An interpretable machine-learning-driven tool (HyWEC) for hydrogen wettability estimation: Implications for underground hydrogen storage

Hydrogen is a viable carbon-neutral or carbon-zero option, with projections indicating it could make up ∼12 % of global energy use by 2050. In this regard, underground hydrogen storage is essential for maintaining a sustainable supply. To reduce hydrogen loss while ensuring sustainability, understanding the behavior of hydrogen in porous media (e.g., wettability) is also vital. However, conventional methods for determining wettability (measured via contact angle (θ)), are resource-intensive and operationally constrained. To surmount these challenges, this study introduces more interpretable, machine-learning-based frameworks: Genetic Programming (GP) and Group Method of Data Handling (GMDH). These are integrated into an elementary, yet effective, user-friendly graphical interface called HyWEC, designed for hydrogen wettability modeling without requiring extensive computational expertise. A comprehensive dataset of experimentally measured hydrogen/cushion gas/brine/quartz θs was compiled. The models were trained on different data partitions and subsets, appraised with rigorous statistical and graphical methods, and deployed to estimate key parameters related to geo-storage integrity. Correlation analysis showed that increasing pressure, salinity, and CO₂ concentration favor hydrogen wettability and withdrawal, whereas rising temperature minimizes these effects. Both models (RMSE∼2.9°) outperformed existing black-box counterparts (RMSE_RandomForest = 5.449°; RMSE_{AdaptiveBoosting} = 6.121°). They also proved highly generalizable and practical, accurately estimating storage height (1029–1164 m) and capacity (274–1807 kg/m²) for a typical site in Saudi Arabia, with deviations as low as −0.5 % from reported data. This paradigm shift presents a more robust, reliable, and accessible framework for estimating θ, with the potential to support or reduce reliance on resource-intensive experiments and optimize underground hydrogen storage parameters.