Research On State-of-charge Estimation Method Of Vehicle Battery Pack Based On Fusion Model

In the situation of increasingly serious environmental pollution and energy shortage,energy conservation and emission reduction has become a major concern of various countries.New energy vehicles are the focus of research and development undoubtedly,the key technologies such as EIC system matching with them are also highly valued and power battery technology is one of them.State of charge(SOC)estimation of battery is one of the core functions of battery management system,the energy utilization rate and safety factor and service life of vehicle battery pack can be improved by estimating SOC accurately.However,as an implicit state of power battery,it is very difficult to estimate SOC accurately.In this paper,relying on the key special projects and taking the lithium-ion power batteries provided by the enterprise as the research object,the modeling and SOC estimation method of vehicle power batteries are studied,which providing theoretical reference and technical support for accurate SOC estimation of battery management system.The specific research content and conclusions are as follows:1.Based on LabVIEW software,the test system of lithium-ion battery is built.The battery characteristic parameters and the main influencing factors of SOC estimation such as capacity,internal resistance,capacitance,terminal voltage and electromotive force are studied.In electromotive force test,the method of combining rapid measurement with coulometric titration is adopted and the law of electromotive force changing with SOC is emphatically analyzed.2.The electrochemistry second-order RC model with one-state hysteresis is established by combining the second-order RC model with shepherd improved model,and the state space equation is derived.The parameters are identified offline and online by HPPC test and EKF algorithm,respectively.The results of model accuracy verification show that the simulation output of the model simulate and track the change trend of battery terminal voltage well in steady-state condition of constant temperature and current and dynamic condition of constant temperature pulse,the model has high prediction accuracy.3.ASRCKF algorithm is used to estimate SOC and model parameters of the power battery.The estimation performance of the algorithm is greatly improved through updating system noise adaptively and compensating estimation results.The SOC estimation comparison between the electrochemistry second-order RC model with one-state hysteresis and the second-order RC model based on ASRCKF algorithm is carried out in the steady and dynamic conditions.The results show that the estimation errors of the two models are kept within ± 1% on the whole in steady-state condition of constant temperature and current,the SOC estimation accuracy of the electrochemistry second-order RC model with one-state hysteresis is higher than that of the second-order RC model,the maximum error of SOC estimation of the second-order RC model is not more than 0.9% and the RMSE is 0.69%,the maximum error of SOC estimation of the electrochemistry second-order RC model with one-state hysteresis is less than 0.87% and the RMSE is 0.6%.The SOC estimation error is distributed in a small area around zero in DST,the RMSE values of the second-order RC model and the electrochemistry second-order RC model with one-state hysteresis are 0.9% and 0.84% respectively,the accuracy is lower than that in constant current condition of the same temperature.4.STHF algorithm is used to estimate the SOC and model parameters of the power battery.The noise distribution does not need to be assumed,the sensitivity of algorithm to the sudden changes of uncertain factors in state and model is improved by integrating the strong tracking algorithm,and the estimation results are compensated.The SOC estimation comparison of two algorithms is carried out in the steady and dynamic conditions.The results show that the estimation error is large before 1300 s in the condition of constant temperature and current,after 1300 s,the error curves converge and the algorithms tend to be stable.The maximum error and the RMSE of SOC estimation of the second-order RC model are 0.02% and 0.07% higher than those estimated by ASRCKF algorithm in the same condition.The maximum error and the RMSE of SOC estimation of the electrochemistry second-order RC model with one-state hysteresis are 0.01% and 0.06% higher than those estimated by ASRCKF algorithm in the same condition.The SOC estimation error is distributed in a small area around zero in DST,the RMSE of the second-order RC model and the electrochemistry second-order RC model with one-state hysteresis are 0.04% and 0.03% higher than those estimated by ASRCKF algorithm in the same condition,respectively.Although the estimation accuracy of ASRCKF algorithm is slightly higher than that of STHF algorithm in good condition without bias noise,STHF algorithm has better robustness and estimation effect in bad condition.The RMSE value of STHF algorithm changes less when bias current exists,it is even smaller than that of ASRCKF algorithm using the same model.The convergence speed of STHF algorithm is about 4 times of ASRCKF algorithm in these two models when SOC initial value error exists.