Empirical modeling of polymer electrolyte membrane fuel cell performance using artificial neural networks

Using an artificial neural network (ANN), a technique for modeling a polymer electrolyte membrane fuel cell is proposed for providing a tool for the design and analysis of fuel cell total systems. The focus of this study is to derive a non-parametric empirical model including process variations to estimate the performance of fuel cells without extensive calculations. ANN models are trained to fit experimental data obtained in a 300 cm2 single cell in H2/air operation using Nafion 115 and Nafion 1135 membrane electrolytes. The models take into account not only the current density but also the process variations, such as the gas pressure, temperature, humidity, and utilization to cover operating processes which are important factors in determining the real performance of fuel cells. All experimental data using Nafion 115 and Nafion 1135 membranes are fitted very well with the ANN models over a wide operating range. The ANN models can be used to investigate the influence of process variables for design optimization of fuel cells, stacks, and complete fuel cell power system.