The prediction of SOC based on multiple dimensioned Support Vector Machine

Traditional method of estimating the residual energy of the battery is based on precise mathematical model which is depended on a large number of modeling assumptions and empirical parameters, so the model accuracy is limited. To improve the accuracy of SOC estimates, use multiple dimensioned Support Vector Machine to achieve the estimates of residual energy of the battery which the Scaling Kernel Function adopts the improved Levenberg-Marquardt (LM) algorithm to optimize data samples under different conditions, achieving the prediction of residual energy of a certain state of the battery during charging and discharging. Experimental results show that the proposed method can make the battery SOC estimates easily and quickly, predict accurately with high practicality.