State Estimation For Ternary Lithium-ion Battery Based On Dynamic Model

With the development of new energy vehicles in recent years,ternary lithium-ion batteries are widely used because of their high energy density.The battery management system that can effectively manage the powe r battery has become the core technology of the new energy vehicle,and the estimation of state-of-charge serves as the primary work of the battery management system.This paper studies the basic characteristics of the battery,dynamic model building and charge-state estimation algorithm based on ternary lithium-ion battery.The main work of this paper is as follows:Firstly,the performance of the battery is tested and analyzed according to the structure and working principle of ternary lithium-ion battery,including the capacity characteristics under the different discharge rate and the discharge state in the previous stage,the characteristics of terminal voltage under different charging and discharging modes,the recovery characteristics of voltage at different discharge rates and different state of charge.These work laid a good foundation for the construction of subsequent models.Secondly,the improved methods of offline and online identification of battery model parameters are presented respectively.The traditional definition of ampere-time method is improved by introducing the actual capacity-discharge rate coefficient.The problem of fixed parameters in the traditional equivalent circuit model of Thevenin is analyzed,and an off-line identification method considering the influence of discharge rate,state of charge,direction of charge and discharge and other factors on the model parameters is studied and the accuracy of the model is verified under different working conditions.Furthermore,the recur sive least squares algorithm with multi-scale forgetting factor is proposed to realize the on-line identification of parameters by analyzing the inconsistent characteristics of the model parameter change rate,the multi-scale forgetting factor method is also verified to have higher output accuracy under the dynamic stress test.The characteristics of the inconsistent rate of model parameters are analyzed,the recursive least squares algorithm with multi-scale forgetting factor is used to realize the online identification of parameters,compared with the single-scale forgetting factor method,the multi-scale forgetting factor method has higher accuracy under the dynamic stress test.Thirdly,the SOC estimation algorithm based on extended Kalman filter is studied.EKF algorithm combined with the identified dynamic model is used to verify the good estimation accuracy under different discharge conditions.In order to improve the convergence speed of EKF algorithm,the strong tracking kalman filter algorithm which introduces fading factor to improve the tracking performance is adopted,the estimation performance of the algorithm and the extended kalman filter algorithm is compared and analyzed under a certain SOC initial error,the results show that the strong tracking kalman filter algorithm has better convergence performance and estimation accuracy.In order to further improve the estimation performance of the strong tracking kalman filter algorithm,multiple fading factors are introduced to better track different states,meanwhile,the better convergence speed and estimation accuracy of the strong tracking kalman filter algorithm with multiple fading factors are verified under different SOC initial errors.Finally,the particle filter based SOC estimation algorith m is compared and analyzed.Because the Kalman filter algorithm can't deal with non-Gaussian noise better,the particle filter algorithm is combined with offline and online identification methods to realize SOC estimation.The particle filter algorithm is verified that has some advantages in estimation accuracy and anti-interference ability under different working conditions,different initial errors and different noise environments.