Degradation Characteristics Prediction for Vehicle PEM Fuel Cell Stack Using a Fusion Prognostic Approach

In order to more accurately predict the degradation characteristic and estimate the state of health of PEM fuel cell stacks (FCS), a fusion model based on LSTM neural network and degradation rate model of driving cycles is proposed to predict the durability decay of vehicle fuel cell stack. Firstly, the accelerated durability tests of the vehicle fuel cell stack were conducted to obtain degradation rates under different driving cycles, and then the durability experiment of 2000 h NEDC driving cycle is used to train and verify the proposed fusion model in this paper. The results show that the proposed fusion model can not only predict the overall trend of performance decay for PEM fuel cell, but also predict the characteristics of local regional performance fluctuation and recovery of fuel cell. The proposed fusion model has obvious advantages in both short-term and long-term prediction, the RMSE value can be reduced significantly and the dynamic characteristic of fuel cell stack can be predicted more accurately.