Control of a PEM fuel cell based on maximum power tracking using radial basis function neural networks

Abstract

This article presents the proposal of a two-level control approach for a type of commercial PEM fuel cell. Thus, in the external control level a model based on neural networks of the FC is used together with a tracking algorithm to follow the maximum efficiency points as a function of the oxygen excess and in the internal level, a PI control strategy is used to guarantee the compressor motor voltage that satisfies the oxygen excess ratio demanded. The neural model of the FC response is developed through the steady-state FC response provided by the physical modelling using a multimodel approach. This approach allows a good relation between the computational cost of the training and the performance that the network offers. The performance of the global controller and the tracking algorithm are evaluated for variable load conditions by simulations and conclusions are drawn.