Designing graded fuel cell electrodes for proton exchange membrane (PEM) fuel cells with recurrent neural network (RNN) approaches

The graded distribution of Pt loading in the catalyst layer (CL) and porosity gas diffusion (GDL) significantly affect spatial distributions electrochemical reaction mass transport rates, thus influencing cell performance durability. A sophisticated physics-based model is established to study influence GDL at cathode, with their function obeying elliptic equation along in-plane through-plane directions, on current density its uniformity a given voltage. To reduce computational time resources, an RNN algorithm-based data-driven surrogate developed assist identification relationship between design parameters objective functions. Latin hypercube sampling (LHS) method implemented for then initial data acquisition conducted training testing model. Results show that machine learning (ML) algorithm could effectively optimal functionally electrode, achieves > 97.9 % prediction accuracy less than 0.13 root mean square error (RMSE) homogeneity. Both individual variation interaction are respectively analysed. also indicate inhomogeneous improves density. On contrary, has greater impact since monotonically increases homogeneous porosity. When both simultaneously considered, homogeneity improved. However, improvement (increases by 54 %) sacrifices maximum (reduces 22 %).