Research On Estimation Of Lithium-ion Battery SOC For Electric Vehicle Based On Multivariate Adaptive Regression Splines

Fossil energy shortage and increasingly serious environmental problems that people have to be engaged in the exploration of new energy vehicle development, the electric vehicle technology has been rapid development. Among the electric vehicle technologies, remaining capacity of power battery affecting the trip range of electric cars and the main factors of driving performance, is also one of the important and difficult of battery management system, so that the accurately estimate of remaining capacity of battery can improve the energy efficiency of the battery, prolong the service life of batteries, thus ensure better electric cars driving.SOC of battery namely the State of Charge. Accurate estimates of the power battery of electric vehicle is the key technology of electric vehicle technologies. The working principle and charge-discharge characteristics of lithium-ion batteries and the influencing factors of the battery SOC are analyzed in this paper as well as the domestic and foreign advanced algorithm. A new estimation method-multivariate adaptive regression splines(MARS) is proposed fully considered the temperature and battery aging degree and battery capacity of the battery SOC.Firstly, the purpose and significance of estimation of SOC are introduced in this paper. In terms of electric vehicles, battery management system and SOC, the importance of accurate estimation of SOC and the feasible of MARS method are interpreted at the same time.Secondly, in respect of the characteristics of lithium ion battery, the principle of lithium ion battery is expounded, the factors (temperature, the rate of charge and discharge, battery life, etc.) influencing SOC are analyzed in detail. Charge-discharge (constant current-constant voltage-constant currant) experiment was carried out at room temperature, charge-discharge characteristics were analyzed then.Thirdly, in this chapter, the several existing estimation methods of SOC and MARS based on theoretical study are systematically summarized. The composition of the MARS model structure, the composed of the basis function and the coefficient corresponding to the function, the features of forward and backward pruning process of the algorithm, and the select of the best model and the interpretability of the important characteristics are expounded.Finally, the required variables for this article are chose based on the analysis of changes of the temperature and voltage variables in the process of charging and discharging as well as the causes of these changes through charge-discharge experiment at three different rate. The MATLAB toolbox is used to program, the data (voltage, current, temperature) getting from experiment are standardized to get a dynamic data set. A model is established based on MARS to estimate battery state of charge achieving the desired effect.