Simulations of one and two-phase flows in porous microstructures, from tomographic images of gas diffusion layers of proton exchange membrane fuel cells

摘要

Hydrogen as an energy carrier is a promising solution for reducing emissions of greenhouse gases. Indeed, hydrogen can be used to store large amounts of energy in a completely carbon-free way. To promote the widespread use of hydrogen energy, it is essential to reduce the cost of fuel cells and increase their durability and performance. The materials in the heart of fuel cells have a strong impact on their performance and durability. In this context, opti-mizing the materials is crucial. We develop in this thesis a modeling approach of porous materials in proton exchange membrane fuel cells. We focus on a specific material that takes part in the gas diffusion layers (GDL). The gas diffusion layers are crossed by gas, electron, heat and water fluxes. To allow such multiple transports, GDL are composed of a fluid phase and a solid phase, itself consisting of several materials. The microstructure of the GDL plays an essential role on the tradeoffs between transports. To model these tradeoffs, we use X-ray tomography to image the microstructure at micrometer scales, and develop digital tools to simulate the transport on tomographic images. We validate the simulations with experimental characterizations and tomographic images of GDL. Great care has been taken in the computer performance of the numerical tools, because tomographic images in three dimensions are a challenge because of the size of the data. The first chapter of this thesis is devoted to modeling of an ex-situ water injection experiment in a GDL. We develop a pore network model extracted from tomographic images, to simulate liquid water flows in GDL in the presence of ca-pillary forces. We validate pore networks simulations using tomographic images showing the liquid water in a GDL dur-ing a water injection experiment. We show that the capillary pressure curves can be determined reliably by pore net-work simulations or full morphology simulations on tomographic images. The second chapter is devoted to one-phase transport simulations in GDL. The first part of this chapter is devoted to the development of pore networks simulations for the diffusivity and the electrical conductivities of the GDL. We de-velop a two-scale simulation methodology, which consists of decomposing the image into elements having simple shapes, and to calibrate physical models on these elements. This method considers the effect of the microstructure on the physical transfers in an economical way, reducing the computing time. We compare the pore network simulations to direct simulation on microstructures and to analytical formulas. The second part is devoted to the comparison of transport simulations with experimental measurements. We show that the transports in the fluid phase can be deter-mined reliably by direct simulations on the tomographic images, while transports in the solid phase require additional information not provided by X-ray tomography. The third chapter is devoted to modeling of the condensation of water in the GDL. The steam produced by the reaction of the hydrogen with the oxygen passes through the GDL and condenses in the cold areas of the GDL. A pore network model coupling diffusion of steam, phase change and capillary forces is developed. We study this model on virtually generated pore networks. The last chapter is devoted to the study of virtually designed microstructures. Virtually exploring new materials designs has advantages over the experimental approach, in terms of speed, cost and control over the microstructures. We show that it is possible to virtually produce microstructures close to those of real materials, to seek optimal microstructures, and control the microstructure to better study some physical effects using simulation.