(Invited) Polymer Electrolyte Fuel Cells Lifetime Prediction By a Full Multi-Scale Modeling Approach

The estimation and increase of the lifetime of PEMFC fuel cell under dynamic conditions is one of a major challenge about PEMFC. Increasing the durability of the fuel cell must be treated by both the development of new material and design but also by optimal strategies and control of the operating conditions of the fuel cell. To validate some new strategies for lifetime and also the durability of new material, experimental tests must be coupled with modeling and numerical simulations. Indeed, the different irreversible degradation mechanisms in the MEA (Membrane Electrode Assembly) (as the platinum dissolution, Ostwald ripening, carbon support corrosion and chemical membrane degradation) and reversible degradation mechanisms (platinum oxidation, liquid water management) are strongly coupled to the local conditions into the MEA. Moreover, the local conditions into the MEA are depending of the material properties and the operating conditions of the fuel cell stack. The strong coupling of the different phenomena and the heterogeneities along the surface of the cell and through the thickness of the MEA can be solved only by numerical simulations. In this talk, to predict the degradation mechanisms along the surface of the cell and function of the dynamic operating conditions, three multi-physics models are linked together (see Figure). The three models are described below. i/ EDMOND model is a 0D double layer model to calculate the local surface potential at the surface of the catalyst as well as the coverage of the various reaction intermediates, based on a dynamic coupling between the local operating conditions and the kinetics of the various reaction steps. Both the surface potential and the coverage are involved in the mechanistic models, which makes the EDMOND framework required for such a modeling approach. ii/ The MEA model is a 2D CFD model cross-section of the MEA. The model is able to compute the differences in operation under the rib and channel of the bipolar plate. It takes into account the gas diffusion, thermal, ionic and electrical transport and electrochemical response based on Butler-Volmer approach. Anisotropy and compression of the materials are also taken account. The catalyst layer is meshed through the thickness. iii/ The fuel cell model (called PS++ code) is a 2D+1D fuel cell model, based on bond graph approach. PS++ is a dynamic multi-physic model taking into account two phases flow and heat transport equations. The model is used to calculate the local conditions along the surface of the cell function of dynamic operating conditions. Degradation mechanisms are added (by bottom-up or top-down approach) to estimate the fuel cell lifetime. The irreversible and reversible degradation mechanism are modeled in the local model and up-scaled into the cell model PS++. The local double layer model (EDMOND) is used to calculate the irreversible degradations of the loss of catalyst (Ostwald ripening). The Ostwald ripening model integrated in EDMOND relies on a multiscale mechanistic approach where parameters come from DFT calculations. It is able to calculate the dependency of both the size and the distribution of the particles, the voltage and the hydration in the degradation rate. The MEA model is used to calculate the reversible degradations. The catalyst oxidations are introduced and are in competition with the electrochemical reactions (ORR and HOR), allowing to compute dynamically the evolution of the active catalyst surface. The upscaling of the degradations and the activity of the catalyst are introduced into the cell model to simulate the loss of performance of the cell under dynamic solicitations and to calculate the effects of local degradations on the repartition of the current density along the surface of the cell. The experimental validation of the approach is realized by two 2000 hours experimental tests in a 30 cells stack. The experimental tests have been carried out in a fuel cell stack durability tests with a permanent current density for the first test and a dynamic solicitation for the second one. Current density measurement card and periodic electrochemical characterizations are realized during the tests and post-mortem characterizations at the end of the tests. The results show a good qualitative agreement with the model simulations, in particular to the current density distribution function of ageing and the reversible and irreversible mechanisms. In conclusion, the approach of lifetime prediction by a full multi-scale approach is demonstrated and open new perspectives about the coupling of modeling and experimental tests to increase the durability of the fuel cell. Figure 1