文献类型：期刊文献

英文题名：Machine learning model based on parallel reaction mechanisms for predicting CO2 capacity of amine solvents

作者：Zhao, Qiyue[1];Cui, Yupeng[1];Liu, Chuanlei[1];Wen, Haiyang[1];Chen, Yuxiang[1];Li, Peicheng[1];Zhou, Yousheng[1];Wang, Yifan[1];Jiang, Hao[1];Wu, Qiumin[1];Shen, Benxian[1];Sun, Hui[1,2]

机构：[1]East China Univ Sci & Technol, Sch Chem Engn, Shanghai 200237, Peoples R China;[2]Xinjiang Univ, Sch Chem Engn & Technol, Key Lab Oil & Gas Fine Chem, Minist Educ, Urumqi 830046, Peoples R China

年份：2026

卷号：320

外文期刊名：CHEMICAL ENGINEERING SCIENCE

收录：;EI(收录号：20253419009163);WOS:【SCI-EXPANDED(收录号:WOS:001584564800005)】；

基金：This work is financially supported by the National Natural Science Foundation of China (Grant 21878097 and 22178109) .

语种：英文

外文关键词：Absorption capacity; Cycling capacity; Machine learning; CO2 capture

摘要：Absorption capacity (Ca) and cycling capacity (Cc) are key indicators for evaluating CO2 capture ability and regeneration energy consumption of solvents in industrial absorption processes. However, their determinations still largely rely on experimental methods, requiring high cost and time consumption. In this study, we propose a machine learning (ML) model that integrates both the zwitterion mechanism and the base-catalysis mechanism to predict Ca and Cc. The model effectively predicts the conversion of structurally diverse amines through different reaction pathways, thereby enabling accurate assessment of absorption and regeneration performance of chemical solvents. The models have an R2 of 0.950, Q2 of 0.856, RMSE of 0.105, and MAE of 0.084 for Ca prediction and R2 = 0.956, Q2 = 0.863, RMSE = 0.059, and MAE = 0.044 for Ccprediction. Dynamic absorption-desorption experiments employing three industrial solvents and two promising candidates screened through the ML model indicate that the coupled model reliably predicts Ca and Cc. Moreover, 30 molecular descriptors were defined to describe local electronic effects of amine groups, steric hindrance effects near amine groups, and the stability of reaction products, largely enhancing the model interpretability. Present study highlights the ML approach for predicting the capacity of gas absorption systems involving complex chemical reactions, offering a strategy for exploring potential solvents for carbon capture and other absorption processes.