Volume 18: Sustainable Energy Solutions for a Post-COVID Recovery towards a Better Future: Part I

Data-driven prognostics for proton exchange membrane fuel cell degradation by deep learning method Songyang Li, Weiling Luan, Chang Wang

Abstract

The durability of Proton exchange membrane fuel cell (PEMFC) is one of the technical challenges restricting its commercial application. In order to enhance the reliability and durability of PEMFC, a feature extraction method based on bi-direction long short-term memory (Bi-LSTM) and bi-direction gated recurrent unit (Bi-GRU) is proposed in this paper, which can effectively extract deeper degradation features. Feature extraction model linked with echo state network (ESN), which form a fusion prognostic framework to realize short-term degradation prediction and remaining useful life (RUL) estimation. For short-term prediction, only the first 200 hours of voltage degradation data were used for training can achieve an acceptable and accurate prediction, which the root mean square error (RMSE), mean absolute error (MAE) and coefficient of determination (R2) are 0.0235, 0.0195 and 0.9822, respectively. Comparing with traditional machine learning methods, proposed fusion prognostic framework shows the best predictive performance. Besides, a 100-step sliding windows method based on the fusion prognostic framework is used to implement RUL estimation. The results show that the percentage error (𝐸𝑟) is only 1.22% with the first 200 hours training data. The proposed method has great significance for guiding online testing and health management of PEMFC.

Keywords PEMFC, Prognostic, Remaining useful life, Bi_x0002_LSTM-GRU, Deep learning