Scholarlink

<p>Lithium-ion batteries are an integral part of numerous smart energy systems. Accurate estimation of battery state of health is vital to ensure the safe and reliable usage of lithium-ion batteries. In this paper, various regression algorithm-based estimation frameworks in combination with sample entropy of battery voltage is implemented to accurately estimate the battery state of health (SOH). The sample entropy, fuzzy entropy, localized area and power spectral density values of charging voltage sequences are utilized to develop the hybrid SOH estimation model and thereby minimizing the estimation error values. The health feature variables based on battery charging attributes are validated as per grey correlation analysis to estimate the battery deterioration trends. Different regression models are compared to illustrate the effectiveness and estimation accuracy of the proposed hybrid model. The results demonstrate that the hybrid model trained on localized voltage area-sample entropy feature variables or power spectral density-sample entropy feature variables are more accurate in estimating battery SOH than other estimation models such as Lasso, and support vector regression models. Despite some batteries following an intricate nonlinear degradation path, the mean absolute error and root mean squared error values of the proposed model do not exceed 0.26% and 0.42% respectively.</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

State of health estimation of Li-ion batteries based on sample entropy and various regression techniques

  • 期刊论文

  • 发表日期：

    2025年3月18日

查看全文
  • Sunil K. Pradhan,
  • Basab Chakraborty

摘要

Lithium-ion batteries are an integral part of numerous smart energy systems. Accurate estimation of battery state of health is vital to ensure the safe and reliable usage of lithium-ion batteries. In this paper, various regression algorithm-based estimation frameworks in combination with sample entropy of battery voltage is implemented to accurately estimate the battery state of health (SOH). The sample entropy, fuzzy entropy, localized area and power spectral density values of charging voltage sequences are utilized to develop the hybrid SOH estimation model and thereby minimizing the estimation error values. The health feature variables based on battery charging attributes are validated as per grey correlation analysis to estimate the battery deterioration trends. Different regression models are compared to illustrate the effectiveness and estimation accuracy of the proposed hybrid model. The results demonstrate that the hybrid model trained on localized voltage area-sample entropy feature variables or power spectral density-sample entropy feature variables are more accurate in estimating battery SOH than other estimation models such as Lasso, and support vector regression models. Despite some batteries following an intricate nonlinear degradation path, the mean absolute error and root mean squared error values of the proposed model do not exceed 0.26% and 0.42% respectively.