Application of group method of data handling and gene expression programming for predicting solubility of CO2-N2 gas mixture in brine

Abstract

The solubility of CO2-N2 gas mixtures in water is important for CO2/flue gas sequestration in aquifer. Having precise thermodynamic measurements as well as credible prediction models to utilize in innovative carbon capture and storage (CCS) systems is critical. In this study, two simple-to-use white-box models, including Gene Expression Programming (GEP) and Group Method of Data Handling (GMDH) models, have been developed using 289 experimental data to predict the solubility of CO2-N2 gas mixtures in aqueous solutions. Four tuned equations of state (EOSs), namely Peng-Robinson (PR), Soave-Redlich-Kwong (SRK), Zudkevitch-Joffe (ZJ), and Redlich-Kwong (RK), as well as the outcomes of the GEP and GMDH, were compared. The results show that the tuned EOSs perform much better than the untuned EOSs. The results show that the GMDH model has the best predictive performance such that the obtained values of root mean square error (RMSE) and coefficient of determination (R2) are 0.000564 and 0.9792, respectively. It should be noted that the GEP model also has acceptable accuracy, with RMSE and R2 values of 0.00081 and 0.9465, respectively. The SRK model obtained the best outcomes among the EOSs for the solubility of the CO2-N2 gas mixture in aqueous solution, with RMSE and R2 values of 0.00128 and 0.9561, respectively. The results also show that the solubility of CO2 in aqueous solutions is much higher than N2, and increasing the pressure increases the solubility of CO2 and N2 in aqueous solutions, while increasing the CO2 content increases and decreases the solubility of CO2 and N2, respectively. Group error analysis also shows that the developed models have less error in low values of temperature, pressure, and CO2 content. Finally, in order to validate the results of the GMDH and GEP models, the leverage technique has been utilized, which illustrated that 95% of the data are in the valid region, thus, the developed models are statistically reliable. The findings of this study can help for better understanding the solubility process of CO2 and N2 in water to overcome thermodynamic and environmental challenges.