Comparative Study On SOC Estimation Of Lithium-ion Batteries Based On EKF And UKF

Today,with the global energy crisis and environmental problems becoming increasingly severe,new energy vehicles represented by electric vehicles have become a trend.At the same time,battery energy storage has also become a key research direction for power grid energy storage due to its rapid response,flexible site selection,and high power density.It has bright application prospects in the fields of peak shaving,valley filling,and auxiliary frequency modulation.The battery is a key component of grid energy storage applications and electric vehicle power sources,and its safe and efficient operation relies significantly on accurate state estimation.The state of charge(SOC)is the most important and widely valued among many states,and its estimation method has gradually transitioned from simple ampere-hour integration method and open circuit voltage method to the state estimation based on the dynamic system state space model represented by Kalman filter(KF)algorithm.Among the various KF variants,the Extended Kalman Filter(EKF)algorithm and the Unscented Kalman Filter(UKF)algorithm are the two main methods for the online estimation of SOC in lithium-ion batteries.In particular,the UKF algorithm using Unscented Transform(UT)has further improvements in estimation accuracy and calculation amount,and is one of the most promising SOC estimation methods.This article focuses on lithium batteries used in power batteries and grid energy storage as research objects,summarizes their operating characteristics through charge and discharge experiments,and builds their equivalent circuit models.Furthermore,the system performance comparison analysis and calculation example verification are carried out on the SOC estimation method based on EKF algorithm and UKF algorithm.Strive to find a SOC estimation method suitable for different battery characteristics and operating conditions,and then provide a reference for the selection and effect evaluation of SOC estimation methods under different applications and operating conditions.Specific research work mainly includes:Firstly,taking lithium iron phosphate batteries and ternary lithium batteries as the experimental objects,based on the battery charge and discharge test platform to carry out charge and discharge tests under different rates and temperature conditions.According to the experimental results,the capacity characteristics,rate characteristics,temperature characteristics and voltage characteristics of the two types of batteries are summarized.The HPPC test at room temperature was designed and carried out.Charge and discharge pulses were used to stimulate the battery voltage response dynamics at different SOC points,and the battery equivalent circuit parameters were identified based on the experimental results.Taking the driving conditions of electric vehicles and the application conditions of energy storage as typical scenarios,the conversion method of the data from the operating conditions of each scenario to the battery charge and discharge current data is designed.Based on the charging and discharging platform,the working condition experiment is carried out to verify the effectiveness of the proposed conversion method.Provide basic data for comparison of SOC estimation methods under typical application conditions.Secondly,it summarizes the mathematical theory of EKF and UKF that meets the needs of state estimation of nonlinear systems,and the implementation process of SOC estimation method based on EKF algorithm and UKF algorithm.Taking the second-order RC equivalent circuit as the dynamic model of the lithium-ion battery,the EKF and UKF algorithms based on the dynamic model are written in the MATLAB environment to complete the online estimation of the battery SOC.Using HPPC experimental data and application operating condition data as input,the test compares the EKF algorithm and UKF algorithm to estimate the battery SOC convergence speed and steady-state accuracy under different operating conditions,and evaluates the adaptability of SOC estimation methods based on EKF and UKF to battery types and operating conditions.The analysis shows that the linearity of parameters such as the open circuit voltage of the ternary lithium ion battery is good,and the SOC estimation results based on EKF and UKF are both good.The lithium iron phosphate battery cell has stronger nonlinear characteristics,and the accuracy and error convergence speed of SOC estimation based on UKF are significantly better than those based on EKF.This difference is more significant under driving conditions and energy storage conditions of electric vehicles.Finally,in view of the superior performance of the UKF algorithm,the mainstream implementation method of the UT transformation of the core module of the UKF algorithm is further detailed in detail.In addition,the Sigma point set generation method and its advantages and disadvantages under four sampling strategies of symmetric sampling,minimal skew simplex sampling,hypersphere simplex sampling and proportional correction sampling are introduced in detail.Comparing the performance differences in convergence speed,accuracy and calculation speed of SOC estimation based on UKF algorithm under different UT transformation methods,the UKF algorithm with different UT transformations is programmed in MATLAB environment.And verify the theoretical analysis results with HPPC experimental data and application conditions data.Analysis shows:the accuracy of SOC estimation by UKF is higher than that of EKF,and the initial error convergence is fast,especially when the non-linear lithium iron phosphate battery and SOC are at the end;Compared with the symmetric sampling strategy,the simplex sampling strategy greatly reduces the amount of calculation,and the numerical stability of the spherical simplex sampling is better;Symmetric proportional sampling is affected by the value of the scale factor a,the effect improvement is not obvious,and the amount of calculation has increased.The effect in the SOC estimation needs further investigation.With the expansion of battery energy storage application scenarios,energy storage batteries may have more operating conditions,and multiple modes may also appear in the same application scenario.This paper mainly considers the driving conditions of electric vehicles and fluctuating smooth conditions in grid energy storage applications,and conducts related tests and performance evaluation methods.In the next step,we can compare the effects of SOC estimation methods based on different Kalman filter variants on a larger scale on this basis,and provide more complete support for the selection of SOC estimation methods in practical applications.