Research On Lithium-ion Battery State Estimation Method Based On Intelligent Algorithm

High capacity and long-life energy storage devices help to improve the performance and efficiency of electric vehicles.The accurate prediction of the state of charge,health state,and remaining service life of lithium-ion batteries have a direct impact on the safety and battery health of energy storage devices.However,the internal characteristics of the battery are nonlinear,and the battery performance is easily affected by working conditions,environment,aging,inconsistency,and other factors,so the battery state cannot be obtained directly.Therefore,the research on accurately predicting the battery state variables has important theoretical significance and application value for the durability,cost,and safety of the battery management system.Aiming at the difficulties of lithium-ion battery state estimation,this dissertation builds the battery test platform,systematically designs the battery test scheme,establishes the battery estimation models,puts forward relevant state estimation methods.The main research work is as follows:The state of charge(SOC)estimation algorithm of lithium-ion battery based on least squares support vector machine(LSSVM)based on gray wolf optimization algorithm is proposed.The sliding window method is used to update the input vector of the LSSVM algorithm,which weakens the interference of the data set,measurement error,and external factors on the estimation ability of the algorithm.Combined with the gray wolf algorithm,the model parameters are identified.The experimental results show that the accuracy of SOC estimation is improved under different working conditions.The SOC estimation method combining the Kalman filter and LSSVM algorithm is proposed.The discrete LSSVM algorithm is used as the observation equation in Kalman filter,which makes up for the problem that the ampere-hour integral method can not get feedback information from the estimated parameters to correct the SOC estimation results and is vulnerable to external factors such as noise;Considering that the accuracy of observation vector estimation will affect the estimation accuracy of SOC,the LSSVM based on gray wolf optimization algorithm is used as the observation equation to further improve the estimation accuracy of observation vector and SOC.The experimental results show that the estimation performance of the fusion algorithm and the accuracy of SOC estimation have been improved.The state of health(SOH)estimation method of lithium-ion battery based on double adaptive extended Kalman filter algorithm is proposed.The Thevenin model is selected as the battery model in SOH estimation,and the estimation algorithm based on SOH,ohmic internal resistance,and SOC is established.The double adaptive extended Kalman filter based on the dynamic window is used to estimate the ohmic internal resistance,and the obtained ohmic internal resistance is used to complete the SOH calculation;The influence of capacity error and open-circuit voltage fitting error on the estimation ability of the algorithm is analyzed,and its error model is introduced to improve the anti-interference ability of the algorithm and the accuracy of the estimated value.The experimental data verify the effectiveness of the proposed method.The prediction algorithm of remaining useful life(RUL)of lithium-ion battery based on Gaussian process regression(GPR)of lightning search method is proposed.The grey correlation analysis method is introduced to select the health factors that can describe the decline of lithium-ion battery capacity,and the RUL prediction is completed combined with the optimized GPR;Considering the low prediction accuracy of a single GPR algorithm and the complexity of particle filter modeling and obtaining new observation data,the prediction results of RUL are improved by integrating particle filter and GPR algorithm,using GPR as the observation equation of particle filter algorithm and combined with capacity decline model.The experimental results show that the accuracy and reliability of the proposed RUL prediction algorithm have been improved.This dissertation proposes a fusion data-driven model and improved Kalman filter algorithm solve the difficulties such as low estimation accuracy of SOC single algorithm,mutual coupling of SOH multi-parameter estimation,and lack of the mathematical model of RUL prediction.Combined with the experimental results,the proposed lithium-ion battery intelligent estimation algorithm can accurately realize SOC,SOH,and RUL estimation,which provides a reference for engineering practice.