This reports investigates how physics-based reservoir simulation an‎d modern artificial intelligence (AI) can be combined to plan, operate, an‎d de-risk underground storage of methane, hydrogen, an‎d carbon dioxide across depleted fields, aquifers, an‎d salt caverns. On the physics side, compositional simulators (e.g., MRST/CMG) are used to quantify key mechanisms gas mixing an‎d CO₂ breakthrough, hydrate risk during CO₂ injection, pressure build-up an‎d geomechanical constraints while establishing safe operating envelopes an‎d design limits. On the AI side, surrogate an‎d physics-informed models (ANN/ensemble learning, stacking, Fourier Neural Operators) emulate high-fidelity responses at orders-of-magnitude lower cost, enabling rapid “what-if” screening, uncertainty analysis, an‎d closed-loop control. A systematic design-of-experiments (≈10³ synthetic runs) is used to train predictive models of clean-production time (e.g., 1% CO₂ cutoff at the well block), achieving high accuracy (R² ≳ 0.99) an‎d validating generalization on Latin-Hypercube samples; operator-learning accelerates long-horizon CCS plume forecasts by ~10⁵×, an‎d a learning-based controller in high-N₂ settings demonstrates small but economically meaningful gains (≈2.4% reduction in produced N₂ an‎d ≈2.4% increase in net energy). Sensitivity consistently ranks reservoir thickness (H) an‎d horizontal permeability (Kh) as first-order levers for well spacing an‎d cycle design. The thesis contributes a practical hybrid workflow simulator as “teacher,” AI as “accelerator” an‎d a digital-twin loop that assimilates new data (VAE + ES-MDA) to keep models aligned with field behavior. The result is a reproducible blueprint that preserves physical credibility while delivering the speed an‎d adaptability required for seasonal UGS, emerging UHS, an‎d long-term CCS, with clear guidance on cushion-gas selec‎tion, integrity monitoring, an‎d market-aware operations

11/10/2025 12:00:00 AM

89261

1404/09/28

Using Simulation an‎d Artificial Intelligence for Gas Storage in Reservoirs

Underground Gas Storage, , Reservoir Simulation, , Artificial Intelligence, , Machine Learning, , CO₂ Storage, , Hydrogen Storage, , Physics-Informed Models.