MEV Journal of Mechatronics, Electrical Power, and Vehicular Technology 13 (2022) 214-221 Journal of Mechatronics, Electrical Power, and Vehicular Technology e-ISSN: 2088-6985 p-ISSN: 2087-3379 mev.lipi.go.id doi: https://dx.doi.org/10.14203/j.mev.2022.v13.214-221 2088-6985 / 2087-3379 ©2022 National Research and Innovation Agency This is an open access article under the CC BY-NC-SA license (https://creativecommons.org/licenses/by-nc-sa/4.0/) MEV is Scopus indexed Journal and accredited as Sinta 1 Journal (https://sinta.kemdikbud.go.id/journals/detail?id=814) How to Cite: Sujito et al., “Long-term forecasting for growth of electricity load based on customer sectors,” Journal of Mechatronics, Electrical Power, and Vehicular Technology, vol. 13, no. 2, pp. 214-221, Dec. 2022. Long-term forecasting for growth of electricity load based on customer sectors Sujito a, *, Ridho Riski Hadi b, Langlang Gumilar a, Abdullah Iskandar Syah b, Moh. Zainul Falah b, Tran Huy Duy c a Intelligent Power and Advance Energy System, Jurusan Teknik Elektro, Universitas Negeri Malang Jl. Semarang No. 5 Sumbersari, Lowokwaru, Malang, East Java, 65145, Indonesia b Electrical Engineering, Electrical Engineering Department, Universitas Negeri Malang Jl. Semarang No. 5 Sumbersari, Lowokwaru, Malang, East Java, 65145, Indonesia c Electrical Engineering Department, Dalat University, Vietnam 1 Đường Phù Đổng Thiên Vương, Phường 8, Thành phố Đà Lạt, Lâm Đồng, Vietnam Received 25 May 2022; 1st revision 18 October 2022; 2nd revision 2 November 2022; 3rd revision 24 November 2022; 4th revision 14 December 2022; Accepted 16 December 2022; Published online 29 December 2022 Abstract The availability of electrical energy is an important issue. Along with the growth of the human population, electrical energy also increases. This study addresses problems in the operation of the electric power system. One of the problems that occur is the power imbalance due to scale growth between demand and generation. Alternative countermeasures that can be done are to prepare for the possibility that will occur in the future or what we are familiar with forecasting. Forecasting using the multiple linear regression method with this research variable assumes the household sector, business, industry, and public sectors, and is considered by the influence of population, gross regional domestic product, and District Minimum Wage. In forecasting, it is necessary to evaluate the accuracy using mean absolute percentage error (MAPE). MAPE evaluation results show a value of 0.142 % in the household sector, 0.085 % in the business sector, 1.983 % in the industrial sector, and 0.131 % in the total customer sector. ©2022 National Research and Innovation Agency. This is an open access article under the CC BY-NC-SA license (https://creativecommons.org/licenses/by-nc-sa/4.0/). Keywords: district minimum wage; gross regional domestic product; long-term forecasting; mean absolute percentage error; multiple linear regression; I. Introduction Electricity is one of many energies that are needed by society to support daily activities [1][2]. Along with the increase in the total electrical loads, the electrical energy required was increased [3][4][5]. Generally, an electric power system is divided into several customer sectors, including a housing sector, a commercial sector, an industrial sector, and a public or general sector [6][7]. One problem in the operation of the electric power system is a power imbalance between the power required and the power generated. The imbalance results in disruption of frequency stability and voltage drop in the system [8]. According to [9], which was explained in the Electricity Supply Business Plan, it was said that the one target of the State Electricity Company or Perusahaan Listrik Negara (“PLN”) is to be able to provide power capacity and electrical energy every year. To keep an electricity demand fulfilled, it is necessary to connect a supply of electrical energy according to demand and load forecasting that will take place in the future [1][10]. Forecasting is a process of predicting a possible structured way that can take place in the future based on data from the past and the current periods to minimize errors [11][12]. The main factors that influence the forecasting of electricity load growth are macroeconomic problems such as economic growth, population, gross regional domestic product (GRDP), etc. [13][14]. According to [15][16], predicting expenses in the future is usually done by analyzing a graph of expenses in the past to future * Corresponding Author. Tel: +62-851-32014085 E-mail address: sujito.ft@um.ac.id https://dx.doi.org/10.14203/j.mev.2022.v13.214-221 http://u.lipi.go.id/1436264155 http://u.lipi.go.id/1434164106 https://mev.lipi.go.id/mev https://mev.lipi.go.id/mev https://dx.doi.org/10.14203/j.mev.2022.v13.214-221 https://creativecommons.org/licenses/by-nc-sa/4.0/ https://sinta.kemdikbud.go.id/journals/detail?id=814 https://crossmark.crossref.org/dialog/?doi=10.14203/j.mev.2022.v13.214-221&domain=pdf https://creativecommons.org/licenses/by-nc-sa/4.0/ Sujito et al. / Journal of Mechatronics, Electrical Power, and Vehicular Technology 13 (2022) 214-221 215 periods, including short-term, medium-term, or long-term. Short-term, mid-term, and long-term models are important to account for the complexity of load data and produce reliable forecasts [15]. Forecasting methods are divided into three types, namely, monitoring, causal, and time series methods [17]. The three types are described in the form of Table 1. Based on Table 1, this study uses a causal method with two techniques in determining forecasting results [18], namely simple regression and multiple regression, where both are required in forecasting for each explanatory variable (population, GRDP, and district minimum wage). This research activity was done at West Kotawaringin Regency because West Kotawaringin Regency is the shaded area of PT PLN (Persero) ULP Pangkalan Bun. According to [19], in 2020, West Kotawaringin Regency was the district with the 4th largest population of 14 regencies or cities in Central Kalimantan Province, with an average population growth rate in the last three years of 2.71 % annually. In addition, according to historical data from PT PLN (Persero) ULP Pangkalan Bun 2021, the growth rate in the number of customers has increased by 0.52 % every month starting from January 2018 to December 2020. Therefore, the availability of electrical energy must be continuously monitored and maintained. Therefore, PT PLN (Persero) ULP Pangkalan Bun needs to maintain a continuity of distribution of electrical energy. One of them by forecasting the growth of the electrical energy load in the future. In this research, long-term forecasting was done in each sector, such as the industrial, household, business, and general sectors. This is to support the maintenance of the availability of electrical energy and planning for the addition of electrical energy in the future by PT PLN [20]. In addition, the data used is a type of linear interpolation because the data always increases and is stable every year, so this research can only be done at the end of each year. The factors that influence a forecasting growth of an electrical load used are the total population, GRDP, and the minimum wage in the West Kotawaringin area. This research uses a linear regression forecasting method. Simple linear regression was used to predict the total population, GRDP, and minimum wage in the West Kotawaringin Regency area. Forecasting results from these variables with multiple linear regression were used to predict an amount of electricity load growth in industrial, household, business, and general sectors. Calculation of the magnitude of the impact of each independent variable on the dependent variable was done using Pearson's product moment analysis (r) and coefficient of determination (r2). Then, the accuracy of measured forecasting is determined by mean absolute percent error (MAPE). II. Materials and Methods A. Correlation theory Correlation is one of the existing methods in statistical analysis as a search for continuity between two variables with quantitative properties, i.e., to see the size of the impact given by an independent variable (independent) to a dependent variable (dependent). Two variables are said to be correlated if a change in one variable will be accompanied by a change in another variable in an organized manner in one direction (positive correlation) or contradictory (negative correlation) [21][22][23]. This research was done using a Pearson product- moment correlation technique and coefficient of determination. 1) Pearson product-moment According to [24][25], in principle, Pearson product-moment correlation is used to determine a correlation between two variables (bivariate model) which has an interval or ratio scale. Pearson product-moment correlation can be interpreted to find whether there is a relationship between the x variable and the y variable. It is useful as an explanation of how much one variable contributes to another variable. The positive (+) and negative (-) signs represent a type of relationship between variables, and the value ranges from -1.0 to 1.0 as a strengthening statement of the relationship. Mathematically, the Pearson product-moment is described in equation (1): 𝑟 = 𝑛 ∑𝑋𝑋 − ∑𝑋 ∑𝑋 �{𝑛∑𝑋2 − (∑𝑋)2} {𝑛∑𝑋2 − (∑𝑋)2} (1) where r is the Pearson product-moment, n is the customers point (X, Y), X is the independent variable, and Y is the dependent variable. The results of correlation calculation using a Pearson product-moment method can represent how strong the relationship between variables was given. Table 2 describes a representation strength of continuity between variables calculated using a Pearson product-moment method [26]. The correlation rate is used as a result of decisions from variables that influence each other. Table 1. Types of forecasting Forecasting methods Description Monitoring Tracking signals Causal Linear regression, multiple regression, ARIMA Time series Simple time series, advanced time series Table 2. Correlation representation using Pearson product moment method Value of r Correlation rate 0.00 – 0.199 Very Low 0.20 – 0.399 Low 0.40 – 0.599 Medium 0.60 – 0.799 Strong 0.80 – 1.000 Very strong Sujito et al. / Journal of Mechatronics, Electrical Power, and Vehicular Technology 13 (2022) 214-221 216 2) Coefficient of determination The coefficient of determination (equation (2)) is symbolized by r2 and is usually represented using a percentage (%). The coefficient of determination is a value used to measure the contribution of the independent variable (X) to the variation (increase/decrease) of the dependent variable (Y). Another explanation is that variable Y can be described by variable X with a magnitude of r2 then the rest was described by other variables. Another variation of y (the rest) was caused by many different factors that influence Y as well and has been included in disturbance error [27][28]. The range coefficient of determination is 0 to 1, with 0 representing no continuity between the independent variables to the dependent variable and 1 representing a perfect relationship between these variables [28][29]. 𝑟2 = ((𝑛)( ∑𝑋𝑋)−(∑𝑋)(∑𝑋))2 ((𝑛(∑𝑋2)−(∑𝑋)2)(𝑛(∑𝑋2)−(∑𝑋)2)) (2) where r2 is the coefficient of determination. B. Linear regression Regression is one of many statistical methods that serve to see the pattern of the relationship between the response variable and predictor variable [30]. 1) Simple linear regression According to [31][32], a simple linear regression is a linear regression that only considers one independent variable (X) and one dependent variable (Y). In linear regression, the variable (Y) can be expressed as a response variable, or in other terms, an output variable and not independent. While the variable (X) can be said to be a predictor variable (used to estimate Y value), it can be expressed as an explanatory variable, input regressor, and independent. The simple linear regression equation model can be explained through equation (3) Y = 𝑎 + bX (3) where a is the constant and b is the coefficient. The value of 𝑎 and b needs to be calculated using equation (4) and equation (5): 𝑎 = (∑𝑋)(∑𝑋2)−(∑𝑋)(∑𝑋𝑋) (𝑛)(∑𝑋2)−(∑𝑋)2 (4) 𝑏 = 𝑛 ∑𝑋𝑋−(∑𝑋)(∑𝑋) 𝑛(∑𝑋2)−(∑𝑋)2 (5) 2) Multiple linear regression According to [33][34], multiple linear regression is a continuation of simple linear regression analysis. Multiple linear regression uses one dependent variable (Y) as the predicted target and several independent variables (X) as the variable used to predict a target. The coefficient on linear regression is an estimated value of the parameter in the regression model for the real condition (true condition). The coefficient for the linear regression model is an average value that has a chance of appearing on the Y variable if a value of X1, X2, and Xn was given. Mathematically, multiple linear regression analysis is explained in equation (6). Y = a + b1X1 + b2X2 + ... ... ... bkXk (6) where bk is the coefficient Xk, X 1 is the 1 st independent variable, b1 is the coefficient X1, X 2 is the 2nd independent variable, b2 is the coefficient X2, and X 3 is the 3 rd independent variable. 3) Forecasting accuracy Forecasting accuracy is a measure of forecast error based on the magnitude of the difference between forecast results and actual demand. The measurement of forecasting accuracy aims to determine the performance or accuracy of forecasting results that have been done using certain methods and techniques. In this research, the mean absolute percentage error (MAPE) method was used to assess forecasting accuracy. According to [35][36], MAPE can be calculated by the following mathematical equation (7) 𝑀𝑀𝑀𝑀 = ∑ |𝐴𝐴−𝐹𝐴|𝑛𝐴=1 𝐴𝐴 𝑛 × 100 (7) where t is the period on data, Ft is the forecasting data in period t, At is the actual data in period t, and n is the total forecasting data. MAPE results can be grouped based on the level of forecasting accuracy [37] as shown in Table 3. Parameter results are evaluated using MAPE and compared with the range value from MAPE to determine whether the accuracy level is high or low. 4) Systematic research Figure 1 shows the flow of research from Table 3. MAPE parameters in forecasting Value of MAPE Prediction accuracy MAPE ≤ 10% High 10% < MAPE ≤ 20% Good 20% < MAPE ≤ 50% Reasonable MAPE > 50% Low Figure 1. Research completion flow Sujito et al. / Journal of Mechatronics, Electrical Power, and Vehicular Technology 13 (2022) 214-221 217 beginning to end. It shows research data that has been collected and analyzed with two different output lines. In the first analysis, an output produced was to identify the correlation between independent variables and the dependent variable using a Pearson product-moment method and coefficient of determination. While the other analysis, a results output, is the forecasting of electrical energy needs and continued by measuring the accuracy of the forecast. III. Results and Discussions A. Effect of correlation of forecasting variables In this study, the data used consisted of two types of variables, namely the independent variable and the dependent variable. Independent variables are macroeconomic factors that can affect the dependent variable, such as population, GRDP, and city minimum wage. The dependent variable is a variable that was influenced by macroeconomic factors; here is electricity data which is divided into four sectors, namely the household, business, industrial, and public sectors. Based on the calculations performed using equation (1) and equation (2), the results of the correlation between the independent and the dependent variables are obtained. The correlation results are shown by the value of the Pearson product moment (r) and the coefficient of determination (r2) listed in Table 4. Thus, we can see how much influence the independent variable has on the dependent variable based on the r and r2 values obtained. The results of the correlation calculation obtained for each independent variable and the dependent variable were corrected with the r value in Table 2. These results indicate that the correlation value obtained is very strong and positive. This signifies that there is a very close relationship between the independent variable and the dependent variable. This means that if there is a change in the independent variable, there will also be a significant change in the dependent variable. B. Forecasting based on variables that affect load demand Based on calculations made using equations (3) to (6), the forecasting results for the growth of electricity loads for each sector are obtained. These sectors are the household, business, industrial, and public sectors. The calculation takes into account macro factors that can affect the growth of electricity expenses, such as population, GRDP, and city minimum wage. The forecasting results are shown in Table 5. Thus, we can estimate the electricity load growth for each sector based on influencing macro factors. In general, each year, the energy growth in each sector has increased (see Figure 2). This is because every year, it is influenced by population growth, as Table 4. Correlation between independent variable and dependent variable Variable Pearson product moment (r) Coefficient of determination (r2) X1 to Y1 0.998 0.995 X2 to Y1 0.984 0.969 X3 to Y1 0.996 0.991 X1, X2, X3 to Y1 0.993 0.999 X1 to Y2 0.978 0.957 X2 to Y2 0.972 0.944 X3 to Y2 0.975 0.950 X1, X2, X3 to Y2 0.975 0.994 X1 to Y3 0.976 0.953 X2 to Y3 0.928 0.861 X3 to Y3 0.978 0.957 X1, X2, X3 to Y3 0.960 0.961 X1 to Y4 0.999 0.997 X2 to Y4 0.974 0.948 X3 to Y4 0.998 0.996 X1, X2, X3 to Y4 0.990 0.997 X1 to Ytot 0.998 0.996 X2 to Ytot 0.984 0.968 X3 to Ytot 0.996 0.992 X1, X2, X3 to Ytot 0.992 0.999 Figure 2. Load demand growth 0 20000 40000 60000 80000 100000 120000 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 Lo ad Time for three years (month) Y1 Household Y2 Business Y3 Indusrty Y4 General Ytot Customers Sujito et al. / Journal of Mechatronics, Electrical Power, and Vehicular Technology 13 (2022) 214-221 218 it is known through Table 4 that the correlation between the two factors is strong. Based on Table 5, it can be seen that the number of electricity customers in each sector has always increased from January 2021 to December 2023. This increase is due to population growth and other sectors which have a high correlation. To measure the accuracy of forecasting the growth of this electrical load, equation (7), which is called the mean absolute percentage error (MAPE), can be used. The results of the calculation of forecasting accuracy according to this method can be seen in Table 6. From these results, it can be seen that forecasting the growth of electricity loads in each sector has fairly high accuracy. In this case, the results of calculating the accuracy of the load forecasting using the linear regression method show that the mean absolute percentage error (MAPE) is quite small. This shows that load forecasting using the linear regression method is quite accurate in contrast to the results of the previous study [14], which showed that the linear regression method for forecasting urban demand loads has a larger MAPE value compared to the K-nearest neighbour (KNN) method they use, which in their research method KNN is more appropriate to the cases they use compared to the linear regression method. In the same case study, the auto-regressive (AR) method was applied to the same dataset to forecast each sector in January and December to ensure the training results. The results of the auto-regressive (AR) test are listed in Table 7. In Table 6, the MAPE value for forecasting the independent variables is 0.005 % for the population and 0.764 % for GRDP. Meanwhile, the MAPE value for forecasting the dependent variable is shown in Table 6, which is 0.142 % for household sector electricity costs, 0.085 % for the business sector, 1.983 % for the industrial sector, and 0.131 % for Table 5. The result of forecasting the growth of electricity load for each sector Year Month Customer Y1 Y2 Y3 Y4 Ytot Household Business Industry General Total customers 2021 January 75377 7770 32 2926 86104 February 75755 7780 32 2943 86510 March 76134 7790 32 2960 86916 April 76513 7800 32 2977 87322 May 76892 7810 33 2993 87729 June 77271 7820 33 3010 88135 July 77650 7830 33 3027 88541 August 78029 7840 34 3044 88947 September 78408 7850 34 3061 89353 October 78787 7861 34 3077 89759 November 79166 7871 34 3094 90165 December 79545 7881 35 3111 90571 2022 January 79924 7891 35 3128 90978 February 80303 7901 35 3145 91384 March 80682 7911 36 3162 91790 April 81061 7921 36 3178 92196 May 81440 7931 36 3195 92602 June 81818 7941 36 3212 93008 July 82197 7951 37 3229 93414 August 82576 7961 37 3246 93820 September 82955 7972 37 3262 94227 October 83334 7982 38 3279 94633 November 83713 7992 38 3296 95039 December 84092 8002 38 3313 95445 2023 January 84471 8012 38 3330 95851 February 84850 8022 39 3347 96257 March 85229 8032 39 3363 96663 April 85608 8042 39 3380 97069 May 85987 8052 40 3397 97476 June 86366 8062 40 3414 97882 July 86745 8072 40 3431 98288 August 87124 8083 40 3447 98694 September 87503 8093 41 3464 99100 October 87881 8103 41 3481 99506 November 88260 8113 41 3498 99912 December 88639 8123 42 3515 100318 Sujito et al. / Journal of Mechatronics, Electrical Power, and Vehicular Technology 13 (2022) 214-221 219 customers as a whole. In addition, in other tests using different methods, as forecasting comparisons are shown in Table 7, the forecasting results prove the same as the calculation of equations (3) to (6). This proves that the cases used are in accordance with the linear regression method, where the data used is time series. Based on these results, forecasting the growth of electricity loads in all sectors can be used with a high accuracy value because the MAPE value is below 10 %, according to Table 3. With the results of forecasting in 2023, it can be used as a comparison in increasing power capacity that is able to support the energy demand needs of consumers from various sectors. IV. Conclusion The population, GRDP, and minimum wage have a very strong (positive) influence on the growth of the electricity load in each customer sector. The results of forecasting the growth of the electrical load showed a steady increase. The results of the measurement of the accuracy of the electricity load Table 6. Forecasting accuracy results with MAPE Year Month Error (%) Y1 Y2 Y3 Y4 Ytot Household Business Industry General Total customers 2021 January 0.169 0.084 3.884 0.549 0.171 February 0.134 0.051 4.970 0.118 0.123 March 0.411 0.005 3.166 0.198 0.347 April 0.234 0.013 2.175 0.271 0.209 May 0.045 0.006 1.184 0.357 0.051 June 0.187 0.052 0.192 0.358 0.145 July 0.268 0.085 3.401 0.276 0.216 August 0.326 0.050 2.451 0.030 0.278 September 0.013 0.096 1.501 0.115 0.024 October 0.343 0.088 0.550 0.118 0.310 November 0.381 0.134 4.765 0.081 0.343 December 0.169 0.008 2.704 0.316 0.134 2022 January 0.019 0.170 1.647 0.367 0.021 February 0.195 0.120 0.590 0.379 0.171 March 0.277 0.102 0.466 0.195 0.257 April 0.202 0.269 1.523 0.091 0.204 May 0.285 0.317 2.580 0.335 0.288 June 0.037 0.102 3.637 0.151 0.029 July 0.030 0.150 0.667 0.482 0.028 August 0.224 0.120 1.684 0.848 0.212 September 0.138 0.037 2.700 0.676 0.140 October 0.037 0.007 3.716 0.206 0.039 November 0.117 0.063 2.749 0.229 0.114 December 0.076 0.080 1.805 0.228 0.065 2023 January 0.050 0.069 0.490 0.153 0.045 February 0.115 0.072 0.826 0.478 0.109 March 0.126 0.049 1.381 0.618 0.127 April 0.104 0.039 0.110 0.436 0.102 May 0.003 0.081 2.212 0.219 0.003 June 0.040 0.058 0.985 0.040 0.032 July 0.076 0.074 2.990 0.138 0.056 August 0.014 0.077 1.802 0.031 0.019 September 0.103 0.066 0.614 0.313 0.106 October 0.103 0.034 0.574 0.487 0.104 November 0.027 0.096 1.762 0.002 0.015 December 0.053 0.132 2.951 0.242 0.065 MAPE 0.142 0.085 1.983 0.281 0.131 Table 7. Forecasting with auto-regressive (AR) method Month (2023) Customer per sector Total customer House Business Industry General January 84470896 8012142 38183383 3329843 95851117 December 8863886 8123124 41202297 3514834 100318076 Sujito et al. / Journal of Mechatronics, Electrical Power, and Vehicular Technology 13 (2022) 214-221 220 growth forecast conducted by MAPE in the household sector are 0.142 %, business is 0.085 %, the industry is 1.983 %, and the number of customers is 0.131 %. MAPE value < 10 %, so the accuracy of forecasting the growth of electricity load in all sectors is high. In further research, the use of other training methods can be used as a comparison for the results of forecasting the electrical loads in Kalimantan. Acknowledgments Thanks are conveyed to the Institute for Research and Community Service (LP2M) State University of Malang for funding this research activity through Basic Research Grants – Thesis Research in 2022. Declarations Author contribution Sujito: Formal Analysis, Validation, Data Curation. R.R. Hadi, L. Gumilar, T.H. Duy: Conceptualization, Formal Analysis, Resources. A.I. Syah, M.Z. Falah: Writing, Formal Analysis, Software Operations, Visualization, Funding Acquisition. Funding statement Thanks are conveyed to the Institute for Research and Community Service (LP2M) State University of Malang for funding this research activity through Basic Research Grants – Thesis Research in 2022 No. 19.5.1259/UN32.20.1/LT/2022. Competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Additional information Reprints and permission: information is available at https://mev.lipi.go.id/. Publisher’s Note: National Research and Innovation Agency (BRIN) remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References [1] M. Subekti, I. A. Rahardjo, and D. Rosyanti, “Forecasting electrical energy demand of PT. PLN (Persero) UP3 Sukabumi using analytical, econometrics, and trends methods,” IOP Conference Series: Materials Science and Engineering, vol. 1098, no. 4, p. 042032, 2021. [2] K. B. Lindberg, P. Seljom, H. Madsen, D. Fischer, and M. Korpås, “Long-term electricity load forecasting: Current and future trends,” Utilities Policy, vol. 58, pp. 102–119, 2019. [3] A. Setiawan, “Losses sharing regulation as a solution for spike losses at PLN’s Distribution system caused by energy over production from independent power producers,” in 2020 International Conference on Technology and Policy in Energy and Electric Power (ICT-PEP), Bandung, Indonesia, Sep., pp. 219–224, 2020. [4] G. Zhang and J. Guo, “A novel method for hourly electricity demand forecasting,” IEEE Transactions on Power Systems, vol. 35, no. 2, pp. 1351–1363, 2019. [5] S. Aslam, H. Herodotou, S. M. Mohsin, N. Javaid, N. Ashraf, and S. Aslam, “A survey on deep learning methods for power load and renewable energy forecasting in smart microgrids,” Renewable and Sustainable Energy Reviews, vol. 144, p. 110992, 2021. [6] Z. A. Khan et al., “Efficient short-term electricity load forecasting for effective energy management,” Sustainable Energy Technologies and Assessments, vol. 53, p. 102337,, 2022. [7] L. Peng, L. Wang, D. Xia, and Q. Gao, “Effective energy consumption forecasting using empirical wavelet transform and long short-term memory,” energy, vol. 238, p. 121756,, 2022. [8] H. Zhou, Y. Zhang, L. Yang, Q. Liu, K. Yan, and Y. Du, “Short- term photovoltaic power forecasting based on long short term memory neural network and attention mechanism,” IEEE Access, vol. 7, pp. 78063–78074, 2019. [9] J. Fahmi, J. Windarta, and A. Y. Wardaya, “Studi awal penerapan distributed generation untuk optimalisasi PLTS atap on grid pada pelanggan PLN sistem jawa bali untuk memenuhi target EBT nasional,” Jurnal Energi Baru dan Terbarukan, vol. 2, no. 1, pp. 1–13, 2021. [10] L. Wen, K. Zhou, and S. Yang, “Load demand forecasting of residential buildings using a deep learning model,” Electric Power Systems Research, vol. 179, p. 106073, 2020. [11] Y. Yu, J. Cao, and J. Zhu, “An LSTM short-term solar irradiance forecasting under complicated weather conditions,” IEEE Access, vol. 7, pp. 145651–145666, 2019. [12] M. Pavlicko, M. Vojteková, and O. Blažeková, “Forecasting of electrical energy consumption in Slovakia,” Mathematics, vol. 10, no. 4, p. 577, 2022. [13] B. Nepal, M. Yamaha, A. Yokoe, and T. Yamaji, “Electricity load forecasting using clustering and ARIMA model for energy management in buildings,” Japan Architectural Review, vol. 3, no. 1, pp. 62–76, 2020. [14] N. J. Johannesen, M. Kolhe, and M. Goodwin, “Relative evaluation of regression tools for urban area electrical energy demand forecasting,” Journal of cleaner production, vol. 218, pp. 555–564, 2019. [15] H. Riahi-Madvar, M. Dehghani, R. Memarzadeh, and B. Gharabaghi, “Short to long-term forecasting of river flows by heuristic optimization algorithms hybridized with ANFIS,” Water Resources Management, vol. 35, no. 4, pp. 1149–1166, 2021. [16] A. Sherstinsky, “Fundamentals of recurrent neural network (RNN) and long short-term memory (LSTM) network,” Physica D: Nonlinear Phenomena, vol. 404, p. 132306, 2020. [17] P. Malhan and M. Mittal, “A novel ensemble model for long- term forecasting of wind and hydro power generation,” Energy Conversion and Management, vol. 251, p. 114983, 2022. [18] N. Talkhi, N. A. Fatemi, Z. Ataei, and M. J. Nooghabi, “Modeling and forecasting number of confirmed and death caused COVID-19 in IRAN: A comparison of time series forecasting methods,” Biomedical Signal Processing and Control, vol. 66, p. 102494, 2021. [19] R. Sihite, “Analisis pengaruh pendapatan perkapita, jumlah konsumsi dan pertumbuhan ekonomi di Kabupaten/Kota Provinsi Kalimantan Tengah,” JEPP: Jurnal Ekonomi Pembangunan Dan Pariwisata, vol. 2, no. 1, pp. 46–57, 2022 [20] P. Mangera, “Perkiraan kebutuhan energi listrik jangka panjang pada PT. PLN (Persero) wilayah Papua dan Papua Barat area Merauke dengan menggunakan metode regresi linier,” Mustek Anim Ha, vol. 7, no. 3, 2018. [21] J. Deng, Y. Deng, and K. H. Cheong, “Combining conflicting evidence based on Pearson correlation coefficient and weighted graph,” International Journal of Intelligent Systems, vol. 36, no. 12, pp. 7443–7460, 2021. [22] H. Zhu, X. You, and S. Liu, “Multiple ant colony optimization based on pearson correlation coefficient,” IEEE Access, vol. 7, pp. 61628–61638, 2019. [23] P. Schober, C. Boer, and L. A. Schwarte, “Correlation coefficients: appropriate use and interpretation,” Anesthesia & Analgesia, vol. 126, no. 5, pp. 1763–1768, 2018. [24] F. Zinzendoff Okwonu, B. Laro Asaju, and F. Irimisose Arunaye, “Breakdown analysis of Pearson correlation coefficient and robust correlation methods,” IOP Conference Series: Materials Science and Engineering, vol. 917, no. 1, 2020. [25] A. Ali et al., “A k-Nearest neighbours based ensemble via optimal model selection for regression,” IEEE Access, vol. 8, pp. 132095–132105, 2020. [26] M. Ausloos, A. Eskandary, P. Kaur, and G. Dhesi, “Evidence for Gross Domestic Product growth time delay dependence over Foreign Direct Investment. A time-lag dependent correlation study,” Physica A: Statistical Mechanics and Its Applications, vol. 527, p. 121181, 2019. https://mev.lipi.go.id/ https://doi.org/10.1088/1757-899x/1098/4/042032 https://doi.org/10.1088/1757-899x/1098/4/042032 https://doi.org/10.1088/1757-899x/1098/4/042032 https://doi.org/10.1088/1757-899x/1098/4/042032 https://doi.org/10.1088/1757-899x/1098/4/042032 https://doi.org/10.1016/j.jup.2019.04.001 https://doi.org/10.1016/j.jup.2019.04.001 https://doi.org/10.1016/j.jup.2019.04.001 https://doi.org/10.1109/ICT-PEP50916.2020.9249826 https://doi.org/10.1109/ICT-PEP50916.2020.9249826 https://doi.org/10.1109/ICT-PEP50916.2020.9249826 https://doi.org/10.1109/ICT-PEP50916.2020.9249826 https://doi.org/10.1109/ICT-PEP50916.2020.9249826 https://doi.org/10.1109/ICT-PEP50916.2020.9249826 https://doi.org/10.1109/TPWRS.2019.2941277 https://doi.org/10.1109/TPWRS.2019.2941277 https://doi.org/10.1109/TPWRS.2019.2941277 https://doi.org/10.1016/j.rser.2021.110992 https://doi.org/10.1016/j.rser.2021.110992 https://doi.org/10.1016/j.rser.2021.110992 https://doi.org/10.1016/j.rser.2021.110992 https://doi.org/10.1016/j.rser.2021.110992 https://doi.org/10.1016/j.seta.2022.102337 https://doi.org/10.1016/j.seta.2022.102337 https://doi.org/10.1016/j.seta.2022.102337 https://doi.org/10.1016/j.seta.2022.102337 https://doi.org/10.1016/j.energy.2021.121756 https://doi.org/10.1016/j.energy.2021.121756 https://doi.org/10.1016/j.energy.2021.121756 https://doi.org/10.1016/j.energy.2021.121756 https://doi.org/10.1109/ACCESS.2019.2923006 https://doi.org/10.1109/ACCESS.2019.2923006 https://doi.org/10.1109/ACCESS.2019.2923006 https://doi.org/10.1109/ACCESS.2019.2923006 https://doi.org/10.14710/jebt.2021.10038 https://doi.org/10.14710/jebt.2021.10038 https://doi.org/10.14710/jebt.2021.10038 https://doi.org/10.14710/jebt.2021.10038 https://doi.org/10.14710/jebt.2021.10038 https://doi.org/10.1016/j.epsr.2019.106073 https://doi.org/10.1016/j.epsr.2019.106073 https://doi.org/10.1016/j.epsr.2019.106073 https://doi.org/10.1109/ACCESS.2019.2946057 https://doi.org/10.1109/ACCESS.2019.2946057 https://doi.org/10.1109/ACCESS.2019.2946057 https://doi.org/10.3390/math10040577 https://doi.org/10.3390/math10040577 https://doi.org/10.3390/math10040577 https://doi.org/10.1002/2475-8876.12135 https://doi.org/10.1002/2475-8876.12135 https://doi.org/10.1002/2475-8876.12135 https://doi.org/10.1002/2475-8876.12135 https://doi.org/10.1016/j.jclepro.2019.01.108 https://doi.org/10.1016/j.jclepro.2019.01.108 https://doi.org/10.1016/j.jclepro.2019.01.108 https://doi.org/10.1016/j.jclepro.2019.01.108 https://doi.org/10.1007/s11269-020-02756-5 https://doi.org/10.1007/s11269-020-02756-5 https://doi.org/10.1007/s11269-020-02756-5 https://doi.org/10.1007/s11269-020-02756-5 https://doi.org/10.1007/s11269-020-02756-5 https://doi.org/10.1016/j.physd.2019.132306 https://doi.org/10.1016/j.physd.2019.132306 https://doi.org/10.1016/j.physd.2019.132306 https://doi.org/10.1016/j.enconman.2021.114983 https://doi.org/10.1016/j.enconman.2021.114983 https://doi.org/10.1016/j.enconman.2021.114983 https://doi.org/10.1016/j.enconman.2021.114983 https://doi.org/10.1016/j.bspc.2021.102494 https://doi.org/10.1016/j.bspc.2021.102494 https://doi.org/10.1016/j.bspc.2021.102494 https://doi.org/10.1016/j.bspc.2021.102494 https://doi.org/10.1016/j.bspc.2021.102494 https://doi.org/10.52300/jepp.v2i1.4433 https://doi.org/10.52300/jepp.v2i1.4433 https://doi.org/10.52300/jepp.v2i1.4433 https://doi.org/10.52300/jepp.v2i1.4433 https://doi.org/10.35724/mustek.v7i3.1736 https://doi.org/10.35724/mustek.v7i3.1736 https://doi.org/10.35724/mustek.v7i3.1736 https://doi.org/10.35724/mustek.v7i3.1736 https://doi.org/10.1002/int.22593 https://doi.org/10.1002/int.22593 https://doi.org/10.1002/int.22593 https://doi.org/10.1002/int.22593 https://doi.org/10.1109/ACCESS.2019.2915673 https://doi.org/10.1109/ACCESS.2019.2915673 https://doi.org/10.1109/ACCESS.2019.2915673 https://doi.org/10.1213/ANE.0000000000002864 https://doi.org/10.1213/ANE.0000000000002864 https://doi.org/10.1213/ANE.0000000000002864 https://doi.org/10.1088/1757-899X/917/1/012065 https://doi.org/10.1088/1757-899X/917/1/012065 https://doi.org/10.1088/1757-899X/917/1/012065 https://doi.org/10.1088/1757-899X/917/1/012065 https://doi.org/10.1109/ACCESS.2020.3010099 https://doi.org/10.1109/ACCESS.2020.3010099 https://doi.org/10.1109/ACCESS.2020.3010099 https://doi.org/10.1016/j.physa.2019.121181 https://doi.org/10.1016/j.physa.2019.121181 https://doi.org/10.1016/j.physa.2019.121181 https://doi.org/10.1016/j.physa.2019.121181 https://doi.org/10.1016/j.physa.2019.121181 Sujito et al. / Journal of Mechatronics, Electrical Power, and Vehicular Technology 13 (2022) 214-221 221 [27] D. Chicco, M. J. Warrens, and G. Jurman, “The coefficient of determination R-squared is more informative than SMAPE, MAE, MAPE, MSE and RMSE in regression analysis evaluation,” PeerJ Computer Science, vol. 7, pp. 1–24, 2021. [28] S. Nakagawa, P. C. D. Johnson, and H. Schielzeth, “The coefficient of determination R2 and intra-class correlation coefficient from generalized linear mixed-effects models revisited and expanded,” Journal of the Royal Society Interface, vol. 14, no. 134, 2017. [29] A. M. da Silva Filho, G. F. Zebende, A. P. N. de Castro, and E. F. Guedes, “Statistical test for multiple detrended cross- correlation coefficient,” Physica A: Statistical Mechanics and its Applications, vol. 562, p. 125285, 2021. [30] T. M. Hope, “Linear regression,” in Machine Learning, Elsevier, pp. 67–81, 2020. [31] M. Sharifzadeh, A. Sikinioti-Lock, and N. Shah, “Machine- learning methods for integrated renewable power generation: A comparative study of artificial neural networks, support vector regression, and Gaussian Process Regression,” Renewable and Sustainable Energy Reviews, vol. 108, pp. 513– 538, 2019. [32] H.-Y. Kim, “Statistical notes for clinical researchers: simple linear regression 3 – residual analysis,” Restorative Dentistry & Endodontics, vol. 44, no. 1, pp. 1–8, 2019. [33] T. A. Trunfio, A. Scala, A. D. Vecchia, A. Marra, and A. Borrelli, “Multiple regression model to predict length of hospital stay for patients undergoing femur fracture surgery at ‘San Giovanni Di Dio e Ruggi d’Aragona’ University Hospital,” in European Medical and Biological Engineering Conference, pp. 840–847, 2020. [34] M. Korkmaz, “A study over the general formula of regression sum of squares in multiple linear regression,” Numerical Methods for Partial Differential Equations, vol. 37, no. 1, pp. 406–421, 2021. [35] N. Kusuma, M. Roestam, and L. Pasca, “The analysis of forecasting demand method of linear exponential smoothing,” International Journal of Educational Administration, Management, and Leadership, pp. 7–18, 2020. [36] S. Prayudani, A. Hizriadi, Y. Y. Lase, and Y. Fatmi, “Analysis accuracy of forecasting measurement technique on random K- nearest neighbor (RKNN) using MAPE and MSE,” in Journal of Physics: Conference Series, vol. 1361, no. 1, p. 012089, 2019. [37] E. Vivas, H. Allende-Cid, and R. Salas, “A systematic review of statistical and machine learning methods for electrical power forecasting with reported mape score,” Entropy, vol. 22, no. 12, p. 1412, 2020. https://doi.org/10.7717/PEERJ-CS.623 https://doi.org/10.7717/PEERJ-CS.623 https://doi.org/10.7717/PEERJ-CS.623 https://doi.org/10.7717/PEERJ-CS.623 https://doi.org/10.1098/rsif.2017.0213 https://doi.org/10.1098/rsif.2017.0213 https://doi.org/10.1098/rsif.2017.0213 https://doi.org/10.1098/rsif.2017.0213 https://doi.org/10.1098/rsif.2017.0213 https://doi.org/10.1016/j.physa.2020.125285 https://doi.org/10.1016/j.physa.2020.125285 https://doi.org/10.1016/j.physa.2020.125285 https://doi.org/10.1016/j.physa.2020.125285 https://doi.org/10.1016/B978-0-12-815739-8.00004-3 https://doi.org/10.1016/B978-0-12-815739-8.00004-3 https://doi.org/10.1016/j.rser.2019.03.040 https://doi.org/10.1016/j.rser.2019.03.040 https://doi.org/10.1016/j.rser.2019.03.040 https://doi.org/10.1016/j.rser.2019.03.040 https://doi.org/10.1016/j.rser.2019.03.040 https://doi.org/10.1016/j.rser.2019.03.040 https://doi.org/10.5395/rde.2019.44.e11 https://doi.org/10.5395/rde.2019.44.e11 https://doi.org/10.5395/rde.2019.44.e11 https://doi.org/10.1007/978-3-030-64610-3_94 https://doi.org/10.1007/978-3-030-64610-3_94 https://doi.org/10.1007/978-3-030-64610-3_94 https://doi.org/10.1007/978-3-030-64610-3_94 https://doi.org/10.1007/978-3-030-64610-3_94 https://doi.org/10.1007/978-3-030-64610-3_94 https://doi.org/10.1002/num.22533 https://doi.org/10.1002/num.22533 https://doi.org/10.1002/num.22533 https://doi.org/10.1002/num.22533 https://doi.org/10.51629/ijeamal.v1i1.3 https://doi.org/10.51629/ijeamal.v1i1.3 https://doi.org/10.51629/ijeamal.v1i1.3 https://doi.org/10.51629/ijeamal.v1i1.3 https://doi.org/10.1088/1742-6596/1361/1/012089 https://doi.org/10.1088/1742-6596/1361/1/012089 https://doi.org/10.1088/1742-6596/1361/1/012089 https://doi.org/10.1088/1742-6596/1361/1/012089 https://doi.org/10.3390/e22121412 https://doi.org/10.3390/e22121412 https://doi.org/10.3390/e22121412 https://doi.org/10.3390/e22121412 Introduction II. Materials and Methods A. Correlation theory 1) Pearson product-moment 2) Coefficient of determination B. Linear regression 1) Simple linear regression 2) Multiple linear regression 3) Forecasting accuracy 4) Systematic research III. Results and Discussions A. Effect of correlation of forecasting variables B. Forecasting based on variables that affect load demand IV. Conclusion Acknowledgments Declarations Author contribution Funding statement Competing interest Additional information References