Degradation prediction of PEM fuel cell stack based on multi-physical aging model with particle filter approach

In this paper, a novel prediction approach for proton exchange membrane fuel cell (PEMFC) performance degradation is proposed based on a multi-physical aging model with particle filter approach. The proposed multi-physical aging model uses aging coefficients to describe fuel cell different physical aging phenomena over time, including membrane conductivity losses, reduction of reactants mass transfer and reaction activity losses. In order to accurately model the activation loss, the implicit Butler-Volmer equation is used. The initial values of the aging parameters are tuned by fitting the fuel cell polarization curve at the beginning of life. Based on the initialized aging model, the first step of prediction approach is to estimate all the aging parameters using Bayesian Monte Carlo-based Particle Filter (PF) during the learning phase of experimental aging test. The suitable fitting curve function is then selected to satisfy the degradation behavior of each trained aging parameter, and further provide the extrapolated values of aging parameters in the validation phase. By applying these extrapolated aging parameters into aging model, the prediction result of fuel cell output voltage in the validation phase can be obtained. The results demonstrate that the proposed approach have good prediction performance for fuel cell degradation. In addition, each obtained aging parameters provides an insight into the different degree of physical aging process over time during the fuel cell operating, which is important to understand degradation mechanisms.