Scholarlink

<p>State of charge (SOC) estimation is accurate if needed, and it is crucial to the proper functioning of the battery-management-system, but model-based Kalman filters are not robust to the process- and measurement-noise covariances. In this paper, a second order RC equivalent circuit battery model is used to build a Kalman filter, and it is proposed to use the Hiking Optimization Algorithm (HOA) to tune the covariance matrix of the Kalman filter offline. The tuning problem is a root-mean-square problem of minimizing the value of the root-mean-square error of the terminal voltage with bounded covariance parameters. The paper combines all the above-mentioned battery-model derivation, discrete filter formulation, HOA search mechanism and a structured evaluation for charging and discharging profiles, convergence, residual behavior, measurement noise, parameter uncertainty, sensitivity, computational cost and a UDDS drive-cycle case. For the respective charging condition, the filter tuned by the HOA achieves a smaller terminal-voltage error metric than the GA and SFO tuned case. The present evidence thus backs up the use of offline covariance tuning, namely HOA.</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

Application of the hiking optimization algorithm for Kalman-filter covariance tuning in battery state-of-charge estimation

  • 期刊论文

  • 发表日期：

    2026年8月4日

查看全文
  • Manoranjan Kumar,
  • Shashi Bhushan Singh

摘要

State of charge (SOC) estimation is accurate if needed, and it is crucial to the proper functioning of the battery-management-system, but model-based Kalman filters are not robust to the process- and measurement-noise covariances. In this paper, a second order RC equivalent circuit battery model is used to build a Kalman filter, and it is proposed to use the Hiking Optimization Algorithm (HOA) to tune the covariance matrix of the Kalman filter offline. The tuning problem is a root-mean-square problem of minimizing the value of the root-mean-square error of the terminal voltage with bounded covariance parameters. The paper combines all the above-mentioned battery-model derivation, discrete filter formulation, HOA search mechanism and a structured evaluation for charging and discharging profiles, convergence, residual behavior, measurement noise, parameter uncertainty, sensitivity, computational cost and a UDDS drive-cycle case. For the respective charging condition, the filter tuned by the HOA achieves a smaller terminal-voltage error metric than the GA and SFO tuned case. The present evidence thus backs up the use of offline covariance tuning, namely HOA.