| A Hybrid Physics-Data-Driven Multi-Scale Prediction Methodology for Leakage Risk Assessment in Underground Hydrogen Storage: Homogenization of Interfacial Tension in the Hydrogen–Cushion Gas–Brine System within Reduced-Order Well Leakage Modeling |
| Paper ID : 1050-UGESR |
| Authors |
|
Ali Dadvar *, Asma Zaree Mousavie Ferdowsi University of Mashhad |
| Abstract |
| Underground hydrogen storage (UHS) is vital for large-scale energy storage in the transition to a hydrogen economy; however, well leakage risk remains a critical safety and environmental concern. This study develops a novel hybrid physics-data-driven multi-scale framework designed for the comprehensive assessment of leakage risks in UHS. The research specifically focuses on the homogenization of interfacial tension (IFT) within the complex H₂–cushion gas–brine system to facilitate more accurate reduced-order well leakage modeling. To achieve this, a physics-informed machine learning (PIML) model is trained using high-fidelity experimental IFT data for various gas mixtures, including H₂–CH₄, H₂–N₂, and H₂–CO₂ in brine. The predicted IFT values are then homogenized into an effective parameter and integrated into a reduced-order leakage model. The results demonstrate high prediction accuracy, with R² > 0.95. Crucially, the study reveals that neglecting the specific effects of cushion gas composition can lead to errors of up to 28% in leakage rate estimates. By identifying cushion gas composition and well permeability as the dominant risk controls, this framework provides a computationally efficient and reliable tool for ensuring the containment security and long-term integrity of UHS facilities. |
| Keywords |
| Underground hydrogen storage, Leakage risk, Interfacial tension, Cushion gas, Physics-informed machine learning. |
| Status: Accepted (Poster Presentation) |