Research On SOC Estimation Method Of Power Battery Based On Data-driven And Kalman Fitter Fusion

In recent years,under the background of “Dual Carbon” with emission peak and carbon neutrality,the new energy vehicle industry has been developed vigorously,and lithium battery has the advantages of high energy density and long cycle life,has been widely used in the field of electric vehicles.The State of Charge(SOC)is one of the key parameters in the battery management system of electric vehicles,which is used to represent the remaining usable capacity of the battery and thus reflect the range of electric vehicles,getting the Power Battery SOC accurately will directly affect the performance and driving safety of the whole vehicle.Due to the strong non-linear variation of the working characteristics of lithium battery,and the actual operation is easily affected by charging and discharging rate,self-discharge,temperature and aging degree.Therefore,accurate estimation of SOC is one of the key and difficult points in power battery management.Based on the comparison and analysis of the current mainstream SOC estimation methods,this paper proposes an estimation method based on data-driven and Kalman filter fusion,which improves the estimation accuracy of SOC.The main research contents are as follows:(1)Aiming at the complex and time-varying characteristics of the battery system,a suitable Thevenin model was selected as the research object through the comparison and analysis of several common equivalent circuit models,and the parameters of the model are identified by Recursive Least Square(RLS)combined with experimental data.Parameter identification results show that the parameters of Thevenin equivalent circuit model based on RLS algorithm can well characterize the dynamic characteristics of the battery.(2)Based on the Thevenin equivalent circuit model,the system state space equations for the Kalman filter algorithm are established,Extended Kalman Filter(EKF)algorithm,Unscented Kalman Filter(UKF)algorithm and Cubature Kalman Filter(CKF)algorithm are used to estimate SOC of battery,and the filtering performance of the three algorithms is analyzed.Simulation results show that the overall error index of CKF algorithm is minimum,and its performance advantage is obvious.(3)In order to improve the estimation accuracy of SOC,a datadriven and Kalman filter fusion estimation method is proposed.Based on the correlation filtering data of CKF algorithm,this paper adopts the Extreme Learning Machine(ELM)algorithm and Kernel Extreme Learning Machine(KELM)algorithm respectively to establish the SOC estimation error prediction model based on data-driven,the output of the error prediction model is used to correct the error of CKF estimation results,and two algorithms,ELM-CKF and KELM-CKF,which are data driven and Kalman filter fusion,are constructed.Finally,the simulation results under various operating conditions show that the fusion algorithm can effectively improve the SOC estimation accuracy,while KELM-CKF maps the input samples to the high-dimensional feature space because of the kernel function,the output instability of ELM-CKF is solved without setting the number of hidden layer nodes,and it has better estimation effect,stronger generalization and robustness in SOC estimation.