Minimally Invasive Characterization of Lithium Iron Phosphate Battery Electrochemical and Health Models using Fisher Information and Optimal Experimental Design.

Battery;Identification;Experiment;Mechanical Engineering;Engineering;Mechanical Engineering

Forman, Joel C.Peng, Huei ;

University of Michigan

关键词: Battery;    Identification;    Experiment;    Mechanical Engineering;    Engineering;    Mechanical Engineering;   

Others  :  https://deepblue.lib.umich.edu/bitstream/handle/2027.42/95950/jcforman_1.pdf?sequence=1&isAllowed=y

瑞士|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

This dissertation bridges an important gap from optimal experimental design and information theory to battery modeling and experimentation.In doing so it creates methods to efficiently model and characterize batteries intended for electrified vehicles.This modeling and characterization focuses on both estimating parameters for a pseudo 2D electrochemical model and the determination of a battery aging model.One of the major goals within the work is to be minimally invasive.For the initial parameter identification and health modeling work this means that the cells are not being disassembled.Later in the dissertation this idea is taken further by using battery health models to minimize the damage caused by experiments.Another theme throughout this work is theuse of Fisher information, which is a measure of how much information a set of data contains for estimating model parameters.This is used for both a posteriori analysis of parameter estimation accuracy and as an a priori goal for the two optimal experimental design problems within this dissertation.This dissertation has three major parts.The first is a parameter identification exercise for the Doyle-Fuller-Newman model, a pseudo 2D electrochemical battery model.In this several experiments based on Plug-in Hybrid Electric Vehicle drive cycles are conducted and the resulting data sets are used to estimate a set of 88 model parameters.This estimation is accomplished using a genetic algorithm.The input-output model results are matched very well and Fisher information is used to quantify the parameter estimation accuracy.Thisestimation demonstrates not only that the model is appropriate for LiFePO4 but also how to quickly parameterize batteries for this model.The second part of the dissertation focuses on battery health modeling.Initially it uses the previously parameterized electrochemical model to create a set of possible constant current constant voltage battery experiments.This set is then acted on by the DETMAX algorithm which creates a locally optimal subset of experimental trials.