Remaining Useful Life Prediction of Proton Exchange Membrane Fuel Cell Based on Deep Learning

Abstract

Proton Exchange Membrane Fuel Cell (PEMFC) undertakes limitations such as insufficient stability and short service life. Hence, it is imperative to predict the remaining service life (RUL) of PEMFC accurately, which is closely related to avoiding accident risks, detecting failures, and maximizing profits. In this paper, a novel deep learning algorithm is proposed for the RUL prediction of PEMFC, which comprises bi-directional long-short-term memory recurrent neural network (Bi-LSTM-RNN), attention mechanism, and deep neural network (DNN). Furthermore, the correlation coefficient analysis method is adopted to determine the stack voltage as the aging index. The proposed algorithm is compared with five other machine learning algorithms on a 1kW PEMFC stack data set provided by FCLAB. Experimental results demonstrate that the proposed algorithm has a significant improvement in prediction accuracy with mean absolute error (MAE), mean squared error (MSE) and root mean squared error (RMSE) on test set are 0.0044, 0.00003 and 0.0056 respectively. This indicates that the proposed RUL prediction method is effective and trustworthy. Finally, the RUL of the PEMFC stack under a constant operation condition is predicted based on the developed algorithm, and the result shows the RUL of this stack is 9467.8 hours.