Data-driven proton exchange membrane fuel cell degradation predication through deep learning method

Proton exchange membrane fuel cells (PEMFCs) is one of the principal candidates to take part of the worldwide future clean and renewable energy solution. However, fuel cells are vulnerable to the impurities of hydrogen and operating conditions, which could cause the degradation of output performance over time during operation. Thus, the prediction of the performance degradation draws attention lately and is critical for the reliability of the fuel cell system. In this work, we propose an innovative fuel cell degradation prediction method using Grid Long Short-Term Memory (G-LSTM) recurrent neutral network (RNN). Long short-term memory cell can effectively avoid the gradient exploding and vanishing problem compared with conventional neutral network architecture, which makes it suitable for the prediction problem for long period. By paralleling and combining the cells, Grid long short-term memory cell architecture can further optimize the prediction accuracy of the fuel cell performance degradation. The proposed prediction model is experimentally validated by three different types of PEMFC: 1.2 kW Ballard Nexa fuel cells, 1 kW Proton Motor fuel cells and 25 kW Proton Motor fuel cells. The results indicate that the proposed Grid long short-term memory network can predict the fuel cell degradation in a precise way. The proposed deep learning approach can be efficiently applied to predict the lifetime of fuel cell in transportation applications.