Physics-guided fault diagnosis method for proton exchange membrane fuel cells based on LSTM neural network
Year: 2024
Type: article
Abstract: In this paper, we propose a novel hybrid method for fault diagnosis of Proton Exchange Membrane Fuel Cells (PEMFCs) based on the combination of a physics-based model and a machine learning (ML) algorithm. By incorporating a physics-based model in the fault diagnosis algorithm, we can access to process variables not directly measured through sensors but related to the state of the PEMFC stack. The model estimates are subsequently combined with signals measured on the PEMFC stack and used as input of a ML algorithm. The performance of the physics-guided method is evaluated on an extensive dataset comprising a thousand hours of operation of a PEMFC stack under dynamic load profiles. The experimental results show that the method allows diagnosing the faults that occur in the system, outperforming purely data-driven approaches. Furthermore, we assessed the generalization ability of the proposed method by using data from the initial phase of stack operation for algorithm training and reserving the data from the aged stack operation for the testing phase. Several algorithms are tested and Long-Short Term Memory (LSTM) neural network demonstrates to be the most efficient, maintaining an accuracy level above 80 %, and, thus, proving its robustness to PEMFC aging.
Source: Journal of Power Sources
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
Estimated APC paid: $4,110
Funder: Agence Nationale de la Recherche
OpenAlex ID: W4404032308