Modeling and optimization of the PEM fuel cell catalyst layer
Date
2014
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
University of Delaware
Abstract
As the demand for clean energy grows rapidly, proton exchange membrane (PEM) fuel cell technology is seen as a viable candidate for an alternative energy conversion device. Despite much research progress in the past two decades, several remaining technical challenges must be overcome for the successful commercialization of PEM fuel cells. In particular, computational models play an important role in PEM fuel cell research by decreasing the dependence on expensive and time-consuming experimental approaches to develop a cost-effective and efficient PEM fuel cell. The goal of this research study is to formulate multiscale catalyst layer (CL) modeling strategies to provide more accurate predictions for the effects of variations in the loading of platinum, carbon and ionomer, as well as porosity, within the catalyst layer. The ultimate goal is to provide guidelines for the design of an improved CL with better performance and lower cost. The well-known agglomerate approach was chosen to model the CL. First, we conducted micro and macroscale numerical studies to investigate the shortcomings of the classical agglomerate approach which lead to unrealistic predictions of the effects of catalyst loading on performance. The investigation showed that the agglomerate model predictions can be improved either by employing relations between agglomerate parameters and compositional variables, or by accounting for discrete catalyst particles to capture local diffusion losses within the agglomerate core. A novel multiscale approach was developed and employed within a 3D cathode model. The new approach incorporates local mass diffusion effects by representing the catalyst as a distribution of discrete particles in the agglomerate core. Unlike the classical approach, the discrete particle approach is shown to provide physically realistic results for the diffusion-limited region when the catalyst loading is varied. In addition, a framework is presented to incorporate numerical results from the discrete particle approach within the 3D cathode model. The agglomerate model was further improved by incorporating a sphere-packing approach to analytically quantify the agglomerate surface area as a function of CL porosity. The importance of an accurate geometric model for the effective surface area was demonstrated by investigating the effect of CL composition on performance. The model was validated against experiments and the results show that the new method provides more realistic predictions for the effects of compositional variations in contrast to the existing approach. Next, the improved agglomerate model was employed to investigate functionally-graded CL composition. In agreement with experimental observations, we found that a higher catalyst and/or ionomer loading at the membrane/CL interface improves performance especially in the ohmic loss regime. In addition, a bidirectionally-graded CL was investigated for the first time and further performance improvement was observed. Finally, the bidirectionally-graded CL constituent (Pt, C, and ionomer) distribution was optimized individually to maximize the current density under different operating regimes. The optimal distributions were found to depend significantly on the operating regime. In comparison to unidirectional grading, significantly higher performance was obtained with optimized CL compositions. In addition, higher performance improvements were obtained with two-component optimization for the joint distributions of Pt with Nafion, and C with Nafion.