Research On Model Parameter Identification And State-of-charge Estimation Methods For Power Lithium Batteries

Lithium-ion batteries have become the preferred battery technology for mobile devices,electric vehicles and renewable energy storage systems due to their excellent characteristics of high specific energy,no memory effect,fast charging capability,long life and low self-discharge rate.However,unlike traditional fuel systems,lithium batteries are a nonlinear system whose internal state is difficult to measure directly,and system parameters are easily affected by external factors such as aging.In recent years,researchers have proposed online parameter identification and state-of-charge(SOC)estimation algorithms based on the equivalent circuit model(ECM)through the working conditions such as current,voltage and temperature monitored by the battery management system(BMS).However,there are still some problems involving the ECM applicability of lithium batteries,the multi-scale effect of the electrochemical reaction process,and the numerical stability of the Kalman filter-based SOC algorithm.In response to the above problems,this paper focuses on the methods of determining the ECM structure and identifying model parameters of power lithium batteries at different time scales,as well as improving the numerical stability of nonlinear Kalman Filtering to achieve SOC estimation.The main research content and specific work of this article are as follows:(1)An online parameter identification method for lithium battery equivalent circuit model based on dual time scales is proposed,which effectively overcomes the limitations of offline SOC-OCV(Open Circuit Voltage)calibration.Based on the battery model of different time scales,the system is divided into fast and slow time scales in the time domain.By separately processing the fast and slow dynamics of the system,and using the OCV of the Nernst model as the slow time scale parameter for collaborative online identification,we can avoid aging and other factors that affect the OCV-SOC relationship and improve the identification accuracy.In addition,a higher-precision adaptive unscented Kalman Filter(AUKF)algorithm is used to estimate the state of charge SOC,and more accurate results are achieved through closed-loop correction.(2)The ECM structures established for different types of lithium battery electrochemical behaviors and dynamic characteristics are different.The distribution of relaxation time(DRT)method is used to deconvolve the lithium battery electrochemical impedance spectrum(EIS)data into each dynamic process.The time scale characteristics of the frequency-based EIS are transformed into the time domain,and the equivalent circuit model structure of the modeled battery is determined based on the characteristic peak of the time constant.At the same time,a multi-time scale parameter identification method is proposed.The battery voltage and current data are analyzed in different frequency bands through the multi-resolution analysis(MRA)method of discrete wavelet transform(DWT),thereby isolating the electrochemical processes occurring on different time scales.This decomposition allows targeted parameter identification in each frequency band,thereby solving the challenge of capturing the multifaceted electrochemical behavior of the battery.Finally,we combine the forgetting factor recursive least squares(FFRLS)and extended Kalman Filter(EKF)methods to complete the estimation of parameters at different time scales.The comparison results show that compared with the traditional FFRLS algorithm based on fixed structure ECM,the proposed method can effectively determine the ECM of the battery and more accurately identify model parameters at different scales.(3)In order to solve the numerical stability problem caused by the truncation error of fixed-point calculation in the process of realizing SOC on resource-constrained hardware,a UKF based on the normalized SigmaRho method is proposed for SOC estimation of lithium batteries.The standard deviation and variance are transferred by decomposing the covariance matrix into a correlation matrix and a standard deviation matrix,and the standard deviation is used to normalize the elements inside the scaling matrix to reduce the condition number when solving the matrix square root and Kalman gain.The test dynamic voltage response sequence of the battery is collected as the data of the state space model,and the condition number test and state SOC estimation simulation of the covariance matrix and correlation matrix under different initial conditions are completed.Experimental results show that UKF based on normalized SigmaRho is effective and superior in achieving SOC state estimation based on fixed-point operations.(4)Through the modeled ECM,model parameter identification and state estimation methods are implemented in the battery management system,including the design and testing of BMS hardware and software.First,an overall architecture scheme for distributed modularization of power lithium battery BMS based on ZYNQ is proposed,and a BMS software architecture that follows the hierarchical structure and methodology of AUTOSAR is designed.Secondly,a BMS hardware system with scalable master-slave units is constructed.By building a complete battery test environment,the parameter identification based on the 2-RC model and the SigmaRho UKF state estimation method are implemented in hardware.The test results verified the effectiveness and reliability of the designed BMS system,providing technical reference and important support for efficient management and reliable operation of power lithium batteries.