the point where a person could be standing while at the same time having a hand in contact with a grounded structure. furthermore, the mesh voltage is the maximum touch voltage within a mesh of ground grid. moreover, the step voltage is the difference in surface potential that could be experienced by a person a distance of 1m with the feet without contacting any grounded object. iii. results three case studies are presented. the case studies 1 and 2 perform the validation of the proposed method by comparing vcm with traditional methods. case study 1 compare the values of the ground resistance of the grids with square mesh by other methods. case study 2 compare the ground resistance, mesh voltage and step voltage with the design procedure in [6]. finally, case study 3 show the results for an unconventional grid. a. case study 1 table i shows the ground resistance values for a square grid (20m x 20m) and a rectangular grid (40m x 10m) in homogeneous soil. the ground resistance values are calculated using the simplified calculations provided in the ansi-ieee std. 80/2013: dwight [7], laurent and nieman [6], sverak [8] and schwarz [9]. in addition to the calculations presented by nahman [10] and chow [11]. the bem method (boundary element method) is obtained from [12] and vcm is computed with the method presented in this paper. the values in parentheses are the percentage differences from the values calculated by vcm [15]. the grounding grid features used as program inputs are: d = 0.01 m (diameter of the conductor) d = 0.5 m (depth of burial) ρ = 100 ωm (soil resistivity) table i ground resistance method square (20mx20m) rectangular (40mx10m) 4 meshes 16 meshes 4 meshes 16 meshes dwight 2.2156 (15.9%) 2.2156 (6.4%) 2.2156 (6.8%) 2.2156 (3.2%) laurent 3.0489 (15.7%) 2.7156 (14.7%) 2.9848 (25.5%) 2.6918 (25.4%) sverak 2.9570 (12.2%) 2.6236 (10.8%) 2.8929 (21.6%) 2.5998 (21.1%) schwarz 2.8084 (6.6%) 2.6035 (10.0%) 2.4690 (3.8%) 2.3211 (8.15%) nahman 3.6367 (38.1%) 3.1491 (33.0%) chow 4.8017 (82.3%) 3.2621 (37.8%) bem 2.6269 (0.3%) 2.3631 (0.2%) 2.2734 (4.4%) 2.0795 (3.1%) vcm 2.6343 2.3669 2.3784 2.1461 b. case study 2 this case study compares vcm with traditional method [6] for two grids in a soil stratified in two layers, rectangular grid and l-shape grid. to calculate the classic method was used the methodology of [13] to find the apparent resistivity. the features of the soil and of two ground grids used as program inputs are: ρ1 = 200 ωm (upper layer resistivity) ρ2 = 400 ωm (bottom layer resistivity) h = 8 m (depth of the upper layer) d = 0.5 m (depth of burial of ground grid) d = 5 mm (wire diameter) ∆l = 5 m (distance between parallel conductors) ig = 1000 a (total current injected into the grid) fig. 2 show a rectangular grid with dimensions 35m x 20m containing 28 meshes. the apparent resistivity seen by grid is 253.33ωm. for the classic method the ground resistance was 4.87ω, the mesh voltage (vm) was 1019.95v and the step voltage (vs) was 687.77v. with vcm the ground resistance was 4.66 ω, the mesh voltage was 927.92v in the corners, the maximum step voltage within the grid was 250.32v and the step voltage in the corners was 509.82v. assuming a t-shaped grid as show in fig. 3 with 18 meshes and dimensions 30m x 25m, the apparent resistivity seen by grid is 246.67ωm. according ieee std. 80-2013 [6], the ground resistance was 6.00 ω, the mesh voltage was 1278.20v and the step voltage was 830.82v. calculating by vcm the ground resistance was 5.40 ω, the mesh voltage was 1168.56v and the step voltage was 330.63v within the grid and 639.80v in the top corners. fig. 2. rectangular grid – 35m x 20m. fig. 3 t-shape grid – 30m x 25m. table ii show the results found to the grids above with the difference of vcm to ansi-ieee std. 80/2013. table ii parameters with ieee std. 80 and vcm grid data method difference std. 80 vcm rectangular 35mx20m rg (ω) 4.87 4.66 4.31% vm (v) 1019.95 927.92 9.02% vs (v) 687.77 509.82 25.87% t-shape 30mx25m rg (ω) 6.00 5.40 10.00% vm (v) 1278.20 1168.56 8.58% vs (v) 830.82 639.80 22.99% c. case study 3 figure 7 show a grounding grid of 120m x 80m, with variable spacing between the conductor. the profiles of the potential at the soil surface in the lines indicated by a,b,c and d obtained by the method proposed in this work are compared with the results of huang [1]. the following input data used: ρa = 200 ωm (apparent ground resistivity) d = 0.6 m (depth of burial of ground grid) d = 8.75 mm (wire diameter) ig = 10000 a (total current injected into the grid) fig. 7 grounding grid with different spacing. figure 8 shows the potential on the soil surface profile obtained. fig. 8 profiles on the soil surface, results obtained by the proposed method. the potential on the soil surface with geographic location of coordinates x = 1.25m and y = 2.0m, obtained in the work of huang [1] is 10.37kv while by the proposed method is 10.40kv. the result obtained for the soil surface potential with geographic location of coordinates x = 52.5m and y = 32.5m in the work of huang [1] is 10.23kv and by the proposed method is 10.34kv. figure 9 shows the distribution of the equipotential through isolines. potential peaks observed at the intersections of the electrodes, except at the border of the grid where potential reduction occurs. the maximum potential at the soil surface occurs in coordinate x = 60m and y = 40m, with a value of 11.33kv. fig. 9 equipotential distributed on the soil surface. the maximum surface potential obtained at the central point of the grid due to the symmetrical distribution of the electrodes around the point. case study 4 it presented a grid composed of conductors at different angles and different lengths as show in the fig. 10. the grid has 16 meters in the x-axis and 17 meters in the y-axis [14]. the following input data were used: ρ1 = 200 ωm (upper layer resistivity) ρ2 = 400 ωm (bottom layer resistivity) h = 8 m (depth of the upper layer) d = 0.5 m (depth of burial of ground grid) d = 5 mm (wire diameter) ig = 1200 a (total current injected into the grid) 6.00 1 7 .0 0 7.088.92 y x 2 .0 0 5 .0 0 0 fig. 10 unconventional grid. fig. 11 shows the voltage profile in three dimensions and contour of the soil surface potential inside the perimeter of the ground grid. fig. 11 surface potential. all voltages calculated for points on the surface located within the perimeter of the mesh. the value obtained for the ground resistance was 8.0ω, for mesh voltage was 2075.98v at the coordinates x = 0m and y = 10m; and the maximum step voltage was 925.04v between the point of coordinates x1 = 16m and y1 = 17m, and the point of coordinates x2 = 15.36m and y2 = 16.23m. the gpr was 9595.60v and the maximum surface voltage (vsurf) is 9245.45v at the coordinates x = 9.8m e y = 10.0m. d. study case 5 the study case presented to verify the influence of the depth of the grounding grid, the ground grid used shown in figure 10, and the depth varied from 0.5m to 3.5m. the potential profiles on the surface were obtained from the cut at y = 11m in the grounding grid shown in figure 10. table iii show the values obtained for the resistance of the grounding grid, gpr, the maximum potential at the ground surface, the touch voltage and the maximum step voltage for different depths of the ground grid. the following input data used: ρ1 = 200 ωm (upper layer resistivity) ρ2 = 400 ωm (bottom layer resistivity) h = 2 m (depth of the upper layer) d = 0.5m – 3.5m (depth of burial of ground grid) d = 5 mm (wire diameter) ig = 1200 a (total current injected into the grid) the table iv show the coordinate maximum of the surface potential and step voltage. fig.12 and fig.13 shows the elevation of the ground resistance values and the gpr of the ground grid, which are directly proportional. table iii grounding grid parameters at different depths d(m) rg (ω) gpr(v) vs (v) vtouch (v) vstep (v) 0.5 9.93 11916.08 11626.38 2377.19 1103.1 1.0 9.64 11565.18 11223.86 2771.30 984.57 1.5 9.48 11371.17 10984.02 3040.44 847.52 1.6 9.46 11350.29 10948.32 3094.98 826.73 1.7 9.45 11337.78 10917.37 3154.29 808.28 1.8 9.45 11337.74 10892.96 3223.90 792.26 1.9 9.47 11363.41 10881.04 3320.62 779.05 2.0 9.68 11620.32 10965.79 3687.02 773.77 2.1 16.89 20270.30 11231.81 12577.11 762.37 2.2 16.85 20219.40 11140.67 12611.32 739.89 2.3 16.78 20138.91 11047.19 12610.62 718.45 2.4 16.71 20050.57 10953.58 12598.41 698.02 2.5 16.63 19960.81 10860.20 12581.82 678.55 3.0 16.29 19549.73 10396.24 12501.50 593.77 3.5 16.02 19224.64 9937.84 12459.86 525.35 table iv coordinate maximum of the surface potential and step voltage. parameter coordinates vstep x = 16.0m and y = 17.0m x = 12.4m and y = 16.3m vs x = 0m and y = 20m fig. 12 resistance (rg) versus depth (d). the boundary between the first and second soil layers occurs exactly in d = 2m. the potential on the soil surface increases in the depths just below to this border (figure 14). figure 15 shows the increase of the touch voltage near the boundary between the soil layers, since the grounding grid when positioned in the second soil layer, which has a higher resistivity (400ω.m) in relation to the first layer that has lower resistivity (200ω.m), produces higher touch potential. figure 16 shows that the pitch voltage decreases smoothly with increasing depth, having a level in the depths near the boundary between the layers. fig. 13 gpr versus depth (d). fig. 14 superficial potential (vs) versus depth (d). fig. 15 touch potential (vt) versus depth (d). fig. 16 surface potential. fig. 17 surface potential. figure 17 illustrates the potential profiles at the soil surface at y = 11 m for the different depths of the grounding grid, where it is observed, reduction of potential with the increase of the depth of the grounding grid, reduction in the number of peaks along the distance. iv. conclusion the method implemented in this paper allows the computation of the ground resistance, grid voltage and step voltage of grids composed by horizontal wire electrodes in shapes that are more complex. wire segments can have any position or displacement among them. the difference between the results obtained with this method and those of the ansi-ieee std. 80/2013 for the grounding resistance was up to 25.5%. for grid voltage was up to 16.6% and 41.9% for step voltage. the individual calculation of the leakage current for each segment leads to a greater precision of the method. this method also proves to be useful for allowing a precise analysis of the voltage on the soil surface, it is possible to calculate the voltage at any desired point. also, the detailed study of any grounding grid at any depth in the soil is possible. v. acknowledgement the authors thank the national counsel of technological and scientific development (cnpq), coordination for the improvement of higher level personnel (capes) and the research support foundation for the state of goias (fapeg) for financial assistance to this research. references [1] l. huang, x. chen, and h. yan, "study of unequally spaced grounding grids," power delivery, ieee transactions on, vol. 10, pp. 716-722, 1995. [2] ieee, "guide for safety in ac substation grounding," in ieee std 80 (revision of ieee std 80-2000/ incorporates ieee std 80-2013/cor 12015), ed, 2015. [3] b. thapar, v. gerez, a. balakrishnan, and d. a. blank, "simplified equations for mesh and step voltages in an ac substation," power delivery, ieee transactions on, vol. 6, pp. 601-607, 1991. [4] a. covitti, g. delvecchio, a. fusco, f. lerario, and f. neri, "two cascade genetic algorithms to optimize unequally spaced grounding grids with rods," in computer as a tool, 2005. eurocon 2005.the international conference on, 2005, pp. 1533-1536. [5] a. j. heppe, "computation of potential at surface above an energized grid or other electrode, allowing for non-uniform current distribution," power apparatus and systems, ieee transactions on, vol. pas-98, pp. 1978-1989, 1979. [6] "ieee guide for safety in ac substation grounding," ieee std 802013 (revision of ieee std 80-2000/ incorporates ieee std 802013/cor 1-2015), pp. 1-226, 2015. [7] h. b. dwight, "calculation of resistances to ground," american institute of electrical engineers, transactions of the, vol. 55, pp. 13191328, 1936. [8] j. g. sverak, "sizing of ground conductors against fusing," power apparatus and systems, ieee transactions on, vol. pas-100, pp. 5159, 1981. [9] s. j. schwarz, "analytical expressions for the resistance of grounding systems [includes discussion]," power apparatus and systems, part iii. transactions of the american institute of electrical engineers, vol. 73, 1954. [10] j. nahman and s. skuletich, "irregularity correction factors for mesh and step voltages of grounding grids," power apparatus and systems, ieee transactions on, vol. pas-99, pp. 174-180, 1980. [11] y. l. chow and m. m. a. salama, "a simplified method for calculating the substation grounding grid resistance," power delivery, ieee transactions on, vol. 9, pp. 736-742, 1994. [12] s. s. m. ghoneim, "optimization of grounding grids design with evolutionary strategies," thesis (phd), faculty of enginnering sciences, univeristät duisburg-essen, duisburg, germany, 2007. [13] j. endrenyi, "evaluation or resistivity tests for design of station grounds in nonunirorm soil," power apparatus and systems, ieee transactions on, vol. 82, pp. 966-970, 1963. [14] pires, thyago g. ; nerys, jose w. l. ; silva, carlos l. b. ; oliveira, diogo n. ; filho, antonio m. silva ; calixto, wesley p. ; alves, aylton j., computation of resistance and potential of grounding grids in any geometry. in: 2016 ieee 16th international conference on environment and electrical engineering (eeeic), 2016, florence. 2016 ieee 16th international conference on environment and electrical engineering (eeeic), 2016. [15] pires, thyago g. ; silva, carlos l. b. ; oliveira, diogo n. ;nerys, jose w. l. ; alves, aylton j.; calixto, wesley p. computation of grounding grids parameter on unconventional geometry. in: 2015 chilean conference on electrical, electronics engineering, information and communication technologies (chilecon), 2015, santiago. 2015. i. introduction ii. methodology a. mutual resistance b. ground resistance c. voltage on soil surface d. touch, mesh and step voltages iii. results a. case study 1 b. case study 2 c. case study 3 case study 4 d. study case 5 iv. conclusion v. acknowledgement references  transactions on environment and electrical engineering issn 2450-5730 vol 2, no 1 (2017) © neha s. shah and hiren h. patel  abstract—non-uniform conditions on the modules of the pv array, especially, partial shading reduces the output of the pv array to a large extent. the shaded module in a string limits the current of the entire string and hence, the output power of the string. the output power under such conditions is reported to be higher for total-cross-tied (tct) configuration. this paper describes two different approaches, one based on current compensation (current equalization) and another based on voltage equalization, to extract higher power from the partially shaded total-cross-tied photovoltaic array. the tct configuration is considered to minimize the number of converters, sensors, cost and complexity involved. the additional converters in the two distinct approaches evaluated here operate only when the partial shading occurs and are controlled to minimize the current and voltage miss-matches. the analysis and the control algorithm are presented. simulation results obtained in matlab/simulink are included to demonstrate the effectiveness of both methods and the relative merits and demerits of these approaches are highlighted. keywords—current compensation, voltage equalization, mppt, partial shading, tct configuration. i. introduction now-a-days, energy generation from renewable energy sources has increased tremendously due to the increasing concern about the environment, technological advancements and decreased cost of renewable technology. amongst all the renewable sources, solar photovoltaic (pv) has emerged as the most popular and prominent source in the last decade, mainly due to bundle of advantages like low maintenance, no operating cost, pollution free, modular nature, etc. the cost of the pv modules has also decreased greatly over the years thereby reducing pay-back period. however, to ensure that the pay-back period is kept to the minimum, it must be ensured that the pv array always operate at or near to its peak power. the peak power available from the pv array is dependent on various parameters like irradiation, temperature, configurations, aging of the modules, module parameters, dusting, miss-match in modules, shading pattern on the array etc. normally, the output power versus voltage (p-v) characteristics of the pv module or an array operating under uniform conditions is characterized by a curve having only one peak. though the peak power available from n. s. shah is with the department of electrical engineering, sardarvallabhbhai patel institute of technology, vasad, india. (e-mail: neha_saurabh_shah@yahoo.com). h. h. patel is with the department of electrical engineering, sarvajanik college of engineering & technology, surat, india. (e-mail: hiren.patel@scet.ac.in). the pv array varies as the irradiation and/or temperature changes, due to the uni-modal (single peak) nature of the p-v characteristics, it is easy to track the maximum power point (mpp) with conventional maximum power point tracking (mppt) techniques[1],[2]. in addition the fact that the mpp occurs corresponding to 75% to 80% of the open-circuit voltage (voc) of the array or corresponding to 90% of the short-circuit current (isc) can also be exploited to locate the mpp. unlike the array operating under uniform conditions, the array operating under non-uniform conditions exhibits much complex output current and voltage (i-v) and p-v characteristics. the i-v characteristic under such case has more than one step, while the p-v characteristic possesses more than one peak: one global peak (gp) and other local peaks. the conventional mppt techniques are no longer effective to track the gp under such conditions. hence, tracking of gp to extract the maximum power from pv array operating in partial shading condition is a challenging task. various global peak power point tracking (gpppt) approaches have been reported[2]-[4]. although these approaches can track the gp, the power is still less than the summation of maximum power that all modules can generate if operating independently. the distributed maximum power point tracking (dmppt) operates on this concept of maximizing the output of each module by applying dedicated mpp tracker with each module. thus, in dmppt the selfcontrol dc-dc power modules connected across each pv module helps to operate individual pv module on their maximum power thereby avoiding scenario which has multiple peaks[5]-[8]. irrespective of the uniform or nonuniform irradiance conditions, the output power of pv array is always processed through the dc-dc converters used as mpp trackers. hence, the same dc-dc converters, which helps in improving the output power and efficiency in partially shaded conditions, results in the decrease in efficiency of the pv system when operating under uniform irradiance. the cost and complexity of the system is also higher. an alternative to overcome the losses occurring in the converter during uniform irradiance condition is current equalization based mppt[9]-[11]. in this method, isolated dc-dc converter is connected across each pv module. unlike dmppt, the converters remain idle during uniform irradiance conditions and come into effect only when partial shading occurs. also, the converters are designed to handle just the miss-match power. however, more sensors are required to measure individual mppt and also system gets periodically disconnected from the load to measure individual maximizing power output of a partially shaded total-cross-tied photovoltaic array neha s. shah and hiren h. patel module i-v characteristics [11],[12]. the approach increases the reliability and output power considerably under severe non-uniform conditions, however at an increased complexity and cost. another limitation of the approach is the losses in the sensing and signal conditioning circuit that occurs continuously, irrespective of status of current equalization control: active or idle. another approach where losses in the converters are eliminated by keeping them ineffective during uniform irradiance conditions is the generation control circuit (gcc). a multi-stage chopper is employed with one chopper connected across each module in a string[13],[14]. to extract maximum power from pv module in partially shaded condition, the off-duty cycle of multi stage choppers are controlled in such a way that the charge equalization (voltage equalization) occurs across the modules connected in strings. the current compensation is applied to the tct configuration [15] with a view to reduce the number of converters. pv modules can be connected in different configuration e.g. series, parallel, series parallel (sp), total cross tied (tct) and bridge link (bl) [16]-[21]. output power in tct configuration is high as compared to sp configuration under partial shading condition [16]-[21]. thus, the current compensation with tct provides a better solution and leads to the reduction in number of converters, sensors and hence, the cost and complexity of the system. also, pv modules always remain connected to the load/ grid [15]. this paper presents and evaluates two different approaches for enhancing the power output of a partially shaded array. with a view to minimize cost of pv system and yet to have maximum possible power from the pv array under uniform and non-uniform insolation conditions, the tct configuration is considered. two different approaches: one based on current compensation (cc) and another that employs concept of generation control circuit (gcc) and is based on voltage equalization principle, are presented and evaluated. ii. current compensation for tct configuration a. system configuration current compensation for tct configuration of ‘n’ ties, where each tie comprises of ‘m’ modules is shown in fig 1(a). the pv module pvij indicates j th module of ith tie. capacitor cout is connected across series connected pv ties which not only acts as a filter to minimize the ripple in the pv array’s output voltage, but also serves as the input to ‘n’ dc-dc converters. the dc-dc converters connected across ties, tap the power from decoupling capacitor and utilize it to minimize the current discrepancies among the series connected ties [15]. as all the dc-dc converters are fed from the same capacitor, cout, it is must to have isolation in the dc-dc converters. hence, flyback converter topology shown in fig. 1(b) is employed for dc-dc converters. also, it handles only the power required to match the difference in the currents of series connected ties. irrespective of the irradiance level on the modules in the tie, as the voltage at which the maximum power point occurs remains nearly the same, the maximum power from all the modules of the tie can be extracted if the tie is made to operate at this specified voltage. if the modules in the tie have different irradiance, they all can be controlled to operate near their individuals mpp by maintaining the voltage across them near to that corresponding to the mpp under uniform insolation condition the current supplied by the different modules in the tie is different (corresponding to their irradiance level). the different ties may be generating different currents depending on the irradiance received by the ties. flyback converters are controlled to output current such that the total current of the tie and its associated flyback converter is equal to the current generated by the least shaded tie (i.e. the tie generating the highest current). the efficiency of flyback convert is considered as 90%. so the input rating of the converter is p/0.9. where p is the output power of converter (1) where icomp_i is the compensating current from i th converter and vconvi is output voltage of i th converter. the flyback converter’s output voltage is controlled by controlling duty cycle of switch swi using vconvi as reference voltage, which is obtained through the algorithm presented in [15]. the output voltage vconvi and current icomp_i of the flyback converters for i th tie are given by: icompconvi ivp _  (a) (b) fig 1(a) current compensation for tct configuration (tct-cc). (b) flyback converter [15] pv11 vpv1 ipv1m ipv pv12pv11 ipv11 pv11 ipv12 pv1m pv11 ipv2m pv22pv11 ipv21 pv21 ipv22 pv2m vpv2 pv11 ipvnm pvn2pv11 ipvn1 pvn1 ipvn2 pvnm vpvn vpv dc dc conv 1 dc dc conv 2 dc dc conv n mppt unit l o a d vo iconv_in c o u t icomp_i 25.05 swj dj dc -dc converter iconv_inicomp_i vpvvconi (2) (3) wheren1 and n2 are the turns of primary and secondary winding of flyback transformer, vpv is the voltage across decoupling capacitor, di is duty cycle of i th converter and iconv_in is the input current of i th flyback converter. total output voltage and current of pv array with the tct configuration are (4) (5) where vpvi is the voltage across i th tie and ipvij is the current generated by the module pvij. total power supplied to flyback converters is (6) and hence the net power supplied is (7) b. control algorithm initially all the flyback converters are inactive and the main boost converter operates to track the mpp, which may not be the optimum operating point. the voltage across each tie is measured and compared to detect whether shading has occurred or not. if all the ties have same voltage, then it indicates that the modules have uniform irradiance. if any mismatch in the voltage across ties is detected, it indicates that shading has occurred. the tie experiencing shading is characterized by lesser voltage than that having less or no shading. flyback converter of the shaded tie is controlled to minimize current mismatch. the duty cycle of main boost converter is kept constant during this phase. the duty cycle of converter of shaded tie is adjusted such that the voltage across shaded tie increases with a condition that the total voltage across all the ties remains constant. this results into increase in the output power from the shaded modules. the detailed control algorithm of the tct-cc method is shown in fig 2. in tct configuration, only one flyback converter is assigned with each tie, resulting in a reduction in the number of flyback converters required compared to that for sp configuration. the main advantage of tct configuration is the reduction in the number of sensors, signal conditioning circuits and the associated losses which occurs even in case when they are not active (i.e. under uniform irradiation on all ties). the main drawback of this method is the low power handling capability of about 150-200w by the flyback converter. flyback converter is operated at high frequency generally 1khz50khz and power rating of the converter is start initialize n = number of pv ties set duty cycle of flyback converter to zero measure the voltage across each pv tie and compare with each other. is mpp tracked? i=1 adjust duty cycle of i th flyback converter with all other flyback converter and main converter duty cycle constant i=i + 1 no yes if all the tie has same voltage? yes no is pk > pk+1 ? is i > n ? find the i th tie with lowest voltage pvi no no yes yes call p&o mppt fig. 2 algorithm of tct-cc method for mppt decided by the voltage at mpp of any of the ties under uniform insolation conditions and the compensating current required to be injected from the converter under the worst possible condition. iii. gcc for tct configuration gcc concept [13], where gcc is used with each module of series connected string is extended for tct configuration with pv i i convi v d d n n v            1 1 2 inconv i i icomp i d d n n i _ 2 1 _ 1               n i pvipv vv 1    m j pvijicomppv iii 1 _ icomp n i conviflyback ivp _ 1    flybackpvpvnet pivp  a view to reduce the number of choppers required. modified control algorithm is presented to extract the maximum power from the tct-gcc configuration shown in fig. 3. a. system configuration of tct-gcc fig 3 shows a tct-gcc configuration having ‘n’ ties. each pv tie consists of ‘m’ pv modules. here only ‘n’ number of multi-stage chopper circuit is required compared to ‘m × n’ for sp configuration. the number of inductors required is n-1. the ‘mppt unit’ shown in fig.3 is a boost converter, which tracks the peak power point. under uniform irradiance conditions when all the ties have same irradiance and generate similar output current, the gate-pulses to switches of tct-gcc circuits are inhibited, making the tct-gcc idle. switch swi along-with the anti-parallel diode provides the bidirectional current capability, thereby allowing the capability of exchanging the charge between the capacitors ci. thus, it provides the feature to achieve voltage equalization amongst the output voltages of all the ties. anti-parallel diodes connected across each switch swi can also work as conventional bypass diodes preventing the large negative voltage across the modules. also, if the switches swi fail, the main converter can still extract higher power from the partially shaded array by tracking the gp. of course the power extracted from the array is less than the case when tct-gcc is in operation, but certainly more than the case when bypass diodes are absent. to extract maximum possible power from the partially shaded pv array, the off-duty ratios of multistage choppers are controlled such that the output voltage of each pv tie is regulated near to the voltage corresponding to the maximum power point. pv11 vpv1 ipv1m ipv pv12pv11 ipv11 pv11 ipv12 pv1m pv11 ipv2m pv22pv11 ipv21 pv21 ipv22 pv2m vpv2 pv11 ipvnm pvn2pv11 ipvn1 pvn1 ipvn2 pvnm vpvn vpv mppt unit l o a d vo sw1 sw2 swn cn c2 c1 l1 l2 ln-1 c o u t fig 3 system configuration for tct-gcc method the relationship between the voltage across each pv module and the off-duty ratios of each switches of the tctgcc are expressed in (8). (8) where is off-duty ratio of switch swi. (9) where ti(off)is the off time of switch swi and tsw is the switching time interval. (10) the off-duty ratio of ith tie depends on generation control voltage of ith tie vpvi and input voltage to boost converter vpv. (11) the output current ipv of the system is given as (12) (13) where ii is the i th pv tie current and ipvij is the current generated by the module pvij. the output power pout of the system is (14) b. control algorithm of tct-gcc configuration the algorithm of the tct-gcc method is shown in fig. 4. initially, all the multi-stage choppers are deactivated and the main boost converter operates and tracks the peak which may not be the optimum operating point. the voltage across each tie of the tct configuration is measured and compared to determine if partial shading has occurred or not. once the partially shaded condition is detected, the duty-cycle of the main(boost) converter is kept constant and the duty ratios of choppers of tct-gcc are adjusted. the off-duty ratio of the tie with lowest output voltage (maximum shading) is calculated and adjusted. the off-duty ratios of all choppers are adjusted so as to meet the conditions mentioned by (8)-(10). continuously the tie with lowest voltage is identified and the off-duty ratios are adjusted to achieve voltage equalization amongst the ties. periodically, the main converter is also activated to ensure that the operation is maintained at or near to mpp to extract the maximum possible output power. to achieve this, a flag is set periodically for a very small duration during which the main converter employs the conventional mppt technique like perturb and observe(p&o) or incremental conductance method. thus, when flag is set duty cycles of multi-stage converters are kept constant and that of boost converter is adjusted as per p&o to set operating point of pv array for maximum output power. pvnpvipvpvni vvvvdddd :....::....:::....::....:: 2121  i d sw offi i t t d )(  1 1   n i i d pv pvi i v v d     n i iipv idi 1    m j pviji ii 1    n i iipvpvout ivivp 1 fig.5 shows the gate pulse sequence of switches sw1-swn in partially shaded condition. at any instant only one switch is turned off and total off-duty ratio of all switches is one. iv. simulation results the algorithms of both the methods are simulated using matlab/simulink software. this section presents results of the simulations and discusses the major observations. solarex-msx60 pv modules are considered for simulations. the specifications of the solarex-msx60 pv module at standard test conditions are shown in table i. table-i pv module specification at 1000 w/m2. 25oc pv module power 60 w open circuit voltage 21 v short circuit current 3.8 a voltage at mpp 17.1v current at mpp 3.5v fig. 6 shows the system configurations considered for the evaluation of two approaches. fig 6(a) shows system configuration for current compensation technique while fig. 6(b) represents system configuration for tct-gcc. in both cases, pv array comprises of two tieseach having two pv modules. initially at t=0s, all pv modules (pv11, pv12, pv21 and pv22) operate under same irradiance of 1000 w/m2. at t=0.5s partial shading occurs and the irradiance on the module pv21 and pv22 of tie-2 decreases to 500 w/m2. at t=3s irradiance on pv12 module decreases to 800 w/m2 with other pv module still having earlier irradiance (that prevailing before t=3s). (a) (b) fig.6 system configurations having two ties for evaluating the two approaches: (a) tct-cc[15] ; (b)tctgcc. pv11 vpv1 ipv12 ipv pv11 ipv11 pv11 pv12 pv11 ipv22 pv11 ipv21 pv21 pv22 vpv2 vpv dc dc conv 1 dc dc conv 2 mppt unit l o a d vo iconv_in c o u t icomp_i vpv1 ipv12 ipv ipv11 pv11 pv12 ipv22ipv21 pv21 pv22 vpv2 vpv mppt unit l o a d voc o u t sw1 sw2 c2 c1 l1 il1 start initalize n = number of pv ties, flag = 0, deactivate gcc control measure tie voltages pvi and compare with each other. is mpp tracked? adjust off duty cycle no yes is shading occurred in any pv tie module? no yes find the tie with lowest voltage call p&o mppt is flag set? yes no ,........ 21 ni dddd  1 1   n i i d i d fig 4. algorithm of tct-gcc technique for mppt. fig 5 gate signals of the multistage chopper for tct-gcc. t1(off) t2(off) tn(off) tsw gsw1 gsw2 gswn till t=0.5s, all pv modules have uniform irradiance of 1000 w/m2. hence, till t=0.5s flyback converters in tct-cc method and multistage choppers in tct-gcc method do not operate. the main dc-dc converter tracks the peak through the p&o algorithm resulting into output power of about 240w in both the cases. at t=0.5s the partial shading occurs on the pv array with lower irradiance on the module of tie-2. during t=0.5s-0.9s, no converters are operated except main boost converter which is tracking the mpp. output current and power is reduced to 3.8a and 130w, respectively compared to 7a and 240w (available before t=0.5s) as shown in fig.7. (a) (b) (c) (d) fig. 7 waveforms of (a) voltages of tie-1 and tie-2, (b) input current of boost converter, (c) pv array voltage and (d) power of tie-1, tie-2 and power fed to boost converter at t=0.9s flyback converters of tct-cc method and multistage choppers of tct-gcc method are activated. in tct-cc method, voltage across tie-2 is increased in small steps and it injects current in tie-2 to minimize the mismatch in current of both ties. this results in to increase in output current and power of pv system to 4.7a and 163w, respectively. fig. 7(d) shows that the total power extracted from the pv array is (power of tie1 + power of tie-2) = 114+58=172w against the maximum 180w that the modules of pv array could generate under the given scenario. thus, the tct-cc method is unable to track power of about 8w. fig.7(d) also shows the power fed to the main dc-dc (boost) converter (referred as pv system power in fig. 7(d)) is 163w, which is further 9w less than that extracted from the pv array. the reduction is due to the losses in the flyback converters. the flyback converters are relatively less efficient as they involve indirect energy transfer. thus, the difference of 17w is due to the losses in the flyback converters and inaccuracy in tracking the mpp. when the irradiance on the pv12 module decreases to 800w/m2 at t=3s, the net output power of the pv system further reduces to 156.5w against the 168w available. in tct-gcc method at t=0.9s multistage chopper is operated and duty cycle of main boost converter is held constant. the duty-cycles of the multi-phase choppers are adjusted to equalize the voltages across the two ties. fig. 7(a) shows that the difference in the tie voltages with tct-gcc is lesser than that with cc approach. initially on the activation of tct-gcc algorithm, the output power of the pv system is increased up to 150w, which is further increased to 171w when the flag activates the boost converter for p&o control. in a pre-set program, the flag for activating the p&o control is periodically set after every one second and stays on for 0.01s duration. during the period when the flag is set, the duty cycle of the boost converter is adjusted as per p&o control while the duty cycles of multistage choppers are held constant. the output power extracted from the pv array and current is thus increased to 173.5w (powers of tie-1 and tie-2 = 115.5w + 58w =173.5w) and 5.1a, respectively. the total power fed to the boost converter is 171w, which is still 9w less than 180w, the sum of maximum power that all modules can generate under the given conditions. the difference of 9w is due to the power lost in multi-phase chopper (2.5w) and power lost due to inaccuracy in tracking the mpp (6.5w). after t=3s, the shading of pv12 module results into the decrease in the net power available from the pv system. under this condition, the power fed to boost converter with tct-gcc approach is 161w against the 156.5w observed for tct-cc approach. thus, compared to tct-cc technique, the tct-gcc approach shows improvement of about 4.5% in the output power during the period t=0.9s-3s and 2.9% after t=3s over. figs. 8(a), (b) and (c) show the zoomed view of tie voltages, input currents of boost converters and the pv array voltages for both the control approaches for the time range t=1.1s – 1.5s. it clearly shows that the tie voltages for tctgcc are close to each other and around the voltage corresponding to mpp. (a) (b) (c) fig. 8 zoomed waveforms: (a) voltages of tie-1 and tie-2, (b) input current of boost converter, (c) pv array voltage each row of table-ii shows maximum output power of pv array with different shading pattern. g_pvij indicates the irradiance on pv module pvij, while po_wo_fc stands for maximum power obtained in absence of flyback converters. po_cc represents the maximum power tracked with tct-cc method and po_gcc represents the maximum power tracked with tct-gcc method. it is observed that the gcc method significantly enhances the amount of output power of the array under partially shaded conditions compared to tct-cc method. table-ii mpp power of pv array under various irradiance conditions for cc and gcc method. g_pv11 (w/m2) g_pv12 (w/m2) g_pv21 (w/m2) g_pv22 (w/m2) po_wo_fc (w) po_cc (w) po_gcc (w) ∑𝑃𝑖𝑗𝑚𝑎𝑥 (w) 1000 1000 1000 1000 235 ------240 1000 1000 1000 500 171 200 203 210 1000 800 500 500 125 156 161 168 1000 1000 500 500 130 163 171 180 1000 800 600 400 125 156 161 168 1000 600 400 400 92 124 126.5 144 v. conclusion two approaches, tct-cc and tct-gcc, are compared with a view to maximize the output power of the partially shaded pv array. the advantage with these approaches is that they operate only when partial shading occurs and remain inactive under uniform irradiance conditions, thereby avoiding the converter losses (except that occurring in the main dc-dc converter). amongst the two approaches, it is observed that the tct-gcc yields more power, about 2.9% to 4.5% than that of tct-cc. the tct-gcc is simple to design as is does not require isolated dc-dc converters. in addition, for tct array configuration, it is more suitable than tct-cc as there are practical constraints in the design of flyback converter with a rating exceeding 250w. further, tct-gcc has an upper hand in terms of reliability. in case if the switches of the multistage choppers fail to operate, the bypass diode connected across the switches serves the function of conventional bypass diodes connected across the pv modules, thereby preventing any damage to modules and simultaneously allowing the mpp tracker to extract higher power by operating near the gp. the highlight of the tctgcc and the control scheme presented is the reduction in the number of converters and the sensors. the reduction in the number of converters is due to the tct configuration adopted. the control scheme relies only on the voltage across the ties and the input current of the main boost converter. thus, the tct-gcc approach presented greatly reduces the cost, complexity and losses. simulation results obtained in matlab/simulink justifies the effectiveness of the tctgcc approach in enhancing the power output of partially shaded array. references [1] h. patel and v. agarwal, “matlab based modeling to study the effects of partial shading on pv array characteristics”, ieee trans. energy conversion, vol. 23, no.1, pp. 302-310, march 2008. [2] h.patel and v. agarwal, “maximum power point tracking system forsystems operating under partially shaded condition”, ieee trans. ind. electron., vol. 55, no.4, pp.1689-1698, april 2008. [3] w.l. chen and c.t. tsai,” optimal balancing control for tracking theoretical global mpp of series pv modules subject to partial shading”, ieee trans. ind. electron, vol. 62, no.8, pp.4837-4848, august 2015. [4] l. gao, r. a. dougal, s. liu, and a. p. iotova, “parallel-connected solarpv system to address partial and rapidly fluctuating shadow conditions,”ieee trans. ind. electron., vol. 56, no. 5, pp. 1548–1556, may 2009. [5] f.blaabjerg, r.teodorescu, m.liserre and a.v.timbus, “overview of control and grid synchronization for distributed power generationsystems”, ieee trans. ind. electron., vol.53, no.5, pp. 13981409, october 2006. [6] n.femia, g. lisi, g. petrone, g. spagnuolo, and m.vitelli, “distributedmaximum power point tracking of photovoltaic arrays : novel approachand system analysis”, ieee trans. ind. electron., vol.55, no.7, pp. 2610-2621, july 2008. [7] z. salam, m. z. ramli, “distributed maximum power point tracker foradditional energy harvesting during partial shading of pv system”, ieee conf. on smart grid and renewable energy (sgre), pp. 1-5, march 2015. [8] g. aadinolfi, g. graditi, p. siano, and a. piccolo, “multiobjectiveoptimal design of photovoltaic synchronous boost converters assessingefficiency, reliability and cost savings,” ieee trans. ind. informat., vol. 11, no. 5, pp. 1038–1048, october 2015. [9] y. nimni, d. shmilovitz, “a retuned energy architecture for improvedphotovoltaic system efficiency”, ieee int. symp. circuits and systems, paris, pp. 2191-2194, june 2011. [10] q. zhang, x. sun, y. zhong and m. matsui “a novel topology for solving the partial shading problem in photovolatic power generation system,” ieee 6th int. power electron. motion control conf. (ipemc), wuhan, pp. 2130-2135, may 2009. [11] p. sharma, p.k. peter and v. agarwal, “exact maximum power point tracking of partially shaded pv strings based on current equalization concept”, 38th ieee photovoltaic specialists conf., austin, usa, pp. 1411-1416, june 2012. [12] p. sharma and v. agarwal, “exact maximum power point tracking of grid-connected partially shaded pv source using current compensation concept,” ieee trans. power electron., vol. 29, no. 3, pp. 4684-4692, september 2014. [13] t. shimizu, m. hirakata, t. kamezawa, and h. watanabe, “generation control circuit for photovoltaic modules,” ieee trans. power electron, vol. 16, pp. 293–300, may 2001. [14] t. shimizu, o. hashimoto, and g. kimura, “a novel high-performance utility-interactive photovoltaic inverter system,” ieee trans. power electro, vol. 18, pp. 704–711, march 2003. [15] n. shah and h. patel, “enhancing output power of pv array operating under non-uniform condition”, 16th ieee int. conf. on environment and electrical engineering (eeeic2016),italy, pp. 1-6, june 2016. [16] m. jazayeri, s. uysal, and k. jazayeri, "a comparative study on different photovoltaic array topologies under partial shading conditions", ieee pes t&d conf. & exposition, pp. 1-5, usa, april 2014. [17] b.j.g.montano, d.j.f.rombaoa, r.a.s.peña and e.q.b.macabebe, “effects of shading on current, voltage and power output of total crosstied photovoltaic array configuration”, 10 region ieee conf. tencon, macao, pp.1-5, november 2015. [18] s. vijayalakshmy, g.r. bindu, s.r. iyer, “performance of partially shaded photovoltaic array configurations under shade dispersion”, ieee int. conf. advances in green energy (icage), trivandrum, pp. 50-55, december 2014. [19] b. i. rani, g.s. ilango and c. nagamani,”enhanced power generation from pv array under partial shading conditions by shade dispersion using su do ku configuration”, ieee trans. on sustainable energy, vol.4, no.3, pp.594-601, april 2013. [20] e.i. batzelis, p.s. georgilakis and s.a. papathanassiou, “energy models for photovoltaic systems under partialshading conditions: a comprehensive review”, iet renew. power generation, vol. 9, no.4, pp. 340–349, 2015. [21] h. sahu, s. nayak, s. mishra, “maximizing the power generation of a partially shaded pv array”, ieee j. emerging sel. topics power electron, vol.4, no.2, pp.626-637.june 2016. neha shah received the b.e. degree in electrical engineering from the s.v. regional college of engineering and technology (now s.v. national institute of technology), south gujarat university, surat, india, in 2000, and the m.e. degree in electrical engineering in 2003 from the m. s. university, baroda, india. she is currently working as an assistant professor in department of electrical engineering at sardar vallabhbhai patel institute of technology, vasad and pursuing the part time ph.d. degree course in electrical engineering affiliated to gujarat technological university, ahmedabad, gujarat, india. her current research interests include distributed generation, energy extraction from photovoltaic arrays, partial shading issues. she is a life member of the indian society for technical education and a member of ieee. hiren patel received the b.e. degree in electrical engineering from the s.v. regional college of engineering and technology (now s.v. national institute of technology), south gujarat university, surat, india, in 1996, and the m.tech. degree in energy systems in 2003 from the indian institute of technology— bombay (iitb),mumbai, india. he has received his ph. d. degree from the indian institute of technology in the year 2009. he is working as a professor at sarvajanik college of engineering and technology, surat. his current research interests include computer aided simulation techniques, distributed generation, and renewable energy, especially energy extraction from photovoltaic arrays. mr. patel is a life member of the indian society for technical education. i. introduction ii. current compensation for tct configuration a. system configuration b. control algorithm iii. gcc for tct configuration a. system configuration of tct-gcc b. control algorithm of tct-gcc configuration iv. simulation results v. conclusion references transactions on environment and electrical engineering issn 2450-5730 vol 2, no 2 (2017) © douglas freire de carvalho, cleber asmar ganzaroli, luiz alberto do couto, rafael nunes hidalgo monteiro dias, wesley pacheco calixto  abstract—this paper presents study about dynamic matrix control (dmc) controller applied to speed control of dc motor. dmc controller parameters (prediction horizon, control horizon and damping rate of reference) are obtained through optimization methods employing heuristic, deterministic and hybrid strategies. the use of advanced control technique combined with using of optimization methods aims to achieve highly efficient control, reducing the transient state period and variations in steady state. these methods were applied on a simulation model in order to verify which one provides better control results. index terms—predictive control, deterministic optimization, heuristic optimization, hybrid optimization, dc motor. i. introduction irect current (dc) motors are used in various situations ranging from residential applications to purposes of industrial scale. utilization of dc motors implies, often, in its speed control. aiming to do a high quality speed control of dc machines several number of techniques has been developed [1]. control systems techniques are employed seeking to promote proper implementation of processes, generally, controlling manipulated system variables to obtain desired values for output system variables [2]. control systems techniques are often applied to control speed of dc machines aiming to promote proper implementation of processes. model based predictive control (mpc) refers to determinate class of control algorithms that seek to obtain optimal control signal minimizing certain objective function, explicitly using process model. by calculating series of actions of manipulated variables, the mpc seeks, overall, that the system output reaches its reference trajectory. mpc have been developed seeking to solve problems of process control in industrial environment, particularly in oil industry, being initially proposed by richalet at 1978, proposing the model predictive heuristic control (mphc) and by cutler & remaker at 1980, proposing the dynamic matrix control authors thank the national counsel of technological and scientific development (cnpq), the research support foundation for the state of goiás (fapeg) and the coordination for the improvement of higher education personnel (capes) for financial assistance to this research. (dmc) [3]. mpc are employed in various areas being widely accepted by industry and academia, applicable to various systems (multivariable, nonlinear, with high dead time, constrained variables, etc.). currently dmc is the most popular mpc algorithm, widely used for control of chemical processes and with good results in several other applications. dmc become widely popular in industry for present intuitive operation and provide significant results [4]. dmc is based on use of step response finite model of the system to be controlled and its control strategy is presented in time domain, being more intuitive than systems modeled in state space. applications of dmc controller for linear systems without restrictions present analytical solutions for objective function minimization problem, reducing computational costs. however, nonlinear systems or with dead time and imposition of restrictions to system are met clearly and efficiently by dmc controller [5]. given the complexity of some problems and the search for efficient and robust controllers, controllers’ optimization is presented as interesting proposal. implementation of controllers’ optimization process, generally, seeks to define optimized values for controller variables, aiming to reduce the error between reference proposed and output of controlled system [6]. the literature presents various optimization methods that can be divided into two distinct groups: i) deterministic and ii) heuristic methods. deterministic methods follow fixed sequence from a defined starting point, returning always the same output value if the starting point is maintained. heuristic methods seek optimal solution promoting stochastic variations from possible solutions sets, this strategy results in unpredictable sequences and return different optimal values each run, even keeping the initial conditions [6]. both optimization methods, deterministic and heuristic, present advantages, being not possible determinate the best method for all cases. in general, heuristic methods present better results when looking for solutions in large search spaces. deterministic methods present better performance in smaller search spaces, close to starting point. aiming to combine features of different methods, hybrid optimization methods are implemented seeking for better solutions that any of individuals methods could present. hybrid optimization methods combines two or more other douglas freire de carvalho, cleber asmar ganzaroli, luiz alberto do couto, rafael nunes hidalgo monteiro dias, wesley pacheco calixto hybrid optimization process applied to tuning of dynamic matrix control: study case with dc motor d optimization methods to solve problems. hybrid methods could either choosing one method depending on the data, or switching between them over the course of algorithm. the hybrid optimization methods combining heuristic and deterministic methods are widely presented in literature. in general, these hybrid methods combines the amplitude and reliability of heuristic methods with accuracy of deterministic methods [9]. ii. methodology a. modeling and simulation of direct current motors the dc motor has mathematical models known in the literature, [8]. generally, models are composed of two parts: the electrical and mechanical. in equating of electrical part, the parameters are resistance ra and inductance la of armature. in equating of mechanical part, we have the moment of inertia j and viscous friction coefficient b. the relation between the two parties is realized through constants of torque kt and back emf kb. the mathematical model of dc motor can be represented, in frequency domain, by the block diagram of fig. 1. figure 1. dc motor block diagram based on diagram shown in fig. 1 and on the multiple systems reduction theory, became possible to obtain two transfer functions. in (1) the speed ω against the armature voltage va and in (2) the speed ω against the load torque tl. 𝜔(𝑠) 𝑉𝑎 (𝑠) = 𝐾𝑡 𝛼𝑠2 + (𝛽 + 𝛾)𝑠 + 𝛿 + 𝐾𝑡 𝐾𝑏 (1) 𝜔(𝑠) 𝑇𝐿 (𝑠) = −𝐿𝑎 ∙ 𝑠 − 𝑅𝑎 𝛼𝑠2 + (𝛽 + 𝛾)𝑠 + 𝛿 + 𝐾𝑡 𝐾𝑏 (2) where: α = la · j, β = la · b, γ = ra · j e δ = ra · b. b. dynamic matrix control dynamic matrix control (dmc) controller enables inclusion of restrictions imposed by system to its basic structure making its application more efficient. the inclusion of restrictions in control law enable the dmc controller to perform actions on the system boundary, really close the restrictions. in systems such as speed control of dc motors, where the breakdown of machine restrictions causes serious injury, the use of dmc controller is highly suggested [5]. dmc controller implements the classic strategy of predictive controllers. classical mpc strategy seeks to select the best possible set of control signals, within predetermined horizon, making explicit use of controlled process model. this application of dmc use the dc motor step response model to predict speed of the dc motor. another important part of classical mpc strategy is the implementation of moving horizon. the use of previous data indicates the current state of the process. in this application the past speed of dc motor and armature voltage are used in controller to increase its accuracy. beyond use of past data, moving horizon strategy limit the application of control signal proposed. this limitation occurs in order to update the data for increased accuracy. repetition of steps mentioned at each sampling instant results in an increase in computational cost and an increase in controller accuracy. moving horizon strategy is completed by updating the data and proposing a new set of control signals at each sampling instant. great care must be taken to enable the controller to perform all operations within the sampling period. fig. 2 illustrates the configuration of mpc algorithm. parts of controller are illustrated, clearing the dmc strategy applied. figure 2. mpc controller block diagram restrictions of peak current ip [a] and nominal armature voltage vn [v] aimed at ensuring proper functioning of dc motor. the non-compliance of these restrictions cause bad engine operation and may cause irreversible damage to the machine. insertion of these restrictions in dmc control law enables the controller to safely operate, close to the limits of machine, providing improved speed control performance. the parameters of dmc controller to be optimized are: prediction horizon r, control horizon l and damping rate of reference signal α c. controllers optimization controllers’ optimization aims, by choosing of values to controller’s optimizable variables, minimize error between output of controlled system and proposed reference [6]. deterministic, heuristic hybrid optimization methods are implemented to dmc controller applied to speed control of dc motor in order to compare them and establish which method provides best results. the best optimization method will provide the controller with less transient period and greater stability in steady state. the deterministic method to be applied will be the quasinewton (qn) method, the heuristic method will be the genetic algorithm (ga). the hybrid method will combine the capacity to cover wide range of values from ga and accuracy of qn [9]. given the fact that optimization methods seek to minimize defined fitness function, having no deep knowledge of system that pretends optimize, such optimization methods may suggest configurations that will bring damage to dc motor. seeking to prevent that optimization methods propose gains to controllers that will bring damage to dc motor some penalties have been imposed to fitness function. these penalties are defined by maximum limits for armature voltage and current to the dc motor to be controlled. the evaluation function to be used by all controllers’ optimizators is the integral of absolute error (iae), calculated basing on error between speed reference and speed developed by dc motor iaeω. taking penalties and evaluation function cited the fitness function is given by: 𝑓(𝑥) = 𝐼𝐴𝐸𝜔 ∙ (𝜅 + 𝜈) (3) where: 𝜅 = 𝐼𝑎𝑚𝑎𝑥 − 𝐼𝑝 and 𝜈 = 𝑉𝑎𝑚𝑎𝑥 − 𝑉𝑁 , to 𝜅 > 0 and 𝜈 > 0 the results given by (3) qualifies the controllers optimized. the best optimization method delivery the process with less error between proposed reference and speed developed by dc motor. the penalties have great influence, taking care of safety operation of dc motor. at end, the best optimization method delivery the process with minor errors, i.e., minor fitness function value. iii. results a. modeling and simulation of direct current motors to perform the simulation, real parameters obtained from a commercial dc motor was used. these parameters are shown in tab. i. table i dc motor parameters j = 0.032000167 kgm 2 la = 0.027089 h b = 0.0022069 nms/rad ra = 6.898 ω kt = 1.073 nm/a kb = 1.073 v/rad/s va = 230.0 v ip = 33.38 a applying the dc motor parameters obtained in (1), becomes possible to verify that the system has two real and distinct poles. then, it is expected that the system response to input of step type is of over-dumped type as illustrated in fig. 3. the fig. 3 present the speed developed by dc motor referred throughout experiment, being the input of the unit step type. the existing variation in instant t = 10 s is caused due to load torque insert with numerical value of 2% of system reference value, being equal to tl = 0.02 n · m for reference of 1.0 rad/s. figure 3. step response of dc motor b. controllers optimization the optimizable parameters of dmc controller, r, l and α, should respect the restrictions of mpc controllers, being 1) prediction horizon r: positive integer value, less than or equal to model horizon n; 2) control horizon l: positive integer value, less than or equal to prediction horizon r and being 3) damping rate reference α: real value between 0 and 1. these restrictions are related to the optimization process, representing constructive restrictions of control technique and being not related to the highlighted restrictions in the system to be controlled. the fitness function presented in (3) is implemented for all optimization methods. taking the parameters of the commercial dc motor presented in tab. i the restrictions are defined as 𝑉𝑁 = 230.0𝑉 to armature voltage and 𝐼𝑝 = 33.38𝐴 to armature current. c. study case 1: deterministic optimization of dmc using the deterministic method, the optimization of dmc controller was realized starting from initial stochastic parameters. fig. 4 present values of speed, armature current and armature voltage developed by dc motor controlled by dmc with parameters values obtained by quasi-newton method. it considered the set-point speed at 100 rad/s and inserted load of 2.0 n · m applied at time t = 10 s. figure 4. dc motor controlled by dmc tuned by deterministic method (ωref =100 rad/s). analyzing fig. 4 is noted higher values for armature voltage and armature current in early moments of the experiment, dc motor starting. is noted also that with the stabilization of speed developed by dc motor in speed reference both the armature voltage and the armature current remain stable. in instant t = 10 s the insertion of load with value 2.0 n causes disturbance to system slowing dc motor speed developed and making it necessary the increase of armature voltage and armature current values to that the speed developed reaches reference speed. again, with stabilization of dc motor speed in speed reference both the armature voltage as the armature current remains stable. fig. 4 yet present the integral of absolute error between reference speed and speed developed by dc motor being iaeω = 26.8101. fig. 5 present values of speed, armature current and armature voltage developed by dc motor controlled by dmc with parameters values obtained by quasi-newton method. it considered the set-point speed at 50 rad/s and inserted load of 1.0 n · m applied at time t = 10 s. figure 5. dc motor controlled by dmc tuned by deterministic method (ωref = 50 rad/s). analyzing fig. 5 is noted similarity to fig. 4. is noted higher values for armature voltage and armature current in early moments of the experiment and stabilization of these values with the stabilization of motor speed. is noted the perturbation of system due to insertion of load with value of 1.0 n at time t = 10 s, similar to the previous experiment, it is evident the increase of armature voltage and armature current values aiming to return and the stabilization of speed developed by dc motor to reference speed. fig. 5 yet present the integral of absolute error between reference speed and speed developed by dc motor being iaeω = 5.9936. tab. ii presents the total error found during the experiments and the optimal values for parameters of dmc controller obtained using the deterministic quasi-newton method. table ii results of dmc controller optimized by quasi-newton method ωref f (x ∗ ) r l α 100 26.81014 15 9 0.00378 50 5.99360 12 10 0.00366 dmc controller tuning by deterministic (quasi-newton) method promote adequate performance of dc motor speed control. note that with different set-point values different parameter are obtained. another observation is that, with lower set-point values smaller fitness function values are obtained; this fact occurs, mainly, because the evaluation parameter is the integral of absolute error of speed iaeω and with the reduction of set-point the error area is reduced in same percentage. d. study case 2: heuristic optimization of dmc genetic algorithm (ga) was implemented with an initial population of 20 individuals. mutation and crossover rates were defined from linear variation where the mutation rate is 30% in the initial generation and 90% in the final generation and the crossover rate of 90% in the initial generation and 30% in the final generation. selection method used was the tournament. the maximum number of generations gmax was set at 100 generations. fig. 6 present values of speed, armature current and armature voltage developed by dc motor controlled by dmc with parameters values obtained by genetic algorithm method. it considered the set-point speed at 100 rad/s and inserted load of 2.0 n · m applied at time t = 10 s. figure 6. speed of dc motor controlled by dmc tuned by heuristic method (ωref = 100 rad/s). analyzing fig. 6 is noted higher values for armature voltage and armature current in early moments of the experiment, dc motor starting. is noted also that with the stabilization of speed developed by dc motor in speed reference both the armature voltage and the armature current remain stable. in instant t = 10 s the insertion of load with value 2.0 n causes disturbance to system slowing dc motor speed developed and making it necessary the increase of armature voltage and armature current values to that the speed developed reaches reference speed. again, with stabilization of dc motor speed in speed reference both the armature voltage as the armature current remains stable. fig. 6 yet present the integral of absolute error between reference speed and speed developed by dc motor being iaeω = 12.2551. fig. 7 present values of speed, armature current and armature voltage developed by dc motor controlled by dmc with parameters values obtained by genetic algorithm method. it considered the set-point speed at 50 rad/s and inserted load of 1.0 n · m applied at time t = 10 s. figure 7. dc motor controlled by dmc tuned by heuristic method (ωref = 50 rad/s). analyzing fig. 7 is noted similarity to fig. 6. is noted higher values for armature voltage and armature current in early moments of the experiment and stabilization of these values with the stabilization of speed. is noted the perturbation of system due to insertion of load with value of 1.0 n at time t = 10 s, similar to the previous experiment, it is evident the increase of armature voltage and armature current values aiming to return and the stabilization of speed developed by dc motor to reference speed. fig. 7 yet present the integral of absolute error between reference speed and speed developed by dc motor being iaeω = 2.7503. with the results obtained using the heuristic method, it is possible to observe the improvement in curves presented in relation to deterministic method. tab. iii presents the total error found during the experiments and the optimal values for parameters of dmc controller obtained using the heuristic genetic algorithm method. table iii optimization of gains using the heuristic method. ωref f (x ∗ ) r l α 100 12.25510 9 5 0.00321 50 2.75025 7 7 0.00299 the optimization using the heuristic method allowed the improvement of the performance of dmc controller in response time and also in the annulment of the error in permanent regime. fitness function values were reduced significantly in relation to values presented by implementation of deterministic method. reduction of fitness function values represent better control of speed developed by dc motor. e. study case 3: hybrid optimization of dmc the implemented hybrid optimization initiates the search for optimized values realizing wide search within the set of possible solutions. to develop wide search hybrid optimization implements genetic algorithm. ga implemented at beginning of hybrid optimization repeats the characteristics of ga implemented isolated presented in iii.d. after wide search performed the hybrid optimization implemented seeks greater precision for solution presented until then. the hybrid optimization implements quasi-newton algorithm for refinement of the solution presented in the first step. the qn algorithm implemented in the hybrid optimization presents repeats the characteristics of qn implemented isolated presented in iii.c. fig. 8 present values of speed, armature current and armature voltage developed by dc motor controlled by dmc with parameters values obtained by hybrid method. it considered the set-point speed at 100 rad/s and inserted load of 2.0 n · m applied at time t = 10 s. figure 8. dc motor controlled by dmc tuned by hybrid method (ωref = 100 rad/s). operation of dc motor controlled by dmc tuned by hybrid optimization method, fig. 8, is similar to operation presented with optimization by heuristic method, fig. 6. in speed developed by dc motor could be noted reduction of overshoot before speed stabilization. note also, reduction in speed decrease caused by insertion of load with value 2.0 n at t = 10s. characteristics evidenced in fig. 8 illustrate reduction in integral of absolute error between reference speed and speed developed by dc motor, resulting in iaew = 10.8909. fig. 9 present values of speed, armature current and armature voltage developed by dc motor controlled by dmc with parameters values obtained by hybrid method. it considered the set-point speed at 50 rad/s and inserted load of 1.0 n · m applied at time t = 10 s. figure 9. dc motor controlled by dmc tuned by hybrid method (ωref = 50 rad/s). fig. 9 presenting operation of dc motor controlled by dmc tune by hybrid optimization method details operation similar to fig. 8. could be noted that even with a reduction of set-point the armature voltage and current still near of safety operation points at dc motor startup. direct relation of armature voltage and current with speed developed by dc motor request high values of these parameters at dc motor startup. high armature voltage and current values at dc motor startup quickly drive speed to desired reference. short period for high values of armature current and voltage guarantees absence of overshoots at speed developed by dc motor. fig. 9 yet present the integral of absolute error between reference speed and speed developed by dc motor being iaeω = 2.031. with the results obtained using the hybrid method, it is possible to observe the improvement in curves presented in relation to heuristic method. tab. iv presents the total error found during the experiments and the optimal values for parameters of dmc controller obtained using the heuristic genetic algorithm method. table iv optimization of gains using the heuristic method. ωref f (x ∗ ) r l α 100 10.89239 7 2 0.00191 50 2.03181 7 1 0.00170 reduction presented in fitness function 𝑓(𝑥) values reflects the improvement of dmc controller with parameters tuned by hybrid optimization method in relation of tuning by other optimization methods. the improvement characterize minor error between speed reference and speed developed by dc motor. f. comparison between optimization methods tab. v presents final values observed after execution of experiments where dc motor speed control was carried out by dmc controller tuned by deterministic method (quasinewton), by heuristic method (genetic algorithm) and by hybrid method (genetic algorithm/quasi-newton). table v dc motor parameters final value optimizator set-point iaeω iamax vamax qn 100 26.8101 22.2045 204.8631 50 5.9936 14.5460 155.5752 ga 100 12.2551 29.4314 227.0795 50 2.7503 27.8424 224.4494 hybrid 100 10.8923 32.6256 229.2125 50 2.5033 29.7023 226.2568 in tab. v iaeω values present the integral of absolute error between speed reference and speed developed by dc motor controlled by dmc tuned by different optimization methods. for smaller values of iaeω smaller are errors occurred in the experiment consequently better is the proposed dmc parameters and better is the optimization method for the analyzed system. for a speed set-point equal to 100 rad/s the controller tuned by hybrid method presents the lowest value of iaeω, characterizing itself as the best method for this system. the value of iaeω presented by the system optimized by hybrid method is 11.12% smaller than the value presented when implemented heuristic method and 56.37% smaller when implemented deterministic method. similarly, for set-point equal to 50rad/s, the implementation of hybrid method presents better performance. the value of iaeω presented by the system optimized by hybrid method is 8.98% smaller than the value presented when implemented heuristic method and 58.23% smaller when implemented deterministic method. yet analyzing the data presented in tab. v, for all the experiments, using parameters obtained from simulations, the armature voltage limits va = vn = 230.0v and armature current ia = ip = 33.38a were respected. in this article, the set of gains r, l and α obtained through system optimization with speed set-point equal to 50rad/s cannot be implemented when it is intended to operate the same plant with speed set-point equal to 100rad/s. to this operating characterizes, the dc motor restrictions, as peak current ip = 33.38 and armature voltage vn = 230.0, are not respected, as shown in the figure below. figure 10. dc motor speed for system controlled by dmc tuned by ag for set-point ωref = 50 rad/s applied for set-point ωref = 100 rad/s. yet analyzing fig. 10, evidence the impossibility of applying the parameters obtained by optimization methods for the dmc controller with reference equal to 50 rad/s on the same controller with the reference equal to 100 rad/s. it note peak current applied to the dc motor exceeding the threshold value, reaching iamax = 55.6820a and armature voltage applied to the dc motor exceeding the limit value, reaching vamax = 448.8387v. iv. conclusion held up the implementation of predictive dmc controller to speed control of dc motor, being employed to tune the controller, optimization techniques deterministic (quasinewton), heuristic (genetic algorithm) and hybrid (genetic algorithm/quasi-newton). the proposed optimized controllers were simulated for the same dc motor speed control in order to compare which optimization method obtain the most efficient controller, searching for reduction of the transient period and variations in continuous operation. to analyze the efficiency of the control developed was used as main criterion the integral of absolute error of speed, presenting the existing error between the speed developed by dc motor and reference speed. analysis of results shows a better performance of dmc controller optimized with hybrid method. attentive to the fact that optimization for a given operating point does not guarantee safe operation of the controller in all parts of the system. note that the tuning of the controller set-point of 50 rad/s afford gains that would outweigh the engine safety restrictions if the same gains are implemented in the same dmc controller, but seeking to reach set-point of 100 rad/s. finally, conclude that the implementation of dmc controller combined with the optimization parameters through the heuristic optimizer using genetic algorithm results in approach that shows promising results, enabling optimized dmc controller to be implemented in systems where search is control with high performance. acknowledgment authors thank the national counsel of technological and scientific development (cnpq), the research support foundation for the state of goiás (fapeg) and the coordination for the improvement of higher education personnel (capes) for financial assistance to this research. references [1] w. goncalves da silva, “speed control of electric drives in the presence of load disturbances,” ph.d. dissertation, university of newcastle upon tyne, 1999. [2] n. nise, control systems engineering. john wiley & sons, 2007. [3] c. r. cutler and b. l. ramaker, “dynamic matrix control?? a computer control algorithm,” in joint automatic control conference, no. 17, 1980, p. 72. [4] s. j. qin and t. a. badgwell, “a survey of industrial model predictive control technology,” control engineering practice, vol. 11, no. 7, pp. 733– 764, 2003. [5] e. f. camacho and c. b. alba, model predictive control. springer science & business media, 2013. [6] w. p. calixto, t. m. pereira, j. mota, a. alves, e. domingues, j. domingos, a. coimbra, and b. alvarenga, “development of mathematical operator to heuristic optimization algorithms applied to problem of geoprospection, in portuguese,” tema (são carlos), vol. 15, no. 2, pp. 177–194, 2014. [7] c. a. e. a. ganzaroli, “heuristic and deterministic strategies applied on a cascade pi controller tuning for speed control of a dc motor,” ieee control systems magazine, 2015. [8] s. chapman, electric machinery fundamentals. tata mcgraw-hill education, 2005. [9] vasant, pandian m., meta-heuristics optimization algorithms in engineering, business, economics, and finance, igi global, 2012. i. introduction ii. methodology a. modeling and simulation of direct current motors b. dynamic matrix control c. controllers optimization iii. results a. modeling and simulation of direct current motors b. controllers optimization c. study case 1: deterministic optimization of dmc d. study case 2: heuristic optimization of dmc e. study case 3: hybrid optimization of dmc f. comparison between optimization methods iv. conclusion acknowledgment references transactions on environment and electrical engineering issn 2450-5730 vol 1, no 4 (2016) © urvi n. patel and hiren h. patel  abstract— in many countries, the grid-code or standards do not allow the photovoltaic (pv) inverters to exchange reactive power with the grid. recently, some countries have relaxed the standards. hence, capacity of the inverters to control reactive power must be utilized. however, the reactive power that a pv inverter can supply is constrained by the maximum power that a pv array generates and changes with the environmental conditions. a reactive power sharing algorithm is proposed that not only ensures proper distribution of reactive power amongst the inverters, but also ensures that the maximum power generated by pv is supplied to the grid. in case of identical pv inverters, the algorithm operates all inverters at nearly equal apparent power leading to nearly equal percentage utilization of the inverters, thereby achieving uniform heating of the similar devices of the inverters. the algorithms are further investigated for power sharing amongst pv inverters of unequal ratings. it is highlighted that the proposed algorithm results into the least change in the utilization factor of a pv inverter, whose power changes due to the change in environmental conditions. the effectiveness of the algorithm over other algorithms in sharing power amongst inverters is displayed through matlab/simulink simulations. keywords — photovoltaic, reactive power, power sharing. i. introduction last couple of decades have experienced significant rise in the electricity generation from non-conventional energy sources like wind and solar. it is attributed mainly by the increased environmental concern, fast depletion of conventional energy sources, increase in cost of conventional energy sources, and decrease in the cost of renewable based energy generation. in recent years, one of the renewable sources that has seen the fastest growth and penetration in the electrical grid is the solar photovoltaic (pv). the reason for the increase in penetration is the reduced cost of pv system and the encouraging feed-in-tariff policies by the governments. however, increased penetration of pv sources has also given rise to several challenges. the challenges are mainly due to the dependence of pv source’s performance on the environment, which makes it intermittent and uncertain in nature. pv source is connected to the grid through the static power converters [1]. thus, it is inertia-less source of energy unlike the conventional rotational generators. hence, if the energy generation in the grid is highly dominated by inertia-less pv (i.e. in a weak grid), the sudden change in output power of pv resulting from the sudden change in irradiation, may affect the stability of grid and the systems connected with the utility. also, if the power electronic u. n. patel is with department of electrical engineering, c. k. pithawala college of engineering and technology, surat, india. (email: urvi.patel@ckpcet.ac.in). converters are not controlled appropriately in such weak electrical grid or a microgrid (mg), they are likely to create issues like harmonic injection, change in voltage levels and power flow, flicker, resonance, mal-operation of protection scheme etc. on the contrary, if the power electronic converters are properly controlled [2]-[3], they can improve the voltage profile and performance of the mg. this can be achieved if pv systems, which are usually commissioned to supply active power, are allowed to inject desired reactive power into the grid. pv systems are usually designed with reasonable margins, and most of the times operate under lightly loaded conditions (in fact inactive at night time). thus, there is a room for reactive power injection to keep the voltage at a desirable level. this objective, along with the transfer of maximum power generated by pv, can be achieved by controlling the amplitude and phase angle of the output voltage of the inverter. the task becomes challenging when several such pv based distributed energy generators are operating in a mg, which even comprises of other types of renewable energy sources. pv inverters are commonly controlled as current controlled source using p-q control strategy to exchange active and reactive p and q respectively, with the microgrid [4]. in islanding mode i.e. when main grid is disconnected, the voltage v and frequency ω are controlled, using p-ω and q-v droop control methods to share active and reactive power amongst the distributed generators (dg) [5]-[7]. battery storage is essential in such system when islanded, in order to maintain power balance in the system. lasseter et al., have presented flexible control and proper coordination amongst dg sources to overcome some problems associated with pv and other non-conventional sources operating simultaneously in a mg [3]. local power management system for coordination of various dg sources to manage active and reactive power successfully is addressed in [8]-[10]. in [8], fundamental algorithm employing hierarchical droop control of power management is presented, where inverter control is considered as primary control whereas microgrid central controller (mgcc) is under secondary control. secondary control focuses on power management and optimization algorithm to optimize performance of mg. power management system plays very important role when mg is having many pv connected inverters, as rapidly varying irradiation condition may cause voltage sags and swells that result in degradation of power quality [16]-[20]. to regulate voltage under such transient condition, pv inverters must have h. h. patel is with the department of electrical engineering, sarvajanik college of engineering & technology, surat, india. (e-mail: hiren.patel@scet.ac.in). power sharing strategy for photovoltaic based distributed generators operating in parallel urvi n. patel and hiren h. patel the capability to match-up the var requirement quickly [11]. as active power delivered by inverter depends on maximum power that pv can generate under given (environmental) conditions, it is necessary to allocate reactive power amongst inverters in a proper way to have uniform loading of the inverters and to also avoid over loading of inverters [12]. an accurate reactive power sharing control that shares reactive power equally amongst inverters is presented [13]. total reactive power of the system is calculated by mgcc and the information is passed to all inverters through communication link. though this method shares reactive power accurately amongst the inverters, in case when active power varies with the change in irradiation, it fails to accurately share the reactive power amongst the pv inverters. it may also cause inverter to work beyond its nominal apparent power transfer capability. in [14], reactive power algorithm is presented which takes into account apparent power limit of each pv connected inverter as well as active power delivered by each pv inverter. optimal reactive power strategy [15] assigns reactive power to each inverter such that entire system can achieve maximum reactive power transfer capability. however, these algorithms are unable to uniformly utilize apparent power capability of each inverter. the paper proposes an approach to overcome these drawbacks. the proposed reactive power algorithm first determines the active power that pv inverters are supplying under given conditions and based on the available margin it assigns the reactive powers to the inverters. section ii introduces system configuration and control scheme employed for operating pv inverters while the secondary control algorithm implemented in mgcc for accurate reactive power sharing is presented in section iii. the results of the simulation study performed in matlab/simulink are included in section iv to demonstrate the performance of algorithm for pv inverters operating in parallel for two different cases: (i) all inverters with equal ratings and (ii) inverters with unequal ratings. ii. system description and control fig.1 shows the system configuration considered for evaluation of the proposed algorithm. the microgrid comprises of four identical distributed energy generators that along with the main grid (or a relatively stiff source) supply the local loads. each dg unit consists of pv as a primary energy source, a three phase inverter and an lc filter. the inverters not only extract the maximum power from the pv but also supply sinusoidal current to the load and grid. pvi shown in fig. 1 represents a pv array with its dc-dc converter operated with maximum power point tracking control. the dgs are connected to the pcc through a transformer, which for the sake of simplicity, is not shown in fig.1. static switch a.c grid pcc s 1 = p 1 + jq 1 l oa d z04z02 z03z01 s 2 = p 2 + jq 2 s 3 = p 3 + jq 3 s 4 = p 4 + jq 4 mgcc pv2 pv3 pv4 pv1 fig.1. system configuration of a microgrid having four dgs the impedances zoi, where ‘i’ represents i th dg, takes into account the impedance of interfacing inductor, the impedance of cable and isolation transformer. active and reactive power management task is performed by mgcc unit using low bandwidth communication links. microgrid hierarchical structure consists of mainly primary, secondary and tertiary control [10], [11]. primary control covers inverters’ control present in microgrid whereas secondary control consists of mgcc unit. tertiary control provides interaction between multiple microgrid and utility grid. primary and secondary controls are used in this paper while tertiary control is not required at this stage. inverter control is achieved by active-reactive power (p-q) control method [4]. p-q method is used to operate inverter as a controlled current source for desired active and reactive power transfer with grid. inverter output current is tightly regulated by inner current control loop. reference currents for current control loop are provided by outer power control loop according to power references provided by mgcc. phase locked loop (pll) used for grid synchronization provides desired angle (ρ) for abc to dq frame transformation. fig.2 shows control circuit diagram for one of the inverters. fig.3 shows the details of the power and current control loops shown in fig. 2. the voltage vdc, across capacitor c is maintained at a desired voltage, vdcref by a voltage control loop. pv grid c inv-1 r l cf ss pwm gate drive current control loop abc/ dq abc/ dq ρ ρρ idq ρ ω vdq prefqref pcc lg+ vdc iabc vsabc vdcref power control loop idref iqref vdc p pll dc bus voltage control ppv q ss=static switch fig. 2. control scheme of pv inverter _ / pi _ pi / lω0 pi +_ lω0 + + + + +_ +_pi+ + + dc-bus voltage control power control current control loop + + vdcref vdc ppv pref qref p q idref iqref id iq vd vq md mq pivdc 2 fig.3. active-reactive power control to maintain this voltage constant it is ensured that the power obtained from pv array, ppv is entirely transferred to the grid side. this is done through the power control loop, which compares actual dg output power (p) with reference power (pref). the reactive power reference (qref) is obtained using the algorithm presented in the next section. pref and qref are used to generate required current references idref and iqref for the current control loop. the direct and quadrature axes components of the inverter output currents id and iq, respectively, are obtained through d-q transformation. the current control loop finally determines the direct and quadrature components of the reference waveform from the direct and quadrature axes modulation indices, md and mq, respectively. iii. proposed reactive power sharing algorithm as the active power that pv inverters supply is directly dependent on the environmental conditions (mainly irradiation), most of the times the inverters do not operate at their rating and hence, their capacity is not utilized fully. the available margin varies with the irradiation, with maximum at night or when irradiation is the least. the reactive power sharing algorithm shown in fig. 4 relies on assigning the reactive power algorithm amongst the inverters based on the margin available with each of them. the algorithm starts with initializing the number of inverters (m) and the apparent ratings of the inverters (sin), where ‘i’ stands for ith inverter. the output power of the pv systems (pi) is obtained from the maximum power point tracker (mppt), which ensures that the pv system operates at its maximum (active) power point. as the apparent power ratings (sin) of the inverters are known and as the inverter must be operated to deliver active power (pi) to the grid side, the available reactive power (qi) is expressed as 22 iini psq  (1) the inverter is capable of supplying and drawing reactive power and it must match the load and grid requirements. accordingly (2) and (3), assigns the reactive power limits for lagging and leading type of reactive demand, respectively. ii qq max (2) ii qq max (3) hence, at a given instant, the total active power (pt), reactive power (qt) and apparent power (st) capabilities that the inverters possess to match the reactive power demand of load and to supply the active power of pv systems to grid are represented by (4), (5) and (6), respectively.    m i it pp 1 (4)    m i it qq 1 (5) 22 ttt qps  (6) if output currents of all the inverters are equal, temperature of similar devices of the different inverters can be made equal. this can be realized if all the inverters operate with the same apparent power. hence, the inverters are made to operate with the reference apparent power (stnew) to have uniform utilization and heating. ])1[( imss ttnew  (7) the algorithm evaluates the condition expressed by (8), and if stnew exceeds sin, the reference apparent and reactive powers are set to values sin and qimax (or qimin), respectively. intnew ss  (8) the algorithm then assigns the reference reactive power qiref and pi for each inverter, where the active power references (pi) for the inverters are obtained from the mppt. once any inverter is assigned the reference active and reactive powers, the total unassigned active and reactive powers to be supplied by the remaining inverters are updated by subtracting the qiref and pi assigned to the earlier inverters from pt and qd, where qd is the reactive power demand of the load. the remaining active power (ptn) to be supplied and reactive power demand to be met (qtn) is calculated as shown in (9), and (10), respectively.     1 0 i i ittn ppp where 0 0 p (9)     1 0 i i irefdtn qqq where 00 refq (10) accordingly, the apparent power (si) that i th inverter must supply is obtained by (11)  imqps tntni  )1(/ 22 (11) hence, the reference reactive power for the ith inverter is 22 iiiref psq  (12) iv. simulation results to demonstrate the effectiveness of the above control strategy, microgrid system shown in fig.1 is simulated in matlab/simulink. in addition to the proposed control algorithm, two more control approaches: (optimal reactive power [15] and equal reactive power sharing [13]) are also evaluated and the results are compared with that obtained with the proposed control algorithm. two different cases are considered for comparing the performance of this algorithm. calculate pt & qt using(4) and (5) respectively is i=i+1 start yes 22 ttt qps  ?intnew ss  initialize sin=nom. apparent power m=no. of pv inverters 22 pisq ini  i=1 is i=m ? no yes b ininew ss  iiref qq  no ])1[( imss ttnew  measure pi=active power of i th inverter 00 p 00 refq measure qd=reactive power demand calculate ptn ,qtn using (9) and (10) calculate si using (11) 22 iiiref psq  i=i+1 is i=m ? end a a b c c no yes i=1 fig. 4. proposed reactive power sharing algorithm in case (i), all the inverters are considered to have the equal ratings while in case (ii), inverters of unequal ratings are considered. case (i): equal dg ratings the parameters considered for evaluating the performance of the algorithms using the system of fig. 1 are mentioned in table i. as shown, all dgs are considered to have the equal nominal apparent power rating of 500 kva. table i ratings and parameters for the system of fig.1 nominal power rating of dg1 (s1n) 500 kva nominal power rating of dg2 (s2n) 500 kva nominal power rating of dg3 (s3n) 500 kva nominal power rating of dg4 (s4n) 500 kva grid voltage(vg), frequency(f) 415v, 50 hz line parameter (z01=z02=z03=z04) l=100µh,r=2.07mω,cf=2500µf load 1.92 mva, 0.78 power factor (lag) no of pv inverters (m) 4 fig. 5 shows the results with the optimal reactive power sharing (orps) algorithm. reactive power references for inverters shown in table ii are calculated using the orps algorithm of [15], while the active power references for the inverters are set at the value equal to the maximum power that the corresponding pv system generates at a given instant. the active power generated by pv arrays pv1, pv2, pv3 and pv4 till t=0.5s are 400kw, 300kw, 250kw and 450kw, respectively. a step change in irradiation on pv array pv1 occurs at t=0.5s, which results in the output of pv1 to decrease to 200kw. at t=1s, step change in irradiation on pv array pv3 occurs, resulting into the change in the output power from 250kw to 400kw. reactive power references for the inverters obtained with optimum reactive power control are mentioned in table ii. fig. 5, shows active, reactive and apparent power of inverters 1 through 4. (a) (b) (c) fig. 5. results with orps algorithm: (a) active power fed by dgs, (b) reactive power shared by the inverters, (c) apparent power of each inverter. table ii utilization factor of each dg for orps algorithm time interval (s) pi (kw) qiref (kvar) si (kva) uti. fac. si/sni t=0-0.5 400 300 500 1.00 300 373 478 0.95 250 310 398 0.79 450 218 500 1.00 t=0.5-1 200 262 329 0.65 300 392 493 0.98 250 325 410 0.82 450 218 500 1.00 t=1-2 200 282 345 0.69 300 400 500 1.00 400 300 500 1.00 450 218 500 1.00 0 0.5 1 1.5 2 0 100 200 300 400 500 p ( k w ) v p p p p 2 1 3 4 0.5 1 1.5 2 0 100 200 300 400 500 q ( k v a r ) v q q q q 4 3 2 1 0.5 1 1.5 2 200 300 400 500 600 time (s) s ( k v a ) v s s s s 1 3 4 2 1 it is observed from figs. 5(a) and (b) that, when p1 is decreased from 400kw to 200kw at t=0.5s, q1 changes from 300kvar to 262kvar. not only q1, but q2 through q4 also changes. similarly at t=1s, when p3 increases to 400kw, q1 through q3 changes. thus, if power generated by any one of the pv array changes, the reactive power references and hence, the reactive power supplied by all the inverters change (except those which are operating at their limits sin). fig. 5(c) shows that inverters 1 and 4 operate at their maximum apparent power limits (s1n and s4n, respectively) till t=0.5s. at t=0.5s, when p1 reduces, q1 also reduces simultaneously and hence, from t=0.5s to t=0.1s only inverter 4 operates at its full capacity. it is observed that the change in pi and qi is such that the ratio pi/qi remains equal for all the inverters that do not reach the rated capacity. an index defined as utilization factor (si/sin) is used to indicate the extent to which the capacity of the inverter is utilized. it is also observed from the table ii that all the inverters are operating at different utilization factors. the utilization factors vary greatly showing that some of the inverters operate much below their rated capacity when some others have already hit their limits. for example, inverter-1 operates with the lowest utilization factor (0.65 from t=0.5s till 1s and 0.69 from t=1s till 2s) while inverter-4 is operating at its limit. the unequal utilization of the inverters, not only results into unequal losses, efficiency and heating of different inverters, but may damage the inverters that continuously operate at their apparent power limits. fig. 6 shows the results obtained with equal reactive power sharing (erps) algorithm [13], according to which reactive power demand is equally shared amongst the inverters. the irradiation pattern on the pv array is considered the same as that considered for the evaluation of orps approach. fig. 6 shows active, reactive and apparent powers respectively, of inverters 1 through 4. if it is intended to meet the total reactive power demand of the load mentioned in table i (1200kvar) through the inverters 1 through 4 using erps control, each inverter must output 300kvar. hence, the reference reactive power for inverter 1,2 and 3 are set equal to 300kvar (reactive power demand of load = 1200kvar) while for inverter4 which hits its apparent power limit, it is restricted to 218 kvar. it is observed from figs. 6(a)-(c), and table iii that, even if the active power supplied by the pv array changes, the effect is not observed in the reactive power sharing. it is also evident from fig. 6(c) that inverter-4 continuously operates at its rated capacity of 500kva. inverters 1 and 3 also operate at their rated capacities for some time. it is also observed that si (for i=1, 2 and 4) remains almost constant for t=0.5s to 2s inspite of the change in p3 at t=1s. the reason being no change in pi and qi (for i =1, 2 and 4) for this period. unlike orps the reactive power demand of the load is not met fully inspite of the fact that many inverters still operate below their rated limits. thus, the inverters are not utilized optimally and also the percentage utilization of all the inverters varies greatly. fig.7 shows performance with proposed algorithm when same pattern of irradiation on the pv array as that considered for orps and erps is maintained. at t=0.5s, when the irradiation of pv1 decreases resulting into the decrease in the (a) (b) (c) fig. 6. results with erps algorithm: (a) active power fed by dgs, (b) reactive power shared by the inverters, (c) apparent power of each inverter (a) (b) (c) fig. 7. results with proposed algorithm: (a) active power fed by dgs, (b) reactive power shared by the inverters, (c) apparent power of each inverter 0 0.5 1 1.5 2 0 100 200 300 400 500 p ( k w ) v p p p p 2 1 3 4 0.5 1 1.5 2 0 100 200 300 400 q ( k v a r ) v q q q q 1 2 4 3 0 0.5 1 1.5 2 200 300 400 500 600 time (s) s ( k v a ) v s s s s 1 2 3 4 0 0.5 1 1.5 2 0 100 200 300 400 500 p ( k w ) v p p p p 2 1 3 4 0.5 1 1.5 2 0 100 200 300 400 500 q ( k v a r ) v q q q q 1 4 3 2 0.5 1 1.5 2 300 400 500 600 time (s) s ( k v a ) v s s s s 1 2 3 4 output power of inverter 1, the reactive power of inverter 1 increases. simultaneously, the reactive powers of all other inverters decrease in spite of the fact that there is no change in the power output from pv arrays pv2, pv3 and pv4. this results into minimizing the gap of percentage utilization of different inverters. similarly, at t=1s when p3 changes from 250kw to 400kw, reactive power of all the inverters changes to achieve better sharing of the active and reactive power amongst them. table iv shows the active, reactive and apparent powers shared by the inverters over the different periods. unlike orps and erps, the utilization factors vary little for all the dgs indicating uniform loading of the inverters. the three algorithms are tested even with a different load having a leading power factor (pf). table v shows the results obtained with a load of 1.16 mva, 0.86 power factor (lead). it is observed that even with leading power factor, proposed algorithm performance is superior. standard deviations of the utilization factors of the various inverters are calculated, to quantify the effectiveness of the algorithm to distribute the apparent power equally amongst the inverters. standard deviations of the utilization factors for the three schemes for the case represented by table v are 0.204, 0.147 and 0.055. the least the standard deviation better is the performance. case (ii): unequal dg ratings the three algorithms are also evaluated for the case when all dgs of the system shown in fig. 1 have unequal ratings. the nominal ratings for the dgs are mentioned in table vi. the load, line parameters, capacitance c and the grid voltage are considered same as that of case (i). in this case the active power generated by pv arrays pv1, pv2, pv3 and pv4 are 200kw, 300kw, 400kw and 500kw, respectively. a step increase in irradiation on pv array pv1 occurs at t=0.5s, which results in the output of pv1 to increase to 300kw. at t=1s, irradiation on pv array pv3 decreases suddenly, resulting into the change in its output power from 400kw to 200kw. the active, reactive and apparent power sharing by inverters 1 through 4 with orps control are displayed in fig. 8 and the results are quantified in table vii. figs. 8(a) and (b) shows that when pv1 is increased from 200kw to 300kw at t=0.5s, q1 also increases from 171kvar to 240kvar. hence its apparent power increases, leading to its utilization factor of 0.96. the reactive powers of inverters 2 through 4 decrease with their active powers still at the same values. thus, s2 through s4 decrease lowering the utilization of inverters 2 through 4. this increases the miss-match in the utilization factors. the miss-match further increases after t=1s, when the output power of pv3 decreases from 400kw to 200kw. the decrease in p3 at t=1s is associated with the simultaneous decrease in q3. hence, to meet the reactive power demand of the load, more reactive power needs to be supplied by inverters 1, 2 and 4. hence, while the utilization factor of inverter-3 decreases, utilization factor of other inverter increases .thus, inverter-3 is the least utilized inverter with utilization factor of 0.45 while inverter-1 is fully utilized with the utilization factor of 1.00. fig. 5(c) also highlights that after t=1s, inverter-1 operates at its apparent power limit (s1n). it is observed from table vii that the percentage change (decrease) in utilization factor of inverter-3 in response to the decrease in output power p3 of inverter-3 is -47%. fig. 9 shows the power shared by dgs (having ratings mentioned in table vi) when operated with erps algorithm. the same shading pattern, adopted earlier for orps of case (ii), is considered. the reference reactive power for all the inverters is set equal to 300kvar to meet the load’s reactive power demand (table viii). fig. 9(c) shows that after t=0.5s, inverter-1 continuously operates at its rated capacity 400kva and hence, is unable to meet its desired reactive power share of 300kvar. like earlier case with erps control, the reactive table iii utilization factor of each dg for erps algorithm time interval(s) pi (kw) qiref (kvar) si (kva) uti. fac. si/sni t=0-0.5 400 300 500 1.00 300 300 424 0.84 250 300 390 0.78 450 218 500 1.00 t=0.5-1 200 300 360 0.72 300 300 424 0.84 250 300 390 0.78 450 218 500 1.00 t=1-2 200 300 360 0.72 300 300 424 0.84 400 300 500 1.00 450 218 500 1.00 table iv utilization factor of each dg for proposed algorithm time interval(s) pi (kw) qiref (kvar) si (kva) uti. fac. si/sni t=0-0.5 400 229 460 0.92 300 354 464 0.92 250 393 466 0.93 450 222 500 1.00 t=0.5-1 200 374 424 0.85 300 311 432 0.86 250 355 434 0.86 450 159 477 0.95 t=1-2 200 405 452 0.90 300 356 465 0.93 400 262 478 0.95 450 175 482 0.96 table v comparison of the various algorithm for leading pf load algorithms pi (kw) qiref (kvar) si (kva) uti. fac. si/sni orps 300 -180 350 0.70 200 -120 233 0.46 150 -90 175 0.35 350 -210 408 0.82 erps 300 -150 335 0.67 200 -150 250 0.50 150 -150 212 0.42 350 -150 380 0.76 proposed 300 0 300 0.60 200 -233 307 0.61 150 -271 309 0.61 350 -96 362 0.72 table vi ratings of dg of the system of fig.1 for case (ii) nominal power rating of dg1 (s1n) 400 kva nominal power rating of dg2 (s2n) 500 kva nominal power rating of dg3 (s3n) 600 kva nominal power rating of dg4 (s4n) 700 kva power demand of the load is once again not met fully. thus, the inverters are not utilized optimally. significant variation in utilization factors is observed. also the percentage change in the utilization factor of inverter-3 due to change in p3 at t=1s is -27.7%. the power sharing, the utilization factors and the variation in the utilization factors are highly dependent on the nominal ratings of the inverters and the load. fig. 10 shows performance of proposed algorithm with same pattern of irradiation on the pv array as considered earlier for erps and orps algorithm of case (ii). it is observed from table ix that during t=0s to t=0.5s, the proposed algorithm tries to share the apparent power equally amongst all the inverters. hence, as the inverter-1 reaches its limit, it is operated at 400kva (100% capacity), while inverters 2, 3 and 4 are operated around 500kva demonstrating the tendency of equalizing the reactive power sharing. at t=0.5s, when the irradiation of pv1 increases resulting into the increase in the output power of inverter 1, the output reactive power of inverter 1 decreases. simultaneously the reactive powers of all other inverters increase. table viii utilization factor of each dg for erps algorithm time interval(s) pi (kw) qiref (kvar) si (kva) uti. fac. si/sni t=0-0.5 200 300 360 0.90 300 300 424 0.84 400 300 500 0.83 500 300 583 0.83 t=0.5-1 300 265 400 1.00 300 300 424 0.84 400 300 500 0.83 500 300 547 0.78 t=1-2 200 265 400 1.00 300 300 424 0.84 200 300 360 0.60 500 300 547 0.78 (a) (b) (c) fig.9. results with erps algorithm (a) active power fed by dgs, (b) reactive power shared by the inverters, (c)apparent power of each inverter this results into minimizing the miss-match in the reactive powers of the inverters and hence, reduces the gap of percentage utilization of different inverters. thus, the algorithm inherently has the feature of minimizing the mismatch. but still the mismatch is relatively large. this is due to the equal apparent power sharing principle of the algorithm, which inspite of the unequal nominal kva rating of the inverters, tries to allocate the apparent power equally amongst the dg inverters. hence, it results into the unequal utilization factor of the dgs. at t=1s when p3 changes from 400 kw to 200kw, q3 increases and q4 and q2 decrease to achieve better power sharing amongst the inverters. the least utilization factor of 0.6 is observed for inverter-3. it is observed from table ix that percentage decrease in utilization factor for inverter-3 (due to 0.5 1 1.5 2 0 100 200 300 400 500 600 p ( k w ) v p p p p 4 3 2 1 0.5 1 1.5 2 100 200 300 400 q ( k v a r ) v q q q q 2 3 4 1 0.5 1 1.5 2 0 200 400 600 800 time (s) s ( k v a ) v s s s s 1 2 3 4 (a) (b) (c) fig. 8. results with orps algorithm: (a) active power fed by dgs, (b) reactive power shared by the inverters, (c) apparent power of each inverter table vii utilization factor of each dg for orps algorithm time interval(s) pi (kw) qiref (kvar) si (kva) uti. fac. si/sni t=0-0.5 200 171 263 0.65 300 257 395 0.79 400 342 526 0.87 500 425 656 0.93 t=0.5-1 300 240 384 0.96 300 240 384 0.76 400 320 512 0.85 500 400 640 0.91 t=1-2 300 276 407 1.00 300 277 408 0.81 200 185 272 0.45 500 462 680 0.97 0.5 1 1.5 2 0 100 200 300 400 500 600 p ( k w ) v p p p p 4 3 2 1 0.5 1 1.5 2 0 100 200 300 400 500 q ( k v a r ) v q q q q 4 3 2 1 0.5 1 1.5 2 200 400 600 800 time (s) s ( k v a ) v s s s s 2 4 1 3 change in p3 at t=1s) is -13.8%, which is relatively smaller than that observed with orps (-47%) and erps (-27.7%). (a) (b) (c) fig.10. results with proposed algorithm (a)active power fed by dgs, (b)) reactive power shared by the inverters, (c)apparent power of each inverter table ix utilization factor of each dg for proposed algorithm time interval(s) pi (kw) qiref (kvar) si (kva) uti. fac. si/sni t=0-0.5 200 346 399 0.99 300 388 500 0.98 400 310 500 0.83 500 154 520 0.74 t=0.5-1 300 265 399 0.99 300 400 500 0.99 400 338 523 0.87 500 197 537 0.76 t=1-2 300 265 399 0.99 300 344 450 0.90 200 412 450 0.75 500 179 500 0.75 v. conclusion in case of renewable energy source (pv or wind) based dg, the reactive power that it can supply varies as the active power supplied by it changes. the conventional algorithm, which relies on the sharing of equal reactive power amongst the inverters, fails under such case. not only the inverter gets overloaded but also the distribution of the total apparent power amongst the inverters vary greatly leading to uneven percentage utilization of the inverters. the optimal reactive algorithm also suffers from similar drawbacks. it is observed that the proposed algorithm maintains operation of all inverters within their nominal ratings and yet they are able to match the total reactive power demand of the load. as the reactive power assigned to the inverters is linked with the available reactive power capabilities, the inverter that supplies lesser active power is controlled to share a greater amount of reactive power. if the dg inverters have equal kva ratings, then with the proposed algorithm, not only the apparent power sharing is better than other algorithms but the utilization factors of the inverters are also nearly similar. however, as the algorithm tries to share the apparent power equally amongst the inverters, the utilization factors are not the same for the inverters of unequal kva ratings. but the algorithm always operates to minimize the miss-match in the utilization factors. hence, with the proposed approach, comparatively better apparent power sharing is observed leading to reduction in the variation of percentage utilization of the inverters. references [1] y. huang, f. z. peng, j. wang and d. yoo “survey of power conditioning system for pv power generation,” power electronics specialist conference, pesc 2006, pp.1-6, june 2006. [2] y. riffonneau, s. bacha, f. barruel, and s. ploix “optimal power flow management for grid connected pv systems with batteries” ieee trans. sustainable energy, vol. 2, no. 3, pp.309-320 july 2011. [3] r. h. lasseter and p. paigi, “microgrid: a conceptual solution,” presented at the ieee power electron. spec. conf. aachen, germany, 2004. [4] n. eghtedarpour and e. farjah, “power control and management ina hybrid ac/dc microgrid,” ieee trans. smart grid, vol. 5, no. 3,pp. 1494–1505, may 2014. [5] c.t. lee, c.c. chu, and p.t. cheng, “a new droop control method for the autonomous operation of distributed energy resource s interface converters,” ieee trans. power electron., vol. 28, no. 4, pp. 1980–1993, april 2013. [6] q. zhong, “robust droop controller for accurate proportional load sharing among inverters operated in parallel,” ieee trans. ind. electronics, vol. 60, no. 4, pp. 3747-3459, april 2013. [7] i. u. nutkani, p. c. loh, and f. blaabjerg, “droop scheme with considering of operating cost,” ieee trans. power electron., vol. 29, no. 3,pp. 1047–1052, march 2014. [8] a. bidram and a. davoudi, “hierarchical structure of microgrids control system,” ieee trans. smart grid, vol. 3, no. 4, pp. 1963– 1976, december 2012. [9] j. he, and y li” an accurate reactive power sharing control strategy for dg units in a microgrid” 8th international conference on power electronics ecce asia,the shilla jeju, korea. may 30june 3, 2011. [10] a. milczarek, m. malinowski and j. m. guerrero” reactive power management in islanded microgrid—proportional power sharing in hierarchical droop control” ieee trans. smart grid, vol. 6, no. 4, pp. 1631-1638, july 2015. [11] k. turitsyn, petrsˇulc , s. backhaus, and m. chertkov ” options for control of reactive power by distributed photovoltaic generators” proceedings of the ieee, vol. 99, no. 6,pp 10631073, june 2011. [12] f. olivier, p. aristidou, d. ernst, and t. v. cutsem, ”active management of low-voltage networks for mitigating over voltages due to photovoltaic units” ieee trans. smart grid, vol. 7, no. 2, pp. 926-93, march 2016. [13] a. micallef, m. apap, j. m. guerrero, and j. c. vasquez,” reactive power sharing and voltage harmonic distortion compensation of droop controlled single phase islanded microgrids” ieee trans. smart grid, vol. 5, no. 3, pp. 1149-1158, may 2014. [14] s. duan, y. meng, j. xiong, y. kang and j. chen. “parallel operationc techqnique of voltage source inverters in ups,” ieee international conference on power electronics and drive systems, , hong kong, 883 –887, peds’99. [15] z.wang. k.m.,passino, j.wang, ”optimal reactive power allocation in large scale grid connected photovoltaic system”, journal of optimization theory and application, november 2015. [16] a. micallef, m. apap and c. spiteristaines, j. m. guerrero zapata” secondary control for reactive power sharing and voltage amplitude restoration in droop-controlled islanded microgrids ” 3rd ieee international symposium on power electronics for distributed generation systems (pedg) 2012. 0.5 1 1.5 2 0 100 200 300 400 500 600 p ( k w ) v p p p p 4 3 2 1 0.5 1 1.5 2 0 100 200 300 400 500 q ( k v a r ) v q q q q 1 2 4 3 0.5 1 1.5 2 300 400 500 600 time (s) s ( k v a ) v s s s s 1 4 3 2 [17] t. l. vandoorn, b. renders, l. degroote, b. meersman, and l. vandevelde, “controllable harmonic current sharing in islanded microgrids: dg units with programmable resistive behavior toward harmonics, ”ieee trans. power del., vol. 27, no. 3, pp. 1405–1414,july 2012. [18] adhikari and fangxing li, “coordinated v-f and p-q control of solar photovoltaic generators with mppt and battery storage in microgrids” ieee trans. on smart grid, vol. 5, no. 3,pp 12701281,may 2014. [19] h. mahmood,, d. michaelson, and j. jiang, “accurate reactive power sharing in an islanded microgrid using adaptive virtual impedances” ieee tran. power electron.vol. 30, no. 3, pp. 16051617,march 2015. [20] j. w. smith, w. sunderman, r. dugan, brian seal, “smart inverter volt/var control functions for high penetration of pv on distribution systems “power system conference and exposition (pace) 2011.. urvi patel received b.e degree in electrical engineering from the s.v. regional college of engineering and technology (now s.v. national institute of technology), south gujarat university, surat, india, in 2000, and the m.e. degree in electrical engineering in 2009 from the m. s. university, baroda, india she is currently working as an assistant professor in the department of electrical engineering at c. k. pithawalla college of engineering and technology, surat and pursuing ph.d. in electrical engineering. her current research interests include distributed generation, renewable energy and microgrid issues. she is a life member of the indian society for technical education and a member of ieee. hiren patel received the b.e. degree in electrical engineering from the s.v. regional college of engineering and technology (now s.v. national institute of technology), south gujarat university, surat, india, in 1996, and m. tech. in energy systems and phd in electrical engineering degree from the indian institute of technology bombay (iitb), mumbai, india in 2003 and 2009, respectively. he is working as a professor and dean, r&d at the sarvajanik college of engineering and technology, surat. his current research interests include computer aided simulation techniques, distributed generation, and renewable energy, especially energy extraction from photovoltaic arrays. he has to his name several publications in international and national journals and conferences. he is a life member of the indian society for technical education and the institute of engineers. 1 unidirectional protection strategy for multi-terminal hvdc grids ataollah mokhberdoran, nuno silva, helder leite and adriano carvalho abstract—protection issue is identified as the main drawback of emerging multi-terminal hvdc grids. multi-terminal hvdc grid demands fast short circuit fault current interruption. fast dc circuit breakers as a promising solution can be implemented as either bidirectional or unidirectional devices. in addition to less implementation cost, the unidirectional dc circuit breakers have less power losses as compared to the bidirectional devices. a protection strategy for multi-terminal hvdc grid based on the unidirectional breaking devices is discussed and assessed in this paper. the performance of unidirectional protection strategy is examined under different fault scenarios in a detailed fourterminal mmc-hvdc grid model. furthermore, the impacts of unidirectional protection strategy on power converters and also current interruption and surge arrester ratings of the dc circuit breakers are discussed. index terms—dc circuit breaker, hvdc, protection device, voltage source converter (vsc), short-circuit. transactions on environment and electrical engineering issn 2450-5730 vol 1, no 4 (2016) © ataollah mokhberdoran, nuno silva, helder leite and adriano carvalho i. introduction i ncreasing penetration of the clean energy resources hasled to a demand for development of more efficient ways to transmit bulk amount of electrical energy over long distances. as a solution, high voltage direct current (hvdc) transmission technology has been employed by different project developers. large offshore wind farms and onshore ac systems can be interconnected through multi-terminal hvdc (mt-hvdc) grid in order to share the harvested energy between various geographical areas and enhance the system reliability [1]. voltage source converter (vsc) offers several technical benefits for application in the future mt-hvdc grid. the vsc technology was introduced by conventional two-level converter and has evolved into multilevel converter topologies [2]. introduction of modular multilevel converter (mmc) paved the way for the application of vsc in hvdc transmission systems. recently, different variants of the half-bridge mmc have been developed and employed in hvdc industry [3]. the conventional vscs and the half-bridge mmc topologies are highly vulnerable against dc side short circuit fault due to behavior of igbts’ antiparallel diodes. although full-bridge mmcs and other fault-tolerant converters can solve this issue, their power losses and lack of protection selectivity have been identified as their application drawbacks. moreover, these converters have not been tested practically in full-scale, yet [1]. the research leading to these results has received funding from the people programme (marie curie actions) of the european unions seventh framework programme (fp7/2007-2013) under rea grant n 317221. ataollah mokhberdoran and nuno silva are with efacec energia, s.a, un. switchgear and automation, rua frederico ulrich, po box 3078 4471907, maia, portugal (e-mail: mokhber@fe.up.pt). ataollah mokhberdoran, helder leite and adriano carvalho are with the department of electrical and computer engineering of university of porto, portugal. hvdc circuit breaker (dccb) as a promising solution may solve the protection problem in the mt-hvdc grids [1], [4], [5], [6]. fast dccbs can be categorized as hybrid dc circuit breakers (hcbs) and solid-state dc circuit breakers (sscbs). the sscbs can interrupt the fault current in tens of microseconds whereas the interruption time in hcb is expected to be less than 3 ms [4], [5], [3]. although the hcb interrupts the current fast enough and has acceptable power losses, its realization cost for mt-hvdc grid can be expensive due to the large number of required semiconductor switches [7]. dccbs are usually considered to be bidirectional and hence interrupt the current in their forward and backward directions [4]. unidirectional dccbs (ucbs) conduct the current in their forward and backward directions whereas interrupts the current only in one direction. the application of a unidirectional sscb in a point-to-point dc connection is investigated in [8] and a unidirectional current releasing dccb has been proposed in [9]. the protection of radial offshore dc grid using ucbs has also been studied [10]. the main concern regarding the application of ucb in the mt-hvdc grid is its inability in interrupting the fault current flowing in its backward direction as it may occur in a dc bus short circuit fault scenario. in this paper, a protection strategy based on the unidirectional hcbs (uhcb) is suggested for mt-hvdc grid. the suggested strategy tries to overcome the main drawback of uhcb. protection logics for dc bus and transmission line faults are investigated. the performance of suggested strategy is examined through different fault scenarios in a four-terminal hvdc grid model. for dc bus fault scenario, two dccb tripping schemes are considered. moreover, the superiorities and limitations of unidirectional protection of the mt-hvdc grid are assessed. the impacts of suggested strategy on the mt-hvdc grid elements and the hcb are also studied. ii. typical protection strategy three different protection strategies for the mt-hvdc grids are identified: 1) handshaking approach with ac breakers, 2) fault-tolerant converters with disconnector switches, 3) fast fault identification relays with fast dccbs [1]. in this paper, the third protection strategy together with the hcbs are considered. the hcbs can be placed at ends of each transmission line and also at dc side of converters. fig. 1 shows the typical bidirectional dccbs (bcbs) arrangement and the protection zones in a three-terminal hvdc grid. cbxy represents the dccb attached to line lxy close to bus bx. also, cbxx represents the dccb attached to vscx at bus bx. 2 ac ac ac converter protection zone bus protection zone line protection zone vsc x bycb yz cbyx cbyycbxx cbxz cbxy bx lxy cbzx vsc y vsc z cbzz cbzy b z fig. 1. bcbs arrangement in a multi-terminal hvdc grid a. dc bus fault typically, a dc bus is protected by bus differential protection scheme. in the dc bus, sum of all incoming and outgoing currents must be zero. if a short circuit fault occurs at the dc bus, the sum of incoming and outgoing currents becomes nonzero. therefore, dc bus trip signal can be generated if the sum of currents exceeds a near zero value. the differential protection scheme is quite fast and selective due to its low computing burden. the fault clearance in the dc bus zone can be done by opening all adjacent dccbs. if the dc bus is connected to a converter, the converter dccb should also be opened. assume n transmission lines are connected to bus bx. the protection logic can be given as (1). fault at bx ⇒ trip(cbx1, ..., cbxx, ..., cbxn) (1) b. transmission line fault when using fast protection schemes transmission line cannot be protected by differential protective relays due to communication delay. therefore, the transmission line is protected by communication-less non-unit protection schemes [11]. the non-unit protection schemes rely on local measurement of current, voltage, current derivative, voltage derivative or their combination [1]. if a fault is detected on the line, both dccbs of faulty line should trip. the protection logic can be given as: fault on lxy ⇒ trip(cbxy, cbyx) (2) iii. unidirectional protection strategy fig. 2(a) depicts the arrangement of ucbs in a threeterminal hvdc grid. the arrow in ucb symbol shows its current interruption direction. a protection strategy covering the dc bus and the transmission line zones based on the ucbs is suggested for the meshed hvdc grids in this section. a. dc bus fault fig. 2(a) shows the fault currents during a short circuit fault at dc bus bz. three fault currents flow though three adjacent ucbs. since the fault current if zz flows in the forward direction of cbzz, it can be interrupted by cbzz. fault vsc x ac ac by cbyz cbyx cbyycbxx cbxz cbxy bx lxy cbzx ac vsc y vsc z cbzz cbzy bz ifzz ifxz ifyz ifxz ifyz vsc x ac ac bycb yz cbyx cbyycbxx cbxz cbxy bx lxy cbzx ac vsc y vsc z cb zz cbzy b z ifxy ifyx (b) dc line fault (a) dc bus fault fig. 2. uhcbs arrangement and directions and fault current directions (a) dc bus fault, (b) dc transmission line fault currents if xz and if yz flow through the adjacent lines to the fault location and are in the backward directions of cbzx and cbzy and cannot be interrupted by them. as shown in fig. 2(a), if xz flows in the forward direction of cbxz and can be interrupted by this ucb, which is placed at the remote end of line. any other fault current flowing from the adjacent lines can be interrupted by the remote ucb. the protection logic for the dc bus fault can be given by: fault at bx ⇒ trip(cb1x, ..., cbxx, ..., cbnx) (3) the trip command for remote dccb can be generated locally or communicated between two buses. in the communicationbased method, fault detection is done locally in the faulted bus and the trip command is communicated to the remote dccbs. on the other hand, the communication-less method relies on the fault detection at the remote buses. a fault at bx can be detected by the remote dccbs at the other buses of system either based on the non-unit protection or overcurrent protection schemes . b. transmission line fault fig. 2(b) shows a short circuit fault in line lxy. two fault currents flow from both ends of the transmission line into the fault location. in any line fault condition, the fault currents flow in the forward directions of corresponding ucbs. therefore, the fault can be cleared by opening the corresponding 3 lcs ufd main breaker surge arrestor dc bus side line side l lcs ufd main breaker surge arrestor dc bus side line side l (a) (b) icb ilcs fig. 3. hybrid dc circuit breaker (a) typical hcb, (b) uhcb ucbs. unidirectional protection logic for transmission line fault is similar to (2). iv. hybrid dc circuit breaker a. typical hybrid dc circuit breaker the configuration of a typical hcb is depicted in fig. 3(a) [12]. the main current conduction path contains a fast mechanical disconnector switch (ufd) in series with the load commutation switch (lcs) [5], [12]. the lcs is not required to block the nominal voltage and therefore, it has a lower voltage rating and can be realized by series connection of few semiconductor switches. hence, the hcb has reasonable power dissipation whereas it is able to interrupt dc fault current quickly (around 2.5 ms) [5]. the parallel solid-state branch is the main breaker unit (mbu) during the fault condition. voltage rating of the main breaker unit can be identified based on the transient recovery voltage (trv) of the circuit breaker, which is usually limited by the reference voltage of surge arrester branch [12]. in order to allow bidirectional current flow and also bidirectional fault current interruption the semiconductor switches are connected in anti-series. b. unidirectional hybrid dc circuit breaker the topology of a unidirectional hybrid dc circuit breaker (uhcb) for the positive pole of hvdc system is depicted in fig. 3(b). in the uhcb topology, two anti-series semiconductor switches are replaced by one switch. therefore, the uhcb is only able to interrupt the fault current in the transmission line side. the normal power flow can be maintained in both forward and backward directions due to presence of antiparallel diodes. the uhcb requires half the number of semiconductor switches in its mbu as compared to the typical hcb [13]. the operation principles of uhcb for the transmission line fault are similar to the typical hcb. however, due to conduction of antiparallel diodes in the lcs unit of uhcb, similar operation sequence cannot be applied for the dc bus fault. suggested operation flowchart for the uhcb is shown in fig. 14. in the flowchart, ilcs and icb represent the lcs unit current and uhcb current as it is illustrated in fig. 3(b). in case of a line fault ilcs is positive and therefore uhcb can activate the mbu for fault interruption. when ilcs is negative, it means that the fault current flows in the backward direction of the uhcb. this case happens for uhcbs attached to a faulty dc bus. in this case, the lcs unit should be opened and commutate whole the fault current into the antiparallel diodes of mbu. thereafter, the ufd can be opened. note mmc3 mmc4 l14 l l24 l34 i13 cb12 cb14 cb13 cb42cb31 cb21 cb24 cb41 cb 43 cb34 mmc1 mmc2l12cb11 cb33 cb44 cb22 m 100 km 200 km 200 km 150 km 100 km 1 b1 b2 b3 b4 13 ac grid ac grid fig. 4. test multi-terminal hvdc grid ll trip rv cbste  trip rv ll cbste  d(a) (b) fig. 5. aggregated models (a) typical hcb, (b) uhcb that the uhcb still conducts the fault current in its backward direction through the antiparallel diodes of mbu until the fault current is interrupted by the remote uhcb. v. test system a. meshed hvdc grid a four-terminal meshed hvdc grid model, which was proposed in [14] is used in this study. the system configuration is shown in fig. 4. the studied model represents a cable-based meshed hvdc grid. the investigated system has a symmetric monopole hvdc configuration and includes four half-bridge mmcs. the mmcs are modeled by a continuous modeling approach with antiparallel diodes representing the blocking capability of the mmcs [14]. in the normal condition, mmcs 1 and 2 inject almost 700 mw into the grid and mmcs 3 and 4 absorb 600 and 800 mw, respectively. the blocking current threshold of mmcs is set to 3.2 ka in order to observe the mmc behavior without blocking during sever fault conditions. the parameters of fourterminal grid are illustrated in table ii. b. dc cable hvdc transmission lines are modeled based on the xlpe insulated cable using frequency dependent modeling approach. the cable cross-sections and properties of material are illustrated in fig. 15 and table iii, respectively [15]. c. circuit breaker model since the internal operation of dccb is not the subject of this paper, the aggregated models of hcb and uhcb are used. fig. 5(a) and (b) depict the aggregated models of hcb and uhcb, respectively. as shown in fig. 5(b) the uhcb model is derived from hcb model by adding one 4 0 2 4 6 8 10 -2 0 2 4 6 8 (a) cb 13 fault on l 13 c u rr e n t (k a ) 0 2 4 6 8 10 -2 0 2 4 6 8 (b) cb 31 fault on l 13 0 2 4 6 8 10 -2 0 2 4 6 8 (c) cb 12 fault on l 12 time (ms) 0 2 4 6 8 10 -2 0 2 4 6 8 (d) cb 21 fault on l 12 0 2 4 6 8 10 -2 0 2 4 6 8 (e) cb 14 fault on l 14 0 2 4 6 8 10 -2 0 2 4 6 8 (f) cb 41 fault on l 14 hcb uhcb t br t id t id t br t id t br t id t br t br t id t br t id fig. 6. fault current for fault at middle of transmission line for bcbs and ucbs (a) fault on l13, cb13 current, (b) fault on l31, cb31 current, (c) fault on l12, cb12 current, (d) fault on l21, cb21 current, (e) fault on l14, cb14 current, (e) fault on l41, cb41 current 0 50 100 150 200 0 2 4 6 8 10 time (ms) e n e rg y ( m j) cb 21 (hcb) cb 12 (hcb) cb 21 (uhcb) cb 12 (uhcb) 0 50 100 150 200 0 2 4 6 8 10 e n e rg y ( m j) time (ms) cb 31 (hcb) cb 13 (hcb) cb 31 (uhcb) cb 13 (uhcb) 0 50 100 150 200 0 2 4 6 8 10 e n e rg y ( m j) time (ms) cb 41 (hcb) cb 14 (hcb) cb 41 (uhcb) cb 14 (uhcb) fig. 7. absorbed energy in surge arresters of bcbs and ucbs (a) cb13 and cb31, (b) cb12 and cb21, (c) cb14 and cb41 parallel diode (d). the ll, rv and tcb represent the limiting inductor, surge arrester and the circuit breaker operation time delay, respectively. the value of limiting inductor (ll) of the line dccbs is set to 100 mh. the ll for the converter station dccbs is set to 10 mh. the reference voltage of surge arresters is set to 480 kv. the hcb operation delay (tcb) is set to 2.5 ms, which includes time delays in the lcs, ufd and mbu operations [5]. the trip command is received by the dccb and breaker component interrupts the current independent of its value after a time delay equal to tcb. d. protection system the system performance is studied based on the typical and the suggested unidirectional protection strategies. 1) typical protection strategy: a non-unit scheme, which has recently been proposed in [11] is employed for transmission line protection. the utilized method uses the local current measurements for line fault detection. the dc bus fault is detected by the differential protection scheme. the dc bus trip signal for fault is generated locally in typical protection strategy. 2) unidirectional protection strategy: the mentioned nonunit scheme is also used in the unidirectional protection strategy. two different schemes for the dc bus fault detection are considered in the unidirectional protection strategy: 1) local fault detection (communication-based), 2) remote overcurrent fault detection (communication-less). in the first method a bus fault is detected by the bus differential relay and the trip command is communicated to the remote uhcbs. communication time is modeled by a time delay block. in the second method a communication-less system is considered. the bus fault is detected at the remote healthy buses when the line current exceeds specific threshold. table i trip times of different types of dc circuit breakers for dc bus and transmission line faults dccb trip time (tid) [ms] fault at b1 fault on l12 fault on l13 fault on l14 bcb ucb bcb ucb bcb ucb bcb ucb cb11 0.1 0.1 cb12 0.1 0.1 0.79 0.79 cb21 5 1.5 0.66 0.66 cb13 0.1 0.1 1.05 1.05 cb31 10 2 0.95 0.95 cb14 0.1 0.1 1.05 1.05 cb41 10 2 0.95 0.95 vi. simulation the results from study of four fault scenarios are presented and compared in this section. a. transmission line fault transmission line fault is studied through three independent permanent pole-to-pole low impedance (rfault = 100 mω) short circuit faults at the middle of lines l12, l13, l14. the line fault is incepted at time 0 s. as discussed in section iii, the line fault clearing process for ucbs and bcbs are the same. therefore, to clear the line fault from the system dccbs at both ends of the faulted line should trip. fig. 6(a)-(f) show the current in dccbs for the short circuit fault in different transmission lines. the currents in hcb and uhcb are the same. tidxy and tbrxy represent the fault identification time and the dccb action time for cbxy, respectively. note that the fault identification and the trip command generation are assumed to be simultaneous. the numerical values of fault identification time are illustrated in table i. due to the longer length of l13 and l14, the midpoint fault is detected later than the similar fault in l12. absorbed energy by the surge arrester of each dccb is 5 0 5 10 15 20 -8 -6 -4 -2 0 2 (a) cb 12 c u rr e n t (k a ) 0 5 10 15 20 -2 0 2 4 6 8 (b) cb 21 0 5 10 15 20 -6 -4 -2 0 2 (c) cb 13 hcb uhcb1 uhcb2 0 5 10 15 20 -2 0 2 4 6 8 (d) cb 31 c u rr e n t (k a ) 0 5 10 15 20 -6 -4 -2 0 2 (e) cb14 time (ms) 0 5 10 15 20 -2 0 2 4 6 8 (f) cb 41 t br,b t id,b t id,u1 t br,b t id,b t id,u1 t br,u2 t id,b t br,b t id,u1 t br,u2 t id,u2 t br,u1 t id,u2 t br,u2 t br,u1 t id,u2 t br,u1 fig. 8. current in dccb during fault at bus (b1) for bcbs and ucbs (a) cb12, (b) cb21, (c) cb13, (d) cb31, (e) cb14, (f) cb41 0 50 100 150 200 0 2 4 6 8 10 e n e rg y (m j) (a) cb 21 (hcb) cb 12 (hcb) cb 21 (uhcb1) cb 12 (uhcb1) cb 21 (uhcb2) cb 12 (uhcb2) 0 50 100 150 200 0 2 4 6 8 10 time (ms) e n e rg y (m j) (b) cb 31 (hcb) cb 13 (hcb) cb 31 (uhcb1) cb 13 (uhcb1) cb 31 (uhcb2) cb 13 (uhcb2) 0 50 100 150 200 0 2 4 6 8 10 e n e rg y (m j) (c) cb 41 (hcb) cb 14 (hcb) cb 41 (uhcb1) cb 14 (uhcb1) cb 41 (uhcb2) cb 14 (uhcb2) fig. 9. absorbed energy in surge arresters of bcbs and ucbs (a) cb12 and cb21, (b) cb13 and cb31, (c) cb14 and cb41 depicted in fig. 7. the difference in dissipated energy in the surge arresters is due to unequal interrupted fault currents in corresponding dccbs and also different fault distances from dccb. dccbs with positive pre-fault current (pre-fault and fault currents are in the same direction) reach higher current than the dccb with negative pre-fault current and therefore, larger amount of energy is dissipated in their surge arresters. b. dc bus fault during a bus fault in the mt-hvdc grid, due to the low inductance between the converter and fault location, high fault current flows in the dc side of the converter. the fault current may exceed the self-protection threshold of the mmc and cause the mmc blocking. protection of converter against dc bus faults can rely on either ac side circuit breaker or the converter station dccb. in this study, the mmc is protected by a dccb at its dc side. a permanent pole-to-pole low impedance (rfault = 100 mω) short circuit fault is incepted at bus b1 at time 0 s. in the typical bidirectional strategy upon fault detection by the differential scheme, cb11, cb12, cb13, cb14 should trip and disconnect the adjacent lines from b1. therefore, terminal 1 of the mt-hvdc grid will be disconnected from rest of the system and consequently, the amount of harvested energy from generation nodes of system will be reduced. hence, mmc 3 and 4 will absorb less power as compared to pre-fault conditions. the remote dccbs of the adjacent lines trip after a longer time delay to disconnect the cables from healthy dc buses. in the unidirectional protection strategy, upon fault detection at bus b1 all adjacent uhcbs including cb11, cb12, cb13, cb14 are opened. the fault currents flow in the backward directions of cb12, cb13, cb14 and these uhcbs are unable to interrupt the fault current. therefore, based on 3, cb21, cb31 and cb41 also should trip. the dc bus fault clearing is studied through two protection schemes as explained in subsection v-d2. in the communication-based method, communication delay is modeled by a time delay block. the time delay block represents sum of propagation and transmitter/receiver delays. the propagation delay is set to 5 µs/km and the transmitter/receiver delay is set to 1 ms [16]. the trip times of remote dccbs are illustrated in table i. the second method is an overcurrent protection scheme. the overcurrent thresholds are set to 3 ka for all lines. fig. 8(a)-(f) show currents of all dccbs attached to the adjacent lines for three discussed protective schemes. in fig. 8 hcb, uhcb1 and uhcb2 represents the bidirectional, communication-based and overcurrent-based unidirectional protective schemes, respectively. in addition, tid and tbr represent fault identification and current interruption times for each protective scheme. absorbed energy in the surge arrester of each dccb is depicted in fig. 9(a)-(c). as can been in fig. 8, current in cb21, cb31 and cb41 reach higher values in overcurrent-based scheme as compared to the communication-based scheme. as seen in fig. 8(a), (c) and (e), the hcbs interrupt the bus fault current before 6 0 25 50 75 100 -2 -1 0 1 2 3 (a) mmc2 c u rr e n t (k a ) 0 25 50 75 100 -2 -1 0 1 2 3 (b) mmc3 time (ms) c u rr e n t (k a ) 0 25 50 75 100 -2 -1 0 1 2 3 (c) mmc4 c u rr e n t (k a ) fig. 10. the arm currents of mmcs in presence of bcbs (a) mmc2, (b) mmc3, (c) mmc4 0 25 50 75 100 -2 -1 0 1 2 3 (a) mmc2 c u rr e n t (k a ) 0 25 50 75 100 -2 -1 0 1 2 3 (b) mmc3 time (ms) c u rr e n t (k a ) 0 25 50 75 100 -2 -1 0 1 2 3 (c) mmc4 c u rr e n t (k a ) fig. 11. the arm currents of mmcs in presence of ucbs (communication-based method) (a) mmc2, (b) mmc3, (c) mmc4 0 25 50 75 100 -2 -1 0 1 2 3 (a) mmc2 c u rr e n t (k a ) 0 25 50 75 100 -2 -1 0 1 2 3 (b) mmc3 time (ms) c u rr e n t (k a ) 0 25 50 75 100 -2 -1 0 1 2 3 (c) mmc4 c u rr e n t (k a ) fig. 12. the arm currents of mmcs in presence of ucbs (overcurrent-based method) (a) mmc2, (b) mmc3, (c) mmc4 reaching higher values. on the other hand, the remote uhcbs in the communication-based unidirectional protective scheme (uhcb1) also interrupt the bus fault current before reaching higher values. due to higher interrupted current and larger cable inductance in unidirectional protection strategy, uhcbs’ surge arresters absorb more energy than the hcbs. fig. 10 shows the mmc arm currents for healthy buses (mmc 2, 3 and 4) in presence of hcbs. in addition, the arm currents of mmc 2, 3 and 4 in presence of uhcbs for communicationbased scheme are depicted in fig. 11. also, the arm currents of mentioned mmcs in presence of uhcbs for overcurrentbased scheme are illustrated in fig. 12. the arm currents of mmc 1 are not included since this converter is isolated from the mt-hvdc grid due to converter station dccb (cb11) action during the dc bus fault. vii. discussion a. circuit breaker current rating the maximum current in possible fault scenarios sets the requirements for dccb current interruption rating. 1) communication-based method: due to lower inductance of short transmission lines, rate of rise of fault current in short lines is higher as compared to the long lines. hence, the remote dccbs in short lines might be required to interrupt higher current as compared to long lines. for instance, as can be seen in fig. 8(b), the current in cb21 reaches almost 5.6 ka with unidirectional strategy while it does not exceed 3.9 ka in cb12 in the bidirectional strategy (see fig. 8(a)). note that the length of line l12 is 100 km. on the other hand, for line l13 (200 km), the current in cb31 reaches almost 3.8 ka with unidirectional strategy and it reaches almost 4 ka in cb13 in the bidirectional strategy (see fig. 8(c) and (d)). the bus fault is cleared in longer time in communication-based unidirectional strategy as compared to the bidirectional strategy. fig. 13(a) provides a comparison between the maximum interrupted current of different dccbs during the dc bus and corresponding transmission line faults. as shown in the figure, the dccbs are required to interrupt higher fault current during the line fault as compared to the dc bus fault in communication-based unidirectional protection strategy. results of this study imply that the communication-based unidirectional scheme does not necessarily require dccbs with higher current rating. despite longer fault detection and trip times in the communicationbased unidirectional scheme, the cable inductance and the dccb current limiting inductor limit the rate of rise of fault current in the remote dccb. 2) overcurrent-based method: as can be seen in fig. 8, the current in cb21, cb31 and cb41 reach higher value in 7 cb_1_2 cb_1_2 cb_2_1 cb_2_1 cb_1_3 cb_1_3 0 2 4 6 8 10 c u rr e n t (k a ) hcb uhcb1 uhcb2 cb_1_2 cb_1_2 cb_2_1 cb_2_1 cb_1_3 cb_1_3 0 2 4 6 8 10 e n e rg y (m j) hcb uhcb1 uhcb2 cb 12 cb 21 cb 13 cb 31 cb 14 cb 41 (b) (a) b 1 l 12 b 1 l 12 b 1 l 13 b 1 l 13 l 14 b 1 l 14 b 1 b 1 b 1 l 13 b 1 l 13 b 1 l 14l 12 l 12 l 14b 1 b 1 cb 12 cb 21 cb 13 cb 31 cb 14 cb 41 fig. 13. (a) maximum interrupted current and (b) absorbed energy in different dc circuit breakers during dc bus and line faults overcurrent-based unidirectional scheme as compared to the current in cb12, cb13 and cb14 in the bidirectional and the communication-based unidirectional schemes. the overcurrent protection threshold may be set to lower values in order to shorten the fault identification time if it is coordinated with the non-unit protection scheme. as shown in fig. 13(a), the maximum interrupted current in cb31 and cb41 is slightly higher for the dc bus fault with overcurrent-based method as compared to the maximum interrupted current for corresponding transmission line faults. hence, the uhcbs might be required to interrupt higher current as compared to the hcbs depending on the overcurrent protection parameters. b. surge arrester energy rating 1) communication-based method: the amount of absorbed energy in the surge arrester of dccb depends on the interrupted current value and the fault location distance from the dccb. the magnitude of interrupted current has higher impact on the amount of absorbed energy. the absorbed energy in surge arresters of dccbs during the transmission line and the dc bus faults are compared in fig. 13(b). as can be seen, the surge arresters of uhcbs dissipate lower amount of energy during dc bus fault in communication-based unidirectional method as compared to corresponding transmission line fault. note that the amount of absorbed energy in cb21 during bus fault is almost equal to the absorbed energy in cb12 during the line fault. these results imply that the energy rating of surge arresters for uhcbs with communication-based unidirectional strategy is not necessarily different than their energy rating for hcbs with the bidirectional strategy. 2) overcurrent-based method: due to the higher fault current during the dc bus fault in overcurrent-based method, the surge arresters dissipate higher amount of energy as compared to the line fault conditions (see fig. 13(b)). hence, the surge arresters in uhcbs with overcurrent-based method are required to be rated for higher energy absorption as compared to the hcbs with the typical strategy. c. impact on the converters as seen in fig. 10, fig. 11 and fig. 12, mmc 2 arm currents reach higher values as compared to the arm currents of other mmcs. mmc 2 is connected to the faulted bus (b1) through a 100 km cable, which is shorter than other adjacent cables. therefore, mmc 2 is more influenced by the fault transient as compared to the other mmcs. furthermore, as can be seen in fig. 12(a), one of mmc 2 arm currents reaches almost 3 ka, which would be higher than self-protection threshold level of mmc. although in this study mmc 2 is not blocked, the application of unidirectional protection strategy may cause blocking of the mmcs connected to the faulted bus by a short transmission line. this issue can be avoided by either slight increase in the inductance of dccbs limiting inductor or using igbts with higher current capability in mmcs. d. impact on dccb the mbu and lcs in the uhcb requires half the number of semiconductors as compared to the hcbs. for instance, an hcb with 320 kv and 9 ka voltage and current ratings requires 1416 igbts with 3.3 kv voltage rating in the mbu [7]. by applying unidirectional protection strategy this number can be reduced to 708 by mean of the uhcb. due to the large number of required igbts by the hcb and considering the peripheral circuits and elements for each igbt, this reduction can significantly decrease the initial investment for implementation of the hcbs. viii. final remarks the uhcbs show technical and economic advantages thanks to their less initial and operational costs as compared to the typical hcbs. a unidirectional protection strategy for mt-hvdc grid is suggested in this paper. the results of study confirm that the unidirectional protection strategy can be utilized for protection of the mt-hvdc grid. two methods for remote dccb tripping are considered: 1) communication-based and 2) overcurrent-based methods. the communication-based method shows better performance as compared to the overcurrent-based method. the results of comparison study for different parameters of dccbs imply that the current rating of dccbs and the size of surge arresters are not necessarily different for the bidirectional and unidirectional strategies. however, the impact of suggested strategy on all converters of the grid, particularly the converters with shorter transmission line between them should be analyzed. in order to avoid blocking of the mmcs at healthy buses, slight increase in dccb limiting inductor size or current rating of mmc’s igbts might be required. application of unidirectional protection strategy may significantly reduce the number of required semiconductor switches in the mbu of hcbs. considering the large number of required semiconductor switches by the hcb and also the peripheral circuits and elements of each semiconductor switch, this reduction can significantly decrease the initial investment for implementation of the hcbs. 8 appendix line fault trigger t_f read trip com. turn off mbu trip? i_cb > i_threshold no yes yes no no yes trigger t_ch turn on mbu v_bus >= v_nominal yes no read the voltages read the currents start v_c = v_bus and i_ch=0 no yes close? yes no read trip com. trip? yes no no start close? yes no lcs opened yes open ufd no close main breaker yes commutation done? yes open ufd yes open main breaker yes failure no no no open lcs i cb = 0 ilcs > 0 ilcs = 0 ilcs = 0 fig. 14. unidirectional hybrid dc circuit breaker operation logic flowchart fig. 15. cross-section and configuration of the xlpe insulated hvdc cable table ii four-terminal hvdc system parameters [14] parameter converter 1, 2, 3 converter 4 rated power 900 mva 1200 mva ac grid voltage 400 kv 400 kv converter ac voltage 380 kv 380 kv transformer, uk 0.15 pu 0.15 pu arm capacitance carm 29.3 µf 39 µf arm reactor larm 84.8 mh 63.6 mh arm,resistance rarm 0.885 ω 0.67 ω bus filter reactor ls 10 mh 10 mh table iii dc cable data [15] layer radius (mm) resistivity (m) rel. permeability rel. permittivity (1) core 25.2 1.72×10−8 1 1 (2) insulator 40.2 1 2.3 (3) sheath 43.0 2.20×10−7 1 1 (4) insulator 48.0 1 2.3 (5) armor 53.0 1.80×10−7 10 1 (6) insulator 57.0 1 2.1 references [1] n. chaudhuri, b. chaudhuri, r. majumder, and a. yazdani, multiterminal direct-current grids: modeling, analysis, and control. john wiley & sons, 2014. [2] a. mokhberdoran and a. ajami, “symmetric and asymmetric design and implementation of new cascaded multilevel inverter topology,” ieee transactions on power electronics, vol. 29, no. 12, pp. 6712–6724, dec 2014. [3] a. mokhberdoran, a. carvalho, n. silva, h. leite, and a. carrapatoso, “application study of superconducting fault current limiters in meshed hvdc grids protected by fast protection relays,” electric power systems research, vol. 143, pp. 292 – 302, 2017. [4] a. mokhberdoran, a. carvalho, h. leite, and n. silva, “a review on hvdc circuit breakers,” in renewable power generation conference (rpg 2014), 3rd, sept 2014, pp. 1–6. [5] j. häfner and b. jacobson, “proactive hybrid hvdc breakers a key innovation for reliable hvdc grids,” in electric system of the future integrating supergrids and microgrids international symposium, italy, sept 2011, pp. 1–8. [6] b. geebelen, w. leterme, and d. v. hertem, “analysis of dc breaker requirements for different hvdc grid protection schemes,” in ac and dc power transmission, 11th iet international conference on, feb 2015, pp. 1–7. [7] g. liu, f. xu, z. xu, z. zhang, and g. tang, “assembly hvdc breaker for hvdc grids with modular multilevel converters,” ieee transactions on power electronics, vol. pp, no. 99, pp. 1–1, 2016. [8] k. sano and m. takasaki, “a surgeless solid-state dc circuit breaker for voltage-source-converter-based hvdc systems,” ieee transactions on industry applications, vol. 50, no. 4, pp. 2690–2699, july 2014. [9] a. mokhberdoran, a. carvalho, n. silva, h. leite, and a. carrapatoso, “a new topology of fast solid-state hvdc circuit breaker for offshore wind integration applications,” in power electronics and applications (epe’15 ecce-europe), 2015 17th european conference on, sept 2015, pp. 1–10. [10] d. jovcic, m. taherbaneh, j. p. taisne, and s. nguefeu, “offshore dc grids as an interconnection of radial systems: protection and control aspects,” ieee transactions on smart grid, vol. 6, no. 2, pp. 903–910, march 2015. [11] s. p. azad and d. v. hertem, “a fast local bus current-based primary relaying algorithm for hvdc grids,” ieee transactions on power delivery, vol. pp, no. 99, pp. 1–1, 2016. [12] a. hassanpoor, j. häfner, and b. jacobson, “technical assessment of load commutation switch in hybrid hvdc breaker,” ieee transactions on power electronics, vol. 30, no. 10, pp. 5393–5400, oct 2015. [13] a. mokhberdoran, n. silva, h. leite, and a. carvalho, “a directional protection strategy for multi-terminal vsc-hvdc grids,” in 2016 ieee 16th international conference on environment and electrical engineering (eeeic), june 2016, pp. 1–6. [14] w. leterme, n. ahmed, j. beerten, l. angquist, d. v. hertem, and s. norrga, “a new hvdc grid test system for hvdc grid dynamics and protection studies in emt-type software,” in ac and dc power transmission, 11th iet international conference on, feb 2015, pp. 1–7. [15] f. mura, c. meyer, and r. w. d. doncker, “stability analysis of highpower dc grids,” ieee transactions on industry applications, vol. 46, no. 2, pp. 584–592, march 2010. [16] s. c. f. behrouz a. forouzan, data communications and networking. mcgraw-hill forouzan networking, 2007. introduction typical protection strategy dc bus fault transmission line fault unidirectional protection strategy dc bus fault transmission line fault hybrid dc circuit breaker typical hybrid dc circuit breaker unidirectional hybrid dc circuit breaker test system meshed hvdc grid dc cable circuit breaker model protection system typical protection strategy unidirectional protection strategy simulation transmission line fault dc bus fault discussion circuit breaker current rating communication-based method overcurrent-based method surge arrester energy rating communication-based method overcurrent-based method impact on the converters impact on dccb final remarks appendix references  transactions on environment and electrical engineering issn 2450-5730 vol 2, no 2 (2017) © uyara f. silva, manoel i. q. júnior, santiago lemos, alan. f. silva, geovanne p. furriel, nikholas segatti and wesley p. cali xto abstract—the purpose of this work is to calculate surge and swab pressures in eccentric wells. analysis of the phenomenon, in which fluid is confined between two eccentric cylinders, are made. conformal mapping calculations is used to lead the original eccentric domain into equivalent concentric domain, since usual models only make calculation for concentric geometries. the results of this study, using the proposed methodology, are presented and discussed. index terms— annulus geometry, conformal mapping, surge pressure, swab pressure, yield power law drilling fluid. i. introduction the usual methodology for pressure calculation, during drilling of wells, assumes that the drill string and well work concentrically. however, the rotary movement of the drill makes the set (drill and well) works eccentrically with the well. therefore, the calculations obtained through the usual methodologies have errors because they do not consider the effect of the eccentricity. surge and swab pressures are problems that can happen during the drilling of oil wells [1]. the prediction of these pressures is essential to determine the appropriate speeds and accelerations for introduction and withdrawal of the drill string in the wellbore [2]. surge pressure is the increase in wellbore pressure, that goes beyond the maximum supported by walls. it can fracture the rock formation and causes fluid loss [3]. swab pressure occurs when pressure inside the wellbore is below the pore pressure, causing fluid penetration from the rock into the wellbore (kick effect). if not controlled, the kick effect can become blowout, which means uncontrolled flow from the hole to surface [2]. to avoid the blowout, it is necessary to control and to manage the pressures, ensuring that they are within desired limits [3]. calculations of surge and swab pressures have extreme importance to avoid kick effect and subsequent blowout. kick effect can occur when the hydrostatic pressure is lower than the pore pressure, fracturing the formation and thus the pore fluid invades the well. the absence of control of the kick effect can results in blowout, fluid inside the well reaches surface uncontrollably and can cause explosions and dump of oil into the environment. the usual methodology to calculate the surge and swab pressures assumes that the casing string and drill work concentrically [4]. however, the rotational movement of the drill causes the geometry of the well to be eccentric. calculations obtained by usual methods do not consider the effect of eccentricity [5]. in order to reduce errors due to the eccentricity it can be used conformal mapping, that takes a domain into another, preserving the angles and physical quantities [6]. the purpose of this work is use conformal mapping to calculate the value of surge and swab pressures taking into account the effect of eccentricity between the casing string and the wellbore. conformal mapping are domain transformations and represent complex analytic functions, so this work assumes that conformal mapping are capable to lead eccentrics geometries into concentric geometries. therefore, the value of the pressures can be calculated at any time of the drilling dynamics. the model is developed for power law drilling fluids under isothermal conditions, considering a constant fluid density. the model can be used in vertical wells, both terrestrial and marine. this work is focused on the problems caused by surge and swab pressures. ii. well drilling process for drilling wells, the drill column with the drill in the tip is introduced into the formation, making the well [7]. the process of well drilling is made by steps, each step is drilled a given depth and thickness [8]. introduction and withdrawing the drill string is called maneuver [9]. for each maneuver is introduced a new column set with new thickness down to a certain depth. during drilling, a hydrostatic fluid is injected through the drill bit into the well with the purpose of carrying gravels out of the wellbore. hydrostatic fluid is also used to lubricate the drill bit and give more support to the wellbore walls. after drilling part of the well, the drill column is removed and then is introduced a casing string into the hole, a metal pipe used to support the wellbore wall which has just been bored, preventing landslides [10]. surge pressure, fig. 1 (a), occurs when the hydrostatic pressure of the fluid is above pores pressure at the time of introduction of the drill string with speed vp estimate of geopressures using conformal mapping in eccentric wells uyara f. silva, manoel i. q. júnior, santiago lemos, alan. f. silva, geovanne p. furriel, nikholas segatti and wesley p. calixto higher than adequate. surge can also occurs when drilling fluid is injected more than necessary, increasing pressure on the rock [11]. hydrostatic pressure becomes greater than the fracture pressure (the rock strength) and can cause loss of drilling fluid. when this occurs, cementing of the well is required, to fill the space where the wellbore was fractured [12]. the swab pressure, fig. 1 (b), happens in the column withdrawal operation with speed vp generating pressure drop by drag of the fluid along the tube walls [13]. when withdrawing the drill, vacuum can be generated causing hydrostatic pressure decrease, thus making pore pressure to become greater than hydrostatic pressure in the well and the kick effect to occur [12]. a. pressures inside the well the concept of pressure within wells is associated with the fluids contained within the rocks, and the result of the loading, which reacts equally in all directions. 1) pore pressure: in rock formation, only part of the total volume is occupied by solid particles, which settle to form the structure. the remaining volume is often called voids, or pores, and is occupied by fluids. pore pressure, often referred as formation pressure or static pressure, is the pressure of fluids contained in porous spaces of the rock. it is a function of the specific mass of the formation fluid and the loads it is bearing. in petroleum areas, the fluid filling a formation can be water, oil or gas [2]. 2) hydrostatic pressure: hydrostatic pressure is exerted by the weight of the hydrostatic column of fluid, being function of the height of the column and the specific mass of this fluid. the drilling fluid has as one of its main objectives to keep the well safe and stable. the pressure provided by the drilling fluid varies if it is inside the drill string or in the annulus. this difference occurs because when the fluid is returning through the annular space, it carry the dirty made by drilling process. the weight of suspended gravels increases the specific mass of the drilling fluid by providing higher pressure at the bottom of the well. another variable that interferes with the hydrostatic pressure generated by the drilling fluid is that the fluid be static or moving [2]. b. kick and blowout effects the pressure inside the well must be between the minimum and maximum boundary (operating window), for this to occur, the amount of fluid injected into the well must be analyzed according to the structure of the formation. the operating window must be determined and respected during maneuvering to prevent the well wall from fracturing, by excessive hydrostatic pressure or by lack of weight to contain the pore pressure. the speed of the maneuver should also be well calculated before the introduction of the drill string. according to the type of formation being drilled, the imbalance between the pressure inside the well and the pore pressure can have different consequences. if the pore pressure becomes larger than the pressure inside the well, the formation fluid may invade the well. this typical undesired occurrence is called kick and can lead to loss of time during drilling. if kick occurs, it should be controlled by a higher injection of hydrostatic fluid through the drill string. in more severe and uncontrolled cases, the kick may hit the surface, resulting in blowout, uncontrolled flow of fluid coming out of the well due to some failure in the pressure control system, which can have disastrous consequences such as total destruction of the platform and the death of workers or damage to the environment. on the other hand, pressures inside the well greater than the fracture pressure, resistance of the formation, can lead to the invasion of the well fluid to the formation, being possible the collapse of the well [2]. c. operating window the operating window determines the allowable pressure variation exerted by drilling fluid inside the well, in order to maintain the integrity of the well, respecting the pore, fracture and collapse pressures. this window should determine the minimum and maximum fluid weight that can be used inside the well. the operating window is used to prevent kick and blowout effects or landslide [2]. fig. 2 illustrates an operating window example where the fluid weight limits should be between the fracture pressure and the pore pressure. fig. 1: geopressures: (a) surge and (b) swab. if the weight of the fluid is less than the pore pressure, there may be a kick effect, if the weight of the fluid is greater than the fracture pressure, there will be loss of fluid for formation and possible landslide [15]. iii. conformal mapping a deep-sea navigator can be guided by the stars and the angle that his course makes between the latitude and longitude lines. since the middle ages, navigators have realized that it was possible to have spherical surface angles conserved on plane maps. when the angles between the curves on earth are equal to the corresponding angles in the plane map, the map is called conform map [1]. the idea of a map conforming to the earth was developed by the belgian mathematician, geographer and cartographer gerardo mercator in 1569, known as mercator projection. although this projection presented distortions in proportions, it revolutionized cartography of the time. the points of the sphere, except the poles, are projected on a cylinder in which the sphere is inscribed, fig. 3, the parallels are parallel straight horizontal lines, where the distance between successive parallels is proportional to their proximity to the equator line, that is, the closer they are to the equator, the smaller the distance between them. meridians are projected in equidistant vertical parallel straight lines [1]. advances in conforming map theory were performed by other scientists, euler in 1777 (sphere in plan), and lagrange in 1779 to obtain all the conforming representations of a part of the earth's surface in a plane where all circles of longitudes and latitudes are represented by circular arcs in the plane. all of these authors, including lambert with conforming conics, used complex numbers, but lagrange's presentation is the clearest and most general. the discovery came in 1851 when riemann gave the fundamental result, known as the riemann's theorem, which was the starting point for all further developments in conformal mapping theory. to prove this theorem, riemann assumed that the variational problem, now known as the dirichlet problem, has a solution. fifty years later, in 1901, hilbert demonstrated the existence of the solution of the dirichlet problem. the validity of riemann's result was rigorously established by schwarz in 1890, using number of theorems coming from the logarithmic potential theory. at the end of the century xix and the beginning of the century xx cauchy, riemann, schwarz, christoffel, bieberbach, carathéodory, goursat, koebe and others have established theoretical aspects about the theory of functions with complex variables a. conformal mapping definition conformal mapping are analytical functions, w=f(z), that carry the points of the domain d in points of the domain i maintaining the property of the angles, capable of converting mathematical problem of difficult solution into a simpler one without changing the physical characteristics of the system. the function has domain and range in the complex plane. under some restrictive conditions, the mapping function can be defined [6]: 𝑤 = 𝑓(𝑥) = 𝑓(𝑥 + 𝑦𝑖) = 𝑢(𝑥; 𝑦) + 𝑣(𝑥; 𝑦)𝑖, (1) where 𝑧 = 𝑥 + 𝑦𝑖 (2) and 𝑣 = 𝑢 + 𝑣𝑖. (3) considering w = f(z) the complex function that takes points from the plane z in points of the plane w, so f(z) is a geometric transformation, since curves in the plane fig. 2: operational window fig. 3: mercator projection. z has as images, curves in the plane w. the curve c0 in the complex plane z has the curve s0 in the complex plane w as its image, fig. 4. considering positive direction of the course along c0, the corresponding positive direction over s0 is determined by the transforming function f. taking the point z0 in c0, its image in w is w0 = f(z0) over s0, as shown in fig. 4. considering also the z0+δz over c0 in the positive sense from z0, the limit of the argument of δz when it tends to zero is the angle of inclination α of the tangent line to c0 in z0, fig. 5 (a). if w0+δw is the image of z0+δz, then the argument of δw tends towards the inclination angle β from the tangent to s0 in w0, when δw tends to zero, with orientation according to fig.5 (b). iv. methodology a. conformal mapping the problem in calculation of pressures consists in obtaining an equivalence between the eccentric and concentric plans. fig. 6 illustrates the transverse section in the annulus formed between the wellbore and the column, by two coaxial cylinders with radii r1 for external cylinder and r2 for internal cylinder in the eccentric plane. the external plate is the well while inner plate is the column (drill). assuming that the plates are circular in the total length and ψ ≠ 0 is the eccentricity of the circles, there is in the fig. 6 (a) the real problem. there is some difficulty to calculate the pressures in devices with this geometry. however, a new geometry that makes possible the pressures calculation can be found. using algebraic manipulation [14], it is possible to develop a conformal mapping, given by (4), that leads two eccentric circles with radii r1 (external cylinder) and r2 (internal cylinder) into two concentric circles fig. 6 (b). 𝑤(𝑧) = 𝑡. 𝑟1 𝑟2 𝑒 𝑖𝜃 . 𝑑(𝑧 − 𝑧𝑎 ) − 𝑠(𝑧𝑏 − 𝑧𝑎 ) 𝑑(𝑧 − 𝑧𝑎 ) − 𝑡(𝑧𝑏 − 𝑧𝑎 ) , (4) where θ, s and t are real, za are the points of the external plate and zb, the points of the inner plate [6], s and t are the roots of the expression (4), given by: 𝑠. 𝑡 = 𝑟1 2 (𝑑 − 𝑠). (𝑑 − 𝑡) = 𝑟2 2. (5) in (4), d = |za zb| = ψ > 0 and za, zb, r1, r2 , r1, r2>0 and za ≠ zb [14], this equation can still write as: 𝑟2 𝑟1 . 𝑡 (𝑑 − 𝑡) = 𝑅2 𝑅1 , − −𝑑2 − 𝑟1 2 + 𝑟2 2 + √−4𝑑2𝑟1 2 + (𝑑2 + 𝑟1 2 − 𝑟2 2)2 2𝑑 = 𝑡, − −𝑑2 − 𝑟1 2 + 𝑟2 2 − √−4𝑑2𝑟1 2 + (𝑑2 + 𝑟1 2 − 𝑟2 2)2 2𝑑 = 𝑠. (6) thus, in this new geometry, circles are concentric and the relationship between the domains d and i can be done. b. rheological parameters fluids are classified according to their rheology, for this reason three parameters are considered: shear stress, shear rate and viscosity [15]. shear stress τ is defined as the force f that, applied to an area a, the interface between the moving surface and fig. 6: plan (a) eccentric in the domain d and plan (b) concentric in the domain i. fig. 5: slope angle (a) α in plane z and (b) β in plane w. fig. 4: curve in z plane and its image in w plane. the liquid, causes flow in the first layer of liquid and the first layer causes flow in the second layer of liquid and so on. this moving surface is the drill string which moves when performing the maneuver. shear rate can be defined as variation of flow velocity with the variation of the distance between the drill string and walls of the well [15]. the viscosity is the ratio between shear stress and shear rate, and is the measure of the fluid resistance. to calculate the shear stress over the shear rate in the well, the hershell model is used. it is also known as yield power law fluid model, whose relationship of the three parameters is given by: 𝜏 = 𝐾𝛾 𝑛 + 𝜏0, (7) where, τ is the shear stress and γ is the shear rate, τ0 is the yield stress, k is the consistency index that indicates the degree of fluid resistance against the flow and n, named behavior index, indicates the distance of the newtonian fluid model [15]. in the ratio γ × τ, for yield power law fluids, when the shear rate is γ = 0, the shear stress is the yield stress τ0. the yield stress is the minimum stress necessary for the fluid to start leaking [15]. c. mathematical model for surge and swab calculation in order to determine surge and swab pressure in concentric plan is necessary to determine the input values for the annular geometry and the rheology of the fluid used in the drilling of the well. for the rheology of the fluid, there are considered three parameters: shear stress, shear rate and viscosity. the shear stress τ is defined as the force f that, applied on an area a of the interface between the moving surface and the liquid, causes flowing in the first liquid layer, which in turn causes, in the second, the second in the third, and so on. this moving surface is the drill string which moves during the performing maneuver. shear rate can be defined as variation in flow rate with the height variation (distance from the surface causing the shear) [15]. the viscosity is the ratio between shear stress and shear rate, and is a measure of resistance to fluid flow. the annular geometry, fig. 7, is the height h, which is the distance between the surface of the drill string and the well, the diameter of the well dh = 2r1, the diameter of the drill string dp = 2r2, and the velocity vp of descent or ascent of the drill string. since the used methodology for calculating the pressures considers the concentricity between the well and the drill string, h=r1 – r2. the rheological parameters are: the consistency index k, the fluid behavior index n and the initial shear stress required for the flow called yield stress τ0 [12]. the values of surge and swab pressure are calculated by [3]: 𝑃𝑠𝑢 = 𝑃𝑒 𝑛 ( 𝑛 𝑛 + 1 ) 𝑛 ( 𝐻 𝑉𝑝 ) 𝑛 ( 𝐻 𝐾 ) . 𝐿, (8) where l is the total length of the drill string and pe is the specific pressure obtained from the ratio of the well geometry and the rheological parameters of the fluid. the difference between surge and swab is that the surge is the increased pressure on the well when the drill string down, and the swab is the decrease of pressure in the well when the drill column rises. for the calculation of (8) is necessary to find expressions that define the speeds in the three regions of fig. 7. the model for calculating the speed was developed by crespo and ahmed [12], for yield power law fluids. vector y assumes values between 0 and h, and is given in meters. for each value of y, there is a value for the velocity of the flow, indicated by the vertical arrows. the constant y1 is the size of region i and y2 is the sum of region i and region ii. region ii is within the limits y1 ≤ y ≤ y2 and has constant speed, the region i is within the limits 0 ≤ y ≤ y1 and region iii is within the limits y2 ≤ y ≤ h. the velocities in the region i and in the region iii fig. 7: annulus geometry. vary according to vector y [12]. for the velocity profiles in the regions of the mathematical model of crespo, dimensionless variables for speed are used as (9): 𝑉𝑖 ′ = 𝑉𝑖 𝑉𝑝 , 𝑖 = 1, 2, 3. (9) this variables are dimensionless because of the division between two quantities with the same unit of measurement. the same is true for the values of widths y1 and y2, as (10): 𝑦𝑖 ′ = 𝑦𝑖 𝐻 , 𝑖 = 1, 2. (10) values of y'i, are always between 0 and 1 because they are the results of the divisions between points in y and their maximum value, h. so, these values are dimensionless. the velocity profile in the region i is given by: 𝑉1 ′ = 𝑃[(𝑦1 ′ − 𝑦′)𝑏 − (𝑦1 ′ )𝑏 ], (11) where 0 ≤ y ' ≤ y ' 1. the velocity profile in the region ii is given by: 𝑉2 ′ = 1 − 𝑃(1 − 𝑦2 ′ )𝑏 , (12) where y ' 1 ≤ y ' ≤ y ' 2. the velocity profile in the region iii is given by: 𝑉3 ′ = 1 − 𝑃[(1 − 𝑦2 ′ )𝑏 − (𝑦′ − 𝑦2 ′ )𝑏 ], (13) where y ' 2 ≤ y ' ≤ 1. the values y ' 1 and y ' 2 are the widths shown in fig. 7 and are associated with regions of different velocity profiles of fluid. the exponent b used in (11), (12) and (13) takes account the fluid behavior index: in (11), (12) and (13) the value of the dimensionless pressure p is given by: 𝑃 = ( 𝑛 𝑛 + 1 ) ( 𝐻 𝑉𝑝 ) ( 𝛥𝑃 𝐿 𝐻 𝐾 ) 1𝑛 . (14) expression (12) is related to the pressure variation δp in the total length l of the drill string, p is a dimensionless vector. the total flow rate is the sum of the flow rate of the three regions given by [3]: 𝑞𝑡 ′ = ∫ (∫ 𝑉1 ′ 𝑑𝑦1 ′ + ∫ 𝑉2 ′ 𝑑𝑦1 ′ + ∫ 𝑉3 ′ 𝑑𝑦1 ′ ) 𝑑𝑥 ′. (15) solving the integral in (15), is obtained: 𝑞𝑡 ′ = −𝑃 [( 𝑏 𝑏 + 1 ) 𝑦1 ′𝑏+1] − [𝑃(1 − 𝑦1 ′ − 𝜒)𝑏 − 1] (1 − 𝑦1 ′ − 𝜒) + 𝑃 ( 𝑏 𝑏 + 1 ) (1 − 𝑦1 ′ − 𝜒)𝑏+1 − 𝑃𝑦1 ′ 𝜒 (16) where: 𝜒 = 2𝜏0 𝐻 𝛥𝑃 𝐿 , (17) and (1 − 𝑦1 ′ − 𝜒)𝑏 − (𝑦1 ′ )𝑏 − 1 𝑃 = 0. (18) in (18), the value of y ' 1 is obtained by iteration by newton-raphson method. relating the geometry of the well and the drill string (geometry of the annulus), it is possible to calculate the specific flow rate using [3]: 𝑞𝑒 ′ = −1 ( 𝑅1 𝑅2 ) 2 − 1 . (19) reconnecting the rheology of the fluid with the geometry of the annular space (p × q ' t), the interpolating polynomial is obtained: 𝑓(𝑞𝑡 ′ ) = 𝑃𝑒 . (20) with interpolating polynomial (20) and (19), is possible to find the value of p for the value of q ' t-=q ' e, that is, p(q ' e). pe is the specific pressure that will be used in (8). the variables v ' 1, v ' 2, v ' 3, y ' , y ' 1, y ' 2, y ' 3 and q ' t are dimensionless. d. algorithm to obtain the values of psu the following algorithm is used: i) the bilinear transformation given by the expression (4) is used to obtain the equivalent concentric plane and thus the constant value of h is obtained; ii) with the input parameters, rheological and geometric, the value of p is obtained from (14), for each given value of δp; iii) with p, the value of y ' 1 in (18) is calculated for the combination of the input values δp and vp; iv) the value of y ' 1 is replaced in (16), so, the value of the total flow q ' t is obtained for each value of l; v) soon, q ' e is obtained from (19); vi) the graph relating p and q ' t is generated, it presents the relation between the geometry of the annular space and the rheology of the fluid. this graph represents the relation between the geometry of the annular space and the rheology of the fluid. from this relation the interpolating polynomial is created (20) and the value of pe is obtained, (q ' e, pe) is the point on the graph (q ' t p) and vii) with the value of pe the values of the pressures surge and swab, psu, can be obtained using (4). the fig. 8 shows the flowchart produced. v. results as results, it is presented the calculation of the surge pressure in a dummy well using the usual methodology and the proposed methodology. a comparison was made between the values obtained for each methodology. a study was performed on the relationship between eccentricity and pressures values. the maximum eccentricity for the well geometry was calculated. a. case study 1: eccentricity × geopressions to relate the eccentricity to the geopressures it is necessary to have the input parameters. however, the thickness h, in the usual models is not uniform throughout the circumference of the coating column (there is eccentricity). in this way, it is proposed that the drilling of the well be monitored, and at each interval of time δt, determined by the user, the concentric geometry equivalent is obtained. all case studies use data from crespo, which are: i) geometry, considering the well being drilled by drilling column with length l=36 m, diameter d2=0.254 m and diameter of the coating column d1=0.508 m and ii) rheological, yield power law fluid with n=0.48, k = 0.74 pa.s n e τ0 = 3.11 n/m 2 . 1) calculation of geopressure considering traditional methodology: in order to carry out this case study, it is considered the drilling of the well whose radius of the coating column and the radius of the drilling column form concentric geometry. for calculation of the surge pressure, generated in the descent of the column with constant speed vp=0.1524 m.s -1 , the usual methodology ignores the value of eccentricity. then the thickness h of the annular space is only the subtraction of the radius of the coating column by the radius of the drilling column h = r1r2 = 0.127m. assuming, for all studies, δp = 1.38×10 7 pa, equivalent to 2000 psi at the bottom of the well, it is possible to obtain the vector p using (14), relating well geometry to fluid rheology. with vector p, it is possible to obtain the vector of the flow rate q ' t. using (19), it is possible to obtain qe=-0.33 and having the interpolating polynomial (20), it is possible to obtain pe=3.68. substituting the pe value in (4), obtains the surge pressure value, psu=1.84×10 5 pa, equivalent to 26,78 psi. in the real well drilling system, the drill string and casing column do not work concentrically. if there is a maximum eccentricity, when the drill string abuts the casing column, they can stick together, damaging drilling the well. table i provides the input and output values for the usual model, which does not consider the value of the eccentricity in the calculations. table i result considering the usual methodology symbol quantity value ψ eccentricity 0.1257 m n behavior index 0.48 k consistency index 0.74 pa.s n τ0 yield stress 3.11 n/m 2 r1 radius of the coating column 0.254 m r2 radius of the drilling column 0.127m psu surge 1.84 × 10 5 pa 2) calculation of geopressure with eccentric geometry: to use the geometry of the previous case considering the eccentricity, before using the mathematical model of crespo, the equivalent concentric geometry must be calculated. the values of r1 e r2 of the concentric fig. 8: algorithm. plane, are the new rays of the coating column and drilling column, respectively. as the conformal mapping do not change the magnitudes of the system, the values of δp remain the same for both eccentric and concentric geometry. for the geometry under study, there is maximum eccentricity ψmax = 0.1257 m. from this eccentricity, the value of the surge geopressure can be calculated using the conformal mapping in (4), which carries the eccentric plane of fig. 9 (a) in the concentric plane of fig. 9 (b). the new radius of the concentric plane are r1 = 0.2806 m for coating column and r2m= 0.2539 m for drill string. from this conformal mapping h = 0.0267 m is obtained. applying the calculations of the mathematical model of the annular space in the new concentric plane, fig. 9 (b), it is possible to get the value of surge geopressure of 1.56×10 6 pa, equivalent to 226.63 psi. using the usual methodology, the value of psu=1.84 ×10 5 pa or psu=26.78 psi. using the proposed methodology and taking into account the eccentricity, it is possible to obtain the value of psu = 226.63 psi. the proposed methodology finds, in this case, a value of approximately 8.46 times higher than the value found by the traditional methodology. a fact that occurs by not considering the eccentricity, which could cause serious problems at the moment of the maneuver. the table ii provides the input and output values for the case study considering the eccentricity and using the proposed methodology. table ii result considering constant eccentricity symbol quantity value ψ eccentricity 0.1257 m n behavior index 0.48 k consistency index 0.74 pa.s n τ0 yield stress 3.11 n/m 2 r1 radius of the coating column 0.254 m r2 radius of the drilling column 0.127 m r1 radius of the coating column after conformal mapping 0.2806 m r2 radius of the drilling column after conformal mapping 0.2539 m psu surge 1.56 × 10 6 pa considering the same input parameters, but with ψ = 0.062 m, the new geometry is calculated. the equivalent concentric plane, has r1 = 0.4664 m and r2 = 0.2539 m for the coating column and the drilling column, respectively, as illustrated in fig. 10. applying the calculations of the mathematical model in the new concentric plane, fig. 10 (b), it is possible to reach the value of psu = 8.642 ×10 4 pa equivalent to psu = 12.51 psi. this result shows the dynamics of the effect of the eccentricity on the pressures at the bottom of the wells, since all the configurations were maintained, varying only the eccentricity. for average eccentricity, psu= 12.51 psi, 2.14 times lower than the value obtained using the traditional methodology. the table iii provides the input and output values for the case study with ψ = 0.062 m and considering the eccentricity and using the proposed methodology. fig. 9: real system transformation (a) eccentric plane e (b) concentric plane. fig. 10: real system transformation (a) eccentric plane and (b) concentric plane. table iii result considering constant eccentricity symbol quantity value ψ eccentricity 0. 062 m n behavior index 0.48 k consistency index 0.74 pa.s n τ0 yield stress 3.11 n/m 2 r1 radius of the coating column 0.254 m r2 radius of the drilling column 0.127 m r1 radius of the coating column after conformal mapping 0.4664 m r2 radius of the drilling column after conformal mapping 0.2539 m psu surge 8.642 × 10 4 pa b. case study 2: calculation of geopressures with variation in eccentricity considering also the situation in which the rotary movement of the drill descending into the well causes the drill string and the casing column to generate different eccentric geometries for each depth, that is, ψ is different for each value of δp/δl, it is possible to calculate the surge geopressure in each section of the well. in this case, it is also considered the speed of descent varying in time, speed increases with depth. for this case it varies from vp = 0 to vp = 0.1524 m.s -1 . for this case, the pressure δp varies from δp = 1.37×10 6 pa, on the surface until δp = 1.38×10 7 pa, at the well bottom. it is observed that there is variation in the value of the eccentricity because the system is dynamic, and the eccentricity varies in time, so it is possible, using the conformal mapping and considering the variation of the eccentricity 0 < ψ < 0.1257m, calculate the new h values for the eccentricity vector. it is possible to produce the surface that relates the eccentricity, the surge geopressure and the descent speed of the drill, as shown in fig. 11. it is observed in fig. 11, that higher the descent speed vp, greater the increase in surge pressure, in this case it would be feasible to establish the maximum velocity of 0.1 m/s, from this point surge pressure begins to grow exponentially. in the same way, another surface can be produced that relates ψ, surge and δp, as shown in fig. 12. fig. 12 shows that the higher the pressure δp inside the well, the higher the surge pressure. fig. 12 indicates whether the injected fluid levels are suitable for well geometry. the traditional method and the proposed method start from the same point when there is concentricity. however, for the input data presented (geometry and rheology) and considering variations in eccentricity values and δp, it can be observed that the values of surge for the traditional method are higher than for the proposed method until a certain value. fig. 13 illustrates the comparison between the traditional method and the proposed method, which considers eccentricity. fig. 12: ψ × δp × psu fig. 11: psu × vp × ψ fig. 13: comparison between surge values obtained if the value of surge pressure is greater than the real one, can cause damage to the formation, therefore, the operating window that contains the pressures and speeds, ideal for safe operation of the maneuvers, has maximum and minimum boundary. if the calculated pressure value is incorrect, hydrostatic fluid will be injected erroneously. the pore pressure may be greater than the rock fracture pressure, and may cause the kick effect. therefore, it is important to know the operating window, the velocity values that the well supports and the actual geopressure values. vi. conclusions it is possible to observe that the eccentricity produces incoherence between the adopted model and the real system. with the proposed methodology it is possible to calculate the maximum and minimum pressures, related to the maximum and minimum eccentricities that alter the values of h in the well drilling. the flow depends on pressure range, length of the drill string, well and drill string radius. in turbulent flows the speed of the drilling column and hydrostatic fluid is of great importance for the calculation of the variation of pressure in the bottom of the well. pressure change increases exponentially with increasing drilling column speed. the increase in pressure also depends on the depth of the well. pressure variation also depends on annular space, flow rate increases with increasing thickness h. to reduce pressure changes it is important to monitor speed and eccentricity between columns as well as amount of hydrostatic fluid injected into the well. the proposed method is able to calculate the surge and swab geopressures for non newtonian fluids at the bottom of the well, considering the variation of the eccentricity of the same. references [1] u. f. silva, m. i. q. júnior, g. p. furriel, a. h. f. silva, a. j. alves, and w. p. calixto, “application of conformal mapping in the calculation of geological pressures,” chilecon, 2015. [2] l. a. s. rocha and c. t. azevedo, geopressures and settlement coatings columns (in portuguese); 2 edition. interciência, 2009. [3] f. crespo and r. ahmed, experimental study and modeling of surge and swab pressures for yieldpower-law drilling fluids. phd thesis, university of oklahoma, 2011. [4] o. n. mbee, “prediction of surge and swab pressure profile for flow of yield power law drilling fluid model through eccentric annuli,” master’s thesis, african university of science and technology, 2013. [5] z. bing, z. kaiji, and z. ginghua, “theoretical study of steady-state surge and swab pressure in eccentricc annulus [j],” journal of southwest china petroleum institute, vol. 1, 1995. [6] w. calixto, “application mapping as at the carter factor calculation (in portuguese),” master’s thesis, federal university of goias, 2008. [7] w. f. rogers, “compostion and properties of oil well drilling fluids, ”gulf publishing co., houston, tx, 1948. [8] w. c. frison, “apparatus and method for treatment of wells,” may 9 1989. us patent 4,828,033. [9] d. amorim júnior and w. iramina, “maneuvers for exchanging drills in oil wells (in portuguese),” tn petróleo, vol. 61, pp. 190–194, 2008. [10] r. caenn and g. r. gray, “drilling fluid and completion; 6 edition (in portuguese),” editora elsevier, pp. 1–5, 2014. [11] a. g. karlsen, “surge and swab pressure calculation: calculation of surge and swab pressure changes in laminar and turbulent flow while circulating mud and pumping,” institutt for petroleumsteknologi og anvendt geofysikk, 2014. [12] f. crespo and r. ahmed, “a simplified surge and swab pressure model for yield power law fluids,” journal of petroleum science and engineering, vol. 101, pp. 12–20, 2013. [13] w. p. calixto, b. alvarenga, j. c. da mota, l. d. c. brito, m. wu, a. j. alves, l. m. neto, and c. f. antunes, “electromagnetic problems solvng by conformal mapping: a mathematical operator for optimization (in portuguese),” mathematical problems in engineering, 2011. [14] h. kober, dictionary of conformal representations, vol. 2. dover new york, 1957. [15] j. c. v. machado, rheology and fluid flow emphasis on oil industry (in portuguese); 2 edition. editora interciência, 2002. i. introduction ii. well drilling process a. pressures inside the well 1) pore pressure: in rock formation, only part of the total volume is occupied by solid particles, which settle to form the structure. the remaining volume is often called voids, or pores, and is occupied by fluids. pore pressure, often referred as fo... 2) hydrostatic pressure: hydrostatic pressure is exerted by the weight of the hydrostatic column of fluid, being function of the height of the column and the specific mass of this fluid. the drilling fluid has as one of its main objectives to keep the... b. kick and blowout effects c. operating window iii. conformal mapping a. conformal mapping definition iv. methodology a. conformal mapping b. rheological parameters c. mathematical model for surge and swab calculation d. algorithm v. results a. case study 1: eccentricity × geopressions 1) calculation of geopressure considering traditional methodology: in order to carry out this case study, it is considered the drilling of the well whose radius of the coating column and the radius of the drilling column form concentric geometry. for... 2) calculation of geopressure with eccentric geometry: to use the geometry of the previous case considering the eccentricity, before using the mathematical model of crespo, the equivalent concentric geometry must be calculated. the values of r1 e r2 ... vi. conclusions references  transactions on environment and electrical engineering issn 2450-5730 vol 2, no 1 (2017) © bruno g. menita , giordani p. medeiros, jose l. domingos, elder g. domingues, aylton j. alves, adriano f. faria, marcio l. s. miguel and wesley p. calixto abstract— the purpose of the present study is to evaluate gains through measurement and verification methodology adapted from the international performance measurement and verification protocol, from case studies involving energy efficiency projects in the goias state, brazil. this paper also presents the stochastic modelling for the generation of future scenarios of electricity saving resulted by these energy efficiency projects. the model is developed by using the geometric brownian motion stochastic process with mean reversion associated with the monte carlo simulation technique. results show that the electricity saved from the replacement of electric showers by solar water heating systems in homes of low-income families has great potential to bring financial benefits to such families, and that the reduction in peak demand obtained from this energy efficiency action is advantageous to the brazilian electrical system. results contemplate also the future scenarios of electricity saving and a sensitivity analysis in order to verify how values of some parameters influence on the results, once there is no historical data available for obtaining these values. index terms—energy efficiency, geometric brownian motion, monte carlo simulation, performance measurement and verification, solar water heating. i. introduction he use of solar energy in residential water heating has growing acceptance as an alternative or supplementary way to the heating provided by electric showers. recently, brazilian government programs have promoted the use of solar water heaters in homes of low-income families, such as energy efficiency projects (eep) of electricity distribution companies. these eep are part of energy efficiency program of brazilian electricity regulatory agency aneel [1]. the application of the international performance measurement and verification protocol (ipmvp) is paper submitted for review on dec 6th 2016, accepted on jan 19th , 2017. 1experimental & technological research and study group (next), federal institute of goias (ifg), goiania, goias state, brazil (e-mails: brunomenita@gmail.com, giordanipmedeiros@gmail.com). 2power quality sector celg distribution s/a (celg-d) goiania, goias state, brazil. 3school of electrical engineering, mechanical and computer (emc), federal university of goias (ufg). mandatory as a reference for measurement and verification (m&v) among other steps involved in evaluation of electricity saving and peak demand reduction of an eep. the ipmvp establishes rigorous criteria that lead many eep to economic unviability, mainly due to long periods of measurement. to solve this problem, the brazilian association of electricity distributors (abradee) developed m&v procedures from ipmvp to apply in eep by final use, with contributions of consultancies and partnerships. thus, a new m&v methodology by end use has been defined and approved by aneel, and passed on to electricity distribution companies in september 2014. the annual consumption of electricity avoided, which represents annual electricity saving by the eep depends on some factors that have random behavior over time such as: the number of residents of the housing project that received the energy efficiency action (eea), the bath habit of these people, changes in family income, and acquisition or replacement of electrical appliances in these houses. this study aims to apply the m&v methodology adapted from ipmvp in order to get results and conclusions in terms of electricity saved and peak demand reduced from the replacement of electric showers by solar water heating systems, as eea. the innovative aspects of this research shall be highlighted, once the application of the simplified methodology is recent and helpful for future improvements of it. this paper presents also the stochastic modeling for generation of future scenarios of electricity saving from these eep. the random variable annual electricity saving is modeled by using the stochastic process called the geometric brownian motion with mean reversion (gbm-mr). ii. project definitions the overall objective of the project consists in obtain results that can express the effects of the use of solar water heating in housing units of low-income families, to residents, to the brazilian electrical system and finally to society. all of it by means of application of m&v adapted from ipmvp in eep of celg distribution s/a (celg-d), the electricity distribution company of goias state. the specific objectives are: a) evaluate the average monthly savings of electricity to the consumers; b) evaluate the impact on demand of electricity at bruno g. menita1, giordani p. medeiros1, jose l. domingos1, elder g. domingues1, aylton j. alves1, adriano f. faria2, marcio l. s. miguel2 and wesley p. calixto1 3 methodology for measurement the energy efficiency involving solar heating systems using stochastic modelling t peak hours to the brazilian electrical system; c) perform diagnosis and contributions to improvement of the m&v procedure from ipmvp adapted; d) generate future scenarios of electricity saving by stochastic modelling associated with the monte carlo simulation technique. three eep of celg-d related with solar water heaters in homes of low-income families in goias state have been selected as case studies: municipality of itumbiara, real conquista residential and orlando de morais residential – both located in goiania that is capital of goias state. for the two first eep it was used data measured by celg-d in september 2008 and june 2010, respectively, prior to the installation of solar heating systems. about the last one case, it was selected because it was in phase of installation of solar heating systems in the second half of 2014. thus it was possible to carry out m&v in the last residences not yet covered by the eea. fig. 1 shows one of the solar water heating systems installed in homes of low-income families. initially some settings for the eea are performed, as established in ipmvp. the following parameters are defined as in [2] and [3]: the measurement border, key performance parameter, independent variable, interactive effect, static factor, baseline period and reporting period. the measurement boundary corresponds to the limits within which it is desired to check the energy savings and the reduction of peak demand. depending on the eea, the measurement boundary may be the whole installation or only equipment(s) or system(s) responsible for eea. for these cases, the measurement boundary is defined as the set of power supply circuits of showers and resistors, in other words, isolating the electric shower. the key parameters are those that are directly related to the consumption of electricity. the key parameters to be measured are the electric power during use of the shower and bath time. independent variables are the variables that, when there is correlation with the energy consumption of the installation or system, explain the variation of this consumption, being used for the necessary adjustments, through a linear regression analysis. the outdoor temperature is defined as independent variable. it is obtained from measurements of the nearest station of brazilian national institute of meteorology inmet. according to the ipmvp, some energy effects affecting eea may occur outside the measurement boundary. these are called interactive effects. according to the m&v guide, it is necessary to define them, if any, and decide whether they will be estimated or ignored. in this context, it is considered that there are no interactive effects, once loss in power supply circuits of the house and of the shower are considered negligible. static factors are those that define the energy consumption pattern of the installation or system of this, such as installation size and number of people. static factors are not considered in this case because of standard size residences and short measurement period, as described below. it is also necessary to define, according to the ipmvp, the baseline and economic determination periods. the first corresponds to the measurement period prior to the installation of the equipment responsible for esa, and should represent a full duty cycle of the components of the measurement boundary. the same applies to the period of determination of the economy, however, considering it after the implementation of eea. the baseline period is: seven days in eep orlando de morais; eleven days in real conquista residential; and sixteen days in itumbiara. it is important to emphasize that these shorts measurement periods are justified by the methodology to be used for m&v adapted from ipmvp to the reality of eep in order to obtain projects economically viable. measurements are not performed in the reporting period because the eea results in installation of solar heaters systems and consequently in the removal of electric showers. the ipmvp offers four options for determining energy savings. option a corresponds to the measurement of at least one of the key parameters that define energy use by eearelated systems isolated, and estimation of the other key parameters. measurement periods can range from a short-term period to continuous measurement. option b corresponds to the measurement of all the key parameters that define energy use by esa-related systems isolated, ranging from a short-term to a continuous period. it can be used, for example, in pilot projects involving new technology or methodology. in option c, the energy consumption assessment involves the whole facility, with continuous measurement in the period of determination of the economy. it is generally the option of lower cost, however, one must be more rigorous in relation to the static factors, since the measurement border is wider. option d involves assessing the energy consumption of the entire facility using simulation calibrated from actual data from distributor power bills. used in new installations, where the model simulates data that does not exist for parameters of the baseline period. from the above definitions, it is chosen option a to determination of electricity savings and peak demand reduction from ipmvp, considering also that one of the key parameters has to be estimated. fig. 1. solar water heating system in home of low-income family iii. methodology the first step related to the methodology used is the definition of sampling. it involves setting the initial number of samples in two ways, one determined according to the relation between the initial population and sample, given by nbr 5426 sampling plans and inspection procedures by attributes considering severe inspection level (level 1), and second calculated by (1) and (2) [3]. 𝑛0 = 𝑧2 × 𝑐𝑣2 𝑒2 (1) 𝑛 = 𝑛0 × 𝑁 𝑛0 + 𝑁 (2) in (1), n0 represents the size of initial sample, z is a standard value of normal distribution as "t-table" available in ipmvp (it shall be adopted value of 1.96, equivalent to 95% confidence), cv corresponds to the coefficient of variation (it shall be adopted from previous projects or 0.5 if no existent) and e is the level of precision desired (it shall be adopted 10%). in (2) n is the size sample for small populations and n is the population size. the results obtained through the two above methods provide a basis for the decision on the initial sample, considering equipment, people, costs and time restrictions. after obtaining the values measured in baseline period, shall be calculated the precision obtained with the initial sample size, for each key parameter. this is calculated by (1), using the coefficient of variation calculated from the data obtained. if the precision of 10% is not reached (if value is greater), it is necessary increase the number of samples, thereby performing iterative process until the desired precision is obtained [5]. after obtaining the keys parameters data in the baseline period and temperatures corresponding to the dates of measurements, it is assessed if there is a correlation between the parameters and independent variable, using linear regression analysis. the first evaluation criterion calculated is the determination coefficient (r2), calculated by (3), where �̂�𝑖 is the key parameter value adjusted by the model to a given point using the corresponding value of the independent variable. 𝑅2 = ∑ (�̂�𝑖 − �̅�) 𝑛 𝑖=1 2 ∑ (𝑦𝑖 − �̅�) 2𝑛 𝑖=1 (3) the second evaluation criterion is the coefficient of variation of root-mean squared error cv(rmse), which measures the forecast accuracy. the calculus is done by (4), dividing the standard error of the estimate by the mean of the electricity consumed. in (4) p is the number of independent variables. 𝐶𝑉(𝑅𝑀𝑆𝐸) = √ ∑ (�̂�𝑖 − 𝑦𝑖 ) 2𝑛 𝑖=1 𝑛 − 𝑝 − 1 �̅� (4) the last evaluation criterion is the t-statistic, a statistical test to determine whether an estimate has statistical significance due to the possibility of variation of regression coefficients. the t-statistic is calculated by (5), dividing the regression coefficient (slope) by the standard error of each coefficient of the regression model. 𝑡 − 𝑠𝑡𝑎𝑡𝑖𝑠𝑡𝑖𝑐 = 𝑏 √ ∑ (𝑦𝑖 − �̅�) 2𝑛 𝑖=1 (𝑛 − 2) ⁄ ∑ (𝑥𝑖 − �̅�) 2𝑛 𝑖=1 (5) where b represents the regression coefficient (for the 1-unit increase of the independent variable there is increase of "b" units of key parameter), 𝑥𝑖 corresponds to the value of independent variable and �̅� is the mean of the values of independent variable. according to [6], maintaining the criteria of the case passed on to electricity companies there is correlation between electricity consumption and outdoor temperature if at least two of the following three criteria are met: r2 greater than 0.75; cv(rmse) less than 5%; t-statistic greater than 2. if correlation is verified and validated, regression is used as a basis for setting baseline and reporting periods, to leveling electricity consumption in the measurement boundary without influence of outdoor temperature. thus, the calculation of electricity savings shall be done by (6) [2]. 𝐸𝑆 = 𝐶𝑏𝑙 − 𝐶𝑟𝑝 (6) where es corresponds to the electricity saving, cbl is the electricity consumption in baseline period adjusted to fixed conditions and crp represents the electricity consumption in reporting period adjusted to fixed conditions. if there is no correlation the same equation is used without adjustments. calculation of reduction of electricity demand at peak hours does not require adjustments, occurring through (7) [2]. 𝐷𝑅 = 𝐷𝑏𝑙 − 𝐷𝑟𝑝 (7) where dr corresponds to the demand reduction in peak hours, dbl represents the mean of peak demand in baseline period and drp is the mean of peak demand in reporting period. several stochastic processes have been used in the brazilian electricity market to model the uncertainties present in this, such as the spot price, affluence, electrical demand and consumption of electricity. these random variables can be modeled as time series by using the monte carlo simulation technique, associated with the stochastic process called random walk [4]. once the annual electricity saving obtained by eep is dependent of variables with random behavior, it is necessary to generate future scenarios of this random variable by using an adequate stochastic process. the geometric brownian motion (gbm) is a particular case of ito's process, which in turn corresponds to the generalization of brownian motion with drift [8]. according to [9], when a random variable follows a gbm, their values tend to diverge from the original starting point, since the variance grows linearly with time. in this context, the process of bgm with mean reversion, also called ornstein-uhlenbeck process, forces the values obtained to the random variable over time to revert in direction of the equilibrium position, i.e., the starting value (mean value, for example). according to [7], there is a force of reversion acting on the random variable pulling it to a long-term equilibrium level. the random variable annual electricity saving by eep (ee) can be obtained by (5) that represents the model by the gbm with mean reversion. 𝐸𝐸𝑡+1 = 𝐸𝐸𝑡 . 𝑒 {[η.(ln ee̅̅ ̅̅ −ln eet)− 1 2 .𝜎2].∆𝑡+𝜎.𝜑.√∆𝑡} (8) where η is the mean reversion speed, 𝐸𝐸̅̅ ̅̅ represents the mean value of the random variable, σ is the constant that represents the percentage volatility random variable, t represents the time and φ corresponds to a random variable with standard normal distribution – n(0,1). the stochastic behavior of the random variable can be represented by curves containing the values obtained for annual electricity saving on the time horizon defined, as a family of time series, using the monte carlo simulation. the stochastic process can be also represented by the evolution of the probability density function (pdf) of the random variable over time [7]. iv. results the step of the methodology regarding to the sampling definition was included only in the case study involving eep orlando de morais residential, once there were definitions before m&v, which there were not possible in relation to the other cases, whereas that measurement data were obtained in past periods. the key parameters electric power and bath time were obtained by measuring energy consumed by electric showers. the meters used were calibrated at the measurement laboratory of celg-d, using reference pattern meter tracked by the brazilian calibration network. these meters were installed directly between the shower and electric installation of the residence, on the other hand, the fig. 2 shows the process of extracting the measurement data from the mass memory of the meter. the initial sample size corresponds to the product of the number of residences with the number of days in the baseline period. thus, based on estimates from the equivalence given by nbr 5426 and (1) and (2), and also considering the restrictions on the available number of meters and the short time until the installation of solar heating systems on the latest residences in orlando de morais residential, the initial sample in this eep consisted of 77 day.residence, which corresponds to 11 homes with measurements in 7 days of the baseline period. in this way, once that the used meters measure energy consumed by the shower over time, it is possible to transform the obtained values in electric power and bath time, that is, the defined key parameters. this procedure is recommended in the methodology in order to have the sampling level of precision obtained, for the variability of power values is smaller than electricity consumed values. it should be pointed out that without this transformation, the range of desired precision level leads to excessive increase in the number of samples, which can make an eep economically unviable. table i shows the results of obtaining the precision level of sampling in the three case studies. in order for a sampling to be statistically valid, the accuracy level must be less than or equal to 10% for at least one of the key parameters, such parameters being considered measurement. the bath-time key parameter is considered as an estimate (not measurement), since the desired level of accuracy is not obtained, besides the fact that the meter used does not perform continuous measurement of bath time, from the records of energy consumed by the shower during the intervals of 5 minutes. the sampling precision levels obtained for key parameters power and bath time are 18% and 27%, respectively, thus not reaching the target of 10%. thus, there would be necessary to increase the number of samples which was not possible due to fig. 2. extraction of measurement data table i data comparison and results of energy economy of pee number of samples orlando de morais real conquista itumbiara number of residences with m & v. 11 8 16 number of measurement days. 7 11 4 number of samples (residence.day). 77 88 64 precision level parameter key: electrical power. 18% 8% 12% precision level parameter key: bath time. 27% 13% 17% the completion of installation of the solar heating systems in the remaining houses. the continuity of methodology phases even not reaching the precision level is justified by the experimental nature of the project. regarding to data relating to m&v in eep real conquista residential and municipality of itumbiara, 8% level of precision was obtained for the key parameter electric power in the first, thereby achieving the desired criteria. however, this did not occur in the case of eep municipality of itumbiara, while that the obtained level of precision was 12%. as the data correspond to measurements taken in the past, in residences where the installation of solar heating systems was completed, the number of samples could not be increased. among the key parameters, electric power has been defined as the parameter most likely to be influenced by the independent variable. in the three case studies it was not found correlation between the electric power on shower use and outdoor temperature in the baseline period, once the criteria have not been met. as an example, the results of calculations relating to the eep orlando de morais residential were: r2 equal to 0.2% (less than 75.0%; cv(rmse) equal to 78.4% (greater than 5.0%); t-statistic equal to 0.35 (less than 2.00). the average daily electricity economy per housing unit was calculated by (3), disregarding the adjustments for not exist correlation between key parameter and independent variable. table i shows the average daily power measured, the average daily bath time, average outdoor temperature in baseline period, averages of daily and monthly consumption with electric showers – which corresponds to savings as result of eea, the main variable for comparison and the annual savings considering all the residences contemplated by eep, for the three case studies. by the current conventional electricity tariff for low-income homes of celg-d – r$ 0,26/kwh – for consumption between 31 and 100 kwh (once the average consumption of low-income homes is 77 kwh) and considering the exchange between brazilian currency and us dollar in july 2015, it is calculated the monetary value of those annual savings, which are presented in table ii. it is possible to observe that the average daily consumption of electricity avoided per housing unities significantly higher in eep real conquista and in itumbiara. such as the average electric power as the average daily bath time are lower in eep orlando de morais. the daily bath time can be a result of the amount of baths in the residences, as well as local routine in relation to bath time. to check these two influences, it would be necessary to consider data related to static factors and habits of the residents of the homes included in the samples, which was not possible for the eep real conquista and itumbiara. thus, the importance of the survey and storage of behavioral information about the families’ routine of the sample of an eep is shown. so it is possible to do comparisons between results, taking into account the conditions and realities of the population of each location evaluated. the difference between the average daily electric power obtained in the eep orlando de morais and those obtained in the others case studies can also be explained by the electric shower position used during the measurement period (power level). while analyzing the average outdoor temperature in days of measurement of each eep, there is large difference between eep orlando de morais and others. in this place, the average outdoor temperature was 30.5°c, while considering the days of measurement in eep real conquista and itumbiara the values were of 22.9°c and 24.0°c, respectively, which may have influenced by the position associated to a higher power used in electric shower and therefore the highest average daily electric power obtained from each eep. it also contributes to the fact that such mean values are obtained considering even samples in which the daily values of these key parameters correspond to zero, that is, where there was no record of electricity consumption by the shower on the day of measurement at the residence. this occurred in 19 of the 77 samples from this case study, either because there was no bath on the day at the residence or more likely because, due to the high temperature, the bath was taken with unheated water, thus not being recorded. these samples with zero-parameter values reduce the mean values obtained, but by disregarding them, the energy savings obtained with the pee can be overestimated. when disregarded, the average power at the pee orlando de morais becomes 2.04 kw, which is closer to the power of a common shower in the "summer" position, and the average bath time becomes 12.46 minutes. that is, when the average values of the key parameters are analyzed, they may seem low, but it is guaranteed that energy savings and reduction in peak demand are not overestimated, thus respecting one of the ipmvp principles. to obtain the average electric power relating to the use of showers in peak hours the average power measured has been multiplied by the average total bath time of this time range, and this value has been divided by three, since this is the table ii results of electricity saving in eep applications results orlando de morais real conquista itumbiara average daily electric power per house (kw) 1.56 2.79 2.79 average total daily bath time per house (minutes) 9.87 33.21 34.57 average daily consumption of electricity avoided per house (wh) 256.38 1545.34 1606.01 average outdoor temperature in baseline period (oc) 30.5 22.9 24.0 average monthly consumption of electricity avoided per house (kwh) 7.69 46.36 48.18 number of homes in the eea 544 478 1080 annual electricity consumption avoided with eep (mwh) 50.21 265.92 624.42 monetary values of annual savings (us$) 4054.10 21471.96 50418.69 number of hours of peak interval. then the average value of electric power taken from peak demand with eea in one residence in each eep has been achieved, and considering the population of the studies, the reductions in peak demand reached from the analyzed eep have been calculated. the results are shown in table iii for the three case studies. in the same way that the average daily economy of electricity, the average reduction in peak demands for one residence is significantly higher with eep real conquista and itumbiara. possible causes of this difference are the same as mentioned in comparing results related to electricity savings (outdoor temperature and power selected for shower operation), since both the average electric power and the average bath time at peak hours are smaller in eep orlando de morais. table iv shows the comparative of the uncertainty calculated for the key parameters used to obtain the avoided electric energy consumption, in each case study. the observed difference reflects the dispersion of the measurement values that impact the sampling uncertainty, especially in relation to the pee orlando de morais, since the modeling uncertainty was not considered and the measurement uncertainty is the same in all cases. the comparison of the uncertainty calculated for the key parameters used to verify the reduction in peak demand in the three case studies is presented in table v. the annual consumption of electricity avoided depends on some factors that have random behavior over time. as already mentioned, the random variable annual electricity saving is modeled by using the stochastic process called the geometric brownian motion with mean reversion (gbm-mr). for generation of future scenarios of electricity saving from an eep, it is defined as the starting value to the random variable electricity saving (in the initial year) the annual electricity consumption avoided by the eep real conquista residential (265.92 mwh). table vi presents the input data for simulation. the volatility (σ) of the random walk is defined as the estimated standard deviation of annual electricity saving, obtained by multiplying the coefficient of variation calculated with the measured values of electric power in the eep real conquista residential (0.39) for the electricity saving in the initial year (mean). for the mean reversion speed (η) it is assigned the value 0.50. both parameters of the random walk should have assigned values based on historical data. however, considering the absence of these data, it was not possible to perform appropriate statistical analysis to obtain these values, which led to the aforementioned assignments, considering the pioneering nature of this study. fig. 3 presents the simulation results for one scenario of the random variable electricity saving, which shows how the random walk can occur along the horizon. fig. 4 shows the pdf for each year of the study horizon, as another way of representation of future scenarios for annual electricity saving obtained by the eep real conquista residential. the red line shown in this figure represents the mean of the values obtained from time series, in each year, which remains around 265 mwh. table iii results of peak demand reduction in eep applications results orlando de morais real conquista itumbiara number of homes in the eea 544 478 1080 average daily electric power per house (kw) 1.96 2.46 3.08 average daily bath time at peak hours per residence (minutes) 7.96 19.82 19.09 average reduction in the peak demand per residence average (kw) 0.09 0.27 0.33 reduction in the peak demand per residence with eep (kw) 47.28 129.29 352.92 table v comparative of the uncertainty calculated for reduction in the demand of tip in the case studies relative uncertainty of key parameters. orlando de morais power (kw) 1,96 ± 10% usage time (min / day) 7,96 ± 15% real conquista power (kw) 2,46 ± 8% usage time (min / day) 19,82 ± 12% itumbiara power (kw) 3,08 ± 8% usage time (min / day) 19,09 ± 11% table vi input data for simulation average value (mwh) 265.92 volatility of the random walk (mwh) 103.71 mean reversion speed 0.50 time horizon of simulation (years) 10 interval between simulation periods (years) 1 number of scenarios 2000 table iv comparison of uncertainty calculated for energy savings in case studies relative uncertainty of key parameters. orlando de morais power (kw) 1,56 ± 18% usage time (min / day) 9,87 ± 27% real conquista power (kw) 2,79 ± 8% usage time (min / day) 33,21 ± 13% itumbiara power (kw) 2,79 ± 12% usage time (min / day) 34,57 ± 17% in the simulation results presented in fig. 4 the values of the mean reversion speed (η) and the volatility (σ) of the annual electricity saving were assigned once there are no historical data of them until this moment. to verify the influence of these parameters on the results, sensitivity analysis was performed. for this objective, it was adopted a variation range of 0.00 to 265.92 mwh with step of 13.30 mwh for the volatility (σ), and range of 0.10 to 10.00 with step of 0.10 for the mean reversion speed (η) , both for the year 3 of the initial horizon of the simulation. fig. 5 shows the behavior of the maximum and minimum values of the annual energy for the 2000 scenarios obtained for the year 3, in function of the variation of the standard deviation and the mean reversion speed. this figure represents the stochastic behavior of the electricity saving for a large range of situations. as expected, there is an increasing of the peak-to-peak amplitude of annual electricity saving with the increasing of both the standard deviation and the mean reversion speed. the behavior of the expected value and the standard deviation of the annual electricity saving for the 2000 scenarios in the year 3, in function of the variation of the volatility and the mean reversion speed is shown in fig. 6. as the variation of the parameters mentioned in this sensitivity analysis results in some extremely high values of annual electricity saving, fig. 5 and fig. 6 have the z axis graphically limited to 1000 mwh/year, in order to highlight the behavior throughout the variations. it can be seen by these figures that increasing the standard deviation leads to greater spread of results, however, it is with the increase of the mean reversion speed that is observed a sharp increase of the mean and standard deviation of the 2000 series. such behaviors are also visualized through fig. 7 and fig. 8. in the first, the reversal velocity was maintained at the constant at 0.5 and the standard deviation representing the volatility was varied. in fig. 8, the standard deviation fig. 3. behavior of the annual electricity saving over time – one scenario 1 2 3 4 5 6 7 8 9 10 240 250 260 270 280 290 300 310 320 year a n n u a l e le c tr ic it y s a v in g [ m w h ] fig. 4. probability density function of the annual electricity saving 0 100 200 300 400 500 0 1 2 3 4 5 6 7 8 9 10 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 annual electricity saving [mwh] year p d f fig. 6. mean and standard deviation of electricity saving by variation of σ and η 0 50 100 150 200 250 300 0 1 2 3 4 5 6 7 8 9 10 0 200 400 600 800 1000 standard deviation volatility [mwh/year] mean reversion speed e le c tr ic it y s a v in g [ m w h /y e a r] mean standard deviation fig. 5. amplitude of electricity saving by variation of σ and η 0 50 100 150 200 250 300 0 1 2 3 4 5 6 7 8 9 10 0 200 400 600 800 1000 standard deviation volatility [mwh/year] mean reversion speed e le c tr ic it y s a v in g [ m w h /y e a r] maximun value minimun value (volatility) was fixed at 103.71 mwh and the rate of reversion was varied to the mean. it is also possible to verify in this that the standard deviation of the 2,000 series reaches the lowest value when the speed is 1, that is, when the slope of the trend line formed by historical data is 45º degree or 135º degree. v. conclusions first of all it is possible to evaluate that these eep resulted in benefits for families. it was found that these savings are approximately 10%, 60% and 62% of the average monthly electricity consumption in low-income residences in the goias state, from eep orlando de morais residential, real conquista residential and municipality of itumbiara, respectively. this saving has great potential to mitigate the increase of the monetary values of electric bills that have recently been taking place in brazil. another evaluated aspect is that the reduction in peak demand obtained with the eep. considering estimate done with all residences with the benefit of the eea, the reduction represents 0.054% of the maximum peak demand in the goias state in 2014. it was found that the cost of this reduction in demand is less than the cost of electricity generation in us$/kw, considering estimation performed with data results of electricity auctions published in august 2015 by electrical energy commercialization chamber (ccee). therefore it can be concluded that, added to the reduction achieved with eaa of other end uses in eep, such as changing light bulbs and appliances for more efficient ones, there are contributions to the postponement or the relocation of investments by electricity companies. the application of the m&v procedure adapted from the ipmvp in the case studies of this work allowed to obtain contributions for future applications in pee of the celg-d and other distributors of electric energy of brazil. first, from the results obtained for the level of precision of the sampling and for the combined uncertainty in the case studies of this work, the importance of the planning of the measurement considering the process of obtaining sampling is highlighted. the adoption of the largest number of samples obtained through the two mentioned estimates, i.e. through nbr 5426 and the expressions provided by the ipmvp, increases the chances of reaching the level of precision required without the need for additional measurement. if there are constraints on the number of meters to comply with the initial defined sampling, it is necessary to consider complementary measurement periods in the m&v process planning. the follow-up of the m&v stages in the orlando de morais pee also made it possible to conclude on the importance of the alignment of the periods of the stages of a pee, more specifically in relation to the measurement periods in the baseline and installation of the solar water heating system in the residences, in order to avoid the loss of samples caused by the installation of the heating system during the period of measurement of the key parameters for the electric shower, which also impairs the level of sampling precision required for the reliability of the results. short measurement periods are essential for the viability of m&v and consequently of the pee, being important adaptations for the current m&v procedure. however, considering the results obtained by checking the correlation between the ambient temperature and the electric power of the shower, it is verified that the variation of the independent variable will hardly explain the variation of the key parameter. however, the comparison between pee results whose measurement occurred and different times of the year, the ambient temperature may be more influential on such results, as could be verified through the case studies. the results presented for pee residencial orlando de morais also reinforce the importance of future work involving the development of propagation methodology for the full year of energy saving and reduction of peak demand verified in a certain period of the year, thus considering seasonality. the measurements performed in october in this case study can be considered atypical, since the mean values of the key parameters were significantly lower than the values obtained in the other case studies. fig. 7. mean and standard deviation of electricity saving by variation of σ 0 50 100 150 200 250 300 0 50 100 150 200 250 300 standard deviation volatility [mwh/year] e le c tr ic it y s a v in g [ m w h /y e a r] mean standard deviation of 2000 series fig. 8. variation of mean and standard deviation of the energy saving values varying η 0 1 2 3 4 5 6 7 8 9 10 0 100 200 300 400 500 600 mean reversion speed e le c tr ic it y s a v in g [ m w h /y e a r] mean standard deviation of 2000 series finally, it is also recommended for future m&v the use of meters whose energy recording is done in intervals of less than 5 minutes, in order to avoid overestimation or underestimation of the key parameters, which contributes to a higher reliability of the economy results. electric power and reduction in peak demand, while taking into account the balance between cost and benefit highlighted by ipmvp. the application of the methodology of stochastic modeling to forecast future electricity saving by eep leads to the conclusion that the reliability of its use is conditioned to obtaining historical data for volatility and mean reversion speed, given the abrupt variation of results. as new results of electricity saving by m&v in eep will be obtained, it will be constituted a sufficient set of historical data for the assignment of these constants used in stochastic modeling. considering these contributions and new information to be used in the stochastic model, as the useful life of equipment, this methodology can be used to obtain prediction of electricity saving by use of solar water heating systems. finally, it is concluded that the society gets benefits from eea, considering the highlighted economies and the reduction of environmental impacts associated with the generation, transmission and distribution of electricity, which contributes to a better use of natural resources of the planet. acknowledgment we thank fapeg for financial resources to participate in the ieee eeeic 2016 conference in florence, italy. references [1] brazilian power plants s/a, solar energy for heating brazil: contributions from eletrobras procel and partners (in portuguese), rio de janeiro: eletrobras, 2012. [2] efficiency valuation organization, international performance measurement and verification protocol: concepts and options for determining energy and water savings, volume 1, toronto: evo, 2012. [3] national electric power agency – energy efficiency program, measurement and verification guide for energy efficiency program regulated by aneel (in portuguese), brasilia: aneel, 2014. [4] e. domingues, risk analysis to optimize portfolios of physical assets in electric power generation, itajubá: doctoral thesis (in portuguese), postgraduate course in electrical engineering, federal university of itajuba, 2003. [5] national electric energy agency energy efficiency program, energy efficiency program procedures – eepp (in portuguese), brasilia: aneel, 2014. [6] national electric energy agency energy efficiency program, the m & v for water heating in low-income (in portuguese), brasilia: aneel, 2014. [7] e. fonseca, comparison between simulations by the brownian geometric movement and the reversal movement to the average in the calculation of the cash flow risk of the downstream department of an oil company (in portuguese), rio de janeiro : master's dissertation, coppead institute of administration, federal university of rio de janeiro, 2006. [8] j. hull, options, futures and other derivatives, prentice-hall, second edition, 1993. [9] a. k.dixit and r. s.pindyck, investment under uncertainty, princeton university press, princeton, new jersey, 1993. bruno g. menita received the bachelor's degree in mechanical engineering from the federal university of santa catarina (2009) and a master's degree in technology of sustainable processes from the federal institute of goiás (2015), being a concentration area technology of clean production systems and a line of research energias renewables and applied economic engineering. now he is a researcher in metrology and quality of the national institute of metrology, quality and technology. giordani p. medeiros is an academic of control and automation engineering by the federal institute of goiás, expected to graduate in 2017. he is currently a researcher at the nucleus of experimental and technological studies and research (next / ifg). jose l. domingos received the master's degree (1998) and a doctorate in electrical engineering from the federal university of uberlândia (2003). professor at the federal institute of education, science and technology of goiás ifg since 1992. he works in the area of industrial networks, intelligent systems (fuzzy logic and neural networks), electric machines (variable reluctance machines) and renewable energy photovoltaic solar and solar thermal conversion). he is a founding member of ifg's center for experimental and technological studies (next / ifg). experience in management as coordinator of the technical course in telecommunications, research and post-graduation coordinator, extension manager, and currently as manager of research, graduate and extension of campus goiânia of the ifg. elder g. domingues received the bachelor’s degree in electrical engineering from the federal university of uberlândia (1993), a master's degree in electrical engineering from the federal university of uberlândia (1996) and a phd in electrical engineering from the federal university of itajubá (2003). he is currently professor of the federal institute of goiás ifg. he has experience in electric power systems and works in the following lines of research: electric energy market, electrical energy quality, applied economic engineering and sustainable process technology. aylton j. alves has a technical degree in electrical engineering from the federal technical school of goiás (etfg1983), a bachelor’s degree in electrical engineering from the federal university of goiás (1990), a master's degree in electrical and computer engineering from the federal university of goiás (2001) and a phd from the federal university of uberlândia (2011). currently a full professor of technology at the federal institute of goiás (ifg) is one of next's leaders nucleus of experimental and technological studies, registered in the cnpq. he has experience in electrical engineering, working mainly in the following subjects: electric machines, electronic automation of electrical and industrial processes, modeling of electrical machines, modeling of environmental systems, electrical drives, orientation of research works in alternative sources of energy: solar , thermoelectric and biogas. wesley p. calixto he holds a degree in physics, a master's degree in electrical and computer engineering from the federal university of goiás (ufg). he completed his phd in electrical engineering by the federal university of uberlândia (ufu) with a period at the university of coimbra (uc), portugal. he is currently professor / researcher at the nucleus of experimental and technological studies and research (next / ifg). i. introduction ii. project definitions iii. methodology iv. results v. conclusions acknowledgment references sustainable adoption of connected vehicles in the brazilian landscape: policies, technical specifications and challenges douglas aguiar do nascimento, yuzo iano, hermes josé loschi, navid razmjooy, robert sroufe, vlademir de jesus silva oliveira, diego arturo pajuelo castro, matheus montagner abstract—this paper addresses the intervehicular communication in connected vehicles (cv) by emphasizing v2v (vehicleto-vehicle) and v2i (vehicle-to-infrastructure) communications in terms of evolution, current standards, state-of-the-art studies, embedded devices, simulation, trends, challenges, and relevant legislation. to accomplish the objective this review is based on studies conducted from 2003 to 2019, government reports about the sustainable deployment of these technologies and their adoption in the brazilian automotive market according to experts. moreover, wave (wireless access in vehicular environment) and dsrc (dedicated short-range communication) standards, the performance analysis of communication parameters and intervehicular available at the market are also described. the current status of its (intelligent transportation system) development in brazil is reviewed, as well as the research institutes and governmental actions focused on introducing the concept of connected vehicles into the society. the brazilian outlook for technological adoption concerning cvs was also discussed. besides those, challenges related to technical aspects, safety and environmental issues, and the standardization for vehicle communication are also described. finally, this review highlights the challenges and proposals from available technologies devoted to the roads and vehicular infrastructure communication, their evolution and upcoming trends. index terms—connected vehicles, intelligent transportation system, vehicular communication, policies, specification, sustainability. i. introduction the number of deaths from traffic accidents has reached about 1.25 million per year around the world, representing the first fatality cause across people aged 15-29 years old [1]. despite brazil’s legislation on the best traffic practices (prohibiting drunk driving and the awareness of using the helmet, seat belt, and child restraint), the total number of fatalities recorded in 2016 was about 34,850 [2]. it is estimated d. a. nascimento, y. iano, h. j. loschi and d. a. p. castro were with the department of communications, school of electrical and computer engineering, university of campinas, campinas, sp, 13083-852 br, e-mail: eng.douglas.a@ieee.org. n. razmjooy is with tafresh university, department of electrical and control engineering, tafresh 39518 79611, iran. r. sroufe is with john f. donahue school of business, duquesne university, rockwell hall 820, 600 forbes avenue, pittsburgh, pa 15282, usa. v. j. s. oliveira is school of exact and technological sciences, mato grosso state university, av. dos ingas, 3001, jardim imperial, sinop, mt, 78555-000. m. montagner is with santa catarina state university, paulo malschitzki st., 200, bairro: zona industrial norte, joinville, sc, cep: 89219-710. manuscript received on february 10, 2019 and revised on march 17, 2019. fig. 1. evolution of mortality from terrestrial transport [4]. that brazil spends us$ 12.3 billion 1 annualy on economic production losses since the victims are absent from work and hospital expenses from traffic accidents on federal and state highways, municipal and urban agglomerate streets [3]. in 2014, for example, the cost to society was us$ 3.16 billion, due to the 167,247 traffic accidents only on federal roads, as shown in table 1. the cost of each accident is the sum of the cost components associated with the control variables of the additive model costs associated with people, vehicles and other costs. as shown in figure 1, the economic production losses arising from labor capacity is the most expensive to society due to accidents involving workers. according to figure 1, there were some sudden death drops from terrestrial accidents (i.e., 1997-2000 and also 20082010), which were caused by the adotion of a more rigid legislation according to the new brazilian traffic code (btc) and the enforcement of the dwi (driving while intoxicated) law, respectively. the long-term trend is the increase in the number of fatalities. the adoption of public policies introduced by the us department of transportation was effective in reducing fatalities and injuries [5]. the adoption of v2v technologies could also help avoid 439,000-615,000 crashes, which represents 13-18% of crashes involving light vechicles, reducing up to 418,000 mais (maximum abbreviated injury scale) within a scale of 1-5 injuries and avoid 746,00 damages 11 usd: brl 4.05 (sept. 22nd, 2018). transactions on environment and electrical engineering issn 2450-5730 vol 3, no 1 (2019) © douglas aguiar do nascimento, yuzo iano, hermes josé loschi, navid razmjooy, robert sroufe, vlademir de jesus silva oliveira, diego arturo pajuelo castro, matheus montagner table i accident costs on federal roads [3]. cost description value ($) % people hospital expenses; attendance; treatment of injuries; removal of victims; and economic production losses 1,963,186,281.98 62.01 vehicle removal of vehicles; damage to vehicles; and loss of production capacity over its lifetime 1,185,294,508.64 37.44 institutions and property damage attendance; prosecution and damage to public and private properties 17,282,788.40 0.55 total 3,165,763,579.02 100.00 to pdovs (property-damage-only vehicles) [6]. technologies based on partial automation levels (e.g. esp electronic stability control ) are responsible for reducing the number of accidents. it is estimated that reducing damages or avoiding crashes in 18% of the total accidents and 34% of those resulting in fatalities is feasible [7]. regarding secutiry applications, such as cicas (cooperative intersection collision avoidance systems), 80% of crashes could be avoided if 50% of the crossings were equipped with rsu devices (road-side unit) from v2i (vehicle-to-infrastructure). moreover, 50% of crashes could be avoided if 20% of the crossings were equipped with rsu v2i [5], [8]. therefore, stakeholders from many countries believe that the systems of connected vehicles (cv – connected vehicle), also known as cooperative intelligent transportation system (c-its), have the ability to enhance traffic in terms of transport security and efficiency [9]. this scenario of connected cars would be associated to smart cities, where several devices are connected in the cloud (data storage network infrastructure) and able to communicate to each other in order to exchange information and data storage. this data, after it was processed and analyzed, would still be useful in providing services and further processing needs. hence, the term “connected car industry” has been widely used and the revenue forecast for the sales of connected cars is shown in figure 2 [10]. fig. 2. global market for connected cars. many market surveys provide a number of 380 million connected vehicles [11]–[14]. the development of connected vehicles is taking place in conjuction with underlying technologies devoted to network and communication infrastructure capable of meeting the bandwidth demand required by connected vehicles, such as the 5g mobile communication and wave (wireless access in vehicular environment) [15]–[18]. therefore, this work proposes an overview on different types of communication methods for connected vehicles and the analysis of prospects and trends for the brazilian scenario. it is out of the scope of this work covering intelligent cooperative transport systems (cits) or intervehicular communication for platooning (convoy), since this approach should describe the routing of vanet networks, cognitive applications and case studies in vanet. readers interested in those technological aspects are encouraged to search elsewhere [19]–[23]. connected vehicles belong to a new model known as intelligent transportation system (its), which is focused on improving traffic safety and efficiency by means of wireless electronic communication. connected vehicles rely on gps (global positioning system) data, connectivity (wireless communication), and data processing to enable vehicles, smart road infrastructure and mobile phones to exchange information for providing warning and security messages to road users [8]. therefore, this work was structured as follows: section 2 introduces iov (internet of vehicles) and associated definitions. section 3 addresses vehicular communication in terms of infrastructure and data exchange, protocols and technologies to connect vehicles. section 4 describes the brazilian adoption of vehicular technologies. section 5 deals with the challenges from technical aspects (e.g., which frequency range to use and allocation issues), safety, and public acceptance within the brazilian scenario. section 6 describes the legislation, public policies and legal perspective. section 7 provides conclusions based on the idea of vehicle communication. ii. internet of vehicles the internet of things (iot) [25] is a paradigm that establishes a world of embedded physical objects, based on sensors and actuators, connected to wireless networks. they can communicate through the internet, which gives rise to a network of intelligent objects able to do various types of data processing, get environmental variables, and react to external event. these objects are connected to each other and to other resources (physical or virtual). they can be controlled through the internet, which allows a plethora of applications that will be able to use new types of data, services and operations available. iot is an example of an emerging technology that contributes to the achievement of fig. 3. iov big data architecture, based on [24]. new fields of application for information and communication technologies (icts). one of these domains can be understood as that from smart cities, wherein the use of communication and sensing technologies provides value-added services to the administrative bodies of such cities and their citizens [26], [27]. one of the applications regarding the ict domain is the intelligent mobility. as described elsewhere [28], mobility is part of the daily life of the modern society, since it requires transportation systems based on different types of vehicles and communication infrastructure. with the recent advances in information and communication technologies, mobility has an important role in providing a better quality of life. the internet of things integrates intelligence into existing research and development areas, such as smart-health, smart-home, smartenergy, smart industry and smart transport or smart mobility as described in [29] (see figure 3). the internet of vehicles (iov) is one of the revolutions mobilized by iot, which involves the concept of vanet vehicular ad hoc networks, to convey the vision of the smartphone to the smart car [30], [31]. vanet is aimed at improving traffic efficiency and safety by means of realtime communication between advanced wireless access technologies, enabling vehicles with or without rsu (road side unit) [31]. vanets are a special class of mobile ad hoc networks (manet) in which the vehicles work as people. these networks are characterized by the high speed of vehicles, low intercontact times among hosts, intermittent connection and real-time data exchange requirements [32]. moreover, these networks rely on a minimum or temporary infrastructure and are characterized by high mobility, fixed road networks, traffic patterns and predictable speed under traffic congestion conditions, low power requirements, and storage limitations. even though other communication systems rely on high message throughput, vanets primarily focus on reliable communication and fast dissemination of safety messages [33]. this way cvs may communicate with each other or with the traffic infrastructure (rsu) by using consolidated technologies, which are divided as [32]: 1. vehicle-to-vehicle (v2v) communication; 2. vehicle-to-infrastructure (v2i) communication; 3. hybrid communication composed of v2v and v2i, and 4. vehicle-to-everything (v2x) communication. v2v communication is also known as c2c (car-to-car), or intervehicle communication (ivc). the term “connected vehicles” refers to the applications, services, and technologies for connecting vehicles. by adopting a similar definition as new cars auto connected, a connected vehicle is basically formed by the presence of devices that connect this vehicle to other services, devices, networks, applications and services outside the vehicle. applications include traffic efficiency and safety, infotainment, parking assistance, roadside assistance, remote diagnostics and telematics for self-driving vehicles, and global positioning system (gps). vehicles with advanced interactive driver-assistance systems (adas) and cooperative intelligent transport systems (c-its) can be considered the typical connected vehicles. safety applications for connected vehicles are designed to increase the awareness and reduce traffic accidents through vehicle to vehicle (v2v) and vehicle to infrastructure (v2i) communications [34], [35]. an increasing number of factories are equiping their vehicles with on-board computers, sensors, and navigating systems able to create mass scale vehicle networks [36]–[38]. by using a myriad of sensors, cameras, computers, and communication services, vehicles are able to harvest, process, analyze, and send information to help drivers [38]. traffic information systems (tiss) allow a better use of road networks by providing real-time traffic conditions and by guiding drivers to make better routing decisions [39], [40]. vehicle network applications may be classified as [38]: 1. safety; 2. entertainment, and 3. driving assistance. safety includes applications intended to provide information to the driver about dangerous road conditions, such as the weather, traffic jams, accidents, etc, with the dissemination of emergency information. the entertainment applications provide the support for internet access, advertising, content sharing, chats and related services. the driver assistance applications will provide the exchange of information for helping drivers to find gas or power stations, restaurants, and toll roads. moreover, the systems of cvs also allow: reduction of greenhouse gas emission and fuel consumption, enhanced safety and protection, higher efficiency, mobility and accessibility, besides the adoption of economical opportunities for advancing investments and research on clean technologies [41]. iii. vehicular communication it is a global network of wat (wireless access technology), which includes vehicles, the internet and other heterogeneous networks, such as the internet of vehicles (iov). the heterogeneous network architecture of iot includes five types of vehicular communication (figure 4): v2v (vehicleto-vehicle), by using wave (wireless acess in vehicular environment) through the 802.11p protocol; v2r (vehicleto-roadside unit) with wave and 802.11p protocol; v2i (vehicle-to-infrastructure) available through wi-fi (wireless fidellity) 802.11b, wi-fi 802.11g and also mobile networks such as 4g, lte (long-term evolution) and 5g [16], [42], [43]; v2p (vehicle-to-personal devices) through technologies such as carplay, oaa and ncf, for example; v2s (vehicleto-sensors), in which the ecu (electronic control unit) from vehicles is able to communicate with sensors installed in the vehicle by using protocols such as ethernet, wifi and most (media oriented systems transport). therefore, each vehicle communication from iov is enabled by using a different wat, which includes, ieee wave for v2v and v2r, wifi and 4g/lte for v2i, carplay/nfc for v2p and most/wifi for v2s. the inclusion of a plethora of devices makes the architecture more complex, even though more market oriented, such as vanets. the structure of a heterogeneous vehicle network of iov has a significant power for guiding and monitoring vehicles [31], [44], [45]. even though the vehicle communication technology is mainly used for traffic safety, connected vehicles are also able to support technologies other than safety, such as telematics and traffic management, which includes the control over road congestion, smart tolling and optimization of routes and directions [8]. a suitable description for better understanding this is proposed elsewhere [46], according to the zones of interest, which may be: personal (communication based on pan), information exchange among vehicles and personal devices; local network (lan), communication among vehicles and vehicle to vehicle and local infrastructure; and regional (by using mobile networks), applications based on broadcasting, in which the traffic center or manager sends information to vehicles in the region. fig. 4. types of iov communication, based on [31]. a. technologies and protocols for vehicular communication within the vehicular communication, each device in the network (vehicle or infrastructure device) is considered a node. the communication among nodes from vehicular networks may be done in three different ways: vehicle to vehicle (v2v), vehicle to infrastructure (v2i), and hybrid (v2x) [47], [48]. the intervehicular communication (ivc), which is part of the its (intelligent transportation system) and its real applications in mobile ad hoc networks, has been researched at the academia and also at industries, most notably in the us, europe and japan. the most important achievement from ivc is its ability to expand the horizon of drivers and on-board devices (radar or sensors, for example), besides improving traffic safety and efficiency on the roads [49]. its enhances not only transport safety and mobility but also the american economic productivity by integrating advanced communication technologies into transport infrastructure and vehicles. its encompasses a myriad of information and electronic technologies based on wired and wireless communication [50]. table 2 shows an overview from the main countries developing research and the implementing its. there are three different categories for automobile applications based on communication [9], [52]: 1. safety-driven: examples include the alert for a parked or slow vehicle and electronic emergency brake lights, v2v warning of an accident, an alert for road resources and cooperative collision warning. 2. comfort-driven: alerts for congested roads, traffic probes, and warning of parking availability, parking lot finder, are some examples. 3. commercial-driven: such as remote table ii overview of its for the main research studies in different communities [51]. japan usa europe standard / committee its-forum ieee802.11p/1609.x cen/etsi en302 663 frequency range 755 – 765 mhz 5850 – 5925 mhz 5855– 5925 mhz number of channels one 10 mhz channel seven 10 mhz channels (two 20 mhz channels formed by combining 10 mhz channels) seven 10 mhz channels modulation ofdm data rate per channel 3 -18mbit/s 3 -27mbit/s 3 27mbit/s output power 20 dbm (antenna input) 23 33 dbm (eirp) 23-33 dbm (eirp) communication one direction multicasting service (broadcast without ack) one direction multicasting service, one to multi communication, simplex communication (broadcast without ack, multicast, unicast with ack) upper protocol arib std-t109 wave (ieee 1609) / tcp/ip etsi en 302 665 (incl. e.g.geonetworking) tcp/udp/ip customization or diagnostics of a vehicle, advertising services, download of contents and real-time video broadcasting. 1) inter-vehicular commmunication: wave (wireless acess in vehicular environment) is a wireless technology primarily developed for harsh environments, wherein it enables the fast communication among vehicles with the advantages of high mobility, threshold delays for security messages with severe qos, optimal energy consumption and respect for privacy and anonymity from roaming users, besides other environmental challenges [53]. according to [47], ieee has started the standadization of vehicular communcation networks within the ieee 802.11 working group. this standard has been under development since 2004 and is called ieee 802.11p – wireless acess in vehicular environment (wave). it is governed by the ieee 1609 [54] and ieee 802.11p [55] standards, which establish vehicular communications. a scheme to show the integration among wave components is shown in figure 5. fig. 5. example of a wave system component, based on [53], [54]. as a global term, wave is nowadays employed to define all wireless vehicular communications. it became arbitrarily used as a general term, such as dsrc. regarding the standardization, wave is used for projecting a set of ieee1609.x standards over the 802.11p standard [53]. therefore, a dsrc/wave offers [8], [54], [56], [57]: 1. low latency 802.11p standard states essencial events (functions and services) for data exchange without establishing a basic service set (bss); 2. data rate – 802.11 defines a 10 mhz-bandwidth channel and eight data rates’ types, i.e. 3, 4, 5, 6, 9, 12, 18, 24, 27 mhz; 3. high reliability – in order to comply within harsh envinronments (sandstorm, rain, etc) ieee 802.11p stablishes a 10 mhz-bandwidth channel rather 20 mhz; 4. security and privaty – dsrc protocol provides securing management and messages of application to overcome attacks from malicious and and untrusted events and softwares such as dod (denial of service), eavesdropping and spoofing. the dsrc (dedicated short-range communication) wireless technology devoted to vehicular communication was designed for its applications within vehicular environments. its primary aim is to offer support to safety applications and the communication among vehicles (v2v), and from vehicle to infrastructure (v2i), this way decreasing the number of accidents. moreover, dsrc also supports its applications, such as managing traffic conditions, information and entertainment (infotainment) [8], [56], [57]. figure 6 shows the dsrc protocol stack. the comercial success of the wifi technology and ieee 802.11 standards led to the development of a new standard, known as ieee 802.11p wave. ieee 802.11p is based on the ieee 802.11a standard, but with improvements on its physical (phy) and medium access control (mac), which are aimed at reaching low latency and high communication reliability in short-range radio connections [8], [54], [56]. the ieee 802.11p standard defines the wave physical layers (phy) and mac, which are extensions to ieee 802.11 standard, in order to communicate outside the bss context. it has also standardized other specifications, such as 5.9 ghz ofdm phy (within 5,850 – 5,925 ghz in the u.s. and 5,855 – 5,925 ghz in europe), channel bandwidth, operating fig. 6. dsrc protocol stack [58]. ranges, transmission power requirements, transmission masks and alternate channel, and alternate adjacent channel rejection requirements [8]. the ieee 1609 family of standards defines the architecture and a set of standardized protocols which provide the foundation for a wide range of dsrc/wave applications. the most important ieee 1609.x protocols are [8], [57]: ieee 1609.3 – defines the using of ipv6 (internet protocol version 6) and employs the wave short message protocol (wsmp) to prevent overhead through the wave short messages (wsm) and the wave service advertisement (wsa), also setting management functions; ieee 1609.2 – defines services related to applications and messages of management for authentication using optimal encryption of dsrc messages based upon signals and digital certificates; ieee 1609.4 – stablishes the multichannel operations by mac extension and address the channel timing and switching considering ieee 802.11 standard. also, it defines the control channel (cch), created to send messages of advertisement and information through service channels (sch); sae j2735 – sae (society of automotive engineers) standard upper layer containing 15 message, frames and data elements types, and the message specifications; sae j2945.1 – also defines an upper layer which stablishes the minimum requirements for communication performance. according to [58] the sae j2735 standard stablishes syntax and semantics of v2x messages and bsm (basic safety message) is the most relevant amongst all other message formats available due bsm address core state information about the broadcasting vehicle and other important informations e.g. location, size, etc. even though bsm is designed to be compact and efficient, additional frame and data elements may extend it, and add-ons may optionally be added within a subset of messages, such as in every second message. all these dsrc/wave functionalities are provided by the communication interface and network technologies as described in the next section. 2) intra-vehicular commmunication: intra vehicular communication allows modules, sensors, and actuators to communicate with each other. it provides the operation for only one vehicle and operates under obd (on-board diagnostics) services. local networks also support aftermarket telematic devices, which can access data through obd’s standardized interface. nowadays, original equipment manufacturers (oems) implement sensors and networks shared over the obd hardware by following oem standards that carry relevant information for vehicle optimization and support applications concerning future communication. in terms of non-obd intravehicle communication, communication systems related to drive-by-wire systems are especially built for robustness and security of critical data, whereas other systems host peripheral data, provide fault tolerance, determinism and flexibility, and support network technologies such as can, lin, most and flexray [59]–[63]. the main traditional intra-vehicular communication networks are [64]: • lin (local interconnect network): it is a low-cost and low speed (20 kbps) serial intra-vehicular communication network. lin is widely used for distributed body control electronic systems in vehicles since it is a user-friendly and a low-cost technology. it is also applied to some comfort functions; • can (controller area network): it is a serial databus communication protocol developed by robert bosch gmbh. nowadays, can has become a standard for transmitting data over intra-vehicular networks with a data rate reaching from 125 kbps up to 1 mbps. it is widely used for automotive communication due to its flexibility and robust nature, which also includes its limited delay, simplicity and low cost; • byteflight: it was developed by bmw and supports a data rate of up to 10 mbps. this network requires broadband services and has been applied to vehicular networks with high level safety requirements (e.g. passive safety); • ttp/c (time-triggered protocol): it offers a data rate of up to 25 mbps and is based on tdma. despite its complex project, it provides low cost applications. a ttp frame may contain 240 bytes of data and 4 bytes of overhead. the ability to schedule the communication makes the ttp/c protocol less flexible, but its timetriggered communication allows it to be predictable; • ttcan (time-triggered controller area network): it was also developed by robert bosch gmbh and is based on the tdma mechanism. ttcan relies on the same standard and message formats from can (supports a data rate from 125 kbps up to 1mbps), but in contrast to can, it has a master node responsible for time synchronization among nodes; • flexray: it is used for high speed and flexible intravehicular communication and offers a data rate of up to 10 mbps. it seems to be the best choice for safety and high speed automotive applications. flexray is based on tdma and ftdma mechanism. it offers star and multiple star topologies. this network exchanges messages with 254 bytes of data and 5 bytes for the header; • most (media-oriented systems transport): it was developed to make information and entertainment (infotainment) and multimidia systems easier to handle, with a data rate up to 24.8 mbps for streaming audio, video, data and control information. this allows such a cost-effective technology to offer an efficient data communication infrastructure; figure 7 shows the main systems and communication protocols that make up the vehicle, in which the ethernet protocol was proposed for transmitting and receiving data from the vehicle. the can protocol is applied to comfort electronics and powertrain systems. the usb protocol is used for information and entertainment systems, whereas flexray is applied to safety systems and chassis control. considering the main wireless technologies devoted to vehicular communication, it is relevant to mention the review on intra-vehicular communication technologies summarized in table 3. according to the work from [68], wireless technologies are preferable over zigbee and bluetooth when taking into account fig. 7. example for the application of an in-vehicle communication, based on [65]. table iii the most commonly used technologies for in-vehicle newtworks. adapted from [66], [67]. standard bluetooth uwb zigbee wi-fi ieee specification 802.15.1 802.15.3 802.15.4 802.11a/b/g in-vehicle applications in-vehicle communication and device connectivity high speed intra-vehicular communication environments in-vehicle communication in-vehicle communication domain telematics and body telematics and powertrain body telematics data rate 1 mbps 100 mbps 250 kbps 54 mbps range 10 m 10 m 10-100 m 100 m power consumption low ultra-low very low high mode (spreading) fhss ds-uwb dsss dsss, cck, ofdm modulation type gfsk bpsk, qpsk bpsk, o, qpsk bpsk, qpsk frequency 2.4 ghz 3.1-10.6 ghz 868 mhz, 915 mhz, 2.4 ghz 2.4 ghz, 5 ghz data protection 16-bit crc 32-bit crc 16-bit crc 32-bit topology star peer-to-peer star/mesh star maximum number of cell nodes 8 8 > 6500 unlimited 2007 (infra-structured) in-vehicle applications relying on a low bitrate and limited power source, as well as its low power consumption that could provide a longer lifetime. on the other hand, high speed data in-vehicle applications could benefit from the use of uwb and wifi due to their low normalized energy consumption. 3) underlying technologies: the need for inter vehicle communication has grown out of some safety issues, such as maintaining or improving road safety for drivers with the ever increasing traffic density. the initial communication was mainly based on the broadcasting type from one central station spreading traffic information. the need for a feedback channel beween vehicles was noticed a long time ago and, consequently, the need for inter-vehicle communication regarding safety applications, such as the stopping distance, for example. the main limitations of a vehicular network are the requirements for developing a system with the restrictions of a real-time system, while keeping network latency to safe reaction time limits [69]. the vehicles are connected through multiple radio access technologies, such as dsrc, ieee 802.11 (wi-fi), cellular technologies like lte and improvements arising from the fifth generation (5g) networks. although 3rd generation partnership program (3gpp) supports telematics and infotainment services for connected cars, cellular networks have received considerable attention for a broader scope with v2x (3gpp release 14) [15] and with 5g providing ultra-high reliability and ultra-low latency demands of tomorrow v2x applications [70]. inter-vehicle communication may be carried out by means of existing standards, such as cellular networks (3g, 4g) or satellite communications [69]. lte is the fourth generation of cellular radio network as defined by 3gpp. even though lte has a centralized architecture (similar to earlier generations of cellular radio network systems and lacking a native ad hoc mode), it may be useful as a potential access technology for cellular radio networls due to serveral reasons: high data rates (¿100 mb/s), which is suitable for information and entertainment services, besides the ability to tolerate high mobility with a low transmission latency. these advantages are relevant for road safety applications. moreover, lte can cover a wider area with a higher penetration rate than 802.11p. major telecommunication companies have been heavily investing in lte infrastructure, which has already been deployed in some markets around the world. lte is a promising technology with the ability to fill in major gaps from ieee 802.11p, such as intermittent coverage and lower penetration rate. major stakeholders have been testing the lte technology for specific road safety applications and traffic efficiency messaging, such as nearby road hazards and traffic alerts over large coverage areas [33]. iv. adoption of its technologies in brazil in brazil, the implemented its covers the dimensions of operation and road services [71]: the traveler; in-transit vehile; support systems; coordination and management systems. the brazilian its communication infrastructure is shown in figure 8. information about general traffic conditions is available to users by means of variable message signs and data centers (wifi services, web portals and mobile applications). the system can also identify travelers through license plate recognition, transponders (rfid tags), mac address from mobile devices and from shared tracking data between private and public organizations in association with road management companies. the vehicular control system can identify and monitor vechicles based on their goals (emergency, commercial use, personal use or road operation management vehicles). besides, the system enables in-vehicle communication through open access networks among vehicles endowed with collision warning systems, which can inform (without human interference) accidents to operational control centers and roadside assistance or emergency vehicles. the structure system to support road operation is composed of field equipments devoted to communication and monitoring, e.g., sensors for counting vehicles, surveillance cameras, general telemetry (such as a weather station), tolls and elements for vehicle classification [71]. brazil has only one occ (operations control center) for managing road operations, emergencies and maintenance. public transport, tolls and commercial fleet management are not under the responsibility of occ, which requires the implementation of communication subsystems through traffic data telematic infrastructure for communication, monitoring and managing protocols to help decision making [71]. cameras, speed controllers, and traffic count are currently widely explored on brazilian highways [72]. the electronic identification of cargo vehicles is under way and the enrolment of drivers is mandatory. drivers must show the national register of road freight carriers (rntr) and in the near future, by means of installed rfid (radio-frequency identification) tags into those vehicles, it will be possible to identify them in order to control overload and minimize cargo theft. the remote operation will be carried out right after the identification and within the occ of each of the road concession companies wherein integrated automation systems (ias) will be employed, such as video surveillance, cameras for the automatic recognition and identification of license plates, identification tag readers, roadside variable message signs to display the mandatory entrance to the weigh station and electronic display panels inside the patio from the surveillance checkpoints. those displays will guide and inform drivers about the overweight and the means to solve the issue [72]. the “semparar” (non-stop) system allows the automatic identification of vehicles, multi-lane free flow, optical character recognition and mobile resources. it works at the frequency range of 915 mhz and 5.8ghz [73]. a standard for communication between field and centralized equipments on brazilian highways has been defined by the national agency of land transportation (antt) through the resolutions 3,323/09, 3,323/09-a and 3,576/10. ntcip (national transportation communications for intelligent transportation system protocol) is a protocol designed from technical standards published by the american association of state highway and transportation officials (aashto). this protocol allows the “center do field” (c2f) or center to center (c2c) communication, which allows the exchange of information between equipments and control centers or among control centers, integrating different agencies. despite being similar to the ip protocol, the ntcip provides a dictionary of specific functional data to be used in smart transport systems, simplifying data communication between devices, as well as the installation and configuration of equipments, which allows these systems to be highly interoperable and scalable [71]. deployment of vanets in the brazilian context can follow different directions. this will depend on the public policy that the government will adopt regarding this technology. possibilities could range from private and public partnerships, which could allow the financial solution for development costs based on strategies to guarantee road safety; or place the radio electromagnetic spectrum for vehicle-to-vehicle communication purposes in a public tender offer. this paper aims at opening academic discussions about these different approaches and bringing to this context current local experiences and success stories from external realities. for example, michigan university has been studying the transition to a new world of connected and automated vehicles and their work goes beyond technology [74]. hence, the concept of vanets is not restricted to the type of medium access technology or routing protocol that may be explored, but the impact that it is going to have on each element of the society. in the brazilian context, there is a traffic engineering company, companhia de engenharia de tráfego (cet) [75], which is owned by the são paulo city hall. it monitors all the intersections with high vehicular load and, this way, can contribute with vehicular data to the civil society for road safety issues, traffic management, etc. nevertheless, this data does not provide vehicular information to drivers in real time. in other words, it is demonstrated that developing cutting-edge technologies with no interaction with other agents involved in a vehicular environment does not solve the real problem, i.e., preventing accidents. on the other hand, the emergence of applications of georeferencing [76], in which drivers interact with each other, by notifying the community that there is an ongoing fig. 8. brazilian its communication infrastructure. adapted from [71]. operation on a street, the existence of radars or even that an accident happened nearby. information generated by this type of technology can be fake or cheat because there is no authority who verifies the veracity of posts from users. again, the lack of communication among users and government agencies is not profitable for the society. v. technical aspects and challenges of vehicular communication a. technical aspects of vehicular communication smart mobility has a massive interaction with technological advances in the future-driven fields of robotics, artificial intelligence, manufacturing (industry 4.0), sustainability and global connectivity (internet of things) and future vehicles will act autonomously and interact permanently. these systems should undergo a profound change from static and monolithic designs towards much more dynamic and compositional concepts with security, safety, real-time, and maintainability in mind, in order to be prepared for use cases with strict demands on dependability. the concepts must support advanced multicore and application-specific processor architectures, protection against environmental perturbation and attacks, dynamic update mechanisms, and simplified software portability for long-term operation [29]. the adoption of vehicular communication environments will include big data, security, privacy, reliability, mobility, and standards. these issues should be addressed to make iov highly reliable and widely adopted. a major challenge is the processing and storage of big data created in iov due to the large number of connected vehicles. for example, driverless cars are expected to process 1 gb of data per second. mobile cloud computing and big data analytics will play important role in handling big data. since iov involves integrating many different technologies, services and standards, there is the need for data security. as an open, public network, iov is a target for intrusions and cyber-attacks that may lead to physical damage and privacy leakages. cars, sensors, and network hardware can malfunction. the system must deal with incorrect data, as well as faulty communications, such as denial of service attacks. as a rule, the safety of the vehicle is more important than entertainment. in a situation where vehicles are moving fast and network topology keeps changing continuously, it is a challenge to keep the nodes connected and provide them with resources to transmit and receive in real time. to accelerate adoption, standardization and interoperability are vital. the lack of standards makes an effective v2v communication more difficult. adopting open standards will enable smooth sharing of information. governments should participate and encourage industries to collaborate on the development of technological best practices and open international standards [77]. as discussed elsewhere [33], those authors describe communication and reliability among the technological challenges regarding vehicle communication: 1. communication: a fully operating vehicular network will require unicast and multicast/broadcast delivery capabilities for v2p, v2v, v2i and applications. however, this will require communication systems able to support both short-range and long-range communications. in this respect, the speed of vehicle trajectories, network density, changes in network topologies, and constrained bandwidth in vehicular networks are some of the challenges not found in smartphone internet applications. open problems have not been fully addressed yet by the technology community, such as ensuring communication systems will not interfere with transmission schemes and how to provide low latency and high data rate for vehicular applications; 2. reliability: it is critical that vehicular communication networks have robust and faulttolerant softwares able to recover from connection downtime and system errors. unlike other electronic devices that are prone to frequent hardware turnover, such as in smartphones, the computing hardware for vehicular networks (onboard units, vehicle sensors, etc) must have much longer usage once the lifespan of a vehicle is relatively high (they can last for more than a decade), whereas repair and maintenance may or may not occur according to manufacturers’ guidelines. the reliability of both computing software and hardware components remains major technical hurdles which need to be solved so that car manufacturers could adopt and deploy them in large scale. it is possible to assign technological challenges to two communication environments: in-vehicle and inter-vehicle. in vehicle connectivity faces the following drawbacks [68]: harsh environment due to severe scattering in a very limited space and often none-line-of-sight. this is the major reason for extensive effort to characterize the intravehicle wireless channels [78]; data transmissions require low latency and high reliability to satisfy the stringent requirement of real-time intra-vehicle control system; interference from neighboring vehicles in a highly densed urban scenario may not be negligible; security is critical to protect the invehicle network and control system from malicious attacks [79]. moreover, a couple of wireless technologies have been studied to find the most convenient technique for connected vehicles. inter-vehicle networks include dsrc, wave, 4g, lte, etc, technologies which may allow the v2v, v2i, and v2r communication, regardless of being broadcast or unicast packets. the most cumbersome challenge is to combat the harsh communication environment [68] and the challenges are mainly related to ways for protocol access due to error estimates and high intensity of nodes and the underlying physics behind vehicle mobility, power consumption, broadcasting support and self-interference, as shown in table 4. in urban scenarios, the line-of-sight (los) path of v2v communication is often blocked by buildings at intersections. on the other hand, on a highway, the trucks on a communication path may introduce significant signal attenuation and packet loss [81]. field tests in [82] demonstrated that multipath fading, shadowing, and doppler effects due to high vehicle mobility and the complex urban environment will lead to severe wireless loss, and with a large scale of vehicles transmitting simultaneously, the mutual interference plays an important role as well. reference [83] presents an overview of the state-of-the-art vehicular channel measurements. it is noteworthy that there is a lack of unified channel model that can be applied for all scenarios (e.g., urban, rural, and highway), and the existing channel models, only for a specific scenario, have their own merits and deficiencies. the authors also provide suggestions for v2v communication systems based on the channel characterization. the adoption of multiple antennas, for example, would enhance the communication reliability [68]. from a network perspective it is possible to point out the following challenges [68]: 1) the network topology changes frequently and very fast due to high vehicle mobility and different movement trajectory of each vehicle. 2) due to the high dynamics of network topology and limited range of v2v communication, frequent network partitioning can occur, resulting in data flow disconnections. 3) surrounding obstacles (e.g., buildings and trucks) can lead to an intermittent link to a mobile vehicle. in addition to the technical challenges, in the [68]: 1) to enable various wireless connectivity, multiple radio interfaces have to be implemented, such as dsrc/wave, wifi, and 3g/4g-lte interfaces, which may incur a high cost and thereby hinder the development of connected vehicles. a unified solution to provide v2x connectivity with low cost might be required. 2) in-vehicle systems have stringent requirements on latency and reliability for control/monitoring purposes. the full adoption of v2s connectivity may not be feasible in the near future unless v2s connectivity can provide the same performance and reliability as the wired communication [84]. 3) connected vehicle offers drivers a variety of information. however, research from [85], [86] suggests an up limit on information provided to the driver. excessive information increases the driver’s workload and hence has a negative impact on safety. therefore, the vehicle information system has to be appropriately designed for offering information to drivers. the brazilian infrastructure for its is still in its infancy and it is not yet integrated to vehicles and rsu. nonetheless, the its communication from center to center through ntcip c2c is already working [87]. thus, there are many challenges to adapt wi-fi technologies to support the unique requirements of vehicular communications, such as achieving high and reliable performance in highly mobile, often densely populated, and frequently none-line-ofsight environments. the automotive and the communication industries, academia, and governments around the world have been devoting tremendous efforts to address these challenges and significant achievements have been made. over the last decade, there have been vigorous joint efforts from the industry, academia and government to validate the dsrc technology and also to identify and address key technical and business challenges. in [10] the author provides the lattice ai which gives a solution on how to enable the sync between multiple companies, which provide connected car solutions so that they can connect with each other on a unanimous network running on machine learning and swarm algorithm, and how to share resources. this remark is in good agreement to what happened in brazil when telecommunication companies deployed cell phone technologies into the national market by the gradual table iv challenges and solutions for inter-vehicular communication. adapted from [80]. medium challenge proposed solution phy layer channel estimation in vehicular environments a) turbo receiver: using tools from modern coding theory to deal with channel estimation errors: e.g. introduce a turbo receiver [83]; b) decision feedback receiver: for the case as already-decoded bits as pilots for the remaining packets to improve channel estimation by tracking the channel variations significantly better than standard non-iterative schemes. time selective fading without time interleaving it can be solved using a better coding scheme (e.g. turbo or ldpc code). however, this indeed requests a standard change and is better to be addressed in future versions of dsrc mac layer csma behavior at high node density pottentialy resulting in congestion control a natural approach to reduce congestion is to reduce the number of transmitters within the carrier sense range of each device [59], [84]–[87]. a typical scheme to balance collisions and channel utilization is to use a distributed congestion control mechanism as described in [59], [87]; another promising method is to use a time-slotted synchronous system with a fixed set of broadcast resources. one can employ a simple mac protocol, [88], [89] to manage which transmitters should use which resources (e.g., time slots); slotted tdm systems are typically well suited for periodic transmissions of roughly equal size packets for prolonged durations so the addition of such “hooks” into the dsrc can be beneficial (not specified). multi-channel operations single radio devices (reduced capacity to support the broadcasting of safety messages) many simulation studies have shown that to support vehicle safety broadcasts in typical vehicle densities, most or all of the sync interval would be required; some studies [80], [81] indicate that even with a fully dedicated 10 mhz channel for safety and control, the channel congestion issues still remain. multiple radio devices (the spillage of power into adjacent bands when transmitting and selfinterference cancellation) certain techniques, which are well studied in the full duplex context, including analog cancelation and digital cancelation can be applied here. and successful adoption in many cities. this has been shown for the implementation of cellular network technologies such as 2g, 3g, and 4g which reached, respectively, a coverage area of 100% of brazilian cities (5570 municipalities), 96.46% (5,373 municipalities) and 74.60% (coverage reaches 4,155 municipalities) [88] through investments from companies such as algar, claro, nextel oi, sercomtel, tim and vivo. they were regulated accordingly by anatel (national telecommunications agency) based on constitutional principles of economic activity, according to art. 126 from the general telecommunications law, through personal mobile service – smp [89]. therefore, it was noticed that the government has to offer tax breaks for private companies so that they can succeed into the automotive market with less restrictions, but without any reduction in the quality and security of services provided. one of the biggest challenges for its is the search for a cooperative and safe environment (c-its), which could share data among all the applications developed by the diverse independent players [90]. the success of its depends on a significant number of players, such as public administration, transport authorities and companies of various segments such as vehicle manufacturers and oem (original equipment manufacturer), telecommunication companies, service providers, etc. within the industrial and technological perspectives, its is considered one of the biggest challenges of the tic community. therefore, the presence of world standards is a key requirement for exploring its entire potential. since 1990 standardization efforts have been evolving from various organizations such as ieee (institute of electrical and electronics engineers), iso (international organization for standardization), cen (european committee for standardization), etsi (european telecommunications standards institute), among others [72]. b. frequency spectrum allocation even though a global harmonization of dsrc standards and its spectral allocation are not available yet, ieee committee for standardization (institute of electrical and electronics engineers) and etsi (european telecommunications standards institute) have been working on the harmonization of standards for the north american and european regions. there is a second band for dsrc applications (915 mhz for the usa and 700 mhz for japan), which has been mainly explored for etc and commercial applications [8]. a complete survey on the history and description of all countries and the allocation spectrum for dsrc is out of the scope of the present study, which is focused on describing only the most important ones. a brief historical study on its communication in japan and in europe was described herein in section 2.1. further information on the architecture and its communication concepts in japan and europe is available elsewhere [33], [91]. differently from wireless networks based on frequency ranges from restricted radiation equipments (ism – industrial, scientific and medical applications), the communication range employed in vehicular networks is exclusive for the dsrc protocol [47]. however, regulations for adotpting dedicated frequency ranges for dsrc in brazil have not been established yet. figure 9 depicts the frequency ranges for vehicular communication explored in the main its development communities: itu-r, europe, and the united states. it may be noticed that the its frequencies chosen in europe and the united states range from 5.8 and 5.9 ghz. fig. 9. spectral range for its applications [55]. regarding spectrum allocation, while fcc (united states federal communication commission) and the canadian industry allocated a radio band of 75 mhz from 5,850 to 5,925 ghz, in europe ecc (european electronic communications commitee) assigned 70 mhz from 5,855 to 5,925 ghz, whereas in japan the allocation was 80 mhz for a band of 5.8 ghz. nevertheless, in japan dsrc is not compatible with the american and european standards due to the japanese development of etc (electronic toll collection). the frequency range of 75 mhz was allocated by the fcc for vehicular communication in the frequency spectrum of de 5.9 ghz – from 5,850 ghz to 5,925 ghz, which is restricted and licensed to vehicular communications, even though it is free of charge and can be used for free as well. according to the european telecommunications standards institute [92], the following frequency ranges have been reserved for vehicular communications: 5,855 ghz to 5,875 ghz (general applications); 5,875 ghz to 5,905 ghz (emergency applications and traffic security); 5,905 ghz to 5,925 ghz (reserved for future needs) [8], [56]. the standard supports four modulation techniques (bpsk, qpsk, 16-qam, and 64-qam). since a transfer rate of 6 mbps (qpsk) seems to provide a suitable balance between channel load and noise-to-signal ratio, many tests involving ieee 802.11p in the united states rely on 6mbps. in order to reach a higher performance in terms of dsrc communication, the system may adapt its data rate according to the distance between the vehicle and rsu. if a vehicle is away from the rsu (more than 150 km away), for example, it is preferred to send data by using a low data rate. in case the vehicle is rather close to the rsu (less than 150 m), commuting is advisable for reaching higher data rates. the sae j2945.1 standard defines the requirements for transmitting data, transmit power control and adaptive message rate control [8] in order to allow minimum performance. in brazil, the organization liable for regulating the automotive sector is abnt (brazilian association of technical standards) by means of the brazilian automotive committee abnt/cb005 [93], which works on the standardization of different automotive areas (cars, trucks, tractors, buses, mopeds, bicycles, motorcycles, autoparts and components), as well as in vehicle repairs and emissions. nevertheless, reports or descriptions of connected vehicles or their regulation and legislation have not been found yet on the abnt database. in terms of the legislation governing telecommunications, anatel (brazilian telecommunications agency) is responsible for, among other duties, managing and supervising the use of the frequency spectrum. when searching its database, regulations for the use of the frequency spectrum according to wave/dsrc (from 5,850 to 5,925 ghz) standards were not found. by analyzing the frequency ranges assignment, destination and distribution plan in brazil [94], it was noticed that the frequency ranges for region 2 (according to itu – international telecommunication union), the geographical location which includes brazil, have similarities to those currently adopted in brazil, as shown in table 5. in table 5 the labels are space-toground (*), ground-to-space(**) and international notes (***). table v assigned services according to frequency ranges [94] ranges (mhz) region 2 (itu) brazil 5725-5830 amateur radio anatel res. n452/2006 (d.o.u., official gov. publication on 12/20/2006) restricted radiation anatel res. n506/2008 (d.o.u., official gov. publication on 07/07/2008) 5830-5850 radiolocation radiolocation amateur radio amateur radio amateur radio over satellite* amateur radio over satellite* 5.150, 5.453, 5.455*** 5.150*** 5850-5925 fixed fixed-satellite** fixed-satellite** amateur radio mobile 5.150*** amateur radio radiolocation 5.150*** the services described for assignment and destination of radiofrequency ranges (see figure 9) are arranged in two columns: primary service (fixed, limited, etc), were written in upper case letters and secondary service (mobile, plain old telephone service – pots, etc) were written in lowercase letters. therefore, it is noticed that the last range (5,650 to 5,925 ghz), which should be assigned to v2x communication, is explored in brazil for primary services (fixed-satellite) and secondary amateur services, whereas the assignment from itu is for primary service “mobile” (cell phones). such terms (fixed, fixed-satellite, radiolocation and mobile) are related to radiocommunication services defined by itu, under its radio regulations (rr), which can be reached at: http://www.itu.int/pub/r-reg-rr-2016. nonetheless, the anatel plan contains an international footnote 5,150, which describes the frequency range of 5725-5875 mhz as also devoted to industrial, scientifical and medical (ism) applications. radiocommunication services operating within this range should accept harmful interference arising from the use of this range for any application. therefore, the use of frequency ranges destined by the dsrc protocol may be adopted within the national territory for vehicular applications, as well as vanet. modeling vanet communication networks should consider which infrastructure would be mounted. re-using existing infrastructure or planning new networks are viable possibilities. lte infrastructure is well deployed in brazil and worldwide. vanets can rely on this network to carry information, despite sharing the same spectrum with other services, such as data, voice or video, which will be a drawback, besides high latency and less reliability [95]. new advances on lte will allow operators to configure a broadcast service, lte broadcasting or embms (evolved multimedia broadcast multicast service) [96]; the operators should give up a portion of the spectrum that will be allocated for vanets communications. the design of how much bandwidth would be allocated for vanets is a challenge for future networks and additional tests focused on delay measurements in this technology are desirable. c. security aspects in vehicular communication vehicular applications from iot have data exchanges involving security and privacy issues [11]. otherwise, attackers could send corrupt or subverted information aiming at wrong warnings for drivers or even wrong automatic reactions from cars in case of autonomous driving, which could give rise to accidents, injuries or fatalities. an example would be the message of electronic emergency brake light that was fabricated or replayed, which would make the receiving vehicle to immediately brake without any obstable ahead. thus, safety mechanisms for intelligent transport systems (its) are crucial for supporting safety applications based on v2x communication [97]. many applications for vehicular networks are directly related to reliable information, non-disclosure of sensitive information and the protection between the receptor and the emitter. many attacks can be done to any kind of electrical vehicle communication, such as those summurized in table 6. table vi vulnerabilities related to connected vehicles [98]. category vulnerability v2s dos, jamming, false data injection, gps deception v2v dos, selfish attack, modification, sybil attack, false data injection, eavesdropping, black hole, gray hole and wormhole attack. v2i replay, router advertisement forgey, privacy, rsu spoofing, dos of the dad v2n mobile femtocell mobike dos, mitm, spoofing femtocell physical attacks, configuration attacks, mitm, dos, privacy, among others 5g mmwave and d2d: eavesdropping, privacy; jamming attack according to [98], the communication attacks can be classified according to the communication category, as: vehicle to sensor (v2s) attacks – attacks that affects the sensitive data between sensor and vehicle and directly impacts on urban platform for connected electric vehicles due to wired communications which are vulnerable to malware, enabling third part to control the communication interface, e.g. can bus. it can result in essential services damaged and system failture, i.e. denial-of-service attack (dos) [99]. for wireless sensor network (wsn) attacks, a malicious node can seriously cause malfunctioning of the vehicle operation by executing attacks on the vehicle sensor network such as jamming attack (malicious nodes block legitimate communication by causing intentional interference in networks [100], false data injection attack (sending false data from sensors to ecu), global positioning system (gps) deception (fase information about vehicle location or gps around) and dos (uninterrupted message bombarding to ecu); vehicleto-vehicle (v2v) attacks: attacks on communication between two or more vehicles and can be classified as selfish attack (refusing to send or receive message with vehicles or server or message not reaching all vehicles if selfish nodes increase), modification attack (modify or alter content of message sent to server), sybil attack (multiple identities and from different positions to inject false information in the network), false data injection attack (a node sending information to vehicles around to affect the prediction in the server), eavesdropping (detect sensitive information by an unauthorized node about a vehicle and leading to privacy attacks), black hole attack (putting a sender node into erroneous condition of shortest path [101], [102]), gray hole attack (extension of black hole attack but dropping messages in a selective way), wormhole attack (by revoking legitimate links between vehicles and hence any data transmitted between legitimate vehicles will pass through the attacker [103]), denial-of-service (when attacker try to cause failure to the system’s operation); vehicle to infrastructure (v2i) attacks: replay attack (through replaying old messages to neighbors resulting in causing connection failure), router advertisement forgery (forging a router advertisementra message with an invalid next hop), privacy attack (assuming that the prefix sent by the rsu in the ra message remains unchanged and prefix is associated to a geographic area, a malicious node on the internet can create a map based on this association), rsu spoofing (spoof the ip address of the rsu and send an invalid prefix), duplicate address detection (this mechanism is vulnerable to a denial of service attack). the attacker responds to all address duplication detection messages that it has already taken the requested ipv6 address [104]). vehicle to network (v2n) attacks: dos on the roadside wifi access point (wifi ap), mobike vulnerability many attacks can be performed on mobike such as denial of service (dos), man-in-the-middle (mitm) and spoofing attacks [105]. besides that, mobile femtocell is based on femtocell (home enodeb (henb) [106]) and it can be subject to several types of attacks as physical attacks (the hacker can modify or replace henb components), configuration attacks (misconfiguration of the access control list – acl of the targeted henb), protocol attacks (e.g. man-in-the-middle attacks on henb), dos on mobile operator’s core network, user data and identity privacy attacks, radio resource management tampering [107]; 5g technology can be perceived as a revolution for vehicle to network communications, even considering vehicles are able to connect to the 5g cellular network through direct links with the mmwave small cells or by via other devices using d2d communication. the mmwave and the d2d can be subject to the eavesdropping attack [108]. the authors in [109] classify next-generation networks (e.g. 5g) as vulnerable to the jamming attack since they are based on the densification (ultra-dense networks udns). spoofing attacks have been reported [110], as well as pii (personally identifiable information) and spi (serial peripheral interface) leaks were some of the threat types for iot (internet of things) in-car wifi smart appliances. such attacks could turn cars into mobile hotspots, connecting passenger devices to the internet and grabbing sensitive information. some studies have shown some security requirements should be found in vehicular communication environments. these requirements are [97]: 1. confidentiality: overall, v2x communication tries to improve the awarness of car’s surroundings, such as the existence of other cars or hazards, for example. therefore, vehicular communications are typically open, since confidentiality represents a minor requirement. only some specific issues such as transactional applications (e-tolling, pay-perview, for example) will require confidentiality. 2. integrity: v2x messages are the basis for decision making, such as in warning drivers or trigerring an automatic response from the vehicle. the latter requires messages in their integrity. assuring messages were not manipulated may be carried out by using integrity mechanisms, such as digital signature, for example, which has been explored in current telecommunications standards [92]; 3. authentication: each message sent over a vehicular network may be authenticated to prevent malicious attackers from inserting messages. the authentication is usually provided by digital signatures and public key infrastructures; 4. availability: v2x communication must be available for real-time safety applications. nonetheless, this is a hard task considering that jamming is likely to happen in wireless communication [111]; 5. protection of privacy: even though there are some requirements for authentication, as described before, vehicles and drivers should not be identifiable, otherwise the location profiling of drivers could be exposed; 6. liability or non-repudiation: this feature requires that any receiving entity could prove to third parties the message was sent by a unique sender. this is somewhat challenging when confronted with privacy protection and the fact that authorized entities, such as those responsible for law enforcement, should identify vehicles or not, is currently under scrutiny. the unique characteristics of moving vehicles have been underlined and some improvement suggestions from literature have been presented elsewhere (improving transport protocols, enhancing mac protocols for mobility related devices etc.) [68]. as described by [112] main constraints under which vanet must operate, as well as the security requirements: authentication, availability, non-repudiation, privacy, real-time guarantees etc. a novel privacy-preserving randomized authentication protocol is proposed by [113]. the principle of homomorphic encryption is used to allow each individual vehicle to generate its own identities. therefore, the public acceptance and the addoption of cooperative v2v safety applications will depend on suitable levels of security as an integral part of the system. differently from other safety technologies, v2v safety applications are cooperative, since both vehicles must send, receive, and analyze data in real time. this cooperative data exchange about potential threats and hazards is crucial for alerting and warning drivers on their decisions and actions towards avoiding imminent accidents. this is a new paradigm which is at odds with stand-alone vehicle systems based on sensors. nevertheless, a cooperative system can only work when participants in that system trust the alerts and warnings broadcasted by other v2v devices relying on messages from other v2v devices [5]. as described in [114], the already existing and/or proposed mitigation strategies depend on the attack type. but despite this still open emerging security problems such as the trustworthiness evaluation of nodes in vanet, data context trust and verification and so on. the dsrc, for example, received more security and privacy improvements than traditional wifi and the connected vehicle environment is guarded by physical controls (physical protection around equipment such as tamper-proof casings), technical controls (technologies designed to protect data, such as firewalls, access management, and encryption) and administrative controls (laws and regulations regarding unauthorized collection, storage, and disclosure of data and fair information practice principles) [5]. moreover, the polytics for data access and its availabitlity in brazil provides the confidence to consumers about iot resources. this may be observed from the laws: 1. the brazilian civil rights framework for the internet (law 12,965/2014) regulates the use of the internet in terms of principles, warranties, rights, duties and government actions in order to ensure the organization and order about the internet; 2. brazilian general data protection law (law 13,709/2018) – this law provides on gathering, processing, protecting, and using data by a natural person or legal entity governed by public or private law, in electronic or physical support the processing of personal data. it is expected to go into effect in february 2020. these laws impose criminal and administrative penalties when the norms are violated. however, despite the existence of laws and safety protocols to be followed by information-sharing communication entities, consumers should rely on best practices regarding information security, such as those described elsehwhere [115], as well as general recommendations on information secutiry, e.g.: ensuring all security and operation systems, and technological devices are up-to-date or automatically updated when their hardware or operating system provider issues new fixes; preventig the execution of malicious code on macros (e.g. by disabling automatic runs), analyzing and improving the configuration setup (firewall, sharing, privilege levels and access permission) and generating strong-type passwords and properly store sensitive information. d. social challenges the organizations liable for managing, regulating, and supervising the national traffic system in brazil are public institutions composed by the entities: national traffic council (contram), national authority for terrestrial transport (antt), national transport infrastructure department (dnit), state traffic department (denatram), federal highway police (prf), state traffic council (cetran), department of roads (der), municipal traffic department, and municipal highway entities [116]. therefore, the adoption of new technologies on municipal roads or federal/state highways must undergo the scrutiny of the public jurisdiction before its implementation. this allows for a higher control over traffic infrastructure (e.g. legislation, signaling, road maintenance, and construction of new roads), monitoring traffic conditions, counting vehicles, weighing and tolling. public institutions must be properly aligned with technological innovations for integrating the smart transport system. this way they would enact laws on the adoption of new technologies and vehicle services, such as the frequency spectrum, homologation of electronic equipments (electronic circuits from control circuit, processing, communication and power boards), besides motivating the integration among many technological assets (e.g. vehicles, rsu, infrastructure and personal devices). therefore, the adoption of a vehicular communication technology will depend on the population, wherein most contributors come from, who should notice the importance of implementing the its traffic infrastructure, the gradual incorporation of vehicular technology and the use of information arising from data processing for social welfare, i.e., the best transport option to choose from (subway, bus or private car), traffic and weather conditions, jamming and route optimization (algorithm for predicting arrival times in mobility applications [117], [118]), so that users could reduce their travel time from origin to destination. an example is given by the progressive use of urban mobility services such as 99 (available in 1705 cities) [119], uber (100 cities) [120] and cabify (8 cities) [121], thanks to the alliance and mobilization of private companies and the population against polytics favoring taxi services aimed at keeping high transport costs and the restriction of using mobility applications [122]. this was possible because the population noticed the benefits from sustainable mobility arising from mobility applications (affordable prices when compared to conventional transport modes for short distances due to their low operating costs, flexibility in terms of service availability and practicality in terms of service use) and kept aligned to the interests of the companies. this achievement may be understood through methodologies and implemented actions by means of laws or resolutions devoted to make people aware of and show the feasibility of using and adopting new technologies, besides successful case studies such as that from mcity driverless shuttle, developed by the university of michigan [74] and the awareness of how sustainable urban mobility is in the city of curitiba [123]. the compelling motivation comes from: 1. gradual investment in preventing accidents by the use of its technologies in order to save 3.1 billion us dollars spent on accidents, as shown in table 1; 2. brazil renewed its commitment towards the united nations, during the decade of action for road safety 2011/2020, which means reducing by half the number of traffic deaths and injuries through the national plan for the reduction of deaths and injuries in road accidents (pnatrans), law 13,614/18, according to [124]. vi. legislation, public policies and legal aspects the adoption of connected vehicles and its technologies in conjuction with the increasing level of vehicular automation has economical, legal, criminal, and administrative consequences. according to [7], the economical consequences impact the allocation of vehicles, the supply chain from automakers, passenger transportation services and the income of those working as drivers, such as taxi services and mobility applications, due to the gradual replacement of the driver; ondemand use of vehicles, instead of a daily rent; and redesign of automakers to adapt themselves to the new consumption requirements without the need of a car per family member. it expected that the productivity of workers increases, once the time for daily trips from home to work and vice-versa (which means 64 minutes on average in sao paulo) will be reverted to production capacity by using technological resources (tablets, laptops, etc). expected legal consequences are the implementation of the social function of property in light of car sharing, which will require usage fees, renting a vehicle and higher traffic efficiency. finally, there is the civil responsibility and the convenience of compulsory insurance: on the basis of negligent entrustment, the brazilian superior court of justice (stj) understood that the losses caused to third parties are the responsibility of the owner of the vehicle, even if it is proven that he/she has not driven it, since the owner has the intellectual command and will be responsible for damages caused to third parties independently of being considered guilty. however, in case the damage was caused by a manufacturing defect, it is legal that the owner asks the manufacturer or the company responsible for the maintenance the reimbursement of the amount of money awarded in compensation, on the basis of prohibiting unjust enrichment, according to article 934 from the brazilian civil code and on the analogical application of article 13, sole paragraph, from the brazilian consumer defense code; the same situation happens when a consumer suffers damages arising from an accident. in such a case the legislation must be adapted for enforcing a mandatory insurance from vehicle owners. therefore, it is noticed that the specific legislation in brazil for the use of its technologies is outdated, even though it was implemented since 2009, it is still restricted to the adoption and implementation of norms, protocols, specifications and prices for highway communication infrastructure, such as antt n. 3,323-a on 11/18/2009 [125]. this regulation provides for the adoption of data communication protocols and libraries of data standards from ntcip (national transportation communications) for its protocols developed by the national electronics manufacturers association (nema), in conjunction with the american association of state highway and transportation officials (aashto), the institute of transportation engineers (ite) and antt n. 3,576 on 09/02/2010 [87]. this resolution provides for the specifications and prices for its systems (intelligent transportation systems) based on vehicular traffic sensing; variable message signs fixed; of variable message signs mobile; weather sensors; surveillance cameras (cftv) and height detection to be adopted on federal concession highways, regulated by antt [125]. however, a transport partnership between brazil and the united states is under way, whose primary aim is to allow a technical exchange in order to boost its and reinforce the implementation of national transportation communications for intelligent transportation system protocols (ntcip), besides fostering the discussion on the evolution of the architecture and support for integrating connected vehicles [126]. vii. conclusion and prospects the vehicle communication technology, mainly driven by the development of v2v, v2i, and the iov concept, is already a reality in the aftermarket and it is possible to acquire equipments to connect vehicles and traffic infrastructure. the connectivity solutions already available in the automotive industry are restricted to v2d communication, via bluetooth and usb connection applications. regarding traffic infrastructure solutions, other resources were found, such as e-toll payment and the opening of gates by using rfid. many researches and practical tests have been carried out since the late 1980s, and today the advance of scientific research is well ahead of the solutions found in the market (such as the use of 5g for v2i communication). in addition, although the current development phase of connected vehicles is not sophisticated, it allows vehicles to communicate with each other and with the traffic infrastructure. it has been found that many countries (from the european union, the united states, japan, and china) have their own vehicle communication standards and are implementing tests in real-life traffic situations for later standardization and description of relevant legislation for connected vehicles. this demonstrates the need for harmonization of international standards and further research on standards, vehicle experiments, development and adoption of connected vehicles more efficient and easier. during the research it was noticed that the automobile market is very restrictive in terms of adoptiing new technologies due to the costs of manufacturing and assembly, both to the automakers and to the customers and to the time required for certification and homologation of new equipments. the solution would be the use of technological resources in the aftermarket, where the user could choose the type of equipment to be used and install it in the vehicle, to make it connected. the present work aimed at the research and discussion of main studies in connected vehicles, emphasizing the state-of-the-art v2v and v2i communications. in this case, the goal for state-of-the-art technologies was reached, since many studies were searched over the main databases from the academic environment and reports from entities related to its. for this purpose, the article objective was accomplished, pointing out the simulation resources, devices used for vehicular connectivity and practical tests developed for communication evaluation. research case studies were presented for academic and governmental entities and it was possible to observe a comprehensive, diversified state of the art in the field with well-defined technical specifications through the standards and norms of major regulatory institutions such as iso, ieee, etsi, usdot, and ec. however, no simulation studies were described, nor was it possible to conduct a case study to obtain all the communication requirements necessary to implement a stand-alone vehicle and to correlate the data achieved with the academic or governmental database. there is still a lot of research and development to be carried out, since the implementation of connected vehicles in real its environments is in the initial phase and there is a shortage of transit infrastructure and telecommunications able to support all the applications. as research and future studies, it is proposed searching for each communication requirement to keep the vehicles connected, perform simulation for operation analysis and practical tests involving vehicular applications and traffic infrastructure. other studies can be developed based on the use of 4g and 5g technologies for broadcasting between the traffic center and the vehicles versus the use of devices connected to the internet through iot. acknowledgment the authors are grateful to: coordenação de aperfeiçoamento de pessoal de nı́vel superior (capes); conselho nacional de desenvolvimento cientı́fico e tecnológico (cnpq); fundação de amparo à pesquisa do estado de são paulo (fapesp); departamento de comunicações (decom); faculdade de engenharia elétrica e de computação (feec); universidade estadual de campinas (unicamp); for their financial, administrative, and infrastructure support for the development of this research. references [1] who, “global status safety on road report 2015,” world health organization, tech. rep., 2015. [online]. available: http: //www.who.int/ [2] v. maciel, “óbitos por acidentes de trânsito caem pelo segundo ano consecutivo,” 2017. [3] institute for applied economic research (ipea), “estimativa dos custos dos acidentes de trânsito no brasil com base na atualização simplificada das pesquisas anteriores do ipea,” relatório de pesquisa do instituto de pesquisa econômica aplicada, p. 20, 2015. [online]. available: http://www.en.ipea.gov.br/ [4] c. h. r. de carvalho, “mortes por acidentes de transporte terrestre no brasil: análise dos sistemas de informação do ministério da saúde,” instituto de pesquisa econômica aplicada, p. 50, 2016. [online]. available: http://www.ipea.gov.br [5] j. harding, g. powell, r. yoon, j. fikentscher, c. doyle, d. sade, m. lukuc, j. simons, and j. wang, “vehicle-to-vehicle communications : readiness of v2v technology for application,” national highway traffic safety administration, washington, dc, tech. rep. august, 2014. [6] nhtsa, “preliminary regulatory impact analysis fmvss no. 150 vehicle-to-vehicle communication technology for light vehicles,” national highway traffic safety administration, tech. rep., 2016. [7] c. e. e. de oliveira and t. a. c. b. leal, “considerations on the autonomous vehicles possible economic, urban and legal impacts,” senado federal, tech. rep., 2016. [online]. available: https://www12.senado.leg.br/publicacoes/ estudos-legislativos/tipos-de-estudos/textos-para-discussao/td214/view http://www.who.int/ http://www.who.int/ http://www.en.ipea.gov.br/ http://www.ipea.gov.br https://www12.senado.leg.br/publicacoes/estudos-legislativos/tipos-de-estudos/textos-para-discussao/td214/view https://www12.senado.leg.br/publicacoes/estudos-legislativos/tipos-de-estudos/textos-para-discussao/td214/view [8] a. chekkouri, a. ezzouhairi, and s. pierre, “connected vehicles in an intelligent transport system,” in vehicular communications and networks: architectures, protocols, operation and deployment, w. chen, ed. elsevier, 2015, vol. 1, ch. 10, pp. 193–. [9] q. hong, e. p. dennis, r. wallace, and j. cregger, “global harmonization of connected vehicle communication standards,” michigan department of transportation and the center for automotive research, tech. rep., 2016. [10] k. ranjan, “the lattice: an intelligent grid for connected car industry,” in 2017 ieee transportation electrification conference (itec-india), ieee, ed., 2017, pp. 1–5. [11] r. a. badea and l. stanciu, “a survey and research model for vehicular communication and security challenges,” 2018 international conference on communications (comm), pp. 291–296, 2018. [12] j. greenough, “the connected car report : forecasts , competing technologies , and leading manufacturers,” 2016. [online]. available: https://www.businessinsider.com.au/ connected-car-forecasts-top-manufacturers-leading-car-makers-2016-4 [13] statista, “projected size of the global connected car market in 2016 and 2021, by segment (in billion euros),” 2016. [online]. available: https://www.statista.com/statistics/297816/ connected-car-market-size-by-segment/ [14] g. v. research, “connected car market size to reach $180.30 billion by 2022,” 2016. [online]. available: https://www.grandviewresearch. com/press-release/global-connected-car-market [15] z. ning, f. xia, n. ullah, x. kong, and x. hu, “enabling mobile and wireless technologies for smart cities: vehicular social networks: enabling smart mobility,” ieee communications magazine, vol. 55, no. 5, pp. 49–55, 2017. [16] r. molina-masegosa and j. gozalvez, “lte-v for sidelink 5g v2x vehicular communications: a new 5g technology for short-range vehicle-to-everything communications,” ieee vehicular technology magazine, no. december, 2017. [17] k. a. rahman and k. e. tepe, “towards a cross-layer based mac for smooth v2v and v2i communications for safety applications in dsrc/wave based systems,” 2014 ieee intelligent vehicles symposium (iv), pp. 969–973, 2014. [18] r. f. atallah, m. j. khabbaz, and c. m. assi, “vehicular networking: a survey on spectrum access technologies and persisting challenges,” vehicular communications, vol. 2, no. 3, pp. 125–149, 2015. [online]. available: http://dx.doi.org/10.1016/j.vehcom.2015.03.005 [19] h. chenguang, z. kaiyu, and w. shouming, “analysis of the channel capacity with shadowing fading in vanet,” 2018 14th international wireless communications & mobile computing conference (iwcmc), pp. 577–581, 2018. [20] p. salvo, f. cuomo, a. baiocchi, and i. rubin, “investigating vanet dissemination protocols performance under high throughput conditions,” vehicular communications, vol. 2, no. 4, pp. 185–194, 2015. [online]. available: http://dx.doi.org/10.1016/j.vehcom.2015.07. 003 [21] d. das and r. misra, “efficient vehicle to vehicle communication protocol for vanets,” engineering and computational sciences (raecs), 2014 recent advances in, pp. 1–6, 2014. [22] m. amoozadeh, h. deng, c. n. chuah, h. m. zhang, and d. ghosal, “platoon management with cooperative adaptive cruise control enabled by vanet,” vehicular communications, vol. 2, no. 2, pp. 110–123, 2015. [online]. available: http://dx.doi.org/10.1016/j.vehcom.2015.03. 004 [23] c. campolo, a. molinaro, a. o. berthet, and a. vinel, “full-duplex radios for vehicular communications,” ieee communications magazine, vol. 55, no. 6, pp. 182–189, 2017. [24] w. xu, h. zhou, n. cheng, f. lyu, w. shi, j. chen, and x. shen, “internet of vehicles in big data era,” ieee/caa journal of automatica sinica, vol. 5, no. 1, pp. 19–35, 2018. [25] l. atzori, a. iera, and g. morabito, “the internet of things : a survey the internet of things : a survey,” computer networks, vol. 54, no. 15, pp. 2787–2805, 2010. [online]. available: http://dx.doi.org/10.1016/j.comnet.2010.05.010 [26] a. zanella, n. bui, a. castellani, l. vangelista, and m. zorzi, “internet of things for smart cities,” ieee internet of things journal, vol. 1, no. 1, pp. 22–32, 2014. [online]. available: http: //ieeexplore.ieee.org/lpdocs/epic03/wrapper.htm?arnumber=6740844 [27] p. f. pires, f. c. delicato, t. batista, and t. barros, “capı́tulo 3 plataformas para a internet das coisas,” in xxxiii simpósio brasileiro de redes de computadores e sistemas distribuı́dos. universidade federal do espı́rito santo, 2015. [28] f. cunha, l. villas, a. boukerche, g. maia, a. viana, r. a. f. mini, and a. a. f. loureiro, “data communication in vanets: survey, applications and challenges,” ad hoc networks, vol. 44, pp. 90–103, 2016. [online]. available: http://dx.doi.org/10.1016/j.adhoc.2016.02. 017 [29] m. baunach, r. m. gomes, m. malenko, f. mauroner, l. b. ribeiro, and t. scheipel, “smart mobility of the future – a challenge for embedded automotive systems,” elektrotechnik & informationstechnik (2018), pp. 304–308, 2018. [online]. available: http://dx.doi.org/10.1007/s00502-018-0623-6 [30] s. q. yang fangchun, wang shangguang, li jinglin, liu zhihan, “an overview of internet of vehicles,” china communications, no. october, pp. 1–15, 2014. [31] o. kaiwartya, a. h. abdullah, y. cao, a. altameem, m. prasad, c. t. lin, and x. liu, “internet of vehicles: motivation, layered architecture, network model, challenges and future aspects,” ieee access, vol. pp, no. 99, 2016. [32] k. ullah, “on the use of opportunistic vehicular communication for roadside services advertisement and discovery,” ph.d. dissertation, university of são paulo, 2016. [33] a. c. regan and r. chen, “chapter 2: vehicular ad hoc networks,” in vehicular communications and networks: architectures, protocols, operation and deployment, w. chen, ed. elsevier, 2015, vol. 1, pp. 29–35. [34] e. uhlemann, “introducing connected vehicles [connected vehicles],” ieee vehicular technology magazine, vol. 10, no. 1, pp. 23–28, 2015. [35] ieee vehicular technology society (vts), “ieee connected vehicles initiative,” p. 2018, 2018. [online]. available: http://sites.ieee.org/ connected-vehicles/ieee-connected-vechicles/connected-vehicles/ [36] a. boukerche, h. a. b. f. oliveira, e. f. nakamura, and a. a. f. loureiro, “vehicular ad hoc networks : a new challenge for localization-based systems q,” computer communications, vol. 31, pp. 2838–2849, 2008. [37] u. lee and m. g. bell, “a survey of urban vehicular sensing platforms,” computer networks, vol. 54, no. 4, pp. 527–544, 2010. [online]. available: http://dx.doi.org/10.1016/j.comnet.2009.07.011 [38] r. i. meneguette, g. p. filho, d. l. guidoni, g. pessin, l. a. villas, and j. ueyama, “increasing intelligence in inter-vehicle communications to reduce traffic congestions: experiments in urban and highway environments,” plos one, vol. 11, no. 8, pp. 1–25, 2016. [39] d. levinson, “the value of advanced traveler information systems for route choice,” transportation research part c, vol. 11, no. 1, pp. 75– 87, 2003. [40] a. grzybek, g. danoy, p. bouvry, and m. seredynski, “mitigating flash crowd effect using connected vehicle technology,” vehicular communications, vol. 2, no. 4, pp. 238–250, 2015. [online]. available: http://dx.doi.org/10.1016/j.vehcom.2015.10.002 [41] f. r. yu, “transportation systems,” ieee transactions on vehicular technology, vol. 65, no. may, pp. 3843–3844, 2016. [42] s. a. a. shah, e. ahmed, m. imran, and s. zeadally, “imminent communication technologies for smart communities: 5g for vehicular communications,” ieee communications magazine, no. january, pp. 111–117, 2018. [43] a. mastrosimone and d. panno, “moving network based on mmwave technology: a promising solution for 5g vehicular users,” wireless networks, vol. 24, no. 7, pp. 1–18, 2017. [44] j. contreras-castillo, s. zeadally, and j. guerrero-ibañez, “internet of vehicles: architecture, protocols, and security,” ieee internet of things journal, vol. 4662, no. c, pp. 1–1, 2017. [online]. available: http://ieeexplore.ieee.org/document/7892008/ [45] z. fantian, l. chunxiao, z. anran, and h. xuelong, “review of the key technologies and applications in internet of vehicle,” in 2017 ieee 13th international conference on electronic measurement & instruments, 2017, pp. 228–232. [46] s. andrews, “section 9: vehicular communications systems,” in handbook of intelligent vehicles, a. eskandarian, ed. springer-verlag london ltd, 2013, vol. 53, no. 9, pp. 1093–1120. [47] v. p. barcelos, “análise e experimentaçãoo do padrão ieee 802.11p em redes veiculares hı́bridas,” ph.d. dissertation, universidade federal de lavras, 2014. https://www.businessinsider.com.au/connected-car-forecasts-top-manufacturers-leading-car-makers-2016-4 https://www.businessinsider.com.au/connected-car-forecasts-top-manufacturers-leading-car-makers-2016-4 https://www.statista.com/statistics/297816/connected-car-market-size-by-segment/ https://www.statista.com/statistics/297816/connected-car-market-size-by-segment/ https://www.grandviewresearch.com/press-release/global-connected-car-market https://www.grandviewresearch.com/press-release/global-connected-car-market http://dx.doi.org/10.1016/j.vehcom.2015.03.005 http://dx.doi.org/10.1016/j.vehcom.2015.07.003 http://dx.doi.org/10.1016/j.vehcom.2015.07.003 http://dx.doi.org/10.1016/j.vehcom.2015.03.004 http://dx.doi.org/10.1016/j.vehcom.2015.03.004 http://dx.doi.org/10.1016/j.comnet.2010.05.010 http://ieeexplore.ieee.org/lpdocs/epic03/wrapper.htm?arnumber=6740844 http://ieeexplore.ieee.org/lpdocs/epic03/wrapper.htm?arnumber=6740844 http://dx.doi.org/10.1016/j.adhoc.2016.02.017 http://dx.doi.org/10.1016/j.adhoc.2016.02.017 http://dx.doi.org/10.1007/s00502-018-0623-6 http://sites.ieee.org/connected-vehicles/ieee-connected-vechicles/connected-vehicles/ http://sites.ieee.org/connected-vehicles/ieee-connected-vechicles/connected-vehicles/ http://dx.doi.org/10.1016/j.comnet.2009.07.011 http://dx.doi.org/10.1016/j.vehcom.2015.10.002 http://ieeexplore.ieee.org/document/7892008/ [48] m. l. sichitiu and m. kihl, “inter-vehicle communication systems: a survey,” ieee communications surveys & tutorials, vol. 10, no. 2, pp. 88 – 105, 2008. [49] j. luo and j.-p. hubaux, “nicht zu warm und nicht zu kalt,” in embedded security in cars. springer berlin heidelberg, 2006, vol. 155, no. 29. [50] usdot, “connected vehicles: cv pilot deployment program,” 2018. [online]. available: https://www.its.dot.gov/pilots/cv{ }pilot{ }apps. htm [51] l. ward and m. simon, “intelligent transportation systems using ieee 802 . 11p application note,” rohde & schwarz, tech. rep., 2015. [52] f. bai, h. krishnan, and v. sadekar, “towards characterizing and classifying communication-based automotive applications from a wireless networking perspective,” proceedings of ieee workshop on automotive networking and applications (autonet), pp. 1–25, 2006. [53] j. m. león-coca, d. g. reina, s. l. toral, f. barrero, and n. bessis, “intelligent transportation systems and wireless access in vehicular environment technology for developing smart cities and wireless access in vehicular smart cities,” in big data and internet of things: a roadmap for smart environments, n. bessis and c. dobre, eds. springer, 2014, pp. 285–313. [54] ieee, “p1609.0/d13, oct 2018 ieee draft guide for wireless access in vehicular environments (wave) architecture,” ieee, tech. rep., 2018. [online]. available: https://ieeexplore.ieee.org/ document/8509668 [55] c. s. ieee, “specific requirements part 11 : wireless lan medium access control ( mac ) and physical layer ( phy ) specifications amendment 5 : television white spaces ( tvws ) operation ieee computer society,” ieee computer society, tech. rep., 2016. [56] iteris, “dedicated short range communication at 5 . 9 ghz standards group,” p. 5590, 2016. [online]. available: https: //local.iteris.com/spc/html/std/stgrdsrc{ }5ghz.htm [57] j. b. kenney, g. bansal, and c. e. rohrs, “limeric : a linear message rate control algorithm for vehicular dsrc systems,” in vanet ’11 proceedings of the eighth acm international workshop on vehicular inter-networking, no. june, 2011. [58] a. festag, “standards for vehicular communication—from ieee 802.11p to 5g,” elektrotechnik & informationstechnik, vol. 132, no. 7, pp. 409–416, 2015. [online]. available: http://dx.doi.org/10. 1007/s00502-015-0343-0 [59] t. noltet, h. hanssont, and l. l. bellot, “automotive communications past, current and future,” in 2005 ieee conference on emerging technologies and factory automation, vol. 1. ieee, 2005, pp. 985– 992. [60] m. faezipour, m. nourani, a. saeed, and s. addepalli, “progress and challenges in intelligent vehicle area networks,” communications of the acm, vol. 55, no. 2, pp. 90–100, 2012. [61] d. paret, multiplexed networks for embedded systems: can, lin, flexray, safe-by-wire... wiley, 2007. [62] s. tuohy, m. glavin, c. hughes, e. jones, m. trivedi, and l. kilmartin, “intra-vehicle networks : a review,” ieee transactions on intelligent transportation systems, vol. 16, no. 2, pp. 534 – 545, 2014. [63] j. e. siegel, d. c. erb, and s. e. sarma, “a survey of the connected vehicle landscape architectures, enabling technologies, applications, and development areas,” ieee transactions on intelligent transportation systems, vol. 19, no. 8, pp. 2391–2406, 2018. [64] y. huo, w. tu, z. sheng, and v. c. leung, “a survey of in-vehicle communications : requirements , solutions and opportunities in iot,” 2015 ieee 2nd world forum on internet of things (wf-iot), 2015. [65] j. swanson and m. serughetti, “using ethernet in automotive networks,” pp. 1–3, 2014. [online]. available: http://www.techdesignforums.com/practice/technique/ using-ethernet-automotive-networks/ [66] a. azman, s. yogarayan, s. leong, w. jian, s. fatimah, a. razak, k. j. raman, m. fikri, a. abdullah, s. z. ibrahim, a. hudaya, m. amin, and k. s. muthu, “comprehensive study of wireless communication technologies for vehicular communication,” 2018 3rd international conference on computer and communication systems (icccs), pp. 314–317, 2018. [67] m. a. haque and m. d. hossain, “technology survey of wireless communication for in-vehicle applications,” skima 2014 8th international conference on software, knowledge, information management and applications, 2014. [68] n. lu, n. cheng, n. zhang, x. shen, and j. w. mark, “connected vehicles : solutions and challenges,” ieee internet of things journal, vol. 1, no. 4, pp. 289–299, 2014. [69] m. ali, “chapter 8: standards of communications in the intelligent transport systems (its),” in autonomous vehicles, n. bizon, l. dascalescu, and n. tabatabaei, eds. nova science publishers, 2014, vol. 3, no. 1, pp. 235–246. [70] ramon dos reis fontes, claudia campolo, christian esteve rothenberg and a. molinaro, “from theory to experimental evaluation : resource management in software-defined vehicular networks,” ieee access, vol. 5, 2017. [71] engimind, “relatório final: desenvolvimento de um manual de referência para concepção, dimensionamento e implementação de praças de pedágio em concessões rodoviárias,” national land transportation agency (antt), tech. rep., 2018. [online]. available: antt.gov.br/backend/galeria/arquivos/2018/10/16/relatorio{ }final.pdf [72] l. r. bencke, a. luiz, f. perez, and o. costa, “smart roads: an overview of the technologies in brazil and in the world,” isys revista brasileira de sistemas de informação, vol. 10, no. 10, pp. 80–102, 2017. [73] semparar, “como funciona,” 2018. [online]. available: https: //www.semparar.com.br/como-funciona [74] d. zhao and h. peng, “from the lab to the street: solving the challenge of accelerating automated vehicle testing,” university of michigan, tech. rep. may, 2017. [75] cet, “companhia de engenharia de trafego,” 2019. [online]. available: http://www.cetsp.com.br [76] waze, “waze,” p. 2019, 2016. [online]. available: http://waze.com [77] m. n. o. sadiku, m. tembely, and s. m. musa, “internet of vehicles: an introduction,” international journal of advanced research in computer science and software engineering, vol. 8, no. 1, p. 11, 2018. [online]. available: http://ijarcsse.com/index.php/ijarcsse/article/ view/512 [78] z. wang and m. hassan, “how much of dsrc is available for nonsafety use ?” in vanet ’08 proceedings of the fifth acm international workshop on vehicular inter-networking, 2008, pp. 23–29. [79] q. chen, d. jiang, and l. delgrossi, “ieee 1609.4 dsrc multichannel operations and its implications on vehicle safety communications,” in 2009 ieee vehicular networking conference (vnc). ieee, 2009, pp. 1–8. [80] x. wu, s. subramanian, r. guha, r. g. white, j. li, k. w. lu, t. zhang, and a. bucceri, “vehicular communications using dsrc : challenges , enhancements , and evolution,” ieee journal on selected areas in communications/supplement, vol. 31, no. 9, pp. 399–408, 2013. [81] m. boban, t. t. v. vinhoza, m. ferreira, j. barros, and o. k. tonguz, “impact of vehicles as obstacles in vehicular ad hoc networks,” ieee journal on selected areas in communications, no. june 2014, 2011. [82] l. cheng, b. e. henty, d. d. stancil, f. bai, and p. mudalige, “mobile vehicle-to-vehicle narrow-band channel measurement and characterization of the 5 . 9 ghz frequency band,” ieee journal on selected areas in communications, vol. 25, no. 8, pp. 1501–1516, 2007. [83] c. f. mecklenbrauker, j. karedal, a. paier, t. zemen, and n. czink, “characterization and its implications for wireless system design and performance,” in proceedings of the ieee, vol. 99, no. 7. ieee, 2011. [84] c. u. bas and s. c. ergen, “ultra-wideband channel model for intra-vehicular wireless sensor networks beneath the chassis : from statistical model to simulations,” 14 ieee transactions on vehicular technology, vol. 62, no. 1, pp. 14–25, 2013. [85] j.-m. girard, n. tricot, k. younsi, and j.-c. popieul, “when does the driver workload reaches its limits?” in proceedings of the 2006 ieee intelligent transportation systems conference. toronto: ieee, 2006, pp. 578–583. [86] c. wu and y. liu, “queuing network modeling of driver workload and performance,” ieee transactions on intelligent transportation systems, vol. 8, no. 3, pp. 528–537, 2007. [87] n. l. t. a. (antt), “resolução 3576 resolução nº 3576 , de 02 de setembro de 2010,” 2010. [online]. available: http://portal.antt.gov.br/ https://www.its.dot.gov/pilots/cv{_}pilot{_}apps.htm https://www.its.dot.gov/pilots/cv{_}pilot{_}apps.htm https://ieeexplore.ieee.org/document/8509668 https://ieeexplore.ieee.org/document/8509668 https://local.iteris.com/spc/html/std/stgrdsrc{_}5ghz.htm https://local.iteris.com/spc/html/std/stgrdsrc{_}5ghz.htm http://dx.doi.org/10.1007/s00502-015-0343-0 http://dx.doi.org/10.1007/s00502-015-0343-0 http://www.techdesignforums.com/practice/technique/using-ethernet-automotive-networks/ http://www.techdesignforums.com/practice/technique/using-ethernet-automotive-networks/ antt.gov.br/backend/galeria/arquivos/2018/10/16/relatorio{_}final.pdf https://www.semparar.com.br/como-funciona https://www.semparar.com.br/como-funciona http://www.cetsp.com.br http://waze.com http://ijarcsse.com/index.php/ijarcsse/article/view/512 http://ijarcsse.com/index.php/ijarcsse/article/view/512 http://portal.antt.gov.br/ [88] ibge (brazilian institute of geography and statistics), “brazil panorama,” brasilia, p. 1, 2018. [online]. available: https: //cidades.ibge.gov.br/brasil/panorama [89] n. t. a. (anatel), “mobile telephony municipalities served,” 2018. [online]. available: http://www.anatel.gov.br/setorregulado/ telefonia-movel/115-universalizacao-e-ampliacao-do-acesso/ telefonia-movel/423-telefonia-movel-municipios-atendidos [90] m. picone, s. busanelli, m. amoretti, f. zanichelli, and g. ferrari, advanced technologies for intelligent transportation systems. springer, 2015. [91] s. k. bhoi and p. m. khilar, “vehicular communication: a survey,” iet networks, vol. 3, no. 3, pp. 204–217, 2014. [online]. available: http: //digital-library.theiet.org/content/journals/10.1049/iet-net.2013.0065 [92] e. t. s. i. (etsi), “intelligent transport systems (its); decentralized congestion control mechanisms for intelligent transport systems operating in the 5 ghz range; access layer part,” european telecommunications standards institute, tech. rep., 2011. [93] brazilian association of technical standards (abnt), “abnt / cb005 brazilian automotive committee,” 2018. [online]. available: http://www.abnt.org.br/cb-05 [94] national telecommunications agency (anatel), “plan of attribution, destination and distribution of frequency bands in brazil,” tech. rep., 2017. [online]. available: http://www.anatel.gov.br [95] l. zhang, y. wu, g. k. walker, w. li, k. salehian, and a. florea, “improving lte embms with extended ofdm parameters and layereddivision-multiplexing,” ieee transactions on broadcasting, vol. 63, no. 1, pp. 32–47, 2017. [96] d. lecompte and f. gabin, “evolved multimedia broadcast/multicast service (embms) in lte-advanced: overview and rel-11 enhancements,” ieee communications magazine, vol. 50, no. 11, pp. 68–74, 2012. [97] f. kargl and j. petit, “chapter 9: security and privacy in vehicular networks,” in vehicular communications and networks architectures, protocols, operation and deployment, w. chen, ed. elsevier, 2015, vol. 1. [98] y. fraiji, l. ben azzouz, w. trojet, and l. saidane, “cyber security issues of internet of electric vehicles,” ieee wireless communications and networking conference, wcnc, vol. 2018-april, pp. 1–6, 2018. [99] t. hoppe, s. kiltz, and j. dittmann, “security threats to automotive can networkspractical examples and selected short-term countermeasures,” reliability engineering and system safety, vol. 96, no. 1, pp. 11–25, 2011. [100] k. grover, a. lim, and q. yang, “jamming and anti-jamming techniques in wireless networks: a survey,” international journal of ad hoc and ubiquitous computing (ijahuc), vol. 17, no. 4, 2014. [101] y. sun, l. wu, s. wu, s. li, t. zhang, l. zhang, j. xu, y. xiong, and x. cui, “attacks and countermeasures in the internet of vehicles,” annales des telecommunications/annals of telecommunications, vol. 72, no. 5-6, pp. 283–295, 2017. [102] y. sun, l. wu, s. wu, s. li, t. zhang, l. zhang, j. xu, and y. xiong, “security and privacy in the internet of vehicles,” proceedings 2015 international conference on identification, information, and knowledge in the internet of things, iiki 2015, pp. 116–121, 2016. [103] b. lipiński, w. mazurczyk, k. szczypiorski, and p. śmietanka, “towards effective security framework for vehicular ad-hoc networks,” journal of advances in computer networks, vol. 3, no. 2, 2015. [104] n. a. sahloul, l. benazzouz, and i. aouini, “towards an ipsec security geonet architecture,” in network of the future nof’12, 2012. [105] s. namal, m. liyanage, and a. gurtov, “realization of mobile femtocells : operational and protocol requirements,” wireless personal communications, pp. 339–364, 2013. [106] c.-k. han, h.-k. choi, and i.-h. kim, “building femtocell more secure with improved proxy signature,” in globecom 2009 2009 ieee global telecommunications conference, 2009. [107] g. mantas, n. komninos, j. rodriguez, e. logota, and h. marques, “chapter 9 security for 5g communications,” in fundamentals of 5g mobile networks, 2015. [online]. available: http://openaccess.city.ac. uk/13047/link [108] m. h. eiza, q. ni, and q. shi, “secure and privacy-aware cloudassisted video reporting service in 5g-enabled vehicular networks,” ieee transactions on vehicular technology, vol. 65, no. 10, pp. 7868–7881, 2016. [109] p. gandotra and r. kumar, “a survey on device-to-device ( d2d ) communication : architecture and security issues,” journal of network and computer applications, vol. 78, no. october 2016, pp. 9–29, 2017. [online]. available: http://dx.doi.org/10.1016/j.jnca.2016.11.002 [110] a. tarabasz, “the internet of things – digital revolution in offline market. opportunity or threat?” handel wewnȩtrzny 2016, vol. 4, no. 363, pp. 325–337, 2016. [111] o. puñal, a. aguiar, and j. gross, “in vanets we trust ? characterizing rf jamming in vehicular networks,” in proceedings of the ninth acm international workshop on vehicular inter-networking, systems, and applications,vanet’12, 2012, pp. 83–92. [112] n. k. chaubey, “security analysis of vehicular ad hoc networks ( vanets ): a comprehensive study,” in international journal of security and its applications, vol. 10, no. 5, 2016. [113] w. jiang, f. li, d. lin, and elisa bertino, “no one can track you : randomized authentication in vehicular ad-hoc networks,” 2017 ieee international conference on pervasive computing and communications (percom) no, 2017. [114] h. hasrouny, a. ellatif, c. bassil, and a. laouiti, “vanet security challenges and solutions : a survey,” vehicular communications, vol. 7, pp. 7–20, 2017. [online]. available: http://dx.doi.org/10.1016/ j.vehcom.2017.01.002 [115] cern, “cern articles on computer security,” 2018. [online]. available: https://security.web.cern.ch/security/training/en/ cernarticlesoncomputersecurity(2018).pdf [116] l. e. d. s. cardoso, “sistema nacional de trânsito,” 2015. [online]. available: http://www.conteudojuridico.com.br/?artigos{&} ver=2.52617{&}seo=1{%}3e [117] y. jia, w. xu, and x. liu, “an optimization framework for online ride-sharing markets.” [118] s. banerjee, d. freund, and t. lykouris, “pricing and optimization in shared vehicle systems : an approximation framework,” in ec ’17 proceedings of the 2017 acm conference on economics and computation, 2017. [119] 99app, “cidades: veja as categorias da 99 em sua cidade,” 2018. [online]. available: https://99app.com/sobre-a-99/cidades/ [120] uber, “descubra quais cidades do brasil têm uber,” 2018. [online]. available: https://www.uber.com/pt-br/blog/ em-quais-cidades-a-uber-esta-no-brasil/ [121] cabify, “cidades cabify. conheça todas as cidades onde operamos.” 2018. [online]. available: https://cabify.com/pt-br [122] a. boadle, “uber, rival apps join forces in brazil to stem tide of regulation,” 2017. [online]. available: https://www.reuters.com/article/us-uber-brazil/ uber-rival-apps-join-forces-in-brazil-to-stem-tide-of-regulation-iduskbn1d71ke [123] h. d. f. miranda and a. n. r. da silva, “benchmarking sustainable urban mobility : the case of curitiba , brazil,” transport policy, vol. 21, pp. 141–151, 2012. [124] j. m. de araujo, “plano nacional de redução de mortes e lesões no trânsito,” 2018. [online]. available: http://www.ctbdigital.com.br/artigo/ plano-nacional-de-reducao-de-mortes-e-lesoes-no-transito-por-julyver-modesto-de-araujo [125] n. l. t. a. (antt), “resolução antt nº 3.323-a de 18/11/2009,” agência nacional de transporte terrestre (antt), tech. rep., 2009. [online]. available: http://www.normasbrasil.com.br/norma/ resolucao-3323-2009{ }109723.html [126] u. d. o. t. (usdot), “the u . s . brazil multimodal intelligent transportation systems ( its ) workshop,” p. 8339, 2017. [online]. available: transportation.gov/office-policy/international-policy-and-trade/ us-brazil-multimodal-intelligent-transportation-systems https://cidades.ibge.gov.br/brasil/panorama https://cidades.ibge.gov.br/brasil/panorama http://www.anatel.gov.br/setorregulado/telefonia-movel/115-universalizacao-e-ampliacao-do-acesso/telefonia-movel/423-telefonia-movel-municipios-atendidos http://www.anatel.gov.br/setorregulado/telefonia-movel/115-universalizacao-e-ampliacao-do-acesso/telefonia-movel/423-telefonia-movel-municipios-atendidos http://www.anatel.gov.br/setorregulado/telefonia-movel/115-universalizacao-e-ampliacao-do-acesso/telefonia-movel/423-telefonia-movel-municipios-atendidos http://digital-library.theiet.org/content/journals/10.1049/iet-net.2013.0065 http://digital-library.theiet.org/content/journals/10.1049/iet-net.2013.0065 http://www.abnt.org.br/cb-05 http://www.anatel.gov.br http://openaccess.city.ac.uk/13047/ link http://openaccess.city.ac.uk/13047/ link http://dx.doi.org/10.1016/j.jnca.2016.11.002 http://dx.doi.org/10.1016/j.vehcom.2017.01.002 http://dx.doi.org/10.1016/j.vehcom.2017.01.002 https://security.web.cern.ch/security/training/en/cern articles on computer security (2018).pdf https://security.web.cern.ch/security/training/en/cern articles on computer security (2018).pdf http://www.conteudojuridico.com.br/?artigos{&}ver=2.52617{&}seo=1{%}3e http://www.conteudojuridico.com.br/?artigos{&}ver=2.52617{&}seo=1{%}3e https://99app.com/sobre-a-99/cidades/ https://www.uber.com/pt-br/blog/em-quais-cidades-a-uber-esta-no-brasil/ https://www.uber.com/pt-br/blog/em-quais-cidades-a-uber-esta-no-brasil/ https://cabify.com/pt-br https://www.reuters.com/article/us-uber-brazil/uber-rival-apps-join-forces-in-brazil-to-stem-tide-of-regulation-iduskbn1d71ke https://www.reuters.com/article/us-uber-brazil/uber-rival-apps-join-forces-in-brazil-to-stem-tide-of-regulation-iduskbn1d71ke http://www.ctbdigital.com.br/artigo/plano-nacional-de-reducao-de-mortes-e-lesoes-no-transito-por-julyver-modesto-de-araujo http://www.ctbdigital.com.br/artigo/plano-nacional-de-reducao-de-mortes-e-lesoes-no-transito-por-julyver-modesto-de-araujo http://www.normasbrasil.com.br/norma/resolucao-3323-2009{_}109723.html http://www.normasbrasil.com.br/norma/resolucao-3323-2009{_}109723.html transportation.gov/office-policy/international-policy-and-trade/us-brazil-multimodal-intelligent-transportation-systems transportation.gov/office-policy/international-policy-and-trade/us-brazil-multimodal-intelligent-transportation-systems introduction internet of vehicles vehicular communication technologies and protocols for vehicular communication inter-vehicular commmunication intra-vehicular commmunication underlying technologies adoption of its technologies in brazil technical aspects and challenges of vehicular communication technical aspects of vehicular communication frequency spectrum allocation security aspects in vehicular communication social challenges legislation, public policies and legal aspects conclusion and prospects references  transactions on environment and electrical engineering issn 2450-5730 vol 2, no 1 (2017) © emanuel j. llanos pérez, laura m. hinestroza olascuaga  abstract—the colombian government introduced a capacity market in order to promote the diversification in the energy matrix and protect users from high prices derived from dry seasonal events. unfortunately, the flaws in the scarcity price definitiona mechanism that activates the capacity market obligation and sets a cap price for the spot markethave led into a market failure. specifically, some generation plants have been forced to be unavailable because their variable costs are significantly higher than the scarcity price. this paper presents a qualitative and quantitative analysis of the colombian electricity system, with a particular emphasis on the definition of the scarcity price. results present lessons and recommendations for policy makers based on the experience of the colombian electricity market, highlighting the need of a new definition of the scarcity price and a different focus on the energy planning scheme. index terms—capacity market, colombian energy system, energy market, energy planning, spot market price i. introduction riggered by a supply shortage provoked by an intense dry season, colombian electrical sector started its a deregulation process during mid-90’s [1]. a new electrical market structure was introduced by the change in the regulatory framework. there were two main objectives of this reform: stabilize the market, and increase the system reliability. the market structure evolved from a regional vertical integrated companies owned by the state to an inclusion of the private sector and a division of companies into four main activities: generation, transmission, distribution, and retail [2]. electricity markets in south america gained broad experience of the process of deregulation [3]. in colombia, the deregulation of electricity started in 1994, and the spot market initiated operations in july 1995, supported by laws 142 and 143. despite some technological similarities with the chilean electricity system, colombia adapted the british model in the mid-1990s. therefore, the colombia’s wholesale electricity e. j. llanos pérez is with university college of london, london, united kingdom (e-mail: emanuel.perez.15@ucl.ac.uk). l.m. hinestroza olascuaga is with instituto superior técnico, university of lisbon, lisbon, portugal (e-mail: laura.hinestroza@tecnico.ulisboa.pt). market-mercado de energía mayorista or mem-is the only one in the region where pool prices are settled in a bidding process [4]. four main transactions occur in the mem: i) transactions in the spot market, where retailers buy the energy they need to meet the requirements of their demand, ii) bilateral financial contracts between retailers and generators, which aim to reduce the agent’s exposure to market risk, iii) auctions to allocate the firm energy obligations -obligaciones de energía firme or oefsunder the scheme known as the reliability chargecargo por confiabilidad or rc and iv) other services to provide ancillary power generation services, such as the automatic generation control (agc) [5]. a. allocation of firm energy obligations in the mem in terms of the colombia’s installed capacity, approximately 64% of it comes from hydro-generation technologies and a minor proportion from thermal-generation plants (31%) [6]. this dependency on hydraulic resources causes that colombia is particularly sensitive to a phenomenon known as “el niño”, characterized by an intensive dry season with an approximate duration of 1 year [7]. as a result, it has been almost imperative for the electric energy sector to have sufficient thermal resources and hydro reservoirs with firm energy to replace hydro-generated energy in dry periods. therefore, in 2006, the colombian commission for the regulation of energy and gas (creg) introduced a new scheme to guarantee the availability of sufficient capacity to meet peak demand during long dry seasons and to ensure the long-term reliability of the electricity supply in colombia [8]. the scheme allocates firm energy obligations to new and existing generation plants at price determined in competitive auctions. the oefs are "option contracts" that commit generators to supply given amounts of energy at a predetermined scarcity price (sp), during scarcity situations. in return for agreeing to supply at the sp, during the commitment period, generators allocated with oefs receive a fixed annual option fee for each kilowatt hour (kwh) contracted. this fee is called the reliability charge. it is paid by consumers in their electricity tariff and is determined in the auction in which the generator sold its firm energy. generators redefinition of the scarcity price: lessons and recommendations from the colombian electricity market emanuel j. llanos pérez and laura m. hinestroza olascuaga t with oefs receive the rc regardless of their actual dispatch and whether the fulfilment of their obligation is required or not. the main benefit of this charge is the contribution to recover a portion of the fixed costs for generation agents, especially for peaking power plants, which does not deliver energy to the grid frequently during regular weather conditions [8]. in case the energy generated is more than the obligation specified in the oef, this additional energy will be paid or rewarded at the spot market price [9]. if generators cannot meet their oefs, they pay a penalty, equal to the difference between the spot market price and the scarcity price on the oef quantity not met in any hour. b. remuneration of the oef: definition of the scarcity price scarcity pricing is a regulatory mechanism used by electrical markets to establish a cap market price sufficiently high during low system reliability periods, before the demand is not covered due to a shortage in supply [10]. properly adjusted sp would help to incentivize: demand reduction, higher generation availability and additional energy imports. with a consistent increase of renewable generation in terms of installed capacity globally, colombia appears as a suitable referent for current and future electrical markets with a predominant renewable energy matrix and a possible exposure of scarcity of resources. in this sense, it is important to explain and recommend solutions for the current flaws of the scarcity price approach in the colombian electrical market. this experience can be valuable in order to establish more robust structures for markets with a high penetration of renewable generation. by definition, in colombia the scarcity situation established when the spot market price (spm) surpasses the sp [9]. it is established by the creg and updated on a monthly basis with the variation of the fuel operational costs of the most inefficient power plant, and other variable costs (ovcs) that generators have to pay to the system. equations (1) to (6) show the methodology to calculate the sp [11]: 𝑆𝑃𝑚 = 𝑆𝑃𝑚 𝑓 + 𝑂𝑉𝐶𝑆𝐼𝑁𝑚−1 + 𝑂𝑀𝐶𝑚−1 (1) where: 𝑆𝑃𝑚 : scarcity price for the month m in cop $/kwh fuel operational cost of the scarcity price (𝑺𝑷𝒎 𝒇 ): it is calculated in usd/mwh and then converted into cop $/kwh according to the exchange currency rate (ecr) of the calculation day. it is calculated using the new york harbor fuel oil no. 6 and depends on the following values: 𝑆𝑃𝑚 𝑓 = (𝑉𝐹𝑂𝑁6)(𝑆𝑃2014.01 𝑓 )(𝐸𝐶𝑅) (2) 𝑉𝐹𝑂𝑁6 = 𝐴𝐹𝑂𝑁6 𝑝𝑟𝑒𝑣𝑖𝑜𝑢𝑠 30 𝐷𝑎𝑦 𝐴𝐹𝑂𝑁62013.12 (3) 𝑆𝑃2014.01 𝑓 = (𝐻𝑅𝑀𝐵𝑇𝑈/𝑘𝑊ℎ )(𝐹𝐶𝐹𝑂𝑁6 ) (4) where: 𝑉𝐹𝑂𝑁6: monthly variation of new york harbor fuel oil no. 6 compared to december 2013 levels 𝐴𝐹𝑂𝑁6 𝑝𝑟𝑒𝑣𝑖𝑜𝑢𝑠 30 𝐷𝑎𝑦 : arithmetic average of the previous 30 days of the new york harbor fon6 𝐴𝐹𝑂𝑁62013.12: arithmetic average of the previous new york harbor fon6 data of december 2013 𝑆𝑃2014.01 𝑓 : january 2014 fuel operational cost of scarcity in cop $/kwh 𝐻𝑅𝑀𝐵𝑇𝑈/𝑘𝑊ℎ : heat rate of the most inefficient power plant 𝐹𝐶𝐹𝑂𝑁6: fuel price (published by ecopetrolthe largest and primary petroleum company in colombia), plus a 1.5% to cover the transportation costs. this price is converted into usd/mbtu according to the er of the last labour day of the month in which the calculation is made (it is certified by the national bank) with a heating power of 0.15 mbtu/gallon. 𝐸𝐶𝑅: exchange currency rate (ecr) of the calculation day other variable costs (𝑶𝑽𝑪𝑺𝑰𝑵𝒎−𝟏 ): they are related to the national interconnected system (sin) for the month of interest minus one, in cop $/kwh, and depend on: 𝑂𝑉𝐶𝑆𝐼𝑁𝑚−1 = (𝐸𝐸𝐶𝑚 ) + ( 𝐴𝐺𝐶𝑚−1) + (𝐹𝐴𝑍𝑁𝐼𝑚−1) + (𝐿𝑎𝑤99𝑚−1) (5) where: eecm: energy equivalent cost for the month of interest in cop $/kwh agcm−1: automatic gain control cost for the month of interest minus one, in cop $/kwh faznim−1: fund for the non-interconnected zones of the country for the month of interest minus one, in cop $/kwh law99m−1: monetary contribution of the law 99-1993 for the thermal plants for the month of interest minus one, in cop $/kwh operation and maintenance costs (𝑶𝑴𝑪𝒎−𝟏): they are calculated for the month of interest minus one in cop $/kwh. this value depends on: 𝑂𝑀𝐶𝑚−1 = 𝑂𝑀𝐶𝑗𝑢𝑛𝑒2006 ( 𝐶𝑃𝐼𝑚−1 𝐶𝑃𝐼𝑗𝑢𝑛𝑒2006 ) (6) where: 𝑂𝑀𝐶𝑗𝑢𝑛𝑒2006 : operation and maintenance costs for june 2006 in cop $/kwh 𝐶𝑃𝐼𝑗𝑢𝑛𝑒2006 : consumer price index for june 2006 in p.u 𝐶𝑃𝐼𝑚−1: consumer price index for the month of interest minus one in p.u. as an example, in september 2015, xm published the estimated value of the sp for october 2015 [12]. the calculation is based on the creg resolution no. 070-2014. https://en.wikipedia.org/wiki/colombia table i shows a detailed explanation on the calculation of this value, using (1) to (6). the obtained value for the sp was about 0.0977 usd/kwh. table i scarcity price calculation for october 2015-example. source: adapted from [12] parameter value 𝑉𝐹𝑂𝑁6 0.0001 𝑆𝑃2014.01 𝑓 0.0001 𝐸𝐶𝑅 3096.98 𝑆𝑃𝑚 𝑓 0.0714 𝐸𝐸𝐶𝑚 0.0164 𝐴𝐺𝐶𝑚−1 0.0037 𝐹𝐴𝑍𝑁𝐼𝑚−1 0.0004 𝐿𝑎𝑤99𝑚−1 0.0009 𝑂𝑉𝐶𝑆𝐼𝑁𝑚−1 0.0214 𝑂𝑀𝐶𝑗𝑢𝑛𝑒2006 0.0034 𝐶𝑃𝐼𝑚−1 0.0397 𝐶𝑃𝐼𝑗𝑢𝑛𝑒2006 0.0280 𝑂𝑀𝐶𝑚−1 0.0049 𝑆𝑃𝑜𝑐𝑡𝑜𝑏𝑒𝑟_2015 ( usd/kwh) 0.0977 despite this scheme designed to provide reliability in the colombia’s hydro-dominated electricity market, current electricity spot market prices in the mem are increasing more than tenfold, setting an energy crisis in colombia (see fig.1). in consequence, the generation plants are paid with the sp which is lower than the variable cost of the generation plant that covers the demand on scarcity or peak periods-it means the plant with the highest variable cost. thus, many generation plants are forced to be unavailable as a result of the extortionate costs of generating at a significantly higher cost than the sp. fig. 1. comparison between the scarcity price and the spot market price in colombia, 2015. note: the prices were converted into usd using the same ecr of spt f (adapted from [13]). in this sense, the current regulatory policy is not solving the problem of “missing money” (the amount of money per mw of capacity that a generator is missing in scarcity hours) [14]. this situation is a consequence of: first, one of the strongest “el niño” periods that the country has faced, and second, a failure in the electricity regulatory framework, regarding the definition of the sp. considering the above evidence, we present an analysis of the scarcity price’ definition, its main flaws, the measurements already taken by the colombian government, some recommendations to overcome the crisis , and the lessons learnt from the colombian electricity market, applicable to other similar markets. the recommendations address the need to include non-conventional energies (not affected by the dry seasons) to provide firm energy, and a new definition of the sp, considering its concept in terms of the generation plants costs and not in terms of the plants’ technical efficiency. these amendments could apply to other countries, which electricity generation relies mainly on hydro or have a similar energy markets and dry climate seasons. ii. scarcity pricing approaches in other electrical markets the relevance and the mechanism which scarcity of generation resources is dealt with, vary depending on the market structure. these approaches were categorized into two main market structures: energy only markets and energy markets with capacity payments. a. energy only markets by definition, in energy only markets generators agents can only obtain revenues through spot market and ancillary services [15]. this market design leads to moderate prices during regular conditions and, occasional high prices during scarcity conditions. in this context, setting adequate scarcity price levels is crucial for a correct market performance. if the scarcity price is too high, excessive generating capacity will be built as a result, thus more power plants would not be operating during regular conditions, increasing the market exposure for the missing money problem. in contrast, if scarcity price is too low, there will be no incentive to build new generation capacity, increasing the possibility for future energy shortages [16]. some scarcity pricing approaches had performed problems in energy only markets. in the electric reliability council of texas (ercot) and australia’s national electrical market (nem), the spoe is set by the highest bid price offer of generator agents, during normal conditions. in this process, the bids of large generator agents are mitigated; the sp is mainly established by small generator agents [17]. there are two main problems with this spoe approach. firstly, spoe calculation relies only on some groups of generator agents to accurately predict the existence and the magnitude of a scarcity period, during normal conditions [18]. second, according to [19], price bids of small-unmitigated generators agents vary widely under identical system condition, and as a result, there is a 0,00 0,10 0,20 0,30 0,40 0,50 0,60 0,70 u s d /k w h month spot market price vs scarcity price spm-spot market price usd spscarcity price usd wide range of spoe calculated under similar conditions. b. energy markets with capacity payments although scarcity situations are necessary less frequent in energy market with an integration of capacity payments, they cannot be avoided completely. scarcity pricing appears as a suitable approach for a better market performance [3]. the benefits of a proper scarcity pricing approach are just not limited to send investment signal. despite the design of the energy market, through a correct definition and method of spoe calculation, resources can be allocated more efficiently and incentivise a higher demand response [20]. in the new england energy market (iso-ne) the spoe is defined as the maximum cost incurred to meet the reserve requirementsalso known as reserve constraint penalty factor (rcpf)-, and the scarcity situation is defined when the sp is higher than rcpf. the rcpf set a cap for the sport market; if the rcpf is lower than the cost of meeting the reserve requirement level, these requirements will not be met (iso-ne, 2009). unlike the colombian approach, during scarcity situations, in the iso-ne, market scarcity revenues are subtracted from the capacity payments, avoiding extra revenues for agents. despite some energy markets do not have a specific mechanism for scarcity pricing, they have established markets structures to deal with scarcity periods. for example, south korea and chile energy markets have a capacity market based on availability. with a fast paced demand growth, energy security is a priority for these countries. therefore, the capacity payment must be high enough to attract the investment in order to avoid energy rationing [3]. the former colombian capacity market was designed based on the chilean initial design, where capacity payments are made based on the installed capacity of the power plant, and the generator availability during peak demands months or capacity shortage situation [21]. moreover, in the south korea capacity market structure, the payment is different for base-load and peaking generation units. this payment is based on the gross capital and the o&m cost of a coal and gas generation unit, for base-load and peaking power plants, respectively [22]. iii. failures of the scarcity price definition in colombia there were two main objectives with the introduction of the sp regulation. first, avoid the peak prices that the demand has to pay in scarcity periods. second, promote a good investment climate by a continuous remuneration to the electrical generators agents. it would guarantee the energy security of the country in terms of supply, especially due to a “gas crush”, motivating most of thermal generation power plants to change their fuel supply from gas to fuel oil based [9]. this section analyses the reasons why the sp is not well defined and describes the basic rationale behind the problem. a. main flaws in the definition of the scarcity price the sp should be based on the variable cost of the peak power plant that attends the demand in the scarcity [23]. instead, in the colombian market the sp is defined based on the variable costs of the most “technically inefficient” plant, which is not always the plant that attends the peak demand. other plants with higher technical efficiencies but also with higher variable costs are the ones who attend the peaks hours. with the latest spot and scarcity prices, the colombian government took over a power generation agent, termocandelaria (314 mw of installed capacity), because it was not generating energy due to the huge debt provoked by the difference between the sp and the variable generation cost of this agent [24]. this situation is particularly critical for other generation agents, and for the electricity market as a whole. there is a big probability that other generation agents could replicate the current situation of termocandelaria, attempting directly to the energy security of the country. mainly, because “el niño” is a seasonal event that impacts colombia’s raining every 2 to 6 years approximately [25] and also, because the current definition of the sp is heavily sensible to the international oil spot price. the sp’s formula has a strong correlation with oil prices (linked to the fuel oil no. 6 index). but the reality is that most colombian thermal plants are backed by fuel oil no. 2, which is more expensive and it’s less correlated with oil. therefore, the current scarcity price’s value comes from using the fuel oil no. 6 as a reference, because it’s much cheaper than the diesel used by plants (fuel oil no. 2).changing the sp represents a dilemma. on one hand, the creg argues that a change from fon6 to fon2 in the sp definition would represent an increase in the sp regardless the time of the year. therefore during normal weather conditions, agents have an incentive to increase spm because it would represent a potential higher revenue. this situation would be reflected in higher electricity bills for consumers. on the other hand, the cap price at which thermal plants can sell is not properly adjusted, increasing the probability of financial unavailability of these plants in scarcity periods, jeopardising the energy security of the country [26]. b. effects on the market’s agents: generators and consumers during el niño” period, the generation plants (especially thermal plants available in dry seasons) have increased their generation price and the final energy prices have also increased, affecting the final consumers’ economy. since september 2015, with the resolution no.178 of 2015, the creg agreed to raise (for six months) the sp from cop $302 to cop $470. this measure applied only for thermal plants that generate with diesel, in order to overcome the scarcity period. since then, the electricity tariff for residential consumers has increased by about 17.54 % [27]. although consumers are protected by the sp, if it increases, the maximum price that the demand has to assume also increases. nevertheless, the deadline for this measure is until april 28. according to the association of the largest consumers of electric energy in colombia (asoenergia), they are not willing to accept any rise in the electricity tariff, after this deadline. the guild stresses the need to adopt measures to correct the structural problems affecting the power system [28]. from the point of view of the generators, due to the millionaire losses they are facing because of the crisis, they are asking for a new agreement. according to [29], operators of the 12 thermal generation plants (which use gas and diesel) are willing to deliver their companies to the government, if by may 1, 2016 the creg does not adjust the scarcity price to more than double. iv. actions taken by the government the colombian government had taken some actions to avoid an energy crisis that could lead a possible blackout of the country in a certain hour of the day. specifically, here we critique the rationale and effects of three measures taken by the colombian government: the creg regulation 172 and 178 of 2015, the energy save program “apagar paga” and other measures also launched by the government. a. creg resolution no. 172 of 2015 given the peaks prices of the spm, almost 7 times greater than the sp (see fig 1), the creg introduced a cap price in the day-ahead bid process by publishing the creg resolution no. 172 of 2015 [30]. the cap price is equal to the 75% of the first level of incremental operational cost of energy blackout, (cro1). currently the cro1 is equal to cop $ 1191.07 [31]; therefore the cap price of the spot market of electricity in colombia is cop $893.31. however this cap price only applies when the variable hsin (hydrology of the national interconnected system) is lower than 90% or when the creg determine that is necessary. the variable hsin measure the amount of energy produced from hydraulic resources compared to the previous month. despite the effectiveness of this measure, there are also some downsides. first the “missing money” problem is not solved. thermal power plant with energy firm obligation will still be paid at the sp. second; the 75% of the cro1 is still lower than the cost of the peaking power plant. it means that some thermal power plants will not recover their variable cost even if they sell their energy at the smp. second, the time of the penalty exposure remains the same. although, the cap price for the day-ahead bid process reduces the magnitude of the penalty exposure of the agents, the spm still higher than the sp level; therefore the time frame of the scarcity situation has not changed. additionally, the price cap introduced in the regulation just depends on the level of the variable hsin to be activated. this variable implies a comparison of at least two months of generation coming from hydro resources. b. creg resolution no. 178 of 2015 the creg decided to establish a temporary floor for six month to the scarcity price. through the creg resolution 178 of 2015, the institution established that the sp would be the greater between, its monthly update and the sp of october 2015 [32]. moreover, for the generation agents that use fuel oil, the remuneration of the firm energy obligation will be approximately cop $470, which is 1.5 times higher than the current sp (approximately cop $302). this regulation also helps to reduce the time of penalty exposure of the agents. however, just like the creg resolution no. 172, the focus is more to reduce the magnitude of the penalty exposure and not the time. the resolution does not solve the “missing money” problem either. it is true that in theory there is lower probability of the event where the smp would be greater than the sp. nevertheless, given the current water reservoir level and the el niño/southern oscillation (enso) forecast [33] and the recent spm levels (see fig. 1), it is highly possible that the smp will be greater than the october sp level. moreover, the resolution is a transitory measure and implies different market conditions for certain agents. regardless the effectiveness of this resolution, the scope is only valid until may 2016, and according to the forecast, the dry season will remain until august of 2016. c. energy save programme “apagar paga” driven by the current intense and long dry season, and the high possibility of an energy black out, the colombian government launched an energy save programme called “apagar paga” or “turn off, pay off” (unofficial translation). the rationale is simple: promote the energy daily saving of at least 5% in order to avoid possible blackouts [34]. behind the widely spread advertising camping through national tv or social media, the programme includes a set of penalty for the users that do not meet the energy savaging targets. through the creg resolution no. 029 of 2016, for each extra kwh compared to the consumption in february of 2015, the users have to pay an additional amount of cop $450. this particular feature of the programme has been polemical. on one hand, the government argues that there must be a price mechanism to incentive the energy saving in the country. on the other hand, normal users argue that the government did not take the correct measure and, now they have to be overcharge for a market failure problem [35], [36]. in addition, the programme was launched late. since september of 2015, the system has been in a “scarcity situation” and the water reservoir level has dramatically dropped since then. in contrast, the energy saving programme was launched in march of 2016, forcing to a dramatic change in behaviour. it is possible that the saving targets are met. however, in order to truly change people behaviour and move them towards a more efficient use of energy, it is necessary the implementation of long-term programmes [37], [38], not just an immediate implementation used as a “safety valve” measure. d. other short-term measures since it seems that “el nino” period is testing the colombia’s electricity framework, the government has had to enact these other temporary measures to preserve the viability of the system: 1) increasing the gas supply for thermal generation plants. however, current gas supply of the country is limited and the amount of gas required by the thermo-electric sector is too high. therefore, it is expected that the regasification plant of liquid natural gas (lng) will start operation by the end of 2016, a few months later of the projected end of the dry season. 2) importing additional energy from ecuador. since december 2015 colombia has imported energy from ecuador, reaching to a peak in march 2016 of about 5.6 gwh. this measure has helped to reduce the stress of the system, although it represents a minor fraction of the country’s electricity demand (5522 gwh) [6], [39]. v. potential measures to overcome the crisis in the long-term a. the need for a better energy planning el niño has left in doubt the colombia's hydroelectric capacity to meet the whole electricity demand, which is increasing as a result of the economic activity and the population growth. all governments in latin america know that unusually dry periods appear, roughly once every 10 years [40] and therefore, there is no reason that justifies why colombia has the possibility of going to an electrical collapse, because it was already known that “el niño” would return. according to the colombia's generation and transmission expansion plan 2006-2020, the electricity demand is expected to continue increasing by between 2.8 % and 3.8 % per year in the "medium" scenario. this increase together with the phasing out of existing power plants (when they reach their end of lifetime), means that there will be a need for establishing new power generation facilities [8]. hence, colombia should consider nonconventional renewable sources of firm energy including wind, solar, biomass and geothermal, such as alternatives to thermal plants. nevertheless, the promotion of an efficient mix of resources and the achievement of a firm energy market that provides reliable electricity at least cost, requires that all resources, including variable resources such as wind power, should be eligible to receive the same reliability payment, based upon the resources ‘ability to provide firm energy [40]. the main benefits of this measure would be: 1) it leads to a more efficient mix of resources and reduces risk by establishing a more diversified portfolio instead of fossil fuels, which also could reduce electricity costs, 2) the long lead times required for large hydro and coal plants contrast with relatively limited externalities and flexibility offered by nonconventional renewable sources of energy, 3) they do not depend on the fossil fuel’s price volatility, and 4) it reduces colombia’s reliance on coal and other fossil fuels to generate electricity during dry periods, hence reducing colombia’s emissions from fossil fuels and the negative effect on the environment and human health. 4) likewise, with the current cop21 agreement there is an expected increase share of renewable energies, especially in the electricity sector [41]. a greater introduction of renewable technology in the generation park would help to achieve the international environmental target. in particular, the wind regime in colombia is among the best in south america. the potential for wind power is 18 gw, which is 900 times as much as the current capacity of 20 mw [42]. unfortunately, according to [39], the rc commitment only considers hydro and thermal power plants’ projects. from the 13 approved projects to entry the market until 2020, 69.23% of them corresponds to hydro power plants. it represents a total installed capacity of about 3769.mw out of 4431.4 mw, the total projected energy firm that will be provided by the projects. thus, it means that none amendments are prevised for a future presence of “el niño” period. in addition, in the last firm energy auction (in 2011), wind power was not eligible for firm energy payment. the main reason is the lack of information about the resource’s ability to supply firm energy. this is already done for hydro and thermal resources, but there is a need of an analogous methodology to estimate firm energy for variable resources [40]. b. new definition of the scarcity price: case study besides the above efforts made by the government, we define another possible proposal that could fix the current problems: taking into account that the current sp does not reflect the actual situation of the market, it is proposed to use the heat rate and the fuel of the most expensive power plant in the market as a reference to calculate the variable cost (instead of the most inefficient one). the intention of this measure is to cover the variable costs of peak power plants and by this avoid a possible market collapse. we simulate the behaviour of the scarcity price by considering the variable cost of the most expensive plant in the market. it corresponds to the thermal plant “termocandelaria”, which has a heat rate of 10 mbtu/kwh. although it is the most expensive plant, it is not the most inefficient. in terms of efficiency (heat rate), the value varies between 6 mbtu/kwh and 12 mbtu/kwh and other thermal plants such as “termoflores 2” (diesel) and “termocartagena 2” (fuel oil), have values of about 12 mbtu/kwh [43]. together with the above assumptions, we use diesel (fuel no.2 instead of fuel no.6) as the operation fuel and the values for other external costs (such as some taxes and other legal deductions) remain the same. also we considered the average spot market price to calculate both, the time and magnitude of penalty exposure. we applied an analysis backwards in time, in a scope of one year (from april 2015 to march 2016), using (1) to (6). fig. 2 shows a new estimation of the scarcity price contrasted with the current scarcity price and the spot market price. fig. 2. proposal for a new definition of the sp: case study. note: the scarcity prices were converted into usd using the same ecr of spt f. the new definition of the sp brings the following benefits: 1) it solves the “missing money problem”, because the plants are paid according to the most expensive plant, therefore if the variable cost of the peak plants is covered, also it does for the rest of the plants. 2) it reduces the time of penalty exposure in 60%, since the number of times the plants are exposed to market (because of the scarcity price) reduces from 196 times (using current sp) to 79 (using new definition of the sp) and 3) it also reduces the magnitude of the penalty exposure in 77%. it means the variation in price ($/kwh) during scarcity periods. vi. conclusions “el niño” period has put to test the colombian energy capacity market, and the regulatory framework as a whole. the combined situation of high spot market prices and low scarcity prices has affected both, generations agents and consumers. the current scarcity price definition is based on the cost of the most technical inefficient power plant in the market, not covering the variable cost of more efficient yet more expensive plants. consequently, some generation agents have been financially unable to operate in the market due to the big debt acquired during the scarcity period. despite the relative effectiveness of the measures taken by the government to mitigate the effects of the current energy crisis, we suggest three main amendments: redefining the scarcity price in terms of the plant with the highest variable costs of the market; redirecting the focus of the energy expansion plan through a more diversified energy matrix, centred in the inclusion of alternative renewable energy sources; and enabling long term energy efficiency programmes. results show that the proposed scarcity price definition mitigates the agent’s penalty exposure in both dimensions: magnitude and time. however, since the new value of the scarcity price is higher compared to the current levels, the consumers could be potentially exposed to higher energy tariffs due to the increase of the spot market cap price. moreover, considering the country’s high potential of solar, wind and geothermal energy, these technologies could provide firm energy, but it requires mainly two conditions: first, more studies with accurate and updated data. second, adjustments in the regulatory framework of the colombian electricity market to support their inclusion. furthermore, aligned with the measures mentioned above, longterm programmes should be implemented to truly change people's behaviour and move them towards to a more efficient use of energy. references [1] j. benavides, and i. fainboim, “private participation in infrastructure in colombia-renegotiation and disputes”, in basañes and r. willig (eds.), second-generation reforms in infrastructure services, chapter 4, pp. 89-132, 2002. [2] j. pérez arriaga, c. solé-martín, c. vázquez martínez, and c. meseguer velasco, “formación de precios en generación y diseño del mercado mayorista en el sistema eléctrico colombiano”. study prepared for the colombian association of power producersacolgen. bogotá, colombia, 1999. [3] s. arango, i. dyner, and e. r. larsen, “lessons from deregulation: understanding electricity markets in south america,” util. policy, vol. 14, no. 3, pp. 196–207, 2006. [4] j. millán, “la segunda generación de bolsas de energía: lecciones para américa latina,” interamerican development bank report, pp. 1–18, april 1999. [5] g. cervigni and d. perekhodtsev, “wholesale electricity markets,” econ. electr. mark. theory policy, pp. 1–17, january 2013. [6] expertos en mercados, xm, “colombia´s installed capacity by generation source”, [online], 2014. available at: http://informesanuales.xm.com.co/2014/sitepages/operacion/2-6capacidad-efectiva-neta.aspx (accessed 15.01.16). [7] p. cramton and s. stoft, “colombia firm energy market: summary,” engineering, february, 2007. [8] cowi, “impact analysis for integration of wind power generation in colombia. 3, market and regulatory aspects,” progress study report no.3, october, 2014. [9] commission for the regulation of energy and gas (creg), “reliability charge: regulatory scheme to guarantee the reliability in the supply of electric energy in colombia, a long-term vision,” pp.1-15., 2007. [10] j. pfeifenberger, k. spees, and a. schumacher, "a comparison of pjm’s rpm with alternative energy and capacity market designs. prepared for pjm interconnection, inc.", [online], 2009. available at: http://www.brattle.com/system/publications/pdfs/000/004/859/original/ a_comparison_of_pjm%e2%80%99s_rpm_with_alt_energy_and_ capacity_mkt_designs_pfeifenberger_et_al_sep_2009.pdf?13790147 89 (accessed: 01.10 2016). [11] asociación nacional de empresas generadoras, “un análisis del precio de escasez,” pp.1-28, 2014. [12] expertos en mercados, xm, “información cálculo del precio de escasez para el mes de octubre de 2015 según resolución creg 070 de 2014”, [online], 2015. available at: www.xm.com.co/_layouts/linkclick/linkclick.aspx?mapid=30 (accessed 15.02.2016). [13] expertos en mercados, xm, “precio de escasez vs precio de bolsa”, [online], 2015. available at: http://informacioninteligente10.xm.com.co/transacciones/paginas/preci o%20escasez%20($kwh).aspx (accessed 14.01.2016). [14] p. cramton, a. ockenfels, and s. stoft, “capacity market fundamentals,” econ. energy environ. policy, vol. 2, no. 2, pp. 1–21, 2013. [15] w. hogan, “on an “energy only” electricity market design for resource adequacy”. california iso. 2005 [16] s, stoft, power system economics. journal of energy literature, vol. 8, pp.94-99. 2002 0,00 0,10 0,20 0,30 0,40 0,50 0,60 0,70 u s d /k w h month spot market price and scarcity price vs proposed scarcity price spm-spot market price usd spscarcity price usd pspproposed scarcity price usd http://informesanuales.xm.com.co/2014/sitepages/operacion/2-6-capacidad-efectiva-neta.aspx http://informesanuales.xm.com.co/2014/sitepages/operacion/2-6-capacidad-efectiva-neta.aspx http://www.brattle.com/system/publications/pdfs/000/004/859/original/a_comparison_of_pjm%e2%80%99s_rpm_with_alt_energy_and_capacity_mkt_designs_pfeifenberger_et_al_sep_2009.pdf?1379014789 http://www.brattle.com/system/publications/pdfs/000/004/859/original/a_comparison_of_pjm%e2%80%99s_rpm_with_alt_energy_and_capacity_mkt_designs_pfeifenberger_et_al_sep_2009.pdf?1379014789 http://www.brattle.com/system/publications/pdfs/000/004/859/original/a_comparison_of_pjm%e2%80%99s_rpm_with_alt_energy_and_capacity_mkt_designs_pfeifenberger_et_al_sep_2009.pdf?1379014789 http://www.brattle.com/system/publications/pdfs/000/004/859/original/a_comparison_of_pjm%e2%80%99s_rpm_with_alt_energy_and_capacity_mkt_designs_pfeifenberger_et_al_sep_2009.pdf?1379014789 http://informacioninteligente10.xm.com.co/transacciones/paginas/precio%20escasez%20($kwh).aspx http://informacioninteligente10.xm.com.co/transacciones/paginas/precio%20escasez%20($kwh).aspx [17] j. pfeifenberger, “resource adequacy requirements, scarcity pricing and electricity market design implications”, 2014. [online]. available at: http://www.brattle.com/system/publications/pdfs/000/005/045/original/ resource_adequacy _requirements__scarcity_pricing__and_electrici ty_market_design_implications.pdf?1404 399120 (accessed 09.10.2016). [18] w. hogan, “electricity scarcity pricing through operating reserves: an ercot window of opportunity”. harvard university. cambridge (ma).2012. [19] protomac economics, “independent market monitor for the ercot wholesale market: state of the market report for the ercot wholesale electricity markets”, 2008. available at: http://www.puc.state.tx.us/wmo/documents/index.cfm (accessed 05.10.2016). [20] w. hogan, “electricity scarcity pricing through operating reserves”. economics of energy & environmental policy, 2(2), 65-86. 2013. [21] h. rudnick and j. montero, “second generation electricity reforms in latin america and the california paradigm”. journal of industry, competition and trade, vol. 2(1-2), pp.159-172, 2002. [22] j. park, n. ahn, y. yoon, k. koh, and d.w bunn, “investment incentives in the korean electricity market”. energy policy, vol. 35(11), pp.5819-5828, 2007. [23] p. cramton and s. stoft, “the convergence of market designs for adequate generating capacity with special attention to the caiso’s resource adequacy problem-a white paper for the electricity oversight board”, pp.1-72, 2006. [24] el tiempo, “millionaire debts to banks and generators, another deficit of termocandelaria”, press, [online], 2015. available at: http://m.eltiempo.com/economia/sectores/millonarias-deudas-a-bancay-generadores-otro-mal-de-termocandelaria-/16437150/1/home (accessed 11.01.2016). [25] national and oceanic atmospheric administration (noaa), “what is the el niño–southern oscillation (enso) in a nutshell?,” [online]. available at: https://www.climate.gov/news-features/blogs/enso/whatel-ni%c3%b1o%e2%80%93southern-oscillation-enso-nutshell (accessed 09.03.2016). [26] d. harbord, “second review of firm energy auctions in colombia,” pp. 1–20, march, 2012. [27] electricaribe s.a. e.s.p., “tariffs, subsidies and contributions”. [online], 2016. (accessed 04.03.2016). [28] el heraldo, “asoenergía refuses possible increase in the energy price,” press, [online], 2016. available at: http://www.elheraldo.co/economia/asoenergia-rechaza-posiblesaumentos-del-gobierno-al-precio-de-la-energia-252521 (accessed 05.04.2016) [29] portafolio, “generation plants threaten to deliver their companies to the government,” press. [online], 2016. available at: http://www.portafolio.co/negocios/empresas/generadoras-termicasamenazan-entregarle-plantas-gobierno-492899 (accessed 05.04.2016) [30] commission for the regulation of energy and gas (creg), “resolution no. 172 of 2015,” regulatory resolution, 2015. [31] unidad de planeación minero energetica (upme). “incremental operational cost of energy blackout,” [online]. available at: http://www.upme.gov.co/costosenergia.asp (accessed 30.03.16). [32] commission for the regulation of energy and gas (creg). “resolution no. 178 of 2015,” regulatory resolution, 2015. [33] national oceanic and atmospheric administration (noaa), “enso forecast,” [online], march 2015. available at: http://www.cpc.ncep.noaa.gov/products/analysis_monitoring/enso_advi sory/ensodisc.html (accessed 30.03.16). [34] ministerio de minas y energía, “energy saving tips,” [online], 2016. available at: https://www.minminas.gov.co/tips-de-ahorro-de-energia (accessed 30.03.16). [35] el colombiano, “late measures taken during energy crisis: experts,” press, [online], 2016. available at: http://www.elcolombiano.com/negocios/apagar-paga-medidas-deahorro-de-energia-son-tardias-expertos-by3809300 (accessed 30.03.16). [36] el tiempo, “recommendations: programed black out in the centre of the country,” press, [online], 2016. available at: http://www.eltiempo.com/economia/sectores/resultados-de-la-campanaapagar-paga/16548415 (accessed 30.03.16). [37] d. southerton, a. mcmeekin, and d. evans, “international review of behaviour change initiatives,” soc. res. (new. york)., pp. 1–46, 2011. [38] g. cooper, “escaping the house: comfort and the california garden,” building research & information., vol. 36, no. 4, pp. 373–380, 2008. [39] ministerio de minas y energía de colombia and unidad de planeación minero energética, “plan de expansión de referencia: generación, transmisión 2006-2020,” 2006. [40] w. vergara, a. deeb, n. toba, p. cramton and i. leino, “wind energy in colombia: a framework for market entry,” world bank study, pp. 1-120, 2010. [41] official cop21 website. “cop what’s it all about?,” [online], 2015. available at: http://www.cop21paris.org/about/cop21 (accessed 07.01.16). [42] f. ferreira, “olade – contribución al desarrollo de energías renovables,” seminario de energía más limpia, may, 2014. [43] commission for the regulation of energy and gas (creg). “análisis del precio de escasez-resolution no. 047,” pp.1-15., 2007. emanuel j. llanos p. was born in barranquilla, colombia, in 1988. he received the b.s degree in electrical and electronics engineering from universidad del norte, in 2011, and an m.s. in economics and policy of energy and the environment from university college of london (ucl), united kingdom, in 2016. from 2010 to 2012, emanuel was working in the electrical transmission and distribution, as an operations intern and a telemetry analyst, respectively. then, emanuel worked in the construction industry in the largest energy project built in colombia, the expansion of cartagena refinery, as a construction planner. after this, he worked in the largest thermo electrical company of colombia, termobaranquilla s.a, as a regulation and electrical/gas ana lyst. mr. llanos was a recipient of the international chevening scholarship award financed by the foreign common wealth in order to study his master degree. during the bachelor degree he received the silver medal, an academic excellence award for performing the best gpa among its classmates. laura m. hinestroza o. was born in barranquilla, colombia, in 1989. she received the b.s. degree in electrical engineering from universidad del norte, barranquilla, colombia, in 2011, and the m.s. degree in electrical engineering from universidad nacional de colombia, bogota d.c., in 2014. she is currently pursuing the ph.d. degree in engineering and public policy at technical university of lisbon-ist, portugal. from 2013 to 2015, she went to the academia as an auxiliary professor at universidad de la costa and universidad del norte, barranquilla, colombia. her research interest includes the study and application of microgrids to enable access to electricity in rural areas and energy planning for sustainable development. mrs. hinestroza’s awards and honors include the roberto rocca fellowship (tenaris-tubocaribe group), in 2011, and the national scholarship to pursue her ph.d., financed by the colombian government through the department of science, technology and innovation-colciencias, in 2015. http://www.brattle.com/system/publications/pdfs/000/005/045/original/resource_adequacy http://www.brattle.com/system/publications/pdfs/000/005/045/original/resource_adequacy http://www.puc.state.tx.us/wmo/documents/index.cfm http://m.eltiempo.com/economia/sectores/millonarias-deudas-a-banca-y-generadores-otro-mal-de-termocandelaria-/16437150/1/home http://m.eltiempo.com/economia/sectores/millonarias-deudas-a-banca-y-generadores-otro-mal-de-termocandelaria-/16437150/1/home https://www.climate.gov/news-features/blogs/enso/what-el-ni%c3%b1o%e2%80%93southern-oscillation-enso-nutshell https://www.climate.gov/news-features/blogs/enso/what-el-ni%c3%b1o%e2%80%93southern-oscillation-enso-nutshell http://www.elheraldo.co/economia/asoenergia-rechaza-posibles-aumentos-del-gobierno-al-precio-de-la-energia-252521 http://www.elheraldo.co/economia/asoenergia-rechaza-posibles-aumentos-del-gobierno-al-precio-de-la-energia-252521 http://www.portafolio.co/negocios/empresas/generadoras-termicas-amenazan-entregarle-plantas-gobierno-492899 http://www.portafolio.co/negocios/empresas/generadoras-termicas-amenazan-entregarle-plantas-gobierno-492899 http://www.upme.gov.co/costosenergia.asp http://www.cpc.ncep.noaa.gov/products/analysis_monitoring/enso_advisory/ensodisc.html http://www.cpc.ncep.noaa.gov/products/analysis_monitoring/enso_advisory/ensodisc.html https://www.minminas.gov.co/tips-de-ahorro-de-energia http://www.elcolombiano.com/negocios/apagar-paga-medidas-de-ahorro-de-energia-son-tardias-expertos-by3809300 http://www.elcolombiano.com/negocios/apagar-paga-medidas-de-ahorro-de-energia-son-tardias-expertos-by3809300 http://www.eltiempo.com/economia/sectores/resultados-de-la-campana-apagar-paga/16548415 http://www.eltiempo.com/economia/sectores/resultados-de-la-campana-apagar-paga/16548415 http://www.cop21paris.org/about/cop21 i. introduction a. allocation of firm energy obligations in the mem b. remuneration of the oef: definition of the scarcity price ii. scarcity pricing approaches in other electrical markets a. energy only markets b. energy markets with capacity payments iii. failures of the scarcity price definition in colombia a. main flaws in the definition of the scarcity price b. effects on the market’s agents: generators and consumers iv. actions taken by the government a. creg resolution no. 172 of 2015 b. creg resolution no. 178 of 2015 c. energy save programme “apagar paga” d. other short-term measures v. potential measures to overcome the crisis in the long-term a. the need for a better energy planning b. new definition of the scarcity price: case study vi. conclusions references