Dynamic neural network based parametric modeling of PEM fuel cell system for electric vehicle applications

The paper is focused on modeling and simulation of artificial intelligent technique based fuel cell driven electric vehicle system. In the first part of this paper, the reliability of the dynamic recurrent network (NARX) and radial basis function network (RBFN) for the output prediction of a PEM fuel cell system in terms of prediction indices such as performance measure (MSE value) and iteration value (number of epochs) is investigated. In the second part, an optimum network is chosen among the two proposed networks to develop a neural network based PEM fuel cell driven electric vehicle that incorporates the modeling of neural network based fuel cell, DC-DC converter system and vehicle dynamics. In this work, modified standard drive cycle (NEDC/ECE_EUDC) is used as the system primary input. The simulation result obtained from the developed model is used to predict the power availability of the vehicle and power required to propel the vehicle.