The Design of Hydrogen Saline Aquifer Storage Processes Using a Machine-Learning Assisted Multiobjective Optimization Protocol

Abstract

The global effort toward decarbonization has intensified the drive for low-carbon fuels. Green hydrogen, harnessed from renewable sources such as solar, wind, and hydropower, is emerging as a clean substitute. Challenges due to the variable needs and instable green hydrogen production highlight the necessity for secure and large-scale storage solutions. Among the geological formations, deep saline aquifers are noteworthy due to their abundant capacity and ease of access. Addressing technical hurdles related to low working gas recovery rates and excessive water production requires well-designed structures and optimized cushion gas volume. A notable contribution of this study is the development of a multiobjective optimization (MOO) protocol using a Kalman filter-based approach for early stopping. This method maintains solution accuracy while employing the MOO protocol to design the horizontal wellbore length and cushion gas volume in an aquifer hydrogen storage project and accounting for multiple techno-economic goals. Optimization outcomes indicate that the proposed multiobjective particle swarm (MOPSO) protocol effectively identifies the Pareto optimal sets (POSs) in both two- and three-objective scenarios, requiring fewer iterations. Results from the two-objective optimization study, considering working gas recovery efficacy and project cost, highlight that extending the horizontal wellbore improves hydrogen productivity but may lead to unexpected fluid extraction. The three-objective optimized hydrogen storage design achieves a remarkable 94.36% working gas recovery efficacy and a 59.59% reduction in water extraction. The latter represents a significant improvement compared to the reported literature data.