Scholarlink

<p>Adaptive unscented Kalman filters (AUKFs) improve the accuracy of state-of-charge (SOC) estimation by dynamically updating process (Q) and measurement (R) noise covariances. However, this approach introduces several new variables and multiple matrix operations, increasing computational time and complexity, which can delay real-time updates. While conventional UKF achieves good results with finely tuned covariances, it lacks real-time covariance update which results in sluggish convergence. To address these challenges, this paper proposes an enhanced UKF (EUKF) algorithm for the SOC estimation of a lithium-ion (Li-ion) cell, where R is updated in real-time using sensor data, and Q is precisely calibrated based on test data to account for errors in the cell model. The proposed algorithm is optimized for embedded code generation and deployed on an Arduino Due board. The generated code was evaluated based on code generation time, build time, and cyclomatic complexity. The EUKF’s performance was tested at different temperatures (T) using urban dynamometer driving schedule (UDDS) and constant current constant voltage (CC-CV) profiles. Results, compared with real data from the SONY VTC6 Li-ion cell, show that even with incorrect initial SOC estimates, the accuracy remains within 0.4% for root-mean-square-error (RMSE), with convergence time less than 5&#xa0;s. The absolute maximum error and RMSE for terminal voltage were below 5&#xa0;mV and 0.3%, respectively, under both UDDS and CC-CV profiles.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Implementing an enhanced unscented Kalman filter for SOC estimation of Li-ion cell on an embedded platform

  • 期刊论文

  • 发表日期：

    2025年3月3日

查看全文
  • Amrit Raj,
  • Rohan Kumar,
  • Kunwar Aditya

摘要

Adaptive unscented Kalman filters (AUKFs) improve the accuracy of state-of-charge (SOC) estimation by dynamically updating process (Q) and measurement (R) noise covariances. However, this approach introduces several new variables and multiple matrix operations, increasing computational time and complexity, which can delay real-time updates. While conventional UKF achieves good results with finely tuned covariances, it lacks real-time covariance update which results in sluggish convergence. To address these challenges, this paper proposes an enhanced UKF (EUKF) algorithm for the SOC estimation of a lithium-ion (Li-ion) cell, where R is updated in real-time using sensor data, and Q is precisely calibrated based on test data to account for errors in the cell model. The proposed algorithm is optimized for embedded code generation and deployed on an Arduino Due board. The generated code was evaluated based on code generation time, build time, and cyclomatic complexity. The EUKF’s performance was tested at different temperatures (T) using urban dynamometer driving schedule (UDDS) and constant current constant voltage (CC-CV) profiles. Results, compared with real data from the SONY VTC6 Li-ion cell, show that even with incorrect initial SOC estimates, the accuracy remains within 0.4% for root-mean-square-error (RMSE), with convergence time less than 5 s. The absolute maximum error and RMSE for terminal voltage were below 5 mV and 0.3%, respectively, under both UDDS and CC-CV profiles.