Real-time optimization of large-scale hydrogen production systems using off-grid renewable energy: Scheduling strategy based on deep reinforcement learning

Abstract

Aiming to amplify the renewable energy consumption capacity, this study delineates the development of an off-grid Renewable Energy Large-Scale Hydrogen Production System (H2-RES). The system was optimized for economic efficiency and safety, promising a reduction in both the investment cost for grid connection and the overall cost of hydrogen production from electrolytic water. We presented a comprehensive mathematical model for each H2-RES unit and designed a control strategy to enhance energy optimization and management. An intelligent energy scheduling policy empowered by the DDPG algorithm is introduced for optimal decision-making in continuous state and action spaces. Comparative analysis with traditional control policies, PSO, and DQN algorithms underscores the superior economic efficiency, enhanced renewable energy consumption, and safe operation facilitated by DDPG. These findings underscore the academic and engineering potential of DDPG in the energy dispatch of H2-RES.