Adaptive Prognostic of Fuel Cells by Implementing Ensemble Echo State Networks in Time-Varying Model Space

Abstract

Prognostic plays an important role in improving the reliability and durability performance of fuel cells (FCs); although it is hard to realize an adaptive prognostic because of complex degradation mechanisms and the influence of operating conditions. In this paper, an adaptive data-driven prognostic strategy is proposed for FCs operated in different conditions. To extract a feasible health indicator (HI), a series of linear parameter-varying models are identified in sliding data segments. Then, virtual steady-state stack voltage is formulated in the identified model space and considered as the HI. To enhance the adaptability of prognostic, an ensemble echo state network is then implemented, given the extracted HI data. Long-term tests on a type of low-powerscale proton-exchange membrane FC stack in different operating modes are carried out. The performance of the proposed strategy is evaluated using the experimental data.