Pareto-based Multi-Objective Optimization of Underground Hydrogen Storage Operational Condition Utilizing Machine Learning-Based Surrogate Models

Underground hydrogen storage (UHS) in depleted hydrocarbon reservoirs, saline aquifers, and salt caverns has emerged as a promising solution for large-scale, long-term energy storage to support the energy transition. However, evaluating the feasibility of UHS requires addressing substantial uncertainties in geological, operational, and economic factors. Physics-based reservoir simulations, though accurate, are computationally expensive and impractical for exploring a wide range of operational scenarios and optimization problems. To overcome these challenges, this study employs an intelligent surrogate model constructed using data collected from Latin hypercube sampling and validated against detailed reservoir simulations to perform sensitivity analysis and formulate a multi-objective optimization framework. The primary objectives are to maximize hydrogen recovery factor and hydrogen purity in the produced gas, while also incorporating net present value (NPV) as an economic criterion. Multi-objective optimization is performed using the Non-dominated Sorting Genetic Algorithm II (NSGA-II), which enables efficient exploration of the trade-offs between conflicting objectives and provides a Pareto front of non-dominated solutions. This Pareto-based decision-making framework offers valuable insights into selecting optimal storage strategies under uncertainty. The results highlight two key findings. First, pure nitrogen (N2) as a cushion gas yields higher hydrogen mole fraction in production stream compared to CO2 and CH4, because of favorable segregation and flow behavior in porous media. Second, the optimal cushion gas mixture was determined to be 45% N2, 40% CO2, and 15% CH4, balancing recovery and purity objectives. Nevertheless, from both economic and operational efficiency perspectives, a CO₂ fraction of 100% emerges as the optimal solution. Additionally, injecting from bottom perforations and producing from top perforations was identified as the most effective configuration due to the gravity override effect of hydrogen. Importantly, re-simulation of the Pareto-optimal solutions confirmed the robustness and reliability of the surrogate-based optimization approach. This study demonstrates the capability of coupling surrogate modeling with evolutionary multi-objective optimization to design cushion gas, optimize injection/withdrawal schemes, and maximize both technical and economic performance of large-scale underground hydrogen storage projects.