Adaptive feed-forward and feedback control for oxygen ratio in fuel cell stacks

Abstract

Automatic control of fuel cell stacks (FCS) using non-adaptive and adaptive radial basis function (RBF) neural network methods are investigated in this paper. The neural network RBF inverse model is used to estimate the compressor voltage for fuel cell stack control at different current demands, reduction in the compressor gain (30% and 20%) and manifold leak (15%) in order to prevent the oxygen starvation. A PID controller is used in the feedback to adjust the difference between the requested and the actual oxygen ratio by compensating the neural network inverse model output. This method is designed and conducted in three stages, starting with the collection of data from the available fuel cell stack model and finished with the non adaptive and adaptive RBF neural network control. RBF neural networks with the K-means and P-nearest Neighbour's training algorithms are used for the investigation. Furthermore, the RBF inverse model is made adaptive to cope with the significant parameter uncertainty, disturbances and environment changes. Simulation results show the effectiveness of the adaptive control strategy.