Automatic design of flow channel for PEMFC based on space-weighted topology and multi-objective optimization

To address the severe concentration polarization in Proton Exchange Membrane Fuel Cell (PEMFC) under high current density, caused by oxygen transport limitation and water management imbalance, this study proposes a three-dimensional automatic flow channel design framework integrating topology optimization and Non-dominated Sorting Genetic Algorithm II (NSGA-II). A multi-physics coupled model is established, with average current density, current density variance, and pressure drop selected as optimization objectives. The weight space is systematically explored through Latin hypercube sampling combined with simplex mapping, and a BP neural network surrogate model is trained based on COMSOL-MATLAB high-fidelity simulation data to enable efficient global optimization. The results reveal a pronounced nonlinear trade-off among the three objectives. Single-objective optimization improves current density but significantly increases pressure drop, while conventional empirical weighting schemes fail to achieve balanced performance. The optimal compromise solution (w 1 = 0.571, w 2 = 0.286, w 3 = 0.143), identified using the TOPSIS decision-making method, achieves a 5.04 % increase in peak power density compared to the conventional structure, while significantly enhancing oxygen distribution uniformity and water management capability. Furthermore, the lateral flow intensity ratio demonstrates a fundamental transition of oxygen transport from diffusion-dominated to convection–diffusion coupled mechanisms. This study provides a systematic and data-driven optimization framework that eliminates empirical bias in weight selection, enables automatic generation of complex flow channel structures, and the optimal structure’s discovery mechanistically reveals the embryonic form of single-channel optima, offering new insights for designing high-performance PEMFC flow fields.