Experimental Validation of State of Charge Estimation by Extended Kalman Filter and Modified Coulomb Counting
The operation of batteries in energy storage systems (SAE) is controlled by the battery management system (BMS). Within the scope of research related to the functions of the BMS, there is attention to the methods of estimating the state of charge (SOC) that use state estimators. Among the estimators...
Gespeichert in:
Veröffentlicht in: | Revista IEEE América Latina 2022-11, Vol.20 (11), p.2395-2403 |
---|---|
Hauptverfasser: | , , |
Format: | Artikel |
Sprache: | eng |
Schlagworte: | |
Online-Zugang: | Volltext bestellen |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Zusammenfassung: | The operation of batteries in energy storage systems (SAE) is controlled by the battery management system (BMS). Within the scope of research related to the functions of the BMS, there is attention to the methods of estimating the state of charge (SOC) that use state estimators. Among the estimators, there is the algorithm known as the extended Kalman filter (EKF). This work proposes the implementation of the EKF for SOC estimation of a lithium ion 18650 single-cell battery, with experimental validation. The algorithm is embedded in BMS composed of Arduino MEGA 2560 microcontroller and auxiliary hardware. The battery is modeled using a simple model, which aims to facilitate implementation in embedded systems. The results revealed that the SOC estimation via EKF embedded in BMS showed maximum errors around 4%, a result compatible with other references in the literature. Based on the EKF approach, an alternative method, called a modified Coulomb counting, was defined, which uses parameters calculated in the EKF to establish a adaptive Coulomb counting to the unknown initial SOC. This new method is also capable of reducing estimation fluctuations, a common feature found in the EKF implementation. The use of the modified counting proved to be useful in several cases, often reducing the maximum estimation error to values less than 1%. Finally, the use of the simple model with EKF proved to be adequate in terms of the balance between precision and simplicity. |
---|---|
ISSN: | 1548-0992 1548-0992 |
DOI: | 10.1109/TLA.2022.9904765 |