A comparative study of multi-objective optimization with ANN-based VPSA model for CO2 capture from dry flue gas

Abstract

A 3-bed-7-step vacuum pressure swing adsorption (VPSA) process was developed for carbon capture from dry flue gas with a silica gel adsorbent. The VPSA process is simulated based on a series of detailed models. This paper aims to solve the multi-objective optimization problem of the VPSA process. A trained artificial neural network (ANN) model with double hidden layers was established to optimize the process performance through the metaheuristic algorithm (NSGA-II, MOPSO, MOEA/D, and NSGA-III). Subsequently, the diversity and convergence performance of the metaheuristic algorithm were analyzed and compared. The optimization results show that on the one hand, the purity of CO2 could reach 80.94% with recovery of 90.61%; and on the other hand, the productivity could reach 0.5233 mol/h/kg with energy consumption of 1004.14 kJ/kgCO2 with the constraint of 70% purity and 90% recovery. The results indicate that, the ANN model can predict the performance metrics and dynamic performance of the VPSA process with very high accuracy. Meanwhile, the NSGA-II can obtain a comprehensive set of trade-off alternatives from different optimization problems, which can be used as a powerful reference for the operation of the carbon capture process.