In “A novel robust optimization framework based on surrogate modeling for underground hydrogen storage in depleted natural gas reservoirs” by Zhilei Han, Zeeshan Tariq, and Bicheng Yan, the authors introduce a powerful new approach to accelerate optimization of Underground Hydrogen Storage (UHS) systems.

Their framework combines advanced compositional reservoir simulation with deep learning–based surrogate models and stochastic optimization. By training CNN-BiLSTM-Attention models on extensive simulation data, the authors created fast, accurate surrogates that were integrated into a genetic algorithm to maximize the net present value (NPV) of UHS projects.

Key results:

• Surrogate models achieved excellent prediction accuracy and scalability
• ~4,878× faster optimization compared to direct reservoir simulations
• Provides a practical pathway for efficient, sustainable UHS design and management

Read the full article in InterPore Journal: https://doi.org/10.69631/ipj.v2i3nr69