Abstract:

Due to the nonlinear and time variant characteristics of Proton Exchange Membrane Fuel Cell (PEMFC), its control is complicated. Thus, a suitable model is needed for PEMFC to gain higher performance stabilization and control. In this paper, the prediction of complicated behaviour of PEMFC is investigated using Artificial Neural Networks (ANN). The averaged cell voltage is regarded as the output; the current density and the cell temperature are considered as the inputs of neural networks. The experimental data are utilized for training and testing the networks. Multilayer perceptron (MLP) with one and two hidden layers and Radial Basis Function (RBF) networks are built, optimized, and tested in MATLAB environment. In order to study the efficiency of the neural network model, a comparison of the results is made through the Support Vector Machine (SVM) model. It is shown that neural model has better and more accurate prediction results than the SVM model of fuel cell, especially in low current region of fuel cell operation. In addition, the performance prediction of PEM fuel cell neural models with noisy data is carried out in order to check the effect of noise on the optimal structure of networks as well as the robustness of neural models

Keywords: