Multi-objective optimization of porous layers for proton exchange membrane fuel cells based on neural network surrogate model

In this paper, the impacts of three important parameters of a proton exchange membrane fuel cell (PEMFC) porous layer: the porosity of the catalytic layer (CL), the electrolyte volume fraction of the CL, and the porosity of the gas diffusion layer (GDL) on the performance of the PEMFC were studied. The electrolyte is composed of platinum catalyst and ionomer. Considering the high cost of platinum catalyst, the goal of parameter optimization of the porous layer is to improve the PEMFC output power density while reducing the electrolyte volume fraction. To achieve this goal, a 3-dimensional two-phase isothermal PEMFC model was built. Under different operating voltages and porous layer parameters, run the PEMFC physical model to obtain a set of data, use the data to train the neural network to replace the physical model, and then use the multi-objective optimization algorithm to optimize the porous layer parameters. The results show that when the operating voltage is 0.4951, the porosity of the CL is 0.2647, the electrolyte volume fraction is 0.4471, and the porosity of the GDL is 0.5043, and the overall performance is good. Compared with the original model, the optimized model improves the maximum output power density by 3.56% and reduces the electrolyte volume fraction by 10.58%.