Ionic liquid for simultaneous desulfurization and dehydration of flue gas: Data-driven thermodynamics and mechanisms

A process of simultaneous desulfurization and dehydration of flue gas with ionic liquid was explored. The selection of ionic liquid was completed based on COSMO-SAC model. Random forest (RF) and one-dimensional convolutional neural network were trained for the prediction of CO2 and SO2 solubility. A feature selection method is designed to remove verbose descriptors. The feature importance was analyzed by SHAP analysis. Quantum chemistry calculation and molecular dynamics (MD) simulation were used to investigate the type and position of intermolecular forces. The results show that the R2 coefficients of the two prediction models for CO2 and SO2 test sets reach 0.981 and 0.983. [EMIM][SCN] was selected as the optimal solvent, and the interaction with SO2 is mainly electrostatic and orbital interaction, while the interaction with H2O is strong hydrogen bond.