Model-Based Remaining Useful Life Prediction of PEMFC Using EMD and ARIMA

The remaining useful life (RUL) prediction is vital to the heath management of Proton exchange membrane fuel cell (PEMFC). However, it is challenging for RUL prediction due to the multiple complex factors such as PEM degradation, bipolar corrosion or diffusion layer collapse etc. In this paper, a model-based RUL prediction method is proposed with particle filter (PF), empirical mode decomposition (EMD) and ARIMA. First, a semi-empirical model is used as the state equation for the PF. The parameters of the semi-empirical model are determined based on the polarization curves of the PEMFC with the PSO. For the acquisition of observation values of the PF, the voltage curves of stacks are decomposed by EMD, and then ARIMA is used to predict intrinsic mode functions (IMFs). The PF makes iterative prediction from the starting point to obtain the RUL of the PEMFC. The simulation results show that the proposed method has an improved prediction accuracy with the RUL prediction errors as low as 1.54%.