Multiple Model Predictive Control for Water Management in PEMFC Based on Recurrent Neural Network Optimization

Water management in proton exchange membrane fuel cells is a nonlinear dynamic system and hard to control. Based on recurrent neural network optimization, a multiple model predictive controller is presented to solve this problem. Moreover, the paper proposes a recurrent neural network optimization method to reduce computational burden of multiple model predictive control. Simulation results show that proposed approach can reduce fluctuation of water concentration in cathode and apply to real-time control.