SOC Estimation Of Lithium-ion Battery For Electric Vehicle Based On Degradation Model

Low battery energy density and battery aging have hampered the development of power batteries.However,accurate SOC estimation can improve battery utilization,avoid over-charging and over-discharging,and increase the mileage and safety of electric vehicles.In view of the problems of inaccurate SOC estimation after battery aging and the sensitivity of Kalman filter to noise,this paper proposes an improved Adaptive Iterative Extended Kalman Filter(I-AIEKF)algorithm based on dualpolarized equivalent circuit model and Kalman filter structure.Based on the algorithm,a SOC estimation strategy considering battery aging is proposed,and the accuracy of SOC estimation during the use of electric vehicles is analyzed and verified through related battery experiments.The main work is as follows:(1)Introduce the experimental object and experimental test platform of this study,and design the battery characteristic test program on the test platform,including static capacity calibration test,capacity characteristic test under different influencing factors,Hybrid Pulse Power Characteristic test,OCV-SOC calibration Test,dynamic working condition test and battery cycle life characteristic test under different influencing factors.According to the experimental results of different characteristics,the operating characteristics of the battery are analyzed,which lays the foundation for accurate modeling of the power battery under different working conditions and provides data support for battery SOC estimation.(2)The advantages and disadvantages of the electrochemical model,equivalent circuit model,black box model,and fractional order model are expounded.The model of this study is weighed from the perspective of model accuracy and computational complexity: the dual-polarization equivalent circuit model.Based on the equivalent circuit physical model and Kirchhoff’s laws,a mathematical model of the battery state space is constructed and the mathematical model is discretized.At the same time,a battery Simscape model and a parameter identification verification model are built in MATLAB / Simulink.Using MATLAB / Simulink’s Parameter Estimation toolbox for offline identification of battery model parameters,the error between HPPC’s voltage experimental data and battery model simulation voltage,is less than 30 mV,indicating that offline parameter identification is accurate.The battery model was identified online based on the recursive least square method.The voltage experimental data of DST and UDDS and the simulated voltage of the battery model were within 30 mV except for individual points,which verified the accuracy of the online identification results.(3)Aiming at the problem of battery capacity degradation after aging,the battery charging and discharging experiment results are analyzed in the SOC interval.When the battery working interval is 100%,the battery charge and discharge capacity is taken as the direct research object,and the battery is estimated based on the gray model(GM)method under the historical scale of 50,100,150,200 different charge and discharge cycles.The experimental results of battery cycle life show that the gray model estimation method has higher accuracy as the historical data scale increases.When the battery SOC working range is not within the range of 0% to 100%.First,extract the characteristic health factors based on the battery charge and discharge data;Second,use the empirical mode decomposition to deal with the fluctuation of the characteristic health factor sequence,and combine particle filtering and polynomial regression to estimate the health factor sequence;Then,based on the Elman neural network,the relationship between the historical health factor and the battery capacity is fitted;finally,the current battery capacity is output based on the estimated health factor and the neural network.The battery cycle life experiment results show that this indirect estimation method also has high accuracy.(4)Aiming at the case where the number of iterations in the Iterative Extended Kalman Filter(IEKF)is uncertain and the algorithm runs for a long time or even does not converge.Firstly,this paper uses voltage difference as an observer to adaptively select local iterations;Secondly,for the noise sensitivity of the Adaptive Iterative Extend Kalman Filter(AIEKF)algorithm,this paper improves the noise variance of the observation equation based on the Sigmoid function and proposes the I-AIEKF algorithm;Finally,for the source of battery SOC error estimated by the model method,this paper combines the I-AIEKF algorithm to propose a SOC estimation strategy considering battery aging.HPPC,DST,and UDDS test conditions data prove that the accuracy of the I-AIEKF algorithm is better than the EKF algorithm when the calculation time is only slightly increased.In addition,the aging data of the battery after 150 cycles and 300 cycles prove that the method of estimating battery SOC considering battery aging has the effect of improving accuracy.The research shows that compared with the EKF algorithm,the I-AIEKF algorithm has a similar time complexity,and can improve the battery SOC estimation accuracy.After the electric vehicle has been charged and discharged for a long time,the strategy of estimating the SOC estimation of battery model parameters can correct the SOC estimation accuracy,so as to timely understand the time when the power battery needs to be replaced,and improve battery utilization and safety.