Impact of gas impurities on CO2 mole fraction: Application in carbon capture and storage (CCS) processes

Accurate and detailed information about phase behavior and vapor-liquid equilibrium (VLE) data of impure CO2 is of great importance in designing and simulation of Carbon Capture and Storage (CCS) processes. In the present study, four computer based models namely multilayer perceptron neural network (MLP-NN), radial basis function neural network (RBF-NN), least square support vector machine optimized by particle swarm optimization (PSO-LSSVM) and adaptive neuro fuzzy inference system optimized by hybrid optimization approach (Hybrid-ANFIS) were developed for prediction of experimental VLE data of CO2 + H2, CO2 + N2 and CO2 + O2 systems. In the case of each computer based model, two models were developed for respective liquid and vapor phases. The performance of the developed models for prediction of CO2 mole fraction in liquid and vapor phases were evaluated by using different statistical quality measure approaches. The outcomes of the developed models were also compared with Peng-Robinson Equation of State (PR-EoS) coupled with different mixing rules. Results show that the developed models are accurate and dependable for prediction of experimental data. In addition, the performance of the developed models is better than the studied thermodynamic models for prediction of experimental VLE data.