Bi-objective optimization of post-combustion CO2 capture using methyldiethanolamine

Process simulation and analyzes based on multiple evaluation indexes are crucial for accelerating the practical use of the post-combustion CO₂ capture process. This study presents a bi-objective optimization of the post-combustion CO₂ absorption process using methyldiethanolamine (MDEA) via machine-learning and genetic algorithm to evaluate CO₂ emissions from the absorption process using life cycle assessment and cost from operating and capital expenditures. An initial dataset was generated by changing eight design variables, and machine-learning models were built using random forest classifier and Gaussian process regression. Pareto solutions were predicted using a genetic algorithm (NSGA-II) with the constraints of purity, recovery, and temperature, and were verified via process simulation. Verified data were added to the dataset, and model building, prediction, and verification were repeated. Eventually, 56 Pareto solutions were obtained after 11 iterations. In the final Pareto solutions, CO₂ emissions increased from 0.56 to 0.6 t-CO₂/t-CO₂ with a decrease in cost from 74 to 66 USD/t-CO₂. The trends and composition of each objective variable were examined, and the optimal structure of the equipment and operation conditions was clarified. The approach of bi-objective optimization in this study is promising for evaluating the CO₂ capture process and individual processes of carbon capture, utilization, and storage.