A PEM Fuel Cell Laminar and Turbulent Models Comparison, Aiming at Identifying Small-Scale Plate Channel Phenomena: A Mesh Independent Configuration 2011-01-1177

Computational Fluid Dynamics is a powerful instrument for PEM fuel cell systems development, testing and optimization. Considering the complication due to the multiple physical phenomena involved in the cell's operations, a good understanding of the micro-scale fluidic behavior in boundary layers is recommended: pressure drop along the reactants gas channels and the cooling channels has a sensible effect on parasite load in fuel cell systems (i.e. the power absorbed by the pump supplying the gases), as well as an important role in thermal transport. A correct thermal and fluid dynamic boundary layer prediction on the channel walls and the other contact surface with porous layers requires usually a dense finite element volumes discretization near wall, especially if laminar flows occur: therefore, the boundary layer computational cost tends to be the major one.

In this paper an electro-chemical, thermal and fluid dynamic model of a PEM fuel cell portion is used to test two different CFD approaches: the standard one, using a laminar model solver, requiring a fine mesh at the wall to adequately capture the boundary layer curves, and a turbulent model, in which, despite of considering turbulent aspects, the availability of wall functions could allow to coarsen the discretization, getting anyway to a realistic prediction of the flow behaviour. The comparison of the two strategies has been performed by considering both the fluid-dynamics and the electrochemistry: the former has been investigated considering a variation of boundary layer cell numbers till reaching a mesh independence, whereas the latter has been analyzed by referring to the geometry that has been used to validate the fuel cell analytical model implemented. The whole process has been carried out through the use of a mesher (Altair Hypermesh) and a CFD software (Cd Adapco Star-CCM+) in conjunction with Mathworks MATLAB (in charge of simulating the electrochemistry).

The model used and described in this paper has been validated through a comparison with analytical results from the currently available literature (mentioned as references) and with experimental data coming from a real fuel cell test performed by the research group.

The experiments carried out have clearly shown that using a turbulent solver from a fluid dynamic point of view might be an interesting solution in order to save some computational time, while preserving a good results quality; however, from the electrochemical point of view, such a strategy revealed a major discrepancy in the polarization curves, with unreliable results both from a quantitative and a qualitative point of view.

In conclusion, the adoption of a turbulent fluid-flow solver for the solution of a laminar fluid-flow inside a PEM fuel cell cannot be considered valid in order to obtain realistic cell performances results from the numerical solution.