Forecasting the remaining useful life of proton exchange membrane fuel cells by utilizing nonlinear autoregressive exogenous networks enhanced by genetic algorithms

The Proton Exchange Membrane Fuel Cell (PEMFC), known for its efficient energy conversion, minimal electrolyte leakage, and low operating temperature, shows great potential as a clean energy source. However, its lifespan is limited due to degradation during normal operation, which, if uncontrolled, can result in dangerous failures such as explosions. Hence, accurately estimating the remaining useful life (RUL) is vital. In this research, a combined prediction method using genetic algorithms (GA) and nonlinear autoregressive neural networks (NARX) with external inputs is proposed. The method's performance was trained and validated using the 2014 IEEE PHM Data Challenge dataset, and it was compared to two commonly used artificial neural network algorithms: GA-based backpropagation neural network (GA-BPNN) and GA-based time delay neural network (GA-TDNN). The findings demonstrate that the proposed approach surpasses the other two artificial neural network algorithms in terms of prediction accuracy. Although GA is known for its computational requirement, optimization is performed offline. Once optimal neural network (NN) hyper-parameters are determined, the optimized NN is used online for RUL prediction.