Remaining useful life prediction of PEMFCs based on mode decomposition and hybrid method under real-world traffic conditions

Accurately predicting the remaining useful life (RUL) of proton-exchange membrane fuel cells (PEMFCs) is essential for the health management of vehicle-oriented PEMFCs. Utilizing operational data of a demonstration PEMFC city bus in Chengdu, China, an RUL prediction method for PEMFCs that combines mode decomposition and a hybrid prediction model is developed. The improved complete ensemble empirical mode decomposition with adaptive noise is used to decompose and reconstruct the data into high- and low-frequency components to cope with the voltage recovery phenomenon. The hybrid prediction model comprises two phases. In the data-driven phase, a bidirectional long short-term memory (BiLSTM) neural network is used to predict the low-frequency components, while a mixed model comprising a convolutional neural network, BiLSTM, and an attention mechanism is used to predict the high-frequency components. The dung beetle optimization algorithm is used to optimize the parameters of the mixed model. The prediction results of the data-driven phase are then used as observed values in the model-driven phase, which involves applying the adaptive unscented Kalman filter to a voltage degradation model and obtaining the final prediction. The results show that the mean absolute percentage error of the proposed method is less than 1 % under real-world traffic conditions.