Research On Model Parameter Identification And State Estimation Method Of Power Lithium Battery For Electric Vehicle

Nowadays,the production of electric vehicles in the global automotive market continues to rise,and advantages have also begun to appear in the competition with the traditional fuel vehicle market.Power lithium battery is widely used in electric vehicles due to their advantages in energy density,cycle life and power density.However,improper management of power lithium battery can cause safety accidents such as fires and explosions,which will discourage consumers' desire to buy and is not conducive to the development of the electric vehicle industry.The State of Charge(SOC)and State of Health(SOH)estimation of lithium battery is of great significance to achieve the effective management of the battery and improve the vehicle performance,which helps to improve the competitive advantage of electric vehicles in the market.The lithium battery equivalent circuit model is the basis for SOC estimation.The accuracy of model parameters is directly related to the accuracy of SOC estimation.There will always be noise errors in current and voltage measurement data during battery operation,and the accuracy of measurement data will affect model parameter identification.However,the existing equivalent circuit model and least-squares identification methods ignore the influence of current and voltage measurement noise on the model parameter identification,which leads to a decrease in the accuracy of parameter identification.In addition to relying on accurate model parameters for lithium battery SOC estimation,the estimation accuracy in practical applications is also be affected by other factors.Voltage sensor drift errors and inaccurate matching of system noise information will interfere with SOC estimation,and the unscented particle filter method has the problems of large amount of calculation and poor numerical stability,which together lead to poor SOC estimation accuracy,robustness and real-time performance.The SOH of lithium battery reflects the attenuation of the battery during usage.The battery capacity is an important indicator of SOH.Based on the mathematical analysis method,the actual battery capacity is obtained by the ratio between the SOC change and the accumulated current over a period of time.The structure is simple.And it is easy to be estimated in real time online,but the error of battery SOC estimation is inevitable,which will cause deviation in capacity estimation and reduce the accuracy of capacity estimation.In view of the above-mentioned problems and gaps in the model parameter identification and state estimation of the electric vehicle power lithium battery,this paper has carried out a lot of related research work,and the main research contents are as follows:Firstly,the online weighted identification method of lithium battery model parameters considering current and voltage measurement noise is studied.Aiming at the problem that the existing equivalent circuit model and least squares identification method have poor parameter estimation performance under the influence of inaccurate current and voltage measurement data.Considering the impact of current and voltage measurement noise error during the operation of the lithium battery,combined with the equivalent the voltage equation of the circuit model,the covariance information of the model is constructed.The covariance matrix can determine the corresponding weight of each measurement data,and fully consider the impact of each measurement data on the model parameters to establish a weighted identification model of lithium battery parameters.Then,the weighted recursive least squares method is used in the battery parameter identification model to compensate for the influence of measurement noise.In addition,in order to improve the tracking ability of the weighted least squares algorithm on the time-varying parameters of the model,an adaptive weighted recursive least squares is proposed,which introduces the strong tracking filtering theory,the adaptive fading factor is obtained according to the orthogonality principle,and the gain matrix is adjusted by adjusting the state variance,so as to keep track of the time-varying parameters of the battery model.Compared with the existing least squares method estimation algorithm,it improves both the estimation accuracy and tracking performance of battery model parameters.Secondly,the lithium battery SOC estimation method based on the noise adaptive square root spherical unscented particle filter is studied.In view of the influence of inaccurate statistics of system noise and voltage sensor drift error on SOC estimation accuracy during battery operation,the maximum posteriori theory is used to estimate and correct the noise mean value and covariance information of the model in real time,which effectively suppresses the influence of voltage sensor parameter drift and system noise inaccuracies on SOC estimation.In addition,in view of the large computational complexity and poor numerical stability of the existing unscented particle filter,the spherical unscented transform sampling method is used to reduce the number of sampling points involved in the calculation,which will reduce the calculation cost,and increase the real-time performance of the algorithm.Then,the state covariance in the state space is propagated in the form of square root during the estimation process,which solves the estimation divergence problem caused by the negative definite covariance matrix caused by factors such as calculation error and noise,ensures the numerical stability of the algorithm,and improves the accuracy and reliability of tracking.The SOC estimation based on the proposed algorithm is used to verify the accuracy and robustness under different operating conditions.Thirdly,the capacity estimation method of lithium batteries based on Rayleigh quotient is studied.The battery capacity is selected as an important characterization parameter of SOH.Considering the influence of current cumulative error and SOC estimation error,a capacity model that can be estimated online is established.The capacity estimation problem is converted into a total least square problem,then a method based on Rayleigh quotient recursive total least squares with variable forgetting factor is proposed to obtain an unbiased estimation of battery capacity by minimizing the Rayleigh quotient cost function,and adaptively changes the forgetting factor to enhance the estimation accuracy and dynamic tracking performance of capacity,and the accuracy of capacity estimation is verified under various working conditions.To sum up,this paper conducts in-depth research on three aspects of battery equivalent circuit model parameter identification method,SOC estimation and capacity estimation.A lithium battery equivalent circuit parameter weighted identification model and an adaptive weighted recursive least square method are proposed to solve the problem of reduced accuracy of parameter identification caused by current and voltage measurement noise;the SOC estimation method based on noise adaptive square root spherical unscented particle filter is proposed to solve the problem of voltage sensor drift error and system noise information mismatch simultaneously,which also can reduce the calculation cost and increase the numerical stability,and improve the accuracy,calculation efficiency and robustness of SOC estimation.A method for capacity estimation of lithium battery based on Rayleigh quotient is proposed to solve reduced estimation accuracy caused by the current cumulative error and SOC estimation error.And the variable forgetting factor is used to increase the tracking estimation performance of the time-varying capacity.The research results obtained in this paper enrich the research content of battery management system and provide a theoretical basis for the efficient,reliable and safe operation of lithium batteries for electric vehicle.