Fuel Cell prognosis using particle filter: application to the automotive sector

Prognosis and Health Management is a powerful approach in the quest to improve fuel cell durability. Prognosis is applied to quantify and predict the fuel cell state of health. By assessing the state of health, fuel cell operating conditions can be adapted to reduce its degradation rate. The objective of this paper is to present a fuel cell prognosis method. An empirical model separating membrane and catalyst degradation is adopted. The model coefficients are estimated using a particle filter. The prediction results are encouraging as the fuel cell voltage can be accurately estimated and predicted. The prognosis method presented in this paper is trained and validated using proton exchange membrane fuel cell stack data, going through a dynamic load profile. This novel experimental data has been generated at Sintef Laboratory, on an automotive fuel cell short stack provided by Symbio.