Artificial intelligence-assisted design and optimization of heterogeneous gas diffusion layers in PEMFCs

To enhance the electrochemical performance and water management of proton exchange membrane fuel cells (PEMFCs) under high-load conditions, this study proposes a zonal gas diffusion layer (GDL) design optimized through artificial intelligence (AI) algorithms. Recognizing the spatial heterogeneity of oxygen demand and water accumulation in serpentine flow fields, the conventional homogeneous GDL (HOG) is divided into four functional regions with independently adjustable porosity and spatial allocation ratios. A multi-physics coupled simulation platform is developed to evaluate the influence of zonal configurations on oxygen transport, liquid water saturation, and current density distribution. An AI-driven optimization framework is implemented to efficiently navigate the high-dimensional design space, identifying optimal structural parameters. Results indicate that the optimized heterogeneous GDL (OHEG) enhances current distribution uniformity by 15.38 %, as evidenced by a significant reduction in relative standard deviation. Additionally, oxygen distribution uniformity improves by 13.0 %, and peak power density increases by 7.63 % compared to the homogeneous design. The zonal GDL effectively mitigates local oxygen starvation and water flooding, especially in downstream regions, by ensuring more balanced reactant delivery and efficient water removal. These findings confirm the viability of zonal architecture and AI-based optimization as a powerful strategy for advancing PEMFC performance, offering a new pathway for intelligent GDL structural design with strong application potential in fuel cell engineering.