SOC Estimation Of Traction Battery Based On Unscented Kalman Filtering

SOC estimation is the most important function of the six essential functions of BMS of Electric Vehicles(EVs).The value of SOC is the key information to predict the range,and have great impacts on the safety performance of EVs.This thesis did the following research work about SOC estimation:Firstly,the theory of UKF algorithm was in-depth studied.The selection of sigma points have two different ways,which were contrasted in the thesis.It was found that selecting twice sigma points can help to converge to the testing value more quickly than selecting once sigma points,although the error of the whole prediction process of these two methods were all the same.There are two important model parameters in UKF algorithm,to get a high estimation accuracy and a rapid convergence speed,appropriate values of Q and R are necessity.The effects of these two parameters on the estimation accuracy were researched,the contrast simulation results with Simulink demonstrate that the selected values of Q and R should lie in a suitable range,they will affect the convergence rate or make the estimated value diverge.Taking these factors into account all the simulations in the thesis select Q=0.0001,R=0.001.Then,the estimation accuracy of UKF were compared under the same working condition with EKF.And the advantages of UKF were verified,it's convergence time only 1/20 of EKF method.The sensitivity of the SOC estimation model is also a important evaluation index.It was found that the UKF model can estimate the SOC rapidly and accurately whenever the discharge starting point is incorrect or the initial value of the computation SOC is incorrect.Finally,the practical application was considered.The same battery model parameters under 25? in 1C were used in different working condition like different temperature,different discharging C-rate,Dynamic Stress Testing(DST)condition,Hybrid Pulse Power Characterization(HPPC)and charging process to estimate the SOCs.At high temperature conditions,high accuracy of SOC estimation could be achieved,and the error of constant current discharging process was always less than 3%.At low temperature condition,the estimation errors increased when the temperature decreased,and the maximum error of constant current discharging process reached 80% without model modification.But it could within 6.3% when the battery model used was modified.At large C-rate condition with modification,HPPC and DST condition,the estimation error were all less than 5%.The error of charging process was less than 6%.