Prediction of hydrogen solubility in aqueous solutions: Comparison of equations of state and advanced machine learning-metaheuristic approaches

Abstract

Hydrogen is the primary carrier of renewable energy stored underground. Understanding the solubility of hydrogen in water is critical for subsurface storage. Accurately measuring the hydrogen solubility in water has implications for monitoring, control, and storage optimization. In this study, two intelligent systems of Radial Basis Function (RBF) and Least Square Support Vector Machine (LSSVM) were used to precisely predict hydrogen solubility in water. These models were optimized using metaheuristic algorithms, namely biogeography-based optimization (BBO), cultural algorithm (CA), imperialist competitive algorithm (ICA), and teaching-learning-based optimization (TLBO). Quantitative and illustrative evaluations revealed that the RBF paradigm optimized using the CA algorithm with a root mean square error of 0.000176 and a correlation coefficient of 0.9792 is the best model for predicting hydrogen solubility in water. Also, to estimate hydrogen solubility in water, the four well-known equations of state (EoSs) of Soave-Redlich-Kwong (SRK), Peng-Robinson (PR), Redlich-Kwong (RK), and Zudkevitch-Joffe (ZJ) were utilized. The results indicated that the SRK has the best performance among EoSs. However, the intelligent models outperformed the EoSs in terms of accuracy. Considering independent factors, pressure and temperature had the greatest effect on hydrogen solubility in water, respectively. The Leverage technique typified that the RBF + CA model has a good degree of validity for forecasting hydrogen solubility in pure and saline water. Finally, the findings of this investigation demonstrated that the RBF + CA model can have industrial applications and accurately predicts the solubility of hydrogen in pure water and saline water under underground storage conditions (high pressure and temperature).