Neural network modeling of proton exchange membrane fuel cell

This paper proposes a neural network model of a 500-W proton exchange membrane (PEM) fuel cell. The nonlinear autoregressive moving average model of the PEM fuel cell with external inputs is developed using the recurrent neural networks. The data required to train the neural network model is generated by simulating the nonlinear state space model of the 500-W PEM fuel cell. It is shown that the two-layer neural network, with a hyperbolic tangent sigmoid function, as an activation function, in the first layer, and a pure linear function, as an activation function, in the second layer can effectively model the nonlinear dynamics of the PEM fuel cell. After model is trained and validated, it is used to analyze the dynamic behavior of the PEM fuel cell. Finally, the effect of measurement noise on the performance of the neural network model is investigated, and the results are shown.