On-line identification of fuel cell model with variable neural network

It is important to predict fuel cells´ behaviors for fuel cell control, power management and other practical applications. In this paper, a Gaussian radial basis function (GRBF) variable neural network is used to on-line identify the PEM (Polymer Electrolyte Membrane) fuel cell model. The structure of the neural network changes over time according to the required accuracy and complexity. Finally, a real test data of fuel cell power system is used to illustrate the effectiveness of the variable neural network for online identification of the fuel cell model. The result shows that this method guarantees the output of the predictive model attains the required accuracy.