(649d) Integration of Polymer Electrolyte Membrane Fuel Cell Subcomponent Models Using Reduced Order Methodology

Smith, R. L., Carnegie Mellon University
Biegler, L. T., Carnegie Mellon University
Jhon, M. S., Carnegie Mellon University

The polymer electrolyte membrane fuel cell (PEMFC) is a green energy device that utilizes hydrogen fuel to convert chemical energy to electrical energy with a higher efficiency than the internal combustion engine producing only water and waste heat.  This device is composed of  several subcomponents on both the anode and cathode sides including the gas channel (GC) where fuel enters, the gas diffusion layer (GDL) which creates a uniform distribution of fuel, the catalyst layer (CL) where the electrochemical reaction occurs, and the PEM through which the protons are conducted.

These subcomponents contain multiphysical phenomena and can be simulated by models on various scales ranging from phenomenological/continuum models to ab initio/molecular scale descriptions.  In a multiscale/ multi-physical device such as the PEMFC, the integration between models must be performed within and across subcomponents. Integration of models within subcomponents requires an innovative methodology to coarse grain high resolution descriptions by reducing complexity while preserving the essential physics.  The reduced order method (ROM) can serve as an excellent tool to integrate models.  Linking across subcomponents requires determining states of two adjacent subcomponents such that continuity of states and fluxes at the interacting boundary are satisfied.  

We examine two different procedures for integrating models for various scenarios.  In the first application of the horizontal procedure, model accuracy is preserved by linking high resolution, complex models for two subcomponents.  However, computational expense often makes this approach intractable.  In an alternative horizontal approach, two reduced, coarse-grained models for each subcomponent can be to a reduction in complexity but also in model accuracy.  In a second procedure, model cross-linking is used to link one high resolution subcomponent model to a coarse-grained, lower resolution model with an adjacent subcomponent model.  This procedure provides a balance between accuracy and computational expense and is one of the most accessible approaches for model integration.

We demonstrate the best procedure by integrating simple GDL and CL subcomponent models with an interactive boundary via a cross-linking procedure. [1], [2]  A ROM for the CL is constructed using principal component analysis and Kriging mapping.[3]  We maximize PEMFC current density by solving the CL model with the GDL boundary as an input.  Our methodology based on simple models will be generalized from 2-D to 3-D as well as integrated with the membrane subcomponent.  This general methodology can be used for other energy areas dealing with multiphysics and multiscale phenomena.