A fusion prognostics strategy for fuel cells operating under dynamic conditions

Abstract

Transportation-oriented proton exchange membrane fuel cells (PEMFC) are attracting much attention, but the strong dynamic operating conditions in transportation applications limit the durability improvement and wide commercialization of fuel cells (FC). Prognostics dedicated to predicting the FC remaining useful life (RUL) can facilitate the early provision of control/maintenance programs to improve durability and reduce costs. However, credible degradation indexes for prognostics are difficult to access or observe directly from the FC operating under dynamic conditions. Moreover, the long-term prediction performance of the state-of-the-art prognostics models is often not satisfactory. This paper proposes a fusion prognostics strategy to address these challenges. Specifically, the system dynamics is identified by using an electrochemical mechanism model and the degradation indexes are extracted based on the identified model parameters. Subsequently, a reduced-dimensional symbolic representation based long short-term memory network is developed for predicting the evolution of degradation. The proposed approach is validated using the long-term accelerated stress test data of a vehicle-oriented PEMFC. The results show that the degradation mechanism model can be used to identify degradation indexes in dynamic operating conditions. Based on the prognostics model, accurate RUL prediction can further be achieved over the extracted degradation indexes.