Gradient catalyst layer design for low-Pt-loading PEM fuel cell based on artificial neural network and multi-objective optimization

The primary challenges facing Proton Exchange Membrane fuel cell include high costs and the limited durability of catalysts. Gradient catalyst layer design has become a promising approach to improve catalyst utilization, but output performance and current density uniformity should be balanced. This study investigates the gradient Pt-loading distribution along the in-plane direction of cathode catalyst layer (CCL) and its structure parameters. Results reveal that gradient Pt-loading distribution exhibits conflict with both output performance and current density uniformity. Furthermore, as the ionomer volume fraction and agglomerate radius decrease, leading to improved output performance but deteriorating uniformity of current density. In order to trade-off between the both, an artificial-neural-network rapid prediction and multi-objective optimization framework are employed to achieve rapid CCL design. Compared with uniform Pt distribution, the two sets of the optimization results indicate that although there is a slight decrease in output performance by 2.45% and 0.61%, the uniformity of current density improves by 48.7% and 31.96%, respectively, which indicates that appropriately gradients distribution with less Pt-loading under channel region and more Pt-loading under the ribs can significantly achieve better durability of the catalyst layer.