Optimization of convergent-divergent flow field for large-scale proton exchange membrane fuel cells based on the numerical model, artificial neural network and genetic algorithm

A well-designed flow field is critical for the performance and durability of large-scale commercial proton exchange membrane fuel cell (PEMFC). In the conventional flow field design and optimization studies, only a few structures are utilized in small-scale PEMFC and the effects caused by the increase of active area are ignored which limit the application of novel flow fields. In this study, the novel converging-diverging (C-D) channel is proposed and optimized based on the multi-objective optimization (MOO) method including numerical model, artificial neural network (ANN) and nondominated sorting genetic algorithm (NSGA-Ⅱ). The results show that the optimized C-D channel can improve water management capacity and net output power density although accounting for the parasitic loss, by facilitating the acceleration of the droplet and forcing the transport of oxygen. The water removal time (t) is additionally introduced and analyzed, which is a key evaluation criterion but always ignored in previous MOO studies. In terms of all criteria, the C-D channel with optimized structure performs better than base case, indicating the success of flow field optimization. This study is expected to provide innovative guidance for design and optimization of future large-scale commercial PEMFC flow field based on numerical model and artificial intelligence.