Scholarlink

<p>To effectively extract information from functional data in lithium-ion battery degradation scenarios, a functional support vector regression algorithm is developed, upon which an ensemble forecasting method is proposed. By employing a specific process architecture for functional features, the functional support vector regression algorithm not only preserves the advantages of support vector regression in handling high-dimensional, small-sample datasets but also enabling global and derivative information extraction from vector-modal factors. Accurate degradation modelling, coupled with the effective resolution of mode mixing, is achieved by the integrated framework through its application of the complementary ensemble empirical mode decomposition with adaptive noise (CEEMDAN) method. An evaluation of the proposed framework is conducted utilizing three open-source lithium-ion battery datasets: National Aeronautics and Space Administration (NASA) battery dataset, Xi’an Jiaotong University (XJTU) battery dataset, and data from Centre for Advanced Life Cycle Engineering (CALCE). Empirical results show that the functional support vector regression-based integrated modelling approach offers superior prediction accuracy, presenting an innovative and reliable method for lithium-ion battery life forecasting with functional inputs.</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

A functional support vector regression for predicting the operating life of lithium batteries

  • 期刊论文

  • 发表日期：

    2026年7月3日

查看全文
  • Yu Zhou,
  • Senhui Wang

摘要

To effectively extract information from functional data in lithium-ion battery degradation scenarios, a functional support vector regression algorithm is developed, upon which an ensemble forecasting method is proposed. By employing a specific process architecture for functional features, the functional support vector regression algorithm not only preserves the advantages of support vector regression in handling high-dimensional, small-sample datasets but also enabling global and derivative information extraction from vector-modal factors. Accurate degradation modelling, coupled with the effective resolution of mode mixing, is achieved by the integrated framework through its application of the complementary ensemble empirical mode decomposition with adaptive noise (CEEMDAN) method. An evaluation of the proposed framework is conducted utilizing three open-source lithium-ion battery datasets: National Aeronautics and Space Administration (NASA) battery dataset, Xi’an Jiaotong University (XJTU) battery dataset, and data from Centre for Advanced Life Cycle Engineering (CALCE). Empirical results show that the functional support vector regression-based integrated modelling approach offers superior prediction accuracy, presenting an innovative and reliable method for lithium-ion battery life forecasting with functional inputs.