Efficient estimation of acid gases (CO2 and H2S) absorption in ionic liquids

This work presents the application of computer based models including radial basis function neural network optimized by particle swarm optimization method (PSO-RBF), multilayer perceptron neural network (MLP-NN), least square support vector machine optimized by coupled simulated annealing (CSA-LSSVM) and adaptive neuro-fuzzy inference system trained by hybrid method (Hybrid-ANFIS) for prediction of solubility of CO2 and H2S in ionic liquids (ILs). A high-valued data base including 2930 data points for 39 CO2-IL systems and 664 data points for 14 H2S-ILsystems were collected to develop the models with acceptable generality. The input parameters of the models are temperature, pressure, molecular weight of IL and several structural related parameters of ILs. The performance of the developed models was evaluated by using statistical quality measure approaches. Moreover, the predictions of the developed models are compared with the predictions of a thermodynamic model based on Peng-Robinson equation of state (EoS). Results show that the developed models provide accurate predictions with reasonable errors. However, the CSA-LSSVM model provides better predictions for both CO2-IL and H2S-IL systems with average absolute relative deviation values of 1.97% and 0.13%, respectively. Moreover, the predictions of the developed models are found better and effective than the predictions of the thermodynamic model. Results of this work also show that the solubility of CO2 and H2S in ILs could be well correlated by structural related parameters of ILs.