Scholarlink

The increasing adoption of lithium-ion batteries in energy storage systems is challenged by safety concerns, particularly self-discharge faults, which can lead to thermal runaway and pose significant risks. Monitoring the remaining charging capacity (RCC) is a typical approach for detecting self-discharge and mitigating potential hazards. However, RCC evaluation in battery systems faces challenges under short-cycle data and multi-stage charging with current switching. A deep learning-based method is proposed to estimate the RCC of individual battery cells under multi-stage charging conditions for self-discharge diagnosis, achieving advancements in three areas: (1) online RCC estimation based on short-cycle voltage/current sequences; (2) an adaptive mechanism for multi-stage charging with current switching; (3) early detection of self-discharge faults. Validation shows that the method achieves a mean absolute error (MAE) of 0.63 Ah in RCC prediction under current switching, successfully identifying self-discharge faults and quantifying its evolution, offering a high-precision, generalizable monitoring solution for dynamic energy storage systems.

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

Self-discharge Fault Diagnosis of Lithium-Ion Battery Packs Based on Remaining Charge Capacity Prediction Under Multi-stage Charging Conditions

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Wenhao Nie,
  • Zhongwei Deng,
  • Jiwei Wang,
  • Chunlin Jiang,
  • Yuxin He

摘要

The increasing adoption of lithium-ion batteries in energy storage systems is challenged by safety concerns, particularly self-discharge faults, which can lead to thermal runaway and pose significant risks. Monitoring the remaining charging capacity (RCC) is a typical approach for detecting self-discharge and mitigating potential hazards. However, RCC evaluation in battery systems faces challenges under short-cycle data and multi-stage charging with current switching. A deep learning-based method is proposed to estimate the RCC of individual battery cells under multi-stage charging conditions for self-discharge diagnosis, achieving advancements in three areas: (1) online RCC estimation based on short-cycle voltage/current sequences; (2) an adaptive mechanism for multi-stage charging with current switching; (3) early detection of self-discharge faults. Validation shows that the method achieves a mean absolute error (MAE) of 0.63 Ah in RCC prediction under current switching, successfully identifying self-discharge faults and quantifying its evolution, offering a high-precision, generalizable monitoring solution for dynamic energy storage systems.