Optimization of baffle and tapering integration in the PEM fuel cell flow field employing artificial intelligence

Abstract

In this study, two surrogate models are developed to investigate and enhance the performance of parallel flow field Proton Exchange Membrane Fuel Cells (PEMFCs) with two modifications. The main channels are tapered, and baffles are inserted to enhance the mass transfer. A data set is generated by the multi-phase, three-dimensional CFD model. Two different data-driven models based on Artificial Intelligence (AI) and Modified Response Surface Methodology (MRSM) provide surrogate models. Subsequently, two multi-objective optimization methods are employed to reveal the optimum case. The results show that the AI model achieves superior accuracy, whereas the MRSM model has greater simplicity. An increased number of baffles and an optimal tapering ratio contribute to higher mass flux and improved uniformity in reactant distribution. The insertion of baffles and the tapering of main channels result in a remarkable 67 % increase in output power density. Optimum tapering also reduces pressure drop, whereas baffles, while improving performance, contribute to increased pressure drop and potential PEMFC degradation. The identified optimum configuration of baffles and tapering, along with the optimum values for voltage, pressure, and anode and cathode stoichiometries, results in an impressive output power density of 0.93 W cm−2 and a parasitic power ratio of 0.17, respectively.