Advanced Smart Models for Predicting Interfacial Tension in Brine-Hydrogen/Cushion Gas Systems: Implication for Hydrogen Geo-Storage

Storing hydrogen (H2) in subsurface formations represents a transformative step toward clean energy solutions and the mitigation of greenhouse gas emissions. A critical challenge in this process lies in accurately predicting interfacial tension (IFT) in the brine-H2/cushion gas systems, which is essential for optimizing hydrogen storage design and injection strategies. This study addresses this gap by leveraging a comprehensive database of IFT values for hydrogen and cushion gases (e.g., CH4, CO2) in brine systems under diverse operating conditions. Advanced machine learning (ML) paradigms, including Light Gradient Boosting Machine (LightGBM), Extra Trees (ET), and Convolutional Neural Network (CNN), were utilized to develop robust predictive frameworks for estimating IFT. To further enhance accuracy, the LightGBM and ET models were hybridized with the Grasshopper Optimization Algorithm (GOA) for hyperparameter optimization. The LightGBM-GOA model emerged as a standout, achieving an average absolute relative deviation (AARD) of 1.5637% and a correlation coefficient of 0.9901, outperforming prior correlations. Its predictive capabilities are closely aligned with the trends observed across varying input features, demonstrating robustness and reliability. This work advances the field by providing a highly accurate, data-driven framework that surpasses existing methodologies, thereby contributing to safer and more efficient hydrogen storage operations. By addressing fundamental challenges in IFT prediction, this study offers a significant step forward in the development of sustainable energy storage technologies.