Weitao Gao

Cathode flow field optimization is crucial for improving the overall performance of proton exchange membrane fuel cells (PEMFCs), yet conventional optimization strategies are hindered by the heavy reliance on iterative experiments and computationally intensive simulations. This work developed an AI-assisted framework for multi-objective optimization of PEMFC cathode flow fields, integrating high-fidelity simulations, surrogate modeling, and optimization algorithms. The proposed framework was demonstrated by optimizing a trapezoidal baffle flow field. First, the key parameters of the three-dimensional (3D) PEMFC model were automatically calibrated using a small set of experimental data. The calibrated model was then used to generate training datasets, from which a physics-informed neural network (PINN) surrogate model was developed to rapidly predict key performance metrics, including the I-V curve, pressure drop, and current density uniformity. To enable system efficiency evaluation, a data-driven air compressor model capturing nonlinear operating characteristics was established and coupled with the PEMFC surrogate model. The flow field was subsequently optimized for improved system efficiency and current density uniformity. The results showed that the PEMFC and air compressor surrogate models exhibited high prediction accuracy, substantially reducing the computational cost of system-level optimization. The multi-objective optimization revealed a trade-off between system efficiency and current density uniformity, and Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) was employed to identify the best-compromise solution from the Pareto-optimal set over the common operating voltage range of 0.6-0.8 V. The proposed framework provides a generalizable approach for advancing PEMFC flow field optimization from single-cell performance evaluation to system-level design.