Symbolic deep learning based prognostics for dynamic operating proton exchange membrane fuel cells

Chu Wang (1) , Zhongliang Li (1) , Rachid Outbib (1) , Manfeng Dou (2) , Dongdong Zhao (2)

1 PECASE - Pronostic-Diagnostic Et CommAnde : Santé et Energie
2 NPU - Northwestern Polytechnical University [Xi'an]

Health indicator of fuel cell under dynamic operating conditions is extracted. • Extracted health indicator has clear trend and can also indicate various faults. • Symbolic-based deep learning captures historical trend and retains it in prediction. • Hybrid approach provides wide prognostic horizon and credible lifetime estimation.

Mots clés: Degenerative behavior model; Symbolic-based long short-term memory networks; Proton exchange membrane fuel cell; Dynamic operating conditions; Prognostic horizon

Domaines: Energie électrique; Automatique / Robotique

Chu Wang, Zhongliang Li, Rachid Outbib, Manfeng Dou, Dongdong Zhao. Symbolic deep learning based prognostics for dynamic operating proton exchange membrane fuel cells. Applied Energy, 2022, 305, pp.117918. ⟨10.1016/j.apenergy.2021.117918⟩. ⟨hal-03362764⟩