Scholarlink

The internal insulation of a capacitor voltage transformer (CVT) is easily degraded, deteriorating its accuracy and operation stability. The dielectric loss factor (DLF) is one of the most critical indicators for assessing the insulation state of a CVT, thus it is important to design anomaly diagnosis of DLF methods for ensuring the safety of the power grid. However, existing methods are impractical for engineering applications due to the requirement of accurate substation configuration or CVT measurements. We propose an unsupervised method that associates DLF with phase angle error to address the above deficiencies. First, the local outlier factor is used to locate the faulty CVTs with abnormal DLFs, and a feature named reference comprehensive deviation is further constructed accordingly. Then, a seasonal-trend decomposition method is applied to extract DLF-related information to enable the evaluation of DLF in faulty CVTs. The proposed method requires no labels of the measurements or additional sensors, thus facilitating timely diagnosis and reducing maintenance costs. Experimental results on a designed hardware platform indicate the effectiveness of the proposed method, where the estimated absolute error of DLF is less than 0.01%.

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

Anomaly Diagnosis of Dielectric Loss in Capacitor Voltage Transformer by Phase Angle Error

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Jing Fang,
  • Ruiming Yuan,
  • Ziqin Gao,
  • Panpan Guo,
  • Cheng He,
  • Chuanji Zhang

摘要

The internal insulation of a capacitor voltage transformer (CVT) is easily degraded, deteriorating its accuracy and operation stability. The dielectric loss factor (DLF) is one of the most critical indicators for assessing the insulation state of a CVT, thus it is important to design anomaly diagnosis of DLF methods for ensuring the safety of the power grid. However, existing methods are impractical for engineering applications due to the requirement of accurate substation configuration or CVT measurements. We propose an unsupervised method that associates DLF with phase angle error to address the above deficiencies. First, the local outlier factor is used to locate the faulty CVTs with abnormal DLFs, and a feature named reference comprehensive deviation is further constructed accordingly. Then, a seasonal-trend decomposition method is applied to extract DLF-related information to enable the evaluation of DLF in faulty CVTs. The proposed method requires no labels of the measurements or additional sensors, thus facilitating timely diagnosis and reducing maintenance costs. Experimental results on a designed hardware platform indicate the effectiveness of the proposed method, where the estimated absolute error of DLF is less than 0.01%.