Format And Type Fonts CHEMICAL ENGINEERING TRANSACTIONS VOL. 39, 2014 A publication of The Italian Association of Chemical Engineering www.aidic.it/cet Guest Editors: Petar Sabev Varbanov, Jiří Jaromír Klemeš, Peng Yen Liew, Jun Yow Yong Copyright © 2014, AIDIC Servizi S.r.l., ISBN 978-88-95608-30-3; ISSN 2283-9216 DOI: 10.3303/CET1439005 Please cite this article as: Sharma I. , Hoadley A., Mahajani S.M., Ganesh A., 2014, Automated optimisation of multi stage refrigeration systems within a multi-objective optimisation framework, Chemical Engineering Transactions, 39, 25-30 DOI:10.3303/CET1439005 25 Automated Optimisation of Multi Stage Refrigeration Systems within a Multi-Objective Optimisation Framework Ishan Sharma a* , Andrew Hoadley b , Sanjay M. Mahajani c , Anuradda Ganesh d a IITB-Monash Research Academy, Indian Institute of Technology Bombay, Powai, Mumbai, 400076, India b Department of Chemical Engineering, Monash University, Clayton, VIC-3168, Australia c Department of Chemical Engineering, Indian Institute of Technology Bombay, Powai, Mumbai, 400076, India d Department of Energy Science and Engineering, Indian Institute of Technology Bombay, Powai, Mumbai, 400076, India 114174001@iitb.ac.in This work demonstrates an automated optimisation of a two stage refrigeration system, which is embedded into an Excel-based Multi-Objective Optimisation (EMOO) framework. The proposed framework has been demonstrated using the Rectisol TM process with CO2 capture as an example. The automated optimisation procedure assesses any opportunities to exploit “pockets” in the process Grand Composite Curve (GCC), besides analysing the GCC for two discrete refrigeration temperature levels. The program uses a Co-efficient of Performance (COP) approximation to estimate the required electrical duty. The results of this sub-program are analysed as part of the wider Multi-Objective Optimisation (MOO) which sets the process decision variables such as the solvent flow-rates and solvent regeneration pressure levels in order to minimise the total electrical power consumption and maximise CO2 capture rate. Two options for increasing the pressure of the captured CO2, i.e. by condensation and pumping of CO2 up to 100 bar (Case-I) and by compression up to 100 bar (Case-II) have also been compared by assessing their respective Pareto plots. This is interesting as the condensation case adds an additional refrigeration duty. 1. Introduction Process plants working in the sub-ambient temperature range require some kind of refrigeration system. Refrigeration systems, because of their high capital and operating costs, are known to dominate the technical and economic aspects of many chemical processes. Process integration methodology may play a significant role in the optimisation of such systems. Most of the industrial refrigeration systems employ vapour compression cycles, where the compressor shaft work is often considered to dominate the operating cost of the system (Lee, 2002). One such process is the acid gas physical absorption process, known as Rectisol™ which uses refrigerated methanol as a solvent to absorb impurities from synthesis gas. The cost of the Rectisol TM process is tightly coupled to the associated refrigeration system. In recent years, there has been a surge in research articles dealing with Multi-Objective Optimisation (MOO) applications in chemical engineering (Rangaiah, 2009). A major milestone was achieved by Bhutani et al. (2007) when they demonstrated the integration of a MOO algorithm with commercial simulation softwares like, Aspen Hysys TM and Aspen Plus TM . Sharma et al. (2012) implemented this strategy in a Microsoft Excel-based MOO framework, which was extended by Harkin et al. (2012) to incorporate heat pinch analysis. In this work, further addition to this framework is developed which incorporates the optimisation of vapour compression refrigeration systems. This addition is demonstrated using the Rectisol TM process with CO2 capture as an example. Rectisol TM process is typically used to produce a very high purity hydrogen product, which is required for applications such as the synthesis of ammonia. Rectisol TM uses chilled methanol at temperatures between -20 and -70 °C (Sun and Smith, 2013) to physically absorb CO2. Gatti et al. (2013) looked at the role of energy integration in optimization of chilled methanol based CO2 capture process. Their analysis was however confined to a particular CO2 capture level, which doesn’t provide the opportunity to analyse the trade-off involved between energy 26 Energy (MW) 0 1 2 3 4 5 6 S h ft e d T e m p e ra tu re ( 0 C ) -60 -40 -20 0 20 40 60 80 100 Energy (MW) 0 1 2 3 4 5 6 S h ft e d T e m p e ra tu re ( 0 C ) -60 -40 -20 0 20 40 60 80 100 Figure 1: (a) Pocket Exploitation and (b) Extra refrigeration level at CO2 condensation temperature consumption and the extent of CO2 capture. Although the authors are interested in the optimisation of the full coal to ammonia plant, in this study, the Rectisol TM plant with CO2 capture is optimised as a stand- alone unit. There are two options for increasing the pressure of the captured CO2, i.e. by condensation and pumping of CO2 up to 100 bar (Case-I) and by compression up to 100 bar (Case-II). These two cases are evaluated in this study, because of the interesting refrigeration synergy in Case I. 2. Refrigeration system optimisation strategy Lee (2002) discussed a number of design options available to improve the performance of a refrigeration system. A typical GCC for the Rectisol TM process involving CO2 condensation is shown in Figures 1(a) and (b). The flat portion in the GCC at around -10 °C corresponds to the CO2 condensation stream. The GCC shows that the two most promising options available to optimise such a refrigeration system are “pocket exploitation”, Figure 1a, and adding another refrigeration level at the CO2 condensing temperature, Figure 1b. Estimating shaftwork is a major challenge encountered while optimizing multi stage refrigeration systems with the help of process GCC. Linnhoff and Dhole (1989) first used the Exergy Grand Composite Curve (EGCC) to estimate the shaftwork requirements. This methodology of estimating shaftwork has been used by many researchers over the years. Recently, Hackl and Harvey (2012) used it, in conjunction with total site analysis, to reduce the shaftwork of the cooling systems in one of the industrial clusters in Sweden. Raei (2011) also used this approach to estimate shaftwork for an industrial case study. However, in this work, the shaftwork for the refrigeration section is estimated by using a simple COP-based formulation (Smith, 2005). An actual performance of 60 % of that of the ideal isentropic performance has been assumed for each stage of the refrigeration system. The pocket exploitation, optimisation problem for the GCC shown in Figure 2(a) is formulated as follows:                                                   minmin min min min min min / *5.015.273 * *5.015.273*5.015.273 TT TT TT TT QQQQ TT TT Q Min E CWS cond cond condcondCWtot cond cond cond iBPAP S S S S (1) Where, cond Q : Heat being given to the pocket  MW , Scond T : Temperature (shifted) on the GCC corresponding to cond Q  C , minT : Minimum temperature (shifted) on GCC  C , minT : Minimum temperature difference for the heat exchanger network  C , totQ : Total Cooling Utility  MW , CWQ : Cooling Water Target  MW and CWST : Cooling water temperature (shifted)  C Energy (MW)0 1 2 3 4 5 6Shfted Temperature (0 C) -60-40-20020406080100 CW (a) 2 nd Stage at CO2 condensing temperature 1 st Stage at minimum temperature (b) 27 cond Q and Scond T are linearly related and the exact relation can be deduced by the decomposition of the GCC. In absence of a pocket below the pinch, the total cooling duty can be provided at two different temperature levels, as shown in Figure 2(b). Such a situation is expected in Case II, i.e. the case in which CO2 is being compressed up to 100 bar. In this case, the optimum intermediate temperature level, Sevap T , can be estimated by solving Eq(2), to minimise the refrigeration compressor work.                                 minmin min min *5.015.273 * *5.015.273 TT TT QQ TT TT QMinE CWS evaptot evap evapCWS evapWP S S (2) Where, Sevap T : Intermediate temperature (shifted) level  C0 , evapQ : Duty corresponding to SevapT (MW) In Case I, a separate refrigeration level can also be provided at the CO2 condensing temperature. The corresponding shaft work is termed as T E . The Optimum shaftwork is thus the minimum value among   iAP E ’s,   iBP E ’s, WPE and TE ; i.e.     TWPiBPiAPoptref EEEEMinE ,,,,  . It is important to note here that exploiting pocket 2 will require an additional refrigeration stage to be introduced in order to satisfy the cooling requirement of the heat source within the pocket. Therefore, in such a case, a three stage refrigeration system would be required. In this work, however, we have restricted our analysis only to two stage refrigeration systems. The proposed MOO framework is depicted in Figure 3. The optimisation algorithm used in this study is Nondominated Sorting Genetic Algorithm-II (NSGA-II) proposed by Deb et al. (2002). Energy (MW) 0 1 2 3 4 5 6 S h ft e d T e m p e ra tu re ( 0 C ) -60 -40 -20 0 20 40 60 80 100 Energy (MW) 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 S h if te d T e m p e ra tu re ( 0 C ) -60 -40 -20 0 20 40 60 Figure 2: (a) Pocket exploitation optimisation and (b) Two stage refrigeration system optimisation in absence of a pocket below pinch Decision variables Decision variables Aspen Plus TM Simulation Two Stage Refrigeration System OptimisationValue for objective functions and constraints Value for objective functions and constraints Optimisation Algorithm NSGA-II Excel-Visual Basic Interface Figure 3: Proposed MOO framework Pocket 1 Pocket 2 (b) (a) 28 Absorber First stage (T 102) Flash drum Second Stage (T 103) Flash drum Third stage (T 104) Flash drum Fourth stage (T 105) Flash drum Intercooler 2 Intercooler 3 Intercooler 4 Intercooler 1 C2 C4 C3C1 Gas Feed 1.6% CO 46.46% CO2 49.32 % H2 0.175% H2S 1.24% Ar 0.472% CH4 2.8112 PPM COS Treated Gas H2S/CO2 Rich Solvent T 101 C6 Recycle Cooler 2 Distillation Column (Stripper) Condenser T 106 Solvent Recycle Pump Solvent Cooler Solvent Pump CW Cooler Bottoms Heater Methanol @ -42 0 C Heater C5 CO2 @ 100 bar (a) (Case II) Methanol Recycle CO2 Condenser T 107 Light Gas Recycle CO2 Pump CO2 Reheater CO2 @ 100 bar (a) (Case I) C7 Recycle Gas Cooler Figure 4: Process flowsheet for Case I and Case II (The dotted section represents the additional unit operations for Case I) 3. Process Description Figure 4 shows the two Rectisol TM configurations considered in this work. Methanol solvent at -42 °C is fed to the absorber where it absorbs the acid gas components from the raw syngas, being fed at -21 °C. The absorber has two liquid side-draw cooling loops. The CO2 rich solvent is heated and flashed to recover the co-absorbed H2 and CO. The flashed gases are then compressed and recycled back to the column. The solvent is then flashed in a series of drums in order to recover CO2 at different pressure levels. A stripper is used downstream in order to complete the regeneration of the solvent with the help of steam to remove H2S. 4. MOO problem The overall MOO problem for the two cases is formulated as follows: Case I : Maximise (%) 2CO CR & Minimise )( 6 1 kWEEEEEPP SSSPRPrefi Ci   (3) BHSolTT VFandFPPtrw ,,:... 105102 ; %8.99%98,10: 22  HCO TG CO RandRPPMxtosubject Case II: Maximise &(%) 2CO CR Minimise )(2 7 1 kWEEEEEEPP SSSPPCORPrefi Ci   (4) condBHSolTT PandVFFPPtrw ,,,:... 105102 ; %8.99%98,10: 22  HCO TG CO RandRPPMxtosubject Where; 2CO CR : CO2 capture rate (%) , PP : Total power penalty associated with CO2 capture )(kW , CiE : Electrical power consumed by compressor i C )(kW , ref E : Optimum electrical power consumed by refrigeration compressor )(kW , RP E : Electrical power consumed by Solvent Recycle pump )(kW , PCO E 2 : Electrical power consumed by CO2 pump )(kW , SP E : Electrical power consumed by Solvent pump )(kW , SS E : Approximate electrical power sacrificed by using LP steam in stripper reboiler )(kW , 102T P : First stage flash pressure )(bar , 105T P : Last stage flash pressure )(bar , SolF : Solvent (methanol) flow rate  hkmol , BH VF : Outlet vapour fraction of Bottoms Heater, cond P : Pressure at which CO2 condensation 29 Figure 7: Pareto fronts obtained for two alternate CO2 pressurisation mechanisms Table 1: Decision variable range for optimisation of Case I and Case II Decision Variable Case I Case II 102T P 10-30 )(bar 10-38 )(bar 105T P 0.1-10 )(bar 0.1-10 )(bar Sol F 2,000-2,900  hkmol 2,000-2,900  hkmol BH VF 0.005-0.035 0.005-0.035 cond P 30-65 )(bar Not Applicable takes place )(bar , TG CO x 2 : Mole fraction of CO2 in treated gas, CO R : Overall recovery of CO (%) and 2H R : Overall recovery of H2 (%) . The electrical power sacrificed by using LP steam in stripper reboiler has been estimated using the Salisbury approximation (Salisbury, 1942). The range for each decision variable is given in Table 1. Results and Discussion The approximate Pareto front for Case I, obtained after 50 generations is shown in Figure 5(a). The fourth stage flash pressure ( 105T P ) had the most significant effect on the final Pareto front. 105T P values corresponding to the optimum objective function values are shown in Figure 5(b). The optimum specific energy penalty (electricity consumed per unit mole of CO2 captured) is an important measure that often gives useful insights in such situations. The optimum specific energy penalty, opt SEP is the ratio of the total power to molar flowrate of captured CO2. Figure 6 shows opt SEP for different CO2 capture rates shown in Figure 5(a), with a minimum at around 75 % capture rate. The total power penalty P T 105 (bar) 0 2 4 6 8 10 C O 2 C ap tu re R at e, C R C O 2 (% ) 40 50 60 70 80 90 100 Figure 5: (a) Approximate Pareto front obtained for Case I and (b) 105T P values corresponding to optimum objective function values CO 2 Capture Rate, CR CO 2 (%) 40 50 60 70 80 90 100 S E P o p t (k W h /k m o l C O 2 ) 3.6 3.8 4.0 4.2 4.4 4.6 4.8 5.0 5.2 5.4 Power penalty (kW) 1500 2000 2500 3000 3500 4000 4500 5000 C O 2 C a p tu re R a te , C R C O 2 ( % ) 40 50 60 70 80 90 100 Case I (Compression + Condensation) Case II (Only Compression) Power penalty, PP (kW) 1500 2000 2500 3000 3500 4000 4500 5000 C O 2 C ap tu re R at e, C R C O 2 (% ) 40 50 60 70 80 90 100 (a) Figure 6: Minimum specific energy penalty at different CO2 capture rates PT 105 (bar)0 2 4 6 8 10CO 2 Capture Rate, CR CO 2 (%) 405060708090100 (b) 30 associated with CO2 capture is composed of the compression and refrigeration penalty. The refrigeration penalty is in turn composed of solvent refrigeration and CO2 condensation. With decreasing CO2 capture rate, the solvent refrigeration requirements remain the same due to the constant solvent flow rate required in order to achieve the quality constraint ( ≤10 PPM) for the treated syngas, which in turn explains the trend observed in Figure 6. A comparison between the Pareto fronts obtained for Case I and Case II is shown in Figure 7. Case I had a slightly lower energy penalty for low capture rates, but the relative difference between the Cases decreased with increasing capture rates. For all the Pareto optimal points obtained for Case I, the optimiser selected a two stage refrigeration system, with refrigeration temperature levels at (1) the CO2 condensation temperature and (2) the minimum temperature on the GCC (i.e. ). By contrast, for every Pareto optimal point obtained in Case II, pocket exploitation was the preferred design option. This demonstrates the value in allowing the software to select the specific method of GCC optimization, rather than selecting it a priori. 5. Conclusions An automated optimisation of a two stage refrigeration system, embedded into an Excel based Multi- Objective Optimisation (EMOO) framework has been proposed. The proposed framework has been applied to a Rectisol TM unit that absorbs and captures CO2 from a water gas shifted syngas stream. The minimum electrical power penalty is thus obtained for different CO2 capture rates, along with the corresponding operating parameters. Two options for increasing the pressure of the captured CO2 have also been compared. The results show that the condensation case performs slightly better than the compression case. However, the relative performance of the condensation case declines with an increase in the CO2 capture rate. Acknowledgement The authors would like to thank Orica Mining Services for funding the project through the IITB-Monash Research Academy. References Bhutani N., Tarafder A., Rangaiah G.P., Ray A.K., 2007, A Multi-platform, multi-language environment for process modelling, simulation and optimisation, International Journal of Computer Applications in Technology, 30(3), 197–214 Deb K., Pratap A., Agarwal S., Meyarivan T.A.M.T., 2002, A fast and elitist multiobjective genetic algorithm: NSGA-II, IEEE Transactions on Evolutionary Computation, 6(2) , 182–197 Harkin T., Hoadley A., Hooper B., 2012, Optimisation of power stations with carbon capture plants – the trade-off between costs and net power, Journal of Cleaner Production, 34, 98-109 Lee G.C., 2001, Optimal Design and Analysis of Refrigeration Systems for Low Temperature Processes, PhD thesis, University of Manchester Institute of Science and Technology, Manchester, UK Linnhoff, B., Dhole, V. R., 1989, Shaftwork Targeting for Subambient Plants, AIChE Spring Meeting, Houston, USA, April, Paper No. 34d Gatti M., Marechal F., Martelli E., Consonni S., 2013, Thermodynamic analysis, energy integration and flowsheet improvement of a methanol absorption acid gas removal process, Chemical Engineering Transactions, 35, 211-216 DOI: 10.3303/CET1335035 Hackl R., Harvey S., 2012, Total site analysis (tsa) and exergy analysis for shaft work and associated steam and electricity savings in low temperature processes in industrial clusters, Chemical Engineering Transactions, 29, 73-78 Raei B., 2011, Optimization in energy usage for refrigeration systems using combimed pinch and exergy analysis, Chemical Engineering Transactions, 25, 135-140 Rangaiah G.P., 2009, Multi-objective Optimization: Techniques and Applications in Chemical Engineering, Singapore: World Scientific Salisbury J.K., 1942, Steam-turbine regenerative cycle – an analytical approach, Trans. ASME 64 (4), 231 Sharma S., Rangaiah G.P., Cheah K.S., 2012 , Multi-Objective Optimization Using MS Excel with an Application to Design of a Falling-Film Evaporator System, Food and Bioproducts Processing, 90(2), 123-134 Smith R., 2005, Chemical Process Design and Integration, England: John Wiley and Sons Sun L., Smith R., 2013, Rectisol wash process simulation and analysis, Journal of Cleaner Production 39, 321-328 .