Remaining useful life prediction for vehicle-oriented PEMFCs based on organic gray neural network considering the influence of dual energy source synergy

Accurately predicting the remaining useful life (RUL) of the fuel cell is essential for effective prognostics and health management of vehicle-oriented fuel cells. Herein, the synergistic influences of proton-exchange membrane fuel cells (PEMFCs) and batteries as dual energy sources for a city bus are considered to develop a data-driven method for predicting the RUL of vehicle-oriented PEMFCs under real-world traffic conditions. First, gray relational analysis is applied to extract relevant features as inputs for a data-driven model. Then, the Pauta criterion and wavelet threshold method are employed to correct gross errors and denoise the data acquired from real-world scenarios. Finally, an organic gray neural network model (OGNNM), which combines four gray models and a backpropagation neural network and uses the output voltage of the PEMFC as a health indicator, is improved to incorporate the battery state of charge (SOC), hydrogen tank pressure, as well as the stack temperature, output voltage, output current, and hydrogen consumption rate of the PEMFC as inputs for the RUL prediction of PEMFCs. The improved OGNNM is validated against two datasets obtained under real-world traffic conditions, and the results demonstrate its high prediction accuracy, surpassing that achieved when the influence of the battery SOC is neglected.