Research On SOC Estimation Method Of Lithium-ion Batteries Based On Data-driven And Filtering Algorithm Fusion

In recent years,lithium-ion batteries have been widely used as the power source of electric vehicles because of their high energy density and power density,long cycle life and good safety.Accurately grasping the state-of-charge(SOC)of the power battery can help predict the vehicle’s driving range,and can effectively control the overcharge and over-discharge of the battery and delay the aging of the battery.However,most of the current mainstream algorithms only focus on improving individual SOC estimation indicators,and ignore the research on improving their comprehensive estimation performance.Therefore,how to achieve real-time high-precision SOC prediction on the basis of taking into account multiple estimation indicators is still an important research direction in today’s academia.Based on the comparative analysis of current mainstream SOC estimation methods,this thesis takes the realization of complementary advantages between estimation methods to improve the comprehensive estimation performance of the algorithm as a breakthrough point,and proposes two methods based on the fusion of data-driven and filtering algorithm to improve the comprehensive SOC estimation effect.The specific research contents of this thesis are as follows:(1)Firstly,taking the battery characteristics as the breakthrough point,analyzing its working mechanism,comparing the advantages and disadvantages of each model,reasonably modeling the battery,and laying a foundation for the subsequent battery state estimation.Then,the advantages and disadvantages of current mainstream SOC estimation algorithms are introduced and analyzed.Based on the investigation of the SOC filtering effect of extended Kalman filter(EKF)algorithm,support vector machine(SVM)algorithm and extreme gradient boosting(XGBoost)algorithm are selected to combine with EKF algorithm for SOC online real-time prediction.(2)In order to improve the comprehensive estimation of SOC of lithium-ion batteries,this thesis proposes two combined algorithms,EKF-SVM and EKF-XGBoost,which are based on the fusion of data-driven and filtering algorithms.Among them,the EKF-SVM algorithm uses the SVM model to learn and predict the SOC estimation error,so as to correct the initial prediction of EKF,thus improving the comprehensive estimation effect of SOC.The proposed EKF-XGBoost algorithm uses the XGBoost model constructed by EKF filtered data to improve the algorithm’s SOC prediction performance.(3)Finally,in order to verify the effectiveness of the proposed algorithm,this thesis investigates and analyzes the comprehensive estimation effect of the proposed algorithm model through various experimental conditions and simulation experiments with different comparison algorithms.The results show that the EKF-SVM algorithm achieves higher estimation accuracy on the basis of low complexity.The EKF-XGBoost algorithm model achieves direct high-precision prediction of lithium-ion battery SOC while taking into account model robustness,generalization and other indicators,and the comprehensive estimation effect is better.