AI-driven optimization of integrated solar systems for hydrogen production and building energy supply

This study presents an integrated photovoltaic-hydrogen energy system designed to supply continuous electricity and on-site medical oxygen for a hospital in Beijing. The novelty of this work lies in the integration of transient TRNSYS simulations with a neural network-based multi-objective optimization framework and a newly developed supervisory control algorithm for real-time management of power flows among photovoltaic panels, the electrolyzer, hydrogen storage, fuel cell, and the utility grid. To address the lack of built-in optimization capabilities in TRNSYS, an artificial neural network surrogate model was trained using annual simulation data, and a genetic algorithm was employed to identify the optimal system configuration in terms of cost, CO₂ emissions, and power supply reliability. The proposed control strategy prioritizes renewable energy utilization by directing surplus solar electricity to hydrogen production, activating the fuel cell during low solar availability, and minimizing grid dependency. The optimized system includes 844 photovoltaic panels, a 720.14 kW electrolyzer, and a 301.45 kW fuel cell, resulting in an annual electricity generation of 3450.4 MWh, approximately 2000 tons of CO₂ emission reduction, and the production of 10,301 medical oxygen cylinders, with an operational cost of about 2.2 USD per hour. The results demonstrate that the proposed system significantly enhances hospital energy resilience, reduces grid reliance, and mitigates environmental impacts, supporting the development of near-zero energy healthcare facilities.