Fuel Cell Ageing Prediction and Remaining Useful Life Forecasting

Fuel cell (FC) lifetime prediction is becoming an integral part of any energy management strategy (EMS) in electrified vehicles since it is one of the key barriers in the commercialization of FCs. Remaining useful life (RuL) estimation is one of the most effective predictive maintenance tools used to develop EMSs. Prognostic health management (PHM) techniques can track the degradation of a FC and predict its RuL. Hence, in this paper, a data-based PHM method based on Long Short-Term Memory Network (LSTM) is proposed to predict the RuL of FC. This manuscript presents a benchmark to compare the prediction of the FC voltage degradation with other works that have used the same dataset. The results indicate an accuracy of 88.13% with RMSE of 0.0079, and RuL is forecasted for 135 hours of operation with an accuracy of 92. 5%. Furthermore, LSTM is used to predict ageing trend of FC by considering a different current profile that the network has been trained with. This generalization has an accuracy of 79.99%. with an RMSE of 0.0351.