Nonlinear predictive control for oxygen supply of a fuel cell system

This paper presents a neural network predictive control strategy to optimize oxygen supply for a proton exchange membrane fuel cell system. We propose using a time varying and local linearization auto-regressive moving average with exogenous (ARMAX) to model the nonlinear system, and employing recurrent neural network to estimate coefficients of the ARMAX model. Then constrained linear model predictive algorithm is presented to optimize oxygen supply of the fuel cell system, which significantly simplifies implementation and can handle multiple constraints. Study results demonstrate that the modeling and control strategy are effective.