Model-based method dedicated to the diagnosis of PEM fuel cells ; Méthode basée modèle pour le diagnostic de l'état de santé d'une pile à combustible PEMFC en vue de sa maintenance

Nowadays, Fuel cells (FCs) are considered as an attractive technological solution for energy storage. In addition to their high efficiency conversion to electrical energy and their high energy density, FCs are a potential candidate to reduce the environmental impact of aircrafts. The present PhD thesis can be located within this context, and especially contributes to the development of methodologies dedicated to the monitoring of the state of health (SoH) of Proton Exchange Membrane Fuel Cells (PEMFCs). FCs are submitted to ageing and various operating conditions leading to several failures or abnormal operation modes. Hence, there is a need to develop tools dedicated to the diagnosis and fuel cell ageing monitoring. One of reliable approaches used for the FC SoH monitoring is based on parametric identification of a model through experimental data. Widely used for the FC characterization, the polarization curve (V-I) and the Electrochemical Impedance Spectroscopy (EIS) coupled with a model describing the involved phenomena may provide further information about the FC SoH. Two models were thus developed: a quasi-static model whose parameters are identified from the polarization curve and a dynamic one identified from EIS data. The need to develop a dynamic model whose formulation may vary over time “without a priori” has been reported in this thesis. The original approach of this thesis is to consider conjointly both characterizations during all the proposed analysis process. This global strategy ensures the separation of the different fuel cell phenomena in the quasi-static and dynamic domains by introducing into each parametrization process (one for the quasi-static model and one for the dynamic model) parameters and/or laws stemming from the other part. The global process starting from the a priori knowledge until the identification of the models parameters was developed during the chapters of this thesis. In addition to the good reproduction of experimental data and the separation of the losses in both static ...