Joint Estimation Of SOC And SOH For Lithium-Ion Batteries Based On Improved Unscented Kalman Filter

As the main type of new energy vehicles,electric vehicles have a good market share.As its power source,power battery is of great importance to the development of electric vehicles.Lithium-ion battery has become the main force of power battery because of its high-power density,long cycle life,no memory and many other advantages.Taking ternary lithium-ion batteries as the object,this paper studied the joint estimation of State of Charge(SOC)and State of Health(SOH)of batteries at0℃ to 45℃.Firstly,the common types of battery models and the estimation methods of SOC and SOH were analyzed.The second-order equivalent circuit model was selected as the battery model,and the accuracy of the two parameter identification methods were compared.Secondly,based on Unscented Kalman Filter(UKF),the multi-information adaptive process is added,and Extended Kalman Filter(EKF)is introduced to construct a joint estimation algorithm,which successfully estimates SOC and SOH.Finally,the scale of joint estimation is discussed,and the cumulative capacity with weight is proposed as the scale of joint estimation,and compared with the common joint estimation with time scale.The main conclusions of this paper are as follows:(1)Reading domestic and foreign literatures related to SOC and SOH estimation,summarizing and analyzing existing researches,and determining the joint estimation method in this paper.According to the test conditions,the battery selection was determined and the battery test bench was set up.The test results show that the battery capacity at 0℃ is reduced by 2.9% compared with that at 25℃.At the same time,the characteristics of the battery are different in different charging and discharging intervals.In the range of 0-0.2 SOC,the extreme value of the open-circuit voltage at different temperatures accounts for 5% of the maximum value,while in other SOC intervals,the value is about 1.2%.(2)Considering the accuracy and complexity,the second-order equivalent circuit model is selected as the battery model.Parameter identification was carried out according to battery characteristic test.The offline identification uses the formula obtained from the equivalent circuit for curve fitting,and the online identification uses the least square method with forgetting factor for identification.Comparing the model accuracy of the two identification methods,it is found that the offline identification effect is better,and the simulated voltage error is about ±0.02 V.(3)The basic principles and steps of UKF and EKF are introduced.By establishing UKF+EKF joint estimation algorithm,the battery parameters can be estimated online and the SOC estimation accuracy can be improved.On the basis of UKF+EKF joint estimation,multi-innovation and adaptive steps are introduced to update the system noise and process noise estimated by SOC,and a joint estimation algorithm combining multi-innovation unscented Kalman filter(MIAUKF)and EKF is established.The Dynamic Stress Test(DST)verifies the accuracy of the algorithm.The results show that the improved MIAUKF+EKF combined estimation can effectively estimate the battery parameters and SOH,and the average absolute error of SOC estimation is reduced by 47% compared with the UKF+EKF algorithm.The root mean square error was reduced by 33%,and the maximum absolute error was also reduced by 20%.The robustness of MIAUKF+EKF joint estimation is verified by common working temperatures,different SOC initial values and different working conditions.The results show that the algorithm has good estimation effect in various environments.(4)Multiple scales are added to the MIAUKF+EKF joint estimation algorithm,and SOC estimation is carried out in DST conditions with different time scales L as variables.The results show that the estimation accuracy does not decrease significantly with the increase of L when L is set low.Considering the updating times of SOH and the estimation accuracy,L is selected as 60.The effect of battery on capacity decline in different SOC usage range was tested.The test results show that the maximum available capacity decay caused by the battery operating in the SOC range of 0-0.2 and 0.8-1 is more obvious,which is twice that of the SOC range of0.2-0.8.According to the relationship between SOC usage interval and capacity decay,a MIAUKF+EKF joint estimation algorithm based on the used capacity with weight was established.The estimation results show that the maximum absolute error and root mean square error of the volumetric joint estimation adopted in this paper are reduced by 17% and 24%,respectively,compared with the time-scale joint estimation under the condition that the number of SOH estimates is the same.