Comparative study of a new semi-empirical model of the proton exchange membrane fuel cell for online prognostics applications

Perez, L. M., Jemei, S., Boulon, L., Ravey, A., Kandidayeni, M. et Solano, J. (2025). Comparative study of a new semi-empirical model of the proton exchange membrane fuel cell for online prognostics applications. Energy Conversion and Management, 331 . Article 119655. ISSN 0196-8904 1879-2227 DOI 10.1016/j.enconman.2025.119655

Prévisualisation

PDF Disponible sous licence Creative Commons Attribution. Télécharger (988kB) | Prévisualisation

Résumé

Abstract

The prognostic of the proton exchange membrane fuel cell is a current topic of research. Consequently, the complexity of its degradation mechanisms has led to the development of semi-empirical models to improve predictive analysis. The accurate estimation of parameters for these models is a challenging task due to their multivariate, nonlinear, and complex characteristics. This work proposes a new semi-empirical model of the proton exchange membrane fuel cell and compares it with a widely used model in the literature. Unlike other similar studies, this comparison not only focuses on minimizing the sum of squared errors in relation to the experimental data but also evaluates the variation in the solution set and the computational effort involved. For both models, the unknown parameters are estimated using the recent Pelican Optimization Algorithm. Four datasets are used to evaluate the development of the proposed model and the selected benchmark model. The first three datasets are open-access and well-recognized in academic literature, whereas the fourth dataset was obtained from a developed experimental test bench. The results show that the proposed model achieves high accuracy, with a mean absolute percentage error lower than 0.89% and the sum of squared errors below 0.9272 for all the studied scenarios. This model reduces parameter variation and decreases the relative standard deviation by over 12.7% compared to the utilized benchmark model for the first three datasets. Hence, the proposed model not only improves the precision of the estimated parameters without a notable increase in error but also reduces the computational load by at least 21.7% across all case studies.

Type de document:	Article
Mots-clés libres:	Proton exchange membrane fuel cell Parameter estimation Pelican optimization algorithm Model identification Prognostic and degradation Filter algorithm
Date de dépôt:	24 mars 2025 15:01
Dernière modification:	24 mars 2025 15:01
Version du document déposé:	Version officielle de l'éditeur
URI:	https://depot-e.uqtr.ca/id/eprint/11754

Actions (administrateurs uniquement)

Éditer la notice