Lithium-ion battery SoC estimation based on online support vector regression

Abstract

Lithium-ion battery is a typical dynamic and nonlinear electrochemical system, and common battery model can't accurately describe its characteristics of dynamic changes, nonlinear and strong coupling. The online support vector regression machine can update the model online in real time under the limited sample, and it has the global optimal and good generalization ability. The working voltage and temperature are selected as input variables, and the state of charge is used as the output variable to train the algorithm model. The simulation results show that the online support vector regression can accurately predict the state of charge of the battery compared with the BP neural network, and has higher SoC prediction accuracy and stability.