Optimization of polymer electrolyte membrane fuel cell flow channels using a genetic algorithm

The design of the flow channels in PEM fuel cells directly impacts the transport of reactant gases to the electrodes and affects cell performance. This paper presents results from a study to optimize the geometry of the flow channels in a PEM fuel cell. The optimization process implements a genetic algorithm to rapidly converge on the channel geometry that provides the highest net power output from the cell. In addition, this work implements a method for the automatic generation of parameterized channel domains that are evaluated for performance using a commercial computational fluid dynamics package from ANSYS. The software package includes GAMBIT as the solid modeling and meshing software, the solver FLUENT, and a PEMFC Add-on Module capable of modeling the relevant physical and electrochemical mechanisms that describe PEM fuel cell operation. The result of the optimization process is a set of optimal channel geometry values for the single-serpentine channel configuration. The performance of the optimal geometry is contrasted with a sub-optimal one by comparing contour plots of current density, oxygen and hydrogen concentration. In addition, the role of convective bypass in bringing fresh reactant to the catalyst layer is examined in detail. The convergence to the optimal geometry is confirmed by a bracketing study which compares the performance of the best individual to those of its neighbors with adjacent parameter values.

• Created a PEM fuel cell model using FLUENT and combined it with a genetic algorithm to efficiently search for the optimal flow geometry that produced the highest net power.
• Explained the reasons for the good performance of the best individual compared to the worst one by examining current density contours, and contours of oxygen and hydrogen concentrations.
• Showed the effectiveness of convective bypass in transporting reactants to the catalyst layer.
• Confirmed by means of a bracketing study that the optimal individual is indeed the best one.