Health State Estimation and Long-Term Durability Prediction for Vehicular PEM Fuel Cell Stacks Under Dynamic Operational Conditions

Abstract

To establish a reliable long-term estimation and prognosis for the state of health (SOH) and voltage degradation prediction of fuel cell stacks (FCSs), this article initiates a fusion prognostic strategy and a rolling prediction framework for long-term SOH estimation for FCSs based on the designed 2500-h prolonged durability experiment on vehicular FCS. Specifically, a time-varying dynamic degradation model is first developed to track the dynamic performance deterioration of FCSs based on the electrochemical mechanism and dynamic equivalent circuit model of the fuel cell. Subsequently, an improved Informer model is proposed for SOH estimation and voltage degradation prediction. The experimental results validate that the proposed model can effectively monitor the dynamic degradation behavior of the proton exchange membrane FCS, exhibiting superior accuracy in forecasting long-term voltage degradation. Moreover, the model can precisely predict the long-term aging trend and voltage periodic recovery of FCSs, with a root-mean-square error ranging from 0.33 to 1.04 V and a mean absolute percentage error below 0.5%. Finally, a rolling prediction framework for SOH estimation of FCSs, applicable to cloud-based implementation schemes, is developed to provide quantitative SOH estimation for each operational period, facilitating the development of FCS design and control strategies.