Integrating multiphysics modeling and machine learning for enhanced efficiency and thermal management in PEM water electrolyzer systems

Proton exchange membrane (PEM) water electrolyzers are a promising technology for sustainable hydrogen production, yet optimizing their performance under varying conditions remains a key challenge. This study formulates an optimization problem to examine how key operating parameters, such as inlet flow rate Qin and temperature Tin, enhance performance in a 5 cm × 5 cm PEM water electrolyzer. The goal is to maximize system efficiency, ensure thermal safety, and minimize energy consumption in the balance of plant (BOP). Firstly, a manifold-type straight-channel PEM water electrolyzer model is introduced, accounting for multiphysics coupling effects, to show how inlet temperature and flow rate influence hydrogen production efficiency and BOP energy consumption. After that, the optimization problem is established to enhance system performance. However, due to the high computational cost of solving the optimization problem with a three-dimensional multiphysics model, an artificial neural network (ANN) model is developed as a surrogate, effectively reducing the computational burden. In the next step, using the ANN model, optimal operating conditions at each input power point are identified through the particle swarm optimization (PSO) algorithm. The results show that as the input electrical power Pin increases from 11 W to 35 W, the optimal efficiency decreases by 7.5 %. To maintain safe and efficient operation, Tin must decrease by 6.1 %, and Qin needs to be tripled. Finally, to validate the optimization, three power points are selected for comparison, confirming the feasibility and reasonableness of the outcomes. This study provides a practical approach for performance analysis of a single-cell PEM water electrolyzer.