Optimization of catalyst layers for polymer electrolyte membrane fuel cells using heterogeneous modeling and machine learning frameworks

The cathode catalyst layer (CCL) of a polymer electrolyte membrane fuel cell (PEMFC) plays a critical role in determining the overall performance and represents a significant cost factor owing to the use of Pt particles. Optimizing the CCL composition to enhance the PEMFC performance while minimizing Pt loading is a promising approach to lowering costs. Therefore, this study proposes an integrated framework that combines a heterogeneous CCL model with machine learning and a genetic algorithm to optimize CCL composition under low-Pt-loading conditions. The optimized CCL achieves a significant reduction in Pt loading of 56.4 %, from 0.25 to 0.14 mg/cm 2 , while maintaining a high performance with only a 6.41 % decrease in the maximum power density. The results reveal that the improvement in the optimized CCL can be attributed to the reduced overvoltage caused by proton resistance in the ionomer. This framework offers a powerful tool for the cost-effective CCL design and optimization of PEMFCs. • A novel framework optimizes CCL with ML and genetic algorithms. • Sensitivity analysis quantifies the importance of each CCL component. • Optimized low-Pt CCL matches high-Pt CCL performance in fuel cells. • Identified key mechanisms enhancing optimized CCL structure performance.