Remaining useful life prediction of vehicle-oriented PEMFCs based on seasonal trends and hybrid data-driven models under real-world traffic conditions

Accurate prediction of the remaining useful life (RUL) of proton exchange membrane fuel cells (PEMFCs) is crucial for effective health management and efficient hydrogen energy utilization in vehicle-oriented PEMFCs. This paper proposes a novel method to predict the RUL of vehicle-oriented PEMFCs using seasonal trends and a hybrid data-driven model based on real-vehicle data from a PEMFC city bus in Chengdu, China, with the relative power loss rate (RPLR) as the health indicator. First, the seasonal trend component is extracted from the preprocessed RPLR data using the Loess-based decomposition method. Then, gray relational analysis is employed to assess the correlation between RPLR and other parameters, extracting features associated with the RUL of PEMFCs. Finally, a bidirectional gated recurrent unit model optimized by the secretary bird optimization algorithm and a least squares support vector machine model are established, respectively. The optimal weighting method is applied to combine the prediction results of each model, forming a comprehensive hybrid data-driven framework for predicting the RUL of PEMFCs. The results show that the mean absolute percentage error of the proposed method is less than 0.8 %, demonstrating high accuracy and reliability in the RUL prediction of vehicle-oriented PEMFCs.