State-of-Charge Estimation in an Embedded Battery Management System for Lithium-Ion Batteries

Abstract

ABSTRACT Lithium-ion batteries power a growing range of devices and vehicles, and a battery management system is needed to operate them safely and to make best use of their capacity, one of its central tasks being to estimate the state of charge, the fraction of capacity remaining, which cannot be measured directly. This study compares two methods of estimating the state of charge in an embedded battery management system, with the aim of identifying an accurate and robust approach. The state of charge was estimated during a discharge by two methods running on an embedded controller: Coulomb counting, which integrates the current drawn over time, and a model-based observer, which combines a model of the battery with measurements of voltage and current and corrects its estimate continuously, the estimates of both being compared with the true state of charge. Coulomb counting tracked the state of charge well at first but drifted away from the true value over time, as small errors in the measured current accumulated in the integration, and it could not correct an error in its starting value. The model-based observer, by contrast, corrected an incorrect initial estimate within a short time and thereafter tracked the true state of charge closely, because its continual correction against the measurements prevented drift. The observer was therefore substantially more accurate and robust, at the cost of greater complexity and the need for a battery model. The study shows that a model-based observer estimates the state of charge more reliably than Coulomb counting in an embedded battery management system, and it discusses the trade-off between accuracy and complexity on which the choice depends. 

KEYWORDS: Battery management system, State of charge, Lithium-ion battery, Coulomb counting, Kalman filter, Model-based observer, Embedded systems, Estimation, Electric vehicles