Research Info

Hydrogen storage is integral to reducing CO2 emissions, particularly in the oil and gas industry. However, a primary challenge involves the solubility of hydrogen in subsurface environments, particularly saline aquifers. The dissolution of hydrogen in saline water can impact the efficiency and stability of storage reservoirs, necessitating detailed studies of fluid dynamics in such settings. Beyond its role as a clean energy carrier and precursor for synthetic fuels and chemicals, understanding hydrogen’s solubility in subsurface conditions can significantly enhance storage technologies. When hydrogen solubility is high, it can reduce reservoir pressure and alter the chemical composition of the storage medium, undermining process efficiency. Machine learning techniques have gained prominence in predicting physical and chemical properties across various systems. One of the most complex challenges in hydrogen storage is predicting its solubility in saline water, influenced by factors such as pressure, temperature, and salinity. Machine learning models offer substantial promise in improving hydrogen storage by identifying intricate, nonlinear relationships among these parameters. This study uses machine learning algorithms to predict hydrogen solubility in saline aquifers, employing techniques such as Bayesian inference, linear regression, random forest, artificial neural networks (ANN), support vector machines (SVM), and least squares boosting (LSBoost). Trained on experimental data and numerical simulations, these models provide precise predictions of hydrogen solubility, which is strongly influenced by pressure, temperature, and salinity, under a wide range of thermodynamic conditions. Among these methods, RF outperformed the others, achieving an R2 of 0.9810 for test data and 0.9915 for training data, with RMSE values of 0.048 and 0.032, respectively. These findings emphasize the potential of machine learning to significantly optimize hydrogen storage and reservoir