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/CET1439186 Please cite this article as: Sharma I., Hoadley A., Mahajani S.M., Ganesh A., 2014, Optimisation of pressure swing adsorption (PSA) process for producing high purity CO2 for sequestration purposes, Chemical Engineering Transactions, 39, 1111-1116 DOI:10.3303/CET1439186 1111 Optimisation of Pressure Swing Adsorption (PSA) Process for Producing High Purity CO2 for Sequestration Purposes 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 Fixed bed adsorption processes such as pressure swing (PSA) and temperature swing (TSA), unlike other chemical engineering separation processes, are dynamic processes which do not produce a continuous or steady flow of either the adsorbate or lean (non-adsorbed) phase. Instead, they operate with multiple vessels in a cyclic operation. These processes, therefore, pose additional challenges which include a much larger array of input parameters that control the individual steps within each PSA or TSA cycle. This work proposes an Aspen Adsorption TM based Multi-Objective Optimisation (MOO) framework for PSA systems. The PSA systems can be optimised against different step times and process parameters such as, blow down pressure, feed pressure, valve co-efficients etc. The proposed framework is demonstrated by considering an example of a PSA based Carbon Capture and Sequestration (CCS) unit, for the removal of CO2 from an entrained flow gasifier synthesis gas stream, downstream of the Water Gas Shift Reactors. The two objective functions maximise the CO2 capture rate and minimise the specific energy penalty associated with CO2 capture. A novel feature of this study is the purification of the CO2 produced by the PSA by condensing it, thereby, allowing it to be pumped up to a pressure of 100 bar. The off-gas from the separation has been constrained to have the same composition as that of the feed, which may be recycled to the PSA process or used for a different purpose. The MOO Pareto curves provide information on the most important variables for both the PSA and the refrigeration system. 1. Introduction A typical PSA or TSA usually has several beds operating in a cyclic manner. Switching from one bed to another is done to optimise the purity of the two product streams, whilst minimising the total volume of adsorbent required in each bed. There is a large array of inputs that need to be specified for each step of the cycle. These specifications may include: step times and process parameters such as, blow down pressure, feed pressure, valve coefficients etc. It is for this reason that optimisation of a PSA process is often a challenging task. Optimisation of a PSA process may also often involve multiple, conflicting objectives. In such situations, a PSA operator may want to examine the trade-offs involved between such objectives. There have been only a few studies that have focussed on Multi-Objective Optimisation (MOO) of PSA processes. A brief summary of these works is given in the following paragraph. Ko and Moon (2002) used a modified Summation of Weighted Objective Functions (mSWOF) method to simultaneously maximise O2 recovery and purity while separating O2 from air. Sankararao and Gupta (2007) applied a modified version of the aJG adaptation of Multi-Objective Simulated Annealing (MOSA- aJG) to maximise the recovery and purity of different product stream produced during separation of air by a PSA. Fiandaca and Fraga (2009) used a custom Multi-Objective Genetic Algorithm (MOGA) to optimise a PSA process, involving air separation, to simultaneously maximise N2 purity and recovery. Cacas et al. (2013) studied the MOO of a PSA process from the point of view of maximising H2 and CO2 recovery from a binary mixture of H2 and CO2. They considered different step times as decision variables. Liu and Sun 1112 (2013) used a meta-model to approximate the performance of a PSA process involving separation of air. Beck et al. (2013) recently reported using a meta-model of the PSA system instead of an accurate and computationally expensive full scale model. This work proposes a generalised Aspen Adsorption TM and Microsoft Excel based MOO framework similar to the one proposed by Sharma et al. (2012). This framework accesses the Aspen Adsorption TM flowsheet variables through an object of ACM (Aspen Custom Modeler™) application. The proposed framework is demonstrated with the help of a PSA based Carbon Capture and Sequestration (CCS) unit for the removal of CO2 from an entrained flow gasifier synthesis gas stream, downstream of the Water Gas Shift Reactors. The objectives that have been considered are maximising CO2 capture rate and minimising the corresponding specific energy penalty. Nondominated Sorting Genetic Algorithm (NSGA)-II, proposed by Deb et al. (2002), is used as the optimisation algorithm. A typical problem associated with using PSA process for carbon capture from synthesis gas is that the H2 mole fraction in the CO2 stream is typically high. The solution to this problem has usually been confined to the introduction of additional steps into the PSA cycle. A common solution to this problem is to follow the adsorption step by a high purity CO2 stream reflux (Sircar (1979)). Chou et al. (2013) used a two stage PSA system to achieve the same goal. In this case, syngas was first passed through a modified activated carbon bed to produce an H2 product stream with desired purity. A higher purity CO2 stream was then produced in the second stage with the help of a zeolite 13X-Ca bed. A novel and alternate way to achieve the same goal has been proposed in this work that involves partial condensation of the CO2 product stream to produce a high purity liquid CO2 stream which can then be pumped to supercritical conditions, required for sequestration. The remaining vapour, constrained to have the same composition as that of feed, may be recycled to the PSA process or used for a different purpose. The proposed strategy also offers an opportunity for energy integration with the rest of the process, for example, in ammonia production by coal gasification, where there is a similar refrigeration requirement. 2. Process Configuration The PSA process under consideration consists of four adsorption beds filled with activated carbon. The PSA system is being operated in a cycle consisting of 12 steps. The PSA system and the time chart for the cycle are depicted in Figure 1 and 2. Figure 1: Flowsheet of the system being studied C1 C2 IC1 IC2 CO2 Condenser FD Recycle Compressor (RC) HX1 Recycle to Feed HX2 Bed 1 Bed 2 Bed 3 Bed 4 VF1 VF2 VF3 VF4 VP1 VP2 VP3 VP4 Feed Tank H2 Product Tank H2 Product VPurge1 VPurge2 VPurge3 VPurge4 VW1 VW2 VW3 VW4 CO2 Tank CO2 Product VPEQ12 VPEQ23 VPEQ34 VPEQ13 VPEQ14 VPEQ24 Pump Captured CO2 (MCO2) Fresh Feed Simulated in Aspen Adsorption TM Simulated in Aspen Plus TM IC3 C3 C4 IC4 C5 1113 PRES AD PED1 PED3 BD PG PEP1 PEP3 PEP3 PRES AD PED1 PED3 BD PG PEP1 BD PG PEP1 PEP3 PRES AD PED1 PED3 PED3 BD PG PEP1 PEP3 PRES AD PED1PED2 PEP2 PED2 PEP2 PEP2 PED2 PEP2 PED2 Figure 2: Time chart for the PSA cycle. Steps are denoted as: PRES: Pressurisation; AD: Adsorption; PED1: First Pressure Equalisation (depressurisation); PED2: Second Pressure Equalisation (depressurisation); PED3: Third Pressure Equalisation (depressurisation); BD: Blow down; PG: Purging; PEP1: First Pressure Equalisation (pressurisation); PEP2: Second Pressure Equalisation (pressurisation); PEP3: Third Pressure Equalisation (pressurisation) The CO2 product stream generated by adsorption system is fed to a compression train to pressurise it to an intermediate pressure. The CO2 is then condensed at this intermediate pressure in such a way that the vapour stream leaving the vessel (FD) has the same composition as that of feed. This vapour stream is then recycled to the PSA section. The high purity liquid CO2 stream from vessel (FD) at intermediate pressure is then pumped to a pressure of 100 bar. By ensuring that the vapour stream from flash drum, FD has the same composition as that of feed, the computation time required to solve the system reduces significantly as the PSA system only needs to be solved once, because recycling the vapour effectively means scaling the PSA operation. The adsorption system is simulated in Aspen Adsorption TM while the compression train and CO2 condensation is simulated in Aspen Plus TM . The stream data for CO2 condenser, HX1 and HX2 are extracted (Harkin et al. (2012)) and the two stage refrigeration system is then optimised with the help of a separate algorithm. Figure 3 depicts the framework pictorially. Figure 3: Proposed MOO framework (The details of the algorithm used to optimise refrigeration system can be found in Sharma et al. (2014)). 3. Adsorption model The following simplifications have been made as part of the PSA model in Aspen Adsorption TM :  Isothermal operation  The flow pattern has been assumed to be plug flow with axial dispersion only  Concentration gradients in the radial direction have been neglected  The overall mass transfer rate is assumed to be described by an overall lumped resistance Table 1 lists the adsorbent and adsorption bed characteristics used in this work. The extended Langmuir Freundlich model is used to predict the multi-component adsorption equilibrium. A linear driving force (LDF) model has been used to estimate the rate of accumulation of adsorbate on the adsorbent. The extended Langmuir Freundlich model parameters and lumped mass transfer co-efficient values have been taken from Jee et al. (2001). 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 Aspen Adsorption TM Simulation CSS Output 1114 Table 1: Adsorbent and adsorption bed characteristics Diameter of adsorption bed 5.82 m Length of adsorption bed Adsorbent Average particle radius Bed void fraction Adsorbent particle density 3 m Activated carbon 0.00115 m (Jee et al. (2001)) 0.433 (Jee et al. (2001)) 850 kg/m 3 (Jee et al. (2001)) Table 2: Decision variable range for optimisation Decision Variable Range iVPurge C 0.000278- 0.00174   barskmol * 3.1 pi F 0.5-8  skmol BD P 0.05-8 )(bar cond P 10, 20, 30, 40, 50 )(bar 4. MOO problem formulation The CO2 capture rate ( 2CO CR , fraction of CO2 being captured) and the specific energy (SE, energy penalty per unit CO2 being captured) have been considered as the two conflicting objectives. Energy is consumed by CO2 compressors, CO2 pump, recycle compressor and refrigeration compressor. Maximise (%) 2CO CR & Minimise   2 5 1 CO pumprefRCi Ci M EEEE SE    (1) w.r.t. iVPurge C , piF , BDP and condP Where; 2CO CR : CO2 capture rate (%) Ci E : Electrical power consumed by compressor i C  kW ref E : Electrical power consumed by refrigeration compressor  kW RC E : Electrical power consumed by recycle compressor  kW pump E : Electrical power consumed by CO2 pump  kW 2CO M : Amount of CO2 captured  kmol iVPurge C ; 4,3,2,1i : Valve co-efficient for purging valves   barskmol * pi F ; 4,3,2,1i : Product flow rate during adsorption steps  skmol (i.e. flow through valves, iVP ) BD P : Blow down pressure )(bar cond P : Pressure at which CO2 condensation is carried out )(bar The range for the decision variables is given in Table 2. 5. Results and Discussion Figure 4(a) shows the approximate Pareto front obtained after 60 generations. The product flow rate and blow down pressure had the most significant effect on the final Pareto front. The pi F and BD P values corresponding to the Pareto optimum objectives is shown in Figures 4(b) and 4(c). The obtained Pareto 1115 front can be thought to consist of two separate regions, i.e. from a CO2 capture rate of ~50 to ~70 % (Region 1) and ~70 to ~95 % (Region 2). Figure 4: (a) Approximate Pareto front obtained after 60 generations of MOO run, (b) Product flow rate, pi F  skmol , (c) Blow down pressure, BD P  )(abar for the Pareto-optimal solutions depicted in (a) The Pareto front in these two regions can be understood as follows: Region 1: In this region blow down pressure is approximately constant, as can be observed in Figure 4(b). The CO2 capture rate, in this region, is increased by decreasing the product flow rate, thereby increasing the residence time of the syngas in the adsorber. This results in an increase in the extent of CO2 adsorption thereby increasing the CO2 capture rate at a constant blow down pressure. Region 2: In this region, the CO2 capture rate cannot be further increased by decreasing product flow rate. Therefore, the blow down pressure needs to be lowered in order to desorb additional CO2 from the adsorbent. Simultaneously, the product flow rate also needs to be increased as the working capacity of adsorbent has increased due to lowering in blow down pressure. 6. Conclusions A generalised Aspen Adsorption TM and Microsoft Excel based MOO framework for optimisation of fixed bed adsorption processes has been proposed. The proposed framework has been demonstrated by optimising a PSA based Carbon Capture and Sequestration (CCS) unit for the removal of CO2 from an entrained flow gasifier synthesis gas stream. A novel strategy has also been proposed to produce high purity CO2 product stream for sequestration, which involves compression and partial condensation of the CO2 rich stream produced from PSA system. This strategy also facilitates the pressurisation of CO2 to supercritical conditions, required for sequestration. The PSA system has been successfully optimised against multiple process parameters. Specific Energy, SE (kWh/kmol CO 2 ) 4.5 5.0 5.5 6.0 6.5 7.0 7.5 C O 2 C a p tu re R a te , C R C O 2 ( % ) 40 50 60 70 80 90 100 Region 1 Region 2 Blow down pressure, P BD (bar(a)) 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 C O 2 C a p tu re R a te , C R C O 2 ( % ) 40 50 60 70 80 90 100 Region 1 Region2 Product flow rate, F Pi (kmol/s) 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 C O 2 C a p tu re R a te , C R C O 2 ( % ) 40 50 60 70 80 90 100 Region 1 Region 2 (a) (b) (c) 1116 Acknowledgement The authors would like to thank Orica Mining Services for funding the project through the IITB-Monash Research Academy. References Beck J., Friedrich D., Brandani S., Fraga E.S., 2013, Surrogate Based Multi-Objective Optimisation for the Design of Pressure Swing Adsorption Systems, AIChE Annual Meeting, San Francisco, 3-8 November Casas N., Schell J., Joss L., Mazzotti M., 2013, A parametric study of a PSA process for pre-combustion CO2 capture, Separation and Purification Technology, 104, 183-192 Chou C.T., Chen F.H., Huang Y.J., Yang H.S., 2013, Carbon dioxide capture and hydrogen purification from synthesis gas by pressure swing adsorption, Chemical Engineering Transactions, 32, 1855-1860 DOI: 10.3303/CET1332310 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 Fiandaca G., Fraga E.S., 2009, A Multi-objective Genetic Algorithm for the Design of Pressure Swing Adsorption, Engineering Optimization, 41(5), 833-854 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 Jee J.G., Kim M.B., Lee C.H., 2001, Adsorption Characteristics of Hydrogen Mixtures in a Layered Bed:  Binary, Ternary, and Five-Component Mixtures, Industrial and Engineering Chemistry Research, 40(3), 868-878 Ko D., Moon I., 2002, Multiobjective optimization of cyclic adsorption processes, Ind. Eng. Chem. Res., 42(2), 93–104 Liu Y., Sun F., 2013, Parameter estimation of a pressure swing adsorption model for air separation using multi-objective optimisation and support vector regression model, Expert Systems with Applications, 40(11), 4496-4502 Sankararao B., Gupta S.K., 2007, Multi-objective Optimization of Pressure Swing Adsorbers for Air Separation, Ind. Eng. Chem. Res., 46(11), 3751–3765 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 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 Sircar S., 1979, Separation of multicomponent gas mixtures, U.S. Patent 4,171,206