Online Modeling And State Estimation Of Equivalent Circuit Model Of LiFePO₄ Power Batteries

With the substantial increase in the number of fuel vehicles,the energy crisis and environmental pollution have attracted wide attention all over the world.Electric vehicles can help to alleviate these two problems,and the major industrial countries of the world have also supported the development of the electric vehicle industry vigorously in terms of policies.Nowadays,lithium-ion power batteries have the advantages of high energy density,long cycle life,and low self-discharge rate,and have become the main battery type for electric vehicle energy supply.However,in recent years,due to the radical technical route of power batteries,unreasonable use and improper management of lithium-ion power batteries,accidents such as electric vehicle fires and explosions have occurred frequently,and their safety has attracted wide attention.Battery Management System(BMS)can provide accurate battery state estimation,which is of great significance for ensuring the safe and efficient use of batteries and promoting the development and popularization of electric vehicles.Battery modeling and state estimation are the core functions of BMS.In this paper,based on the data set obtained from the performance test experiment of LiFePO4 power battery completed by ourselves,the online modeling and state estimation of the equivalent circuit model of LiFePO4 power battery are realized.It can provide the basis of ensuring the safety and reliability of battery use,optimizing the battery energy management strategies,improving battery usage efficiency and extending battery life provide the basis.The main research work of this thesis can be summarized as follows:(1)LiFePO4 power battery performance test.There are many shortcomings of the existing public data sets of LiFePO4 power batteries,such as too old,poor performance,poor practicability,no mass-produced vehicle installation and no largescale application in the field of electric vehicle power batteries.A certain LiFePO4 power battery launched by a well-known domestic power battery manufacturer in 2018 is selected to carry out battery performance test experiments.The maximum usable capacity test,open circuit voltage test,constant power test and dynamic condition test data set are established at 10℃,25℃ and 40℃ of four LiFePO4 power batteries with different aging degrees.The effects of ambient temperature and battery aging on the test results are analyzed.(2)Online modeling of LiFePO4 power battery.LiFePO4 power battery itself is a complex electrochemical system.Based on the analysis of the internal mechanism and external behavior of the battery,the accurate description of battery behavior is realized.Aiming at the problems that the offline parameter identification method has poor adaptability to the aging state of the battery and the operating environment,the noise model of the conventional online parameter identification method is not consistent with reality and the ability of the algorithm to track model parameter changing is contradictory with the stationarity of the identification results,Adaptive Forgetting Factor Recursive Extended Least Squares(AFFRELS)is proposed and the theoretical properties and application value of it are studied.Meanwhile,the identification intelligent supervision level was designed to ensure the authenticity and validity of the identification data,ensure the normal operation of the identification algorithm,judge the convergence of the parameter identification results and check the rationality of the convergence results.The battery model which can accurately characterize the battery characteristics is established to lay the foundation for the battery state estimation.(3)State estimation of LiFePO4 power battery,including state of charge estimation,maximum available capacity estimation and state of power estimation.The effects of ambient temperature and battery aging degree on Coulomb efficiency are analyzed.Aiming at the nonlinear dynamic characteristics of the LiFePO4 power battery model,the noise is colored noise and the noise statistical characteristics are time-varying,a model parameter-state of charge adaptive estimator is proposed to update the battery model parameters online during the state of charge estimation process.The adaptive estimator improve the battery state of charge estimation accuracy and online modeling accuracy,as well as robustness.Compared with the traditional Extended Kalman Filter,the mean value of the average absolute error of SOC estimation results under the DST condition is reduced by 3.61%,the average absolute error of terminal voltage estimation results is reduced by 0.62mV.The mean value of the average absolute error of SOC estimation results under the FUDS condition is reduced by 5.22%,and the average absolute error of terminal voltage estimation results is reduced by 0.62mV.The mean value of the average absolute error of SOC estimation results in the charging process is reduced by 4.30%,and the average absolute error of terminal voltage estimation results is reduced by 0.41mV.The mean value of the average absolute error of SOC estimation results under the CLTC-P condition simulation test is reduced by 2.71%,and the average absolute error of terminal voltage estimation results is reduced by 5.28mV.The main causes of battery aging are analyzed.Aiming at the problem of estimating the maximum available capacity of batteries with noise in both input and output,Forgetting Factor Recursive Total Least Squares(FFRTLS)is proposed.In order to avoid the sudden change of the estimated value of the maximum available capacity of the battery under normal use,the capacity convergence coefficient is introduced and the model parameter-state of chargemaximum available capacity joint estimation method is proposed to achieve a more accurate battery state estimation results.Even if the initial values of SOC and maximum available capacity are both wrong,the joint estimation method can achieve accurate estimation of the terminal voltage,the SOC and the maximum available capacity as well as reasonable estimation of the polarization voltage at the same time,and has strong robustness.Compared with the estimation results of the maximum available capacity without updating,the average absolute error of SOC estimation results under the DST condition is reduced by 1.15%,under the FUDS condition is reduced by 1.18%,and under the CLTC-P condition simulation test is reduced by 0.90%.The main limiting factors that affect the state of power of the battery are analyzed.Aiming at the fact that the rationality description of the estimation method and the experimental verification are lacked of most of the existing state of power estimation research,a multi-constraint state of power estimation method of continuous charging and discharging is proposed and a state of power reference value model of continuous charging and discharging is established.Experiments are carried out from two aspects including the test point test and dynamic condition test.The influence of the duration of continuous charging and discharging on the state of power estimation results is studied.The average absolute errors of the state estimation results of the continuous charge and discharge power of the test points are 3.46W and 6.29W respectively.With the increase of the duration of the continuous charge and discharge,the absolute values of the state estimation results of the continuous charge and discharge power decrease.