Multi-objective optimization of the drainage performance of dual-flow channel proton exchange membrane fuel cells driven by machine learning surrogate model

To enhance the drainage performance of fuel cells under high load conditions, this study proposed a helical baffle flow field (FF). This FF is based on a semi-circular cross-section design and generated along a helical trajectory, with four structural parameters (pitch, radius, gap, and shear angle). To optimize the drainage performance of the novel FF, a multi-objective optimization framework is established with four structural parameters as the design variables and pressure drop in the gas flow channel and water coverage ratio on the gas diffusion layer surface as the optimization objectives. The optimization process is divided into three steps: First, initial sample data is collected through experimental design, and the simulation model considered the coupled effects of multichannels, dynamic contact angle effects of liquid droplets in high-speed motion, and linked the inlet gas and water velocities with the operating current density to enhance the simulation accuracy. Second, artificial neural network surrogate models are trained based on the obtained simulation data to establish a high-precision mapping relationship between the performance indicators and the design variables and ensure good generalization ability. Finally, based on the surrogate models, the multi-objective optimization is carried out using genetic algorithms to generate the Pareto front, and the optimal FF structure parameter combination is obtained. The research results show that the optimized FF structure improves the comprehensive drainage performance by 7.8%. Among them, the diameter of the baffle has the greatest impact on the drainage performance, while the shear angle has the least impact. This study provides a new perspective for optimizing the drainage performance of fuel cells under high load conditions.