An effective computational approach to the parametric study of the cathode catalyst layer of PEM fuel cells

We propose an integrated modeling, prediction, and analysis framework for the parametric study of the cathode catalyst layer (CCL) of PEM fuel cells. A parametric study is performed on a macro-homogeneous film model of the CCL. An artificial neural network (ANN) is then used in order to model and predict the effect of various structural parameters on the activation overpotential of the CCL. The application of the ANN approach is an asset to deal with the complexity of this problem and leads to considerably save the computational time and cost and to remove undesired computational errors. The proposed computational approach shows that an increase in the platinum mass loading causes a decrease in the activation overpotential or equivalently an increase in the CCL performance. The main effects of increasing the carbon mass loading, gas diffusion layer (GDL) volume fraction in the CCL, and CCL thickness are that the activation overpotential is going up. GDL porosity has almost no effect on the CCL performance while the CCL performance has a quadratic behavior with respect to the membrane volume fraction in the CCL. Further investigation is done in order to quantify these effects as well as the combined effects of these parameters.