Accurate estimation of the methane hydrate equilibrium in brines based on advanced modeling tools.

Knowledge of hydrate equilibrium conditions is of significant importance for processes associated with seawater purification, energy storage, and gas separation. Therefore, it is vital to design reliable models for determining such conditions. This investigation aimed to construct robust machine learning algorithms for assessing the equilibrium state of methane hydrates within aqueous salt solutions. A substantial dataset, comprising 1051 individual data points, was compiled from available literature. This dataset contained methane hydrate equilibrium across 26 distinct brines. Data-driven modeling was executed via the application of Support Vector Machine (SVM) and Decision Tree (DT) methodologies. Various graphical and statistical tools were employed to evaluate the validity of the models. It was found that both SVM and DT models exhibit strong capabilities, achieving mean absolute percentage errors (MAPEs) of 0.36 % and 0.48 %, and relative root mean square errors (RRMSEs) of 0.89 % and 0.83 %, respectively, in the testing stage. The intelligent models also adeptly captured the relationships between hydrate equilibrium and operational parameters. A sensitivity analysis ultimately elucidated the relative importance of the factors influencing the hydrate equilibrium phenomenon.