Deep optimization of catalyst layer composition via data-driven machine learning approach 2020-01-0859
Proton exchange membrane fuel cell (PEMFC) is considered as a promising automotive powertrain by many governments and auto factories. However, the commercialization of PEMFCs is still limited by its high cost and insufficient lifetime at present. Catalyst layers (CLs) are the electrochemical reaction region of PEMFCs which cause a large proportion of the total cost, and the composition proportion of CLs significantly influences output performance of PEMFCs. However, traditional experimental and numerical methods cause the large economic and time cost and infeasibility in the CL composition proportion optimization. In this study, we propose a data-driven surrogate modeling framework to achieve the CL composition proportion optimization. A few simulation results provided by a three-dimensional full cell physical model as dataset to construct the representation between the CL composition proportion and output performance via data-driven approach. Genetic algorithm is employed to search the optimum CL composition proportion to obtain the maximum output performance.
Bowen Wang, Biao Xie, Jin Xuan, Wen Gu, Dezong Zhao, Kui Jiao
Tianjin University, Loughborough University