Support Vector Machines Used to Estimate the Battery State of Charge

Abstract

The aim of this study is to estimate the state of charge (SOC) of a high-capacity lithium iron manganese phosphate (LiFeMnPO <formula formulatype="inline" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex Notation="TeX">$_{4}$</tex></formula> ) battery cell from an experimental dataset using a support vector machine (SVM) approach. SVM is a type of learning machine based on statistical learning theory. Many applications require accurate measurement of battery SOC in order to give users an indication of available runtime. It is particularly important for electric vehicles or portable devices. In this paper, the proposed SOC estimator extracts model parameters from battery charging/discharging testing cycles, using cell current, cell voltage, and cell temperature as independent variables. Tests are carried out on a 60 Ah lithium-ion cell with the dynamic stress test cycle to set up the SVM model. The SVM SOC estimator maintains a high level of accuracy, better than 6% over all ranges of operation, whether the battery is charged/discharged at constant current or it is operating in a variable current profile.