State-of-charge estimation of lithium-ion battery using square root spherical unscented kalman filter (Sqrt-UKFST) in nanosatellite

State of charge (SOC) estimation is an important aspect for modern battery management system. Dynamic and closed loop model-based methods such as extended Kalman filter (EKF) have been extensively used in SOC estimation. However, the EKF suffers from drawbacks such as Jacobian matrix derivation and...

全面介紹

Saved in:
書目詳細資料
Main Authors: Aung, Htet, Low, Kay Soon, Goh, Shu Ting
其他作者: School of Electrical and Electronic Engineering
格式: Article
語言:English
出版: 2015
主題:
在線閱讀:https://hdl.handle.net/10356/96192
http://hdl.handle.net/10220/38482
標簽: 添加標簽
沒有標簽, 成為第一個標記此記錄!
實物特徵
總結:State of charge (SOC) estimation is an important aspect for modern battery management system. Dynamic and closed loop model-based methods such as extended Kalman filter (EKF) have been extensively used in SOC estimation. However, the EKF suffers from drawbacks such as Jacobian matrix derivation and linearization accuracy. In this paper, a new SOC estimation method based on square root unscented Kalman filter (Sqrt-UKFST) using spherical transform with unit hyper sphere is proposed. The Sqrt-UKFST does not require the linearization for nonlinear model and uses fewer sigma points with spherical transform, which reduces the computational requirement of traditional unscented transform. The square root characteristics improves the numerical properties of state covariance. The proposed method has been experimentally validated. The results are compared with existing SOC estimation methods such as Coulomb counting, portable fuel gauge and extended Kalman filter. The proposed method has an absolute root mean square error (RMSE) of 1.42% and an absolute maximum error of 4.96%. These errors are lower than the other three methods. When compared with EKF, it represents 37% and 44% improvement in RMSE and maximum error respectively. Furthermore, the Sqrt-UKFST is less sensitive to parameter variation than EKF and it requires 32% less computational requirement than the regular UKF.