Research On Dynaminc Modeling And Optimization Of Fuel Cell Start-up Process Based On Segmented Cell Measurement

Proton Exchange Membrane Fuel Cell（PEMFC）is considered as a significant technological approach for future power source of new energy vehicles,due to its advantages such as high power-density,high efficiency and zero emissions.Although the PEMFC technology has been improved significantly in the past decades,issues concerning the lifetime and degradation remain the major challenge which hinders the large-scale commercialization of fuel cell products.The degradation of PEMFC system is mainly related to the four kinds of operations including the load changing,start-stop,heavy loading,and idling operations.To note,the start-stop operation contributes about33% to the overall degradation of PEMFC in practical applications.For the experimental work,this thesis quantitatively studies the dynamic variations of the local internal current of fuel cell stack during the start-up process with a selfdesigned high-resolution segmented cell device.For the modelling work,a novel equivalent circuit based dyniamic model is proposed to investigate the dynamic timedelay characteristics of the internal current with optimized electrical circuit network and physical parameters indentifications.The major contents of this thesis are as follows:Firstly,a fuel cell test system as well as a high-resolution segmented current detection device are applied to measure the local current of the commercial-size fuel cell stack.The corresponding comparative experiments are designed to simulate the start-up operations of the fuel cell vehicle.The dynamic behaviors of the internal current distributions and fuel cell voltage are tested and analyzed with different operation conditions of hydrogen flow rates.Secondly,five dynamic equivalent circuit models are proposed considering the double-capacitive layer effect during the start-up process of the fuel cell.The electrical parameters of the dynamic equivalent circuit models are obtained from the experimental tests and mechanism analysis.The five models are simulated with MATLAB/Simulink software and the equivalent circuit structure is optimized according to the physical mechanism analysis and the accuracy of simulation results.Thirdly,the electrical parameters of the dynamic equivalent circuit model are iteratively estimated based on three parameter identification methods: gradient descent method,trust region reflection algorithm and Levenberg-Marquardt algortithm.Validated by the experimental results of the segmented cell detection,the key physical parameter matrix of the dynamic equivalent circuit model is iteratively optimized.The parameter identification methods are compared in terms of performance and efficiency of the simulation indicating that the L-M algorithm performs efficiently with sufficient precision for model prediction.Finally,the optimized L-M algorithm is used to identify the key parameters of the equivalent ciucuit model and a quasi two-dimensional equivalent circuit network model is designed according to the fuel cell stack design.The numerical results of the dynamic model accord well with the the two-dimensional segmented current measuremnts.The dynamic equivalent circuit model as well as the parameter identification method are benefical for the investigation of local current behavior during start-up process and optimization of the relevant control designs.