A Hybrid Health Prognostics Method for Proton Exchange Membrane Fuel Cells With Internal Health Recovery

Existing health prognostics methods often omit the internal health recovery of proton exchange membrane fuel cells (PEMFCs), although this phenomenon commonly exists, especially in the long-term usage of PEMFCs for hydrogen fuel cell vehicles. To this end, a novel hybrid method for PEMFCs is proposed, and internal recovery effects and external health data are collaboratively leveraged to achieve high-accuracy health prognostics. Aiming at characterizing health degradation in detail, the health prognostics of PEMFCs is addressed as voltage prediction with recovery identification, trend prediction, and fluctuation prediction. Notably, the internal impedance extracted from electrochemical impedance spectroscopy (EIS) is used to identify the internal recovery effects. This model-based recovery identification is further incorporated with particle filter for the trend prediction and with random forest regression for the fluctuation prediction by using external health data. Equipped with this hybrid method, simultaneous long- and short-term health assessment and prognostics are realized. Durability test data of two PEMFCs are used to demonstrate the proposed method. The RMSE of the proposed method can reach 0.0090, 0.0088, and 0.0094 for the long-term predictions at 550, 600, and 650 h, respectively, which are smaller than conventional model-based, data-driven, and extended Kalman filter (EKF)-long shortterm memory (LSTM) hybrid methods.

Automated summary

This study proposes a novel hybrid method for high-accuracy health prognostics of PEMFCs by leveraging internal recovery effects and external health data. The health degradation of PEMFCs is characterized through voltage prediction, trend prediction, and fluctuation prediction. Experimental results demonstrate that the proposed method achieves accurate long-term predictions with small errors.