Predicting Hydrogen Adsorption in Depleted Oil Reservoirs Using Machine Learning: Application for Underground Storage

This study evaluates the performance of three machine learning (ML) models- Linear Regression, Random Forest, and XGBoost-in predicting hydrogen adsorption in oil-wet Berea sandstone under varying pressure and temperature conditions. Using laboratory data, XGBoost demonstrated superior accuracy based on evaluation metrics, outperforming the other models in capturing nonlinear adsorption behavior. Unlike previous studies, which were primarily experimental or simulation-based and focused on coal, shale, or clay minerals, this work employs ML techniques to address a knowledge gap in conventional reservoir rocks. The results highlight data-driven models' potential to improve underground hydrogen storage (UHS) and enable scalable, practical solutions.

Keywords:

Underground Hydrogen Storage (UHS) , Hydrogen Adsorption , Depleted Oil Reservoirs , Oil-Wet Berea Sandstone , Machine Learning (ML)