Lifespan prediction model for proton exchange membrane fuel cell vehicle based on time series information feature extraction and optimization

The accurate prediction of Proton Exchange Membrane Fuel Cells (PEMFCs) lifespan is crucial for ensuring vehicle safety. However, Long Short-Term Memory (LSTM) models have limitations in long-term predictions due to their data dependence. To address this, a hybrid model is proposed. For this study, a customized experimental fuel cell stack is used to mimic real vehicle performance. The stack undergoes an aging test to obtain experimental results. A Convolutional Neural Network (CNN) is then employed for data feature extraction. The extracted data is fed into an LSTM network layer. To optimize the hyperparameters of the LSTM model and accelerate convergence, the weighted mean of vectors algorithm (INFO) is introduced. The effectiveness of the proposed algorithm is validated using a dataset from the French FCLAB research center. The results show that combining the two models improves prediction accuracy by 7.42 % and 18.21 % in predicting the short-term trend of reactor voltage decline. In addition, the maximum life prediction errors of the two sets of data were reduced by 38.53 % and 52.177 % respectively when estimating the remaining useful life of fuel cell stack. This proves the effectiveness and applicability of the proposed hybrid model.