Data-driven optimization of hybrid membrane electrode assembly configurations for enhanced proton exchange membrane fuel cell performance

• Developed hybrid membrane electrode assembly (MEA) configurations • Incorporated experiments with machine learning to guide MEA structure design • Showed much better performance for hybrid MEAs than the conventional designs • Achieved high peak power density of 1.15 W/cm 2 under low operating conditions • Reached over 0.99 R² value between the predicted and experimental results The design of membrane electrode assemblies (MEAs) is crucial to proton exchange membrane fuel cell (PEMFC) performance, and the conventional catalyst-coated membrane (CCM) and catalyst-coated substrate (CCS) configurations face limitations in meeting practical performance requirements. In this study, hybrid MEA configurations combining CCM and CCS structures are systematically investigated for optimal MEA design, considering the excellent proton transport and effective catalyst–membrane interface provided by the CCM structure, while the CCS configuration offers high porosity that facilitates reactant transport and water removal. MEAs with various CCM-to-CCS ratios are prepared with a commercial catalyst, a Nafion 211 membrane and a 45 cm 2 active area, and evaluated under various operating conditions. Experimental results indicate that hybrid MEA configurations significantly outperform pure CCM and CCS designs. At a 2:1 ratio, the MEA achieves a peak power density of 1.02 W/cm² under the flow condition of 2.00 nlpm (H₂), 4.00 nlpm (air) and 35 kPag backpressure, compared with 0.76 W/cm² and 0.69 W/cm² for pure CCM and CCS, respectively. To accelerate optimization, a data-driven machine learning approach is employed to predict the optimal CCM-to-CCS ratio, identifying a 4:1 ratio that delivers higher peak power densities of 1.04 W/cm 2 at 2.00 nlpm/4.00 nlpm/35 kPag and 1.23 W/cm 2 at 4.45 nlpm/9.00 nlpm/70 kPag. Experimental validation shows strong agreement with the model predictions, with an R² value exceeding 0.99. This study demonstrates a novel MEA configuration design with data-driven optimization, providing a practical and application-focused strategy for improved PEMFC performance.