Scholarlink

<p>Accurate, efficient, and robust estimation of lithium-ion battery state of health (SOH) is essential for the safe and reliable operation of electric vehicles and energy storage systems. However, the complexity of real-world operating conditions, the challenge of extracting physically meaningful health features, and the scarcity of labeled SOH data under dynamic conditions limit the accuracy and generalizability of conventional methods. To address these challenges, we propose a data-driven, lightweight machine learning framework that integrates both model-domain and data-domain features, significantly enhancing SOH estimation under practical scenarios. Model-domain features are derived from equivalent circuit model parameter identification, whereas the data-domain feature is computed as the fuzzy entropy of the discharge voltage signal. Furthermore, we propose a pseudo-labeling strategy based on the piecewise cubic Hermite interpolating polynomial (PCHIP).The proposed framework was validated on a dynamic-condition battery dataset, where the implementation of the Categorical Boosting-Whale Optimization Algorithm (CatBoost-WOA) achieved Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) values below 0.5%. The method also demonstrates strong robustness and inference efficiency. This study presents a novel approach for SOH estimation under dynamic conditions and limited labeling, providing a theoretical and practical foundation for the development of efficient and reliable battery management systems in real-world applications.</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 lithium-ion battery under dynamic conditions using a dual-domain feature fusion machine learning approach

  • 期刊论文

  • 发表日期：

    2026年4月15日

查看全文
  • Jin Xiong,
  • Minlei Xia,
  • Shengtao Xu,
  • Peng Ding,
  • Hui Pan,
  • Qingwei Gao,
  • Wenyao Guo,
  • Yulin Min

摘要

Accurate, efficient, and robust estimation of lithium-ion battery state of health (SOH) is essential for the safe and reliable operation of electric vehicles and energy storage systems. However, the complexity of real-world operating conditions, the challenge of extracting physically meaningful health features, and the scarcity of labeled SOH data under dynamic conditions limit the accuracy and generalizability of conventional methods. To address these challenges, we propose a data-driven, lightweight machine learning framework that integrates both model-domain and data-domain features, significantly enhancing SOH estimation under practical scenarios. Model-domain features are derived from equivalent circuit model parameter identification, whereas the data-domain feature is computed as the fuzzy entropy of the discharge voltage signal. Furthermore, we propose a pseudo-labeling strategy based on the piecewise cubic Hermite interpolating polynomial (PCHIP).The proposed framework was validated on a dynamic-condition battery dataset, where the implementation of the Categorical Boosting-Whale Optimization Algorithm (CatBoost-WOA) achieved Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) values below 0.5%. The method also demonstrates strong robustness and inference efficiency. This study presents a novel approach for SOH estimation under dynamic conditions and limited labeling, providing a theoretical and practical foundation for the development of efficient and reliable battery management systems in real-world applications.