AI-assisted design and optimization of novel asymmetric microchannel flow fields for proton exchange membrane fuel cells

Proton exchange membrane fuel cells (PEMFCs) are promising candidates for zero‑carbon power generation; however, the complexity of flow field design remains a key barrier to large-scale application. In this work, a novel asymmetric microchannel flow field (AMFF) with under-rib channels is proposed to enhance oxygen distribution and improve water removal, particularly at high current densities. To optimize the proposed structure, an AI-aided auto fast design system (AAFDS) was applied. This framework integrates a multi-objective AI optimization algorithm directly with a multi-physics simulation model, enabling fully automated and simultaneous optimization of multiple geometric parameters without the need for pre-generated datasets or surrogate model training. Nineteen parameters of the AMFF were optimized to maximize current density, improve gas distribution uniformity, and reduce pressure drop. The optimized AMFF achieved a 7.22 % increase in peak power density, and the uniformity of current density and gas distribution at the peak-power operating point improved by 20.9 % and 24.7 %, respectively. Moreover, the system completed 400 design iterations within 140 h, representing an 8.6-fold increase in efficiency compared with manual optimization. Validation through simulations and full-scale experiments confirmed the robustness of the design, especially under high current density, high temperature, and high humidity conditions. This study introduces a novel structural approach for PEMFC flow field design and demonstrates the potential of AI-assisted optimization to accelerate the development of high-performance fuel cell systems.