Residual Life Prediction Of Pure Electric Vehicle Li-ion Battery

With the increasingly depletion of energy and the continuous deterioration of environment,electric vehicles have shown more and more prominent advantages on energy saving and environmental protection.Battery technology is not only the key of the development of electric vehicles,but also the intersection between power industry and the automotive industry.Recent researches on electric vehicle Li-ion battery management technology have been focused on the reliability of battery,in which the study of Remaining Useful Life(RUL)has attracted much attention and has developed into one of the hot issues.Exact prediction of the remaining life of Liion batteries is of great significance to improve the safety and reliability of electric vehicles.According to the category of current Li-ion battery used in the electric vehicle,Li-ion battery is used in this subject.In this paper,NASA's 18650 Li-ion battery experiment data was analyzed,and the characteristics related to battery degradation were extracted.These characteristics were fused to obtain the health factor which could reflect the residual life.Relevance Vector Machine(RVM)algorithm for life prediction was used.In the past,many studies have extracted only one single value as the health factor.While in this paper,multiple feature parameters are extracted and the dimensionality reduction is used as a health factor for life prediction.Consequently more comprehensive and accurate data is provided.The main content of this article is as follows:(1)Research on Li-ion battery characteristicsNASA's 18650 Li-ion battery data is described and the feature acquisition method for Li-ion batteries is described in detail.(2)Extraction of Li-ion battery performance characteristics and parametersIn view of the problem that the degradation parameters of Li-ion battery capacity or impedance are not easy to measure under the practical application conditions,the characteristic parameters related to the voltage that can be detected on-line are extracted.The correlation of these parameters is analyzed,aiming to find the first three characteristic parameters with high correlation.Thus model of Li-ion health factor is built with these online detected parameters.(3)Health factor buildingWith the fusion of multiple information and vector data description algorithm,health factor is finally obtained with the three extracted parameters,which can directly reflect the battery's remaining life.(4)Li-ion battery remaining life predictionThe obtained health factor sequence is used as the forecast input.Thus the residual life of the Li-ion ion battery is predicted by the correlation vector machine algorithm.Simultaneously the prediction result is expressed by the interval range.Through the evaluation of the satisfaction of the forecast results,the applicability and accuracy of the whole forecasting method are described,and the prediction of remaining battery life is realized.