Research On SOC Estimation Of Lithium-ion Battery With Wide Temperature Range

As the power source of electric vehicle,power battery has an important effect on the power performance and safety of the whole vehicle.LiFePO₄ battery has become the main choice of power battery because of its excellent performance.The accurate estimation of the battery SOC can improve its efficiency greatly and prevent it from overcharging or overdischarging,its useful life is prolonged.The SOC of battery can not be measured directly and is greatly affected by temperature,to solve the problem of accurate estimation of SOC with wide temperature range,the following aspects are researched in this thesis.Firstly,the structure and principle of the LiFePO₄ battery are analyzed,and the voltage,capacity,temperature and cycle characteristics of the LiFePO₄ battery are quantitatively analyzed through the characteristic experiments of the APR18650battery,and the definition of SOC considering the influence of temperature is put forward according to the temperature characteristics.this thesis compares and analyzes the advantages and disadvantages of various SOC estimation methods,and selects the fusion method of Kernel Extreme Learning Machine（KELM）and Unscented Kalman Filter（UKF）to estimate the SOC of LiFePO₄ battery with wide temperature range.Secondly,the structure and learning method of KELM are analyzed,and the KELM model is established to estimate the SOC.The training data is obtained by conducting discharge experiment of DST condition at different temperatures,then the data is sampled and normalized.The input vector structure of KELM model is determined by heuristic method,and the parameters of KELM model are optimized by genetic algorithm.The KELM model is trained on MATLAB,and the feasibility of estimating battery SOC by KELM model is verified with the discharge data of US06condition and FUDS condition at-10℃to 50℃.The results show that the KELM model has a high accuracy for SOC estimation at different temperatures,and the root mean square errors are within 4%.However,there are large fluctuations of the SOC estimation value at some times,and the maximum errors are more than 10%.Finally,in order to reduce the fluctuation of the SOC estimation value and reduce the maximum error,the UKF and KELM models are combined,and the state function of UKF is established by Coulomb counting,the measurement function is established by KELM model,thus KELM-UKF model is established to estimate SOC.The results show that compared with the KELM model,the root mean square error and the maximum error of KELM-UKF model for battery SOC estimation in FUDS and US06 conditions at-10℃and 50℃are obviously reduced,which are below2.5%and 4%respectively.The estimation accuracy of SOC is improved effectively.The KELM-UKF model presented in this thesis can accurately estimate the SOC of LiFePO₄ battery in the range of-10℃to 50℃,and can provide reliable data for the battery management system.It is of a certain theoretical and practical value for the SOC estimation of lithium battery in a wide temperature range...