To optimize the cathode catalyst layer of MEAs for polymer electrolyte fuel cells, we developed a data-driven model to predict the performance of MEAs by using the physical properties of the catalyst layer as a property variable. To use the physical properties of the catalyst layer as an input variable, feature extraction through image processing and measurement data profiles from electron microscopy images, and XRD, were automated to eliminate variation due to differences among analysts. Furthermore, a sensitivity analysis using a data-driven model extracted important features and revealed that diffusion in the interparticle voids within the catalyst layer and intra-particle diffusion of carbon aggregates were the key factors. These results are consistent with the findings of previous numerical and experimental approaches for cathode catalyst layers and demonstrate that data-driven modeling is a powerful tool for developing cathode catalyst layers for MEAs. Furthermore, the combination of data-driven modeling and genetic algorithm optimization techniques guides the optimization of material properties to improve the performance of membrane electrode assemblies. These results are expected to provide valuable insights into improving the performance of polymer electrolyte fuel cells, promoting their widespread use, and realizing CO2 recycling in a carbon-neutral and green manner.
Read full abstract