Research On Health State Evaluation Algorithm Of Power Battery

With the rapid development of the automotive industry,the problem of automobile exhaust pollution has attracted wide attention.Due to the advantages of high energy density,low energy self-discharge effect and long cycle life,lithium-ion batteries are the most widely used batteries in electric vehicles.The estimation of battery health can make control strategies for the battery management system and effectively avoid the safety risks caused by excessive use of batteries.However,how to accurately estimate the state of health(SOH)of battery has always been the main problem of the battery management system.The thesis takes lithium-ion power battery as the research object,and researches the method of estimating the state of health of electric vehicle lithium-ion power battery.The main research work is as follows:(1)The paper selects capacity as a direct parameter to characterize the SOH of the battery.Aiming at the feature that the capacity is difficult to obtain directly during battery operation,a method for estimating the SOH of the lithium-ion battery based on the charging voltage sequence is proposed.Considering the fluctuation of the charging voltage rise process,in order to reduce the interference of voltage fluctuation,the average charging voltage rise sequence(original HI)is used to characterize the capacity change.To further improve the linearity of HI and actual capacity,Box-Cox transformation is used to optimize the original health indicators,and the optimized data is used to estimate the battery capacity.Through comprehensive analysis,the newly established health indicators have a high accuracy in estimating the capacity of lithiumion batteries,which verifies the effectiveness of the method.(2)Considering the limitation of single health index extraction,the paper attempts to propose a multi-index fusion least squares support vector machine to evaluate the battery health status based on a single HI framework.The lithium ion battery charging data of NASA Ames database was used to extract features,and gray correlation analysis was used to measure the correlation between each extraction amount and capacity,and a more general multi-operating battery SOH estimation model was established.(3)Considering the importance of kernel function selection,the local kernel function has a strong learning ability in the local range,and the generalization ability is weak.The global kernel function has good generalization ability and poor local learning ability.Therefore,the LSSVM model based on the hybrid kernel function is established to evaluate the health status of the battery.Comparative analysis of standard SVM proves that the battery SOH online estimation of the general model of least squares support vector machine can be established under complex conditions,and has high accuracy and adaptability.