Parameter optimization and performance enhancement of anode fishbone-rib flow channels in proton exchange membrane fuel cells based on LHS-NSGA-II

This study presents an AI-driven optimization strategy for the anode fishbone flow field design in high-temperature Proton Exchange Membrane Fuel Cells (PEMFCs), targeting enhanced power output, durability, and reactant distribution uniformity. Four geometric parameters—channel depth, width, angle, and length—are simultaneously optimized using Latin Hypercube Sampling (LHS) and a Kriging surrogate model. Model accuracy is validated with R2 values exceeding 0.95 and low mean squared error, ensuring predictive reliability. The optimization process, compared to conventional methods, reduces computational workload by approximately 90%. A combination of LHS and the Non-dominated Sorting Genetic Algorithm II (NSGA-II) is employed to construct the Pareto front and determine optimal design parameters: 47.3° angle, 1.961 mm length, 0.338 mm width, and 0.6 mm depth. Under these conditions, power density, hydrogen distribution uniformity, and membrane current density uniformity are improved by 1.02%, 1.23%, and 2.28%, respectively. The optimized flow field significantly improves mass transport and enhances operational stability across varying conditions. This approach demonstrates strong generalizability, offering a robust framework for the optimization of flow fields in PEMFCs and other multiphase flow systems.