Remaining useful life estimation for PEMFC in dynamic operating conditions

In this paper, the topic of prognosis for proton exchange membrane fuel cell (PEMFC) is talked about. The objective is to address the residual life estimation problem in consideration of the various operating conditions and system dynamics. To achieve this, a prognosis approach is designed by learning in model space. Specifically, model identification is firstly implemented by fitting a series of LPV models using a series of signal segments. The prognosis oriented features are then extracted from the model parameters. The prognosis is finally designed using the features. The proposed approach is validated using experimental data acquired from long-term test.