Scholarlink

The distribution transformer meter is an instrument used in distribution systems to measure various electrical parameters. Identifying its faults promptly and ensuring its proper functioning is critical for the daily management of the distribution network. To address the issue of fault identification in distribution transformer meters, this study applies a classification method based on Circular Dilated Convolutional Neural Network (CDIL-CNN) using 15-dimensional time-series measurement data from the meters. This paper analyzes the differences between CDIL-CNN and traditional convolutional neural networks (CNN) as well as temporal convolutional networks (TCN). Furthermore, the study compares the performance of CDIL-CNN with several existing models, highlighting its advantages in long-sequence classification tasks. Experimental results demonstrate that CDIL-CNN significantly outperforms other methods in classification accuracy, especially for long-sequence data.

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

Fault Identification of Distribution Transformer Meter Based on CDIL-CNN

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Yunpeng Guo,
  • Runlong Liu

摘要

The distribution transformer meter is an instrument used in distribution systems to measure various electrical parameters. Identifying its faults promptly and ensuring its proper functioning is critical for the daily management of the distribution network. To address the issue of fault identification in distribution transformer meters, this study applies a classification method based on Circular Dilated Convolutional Neural Network (CDIL-CNN) using 15-dimensional time-series measurement data from the meters. This paper analyzes the differences between CDIL-CNN and traditional convolutional neural networks (CNN) as well as temporal convolutional networks (TCN). Furthermore, the study compares the performance of CDIL-CNN with several existing models, highlighting its advantages in long-sequence classification tasks. Experimental results demonstrate that CDIL-CNN significantly outperforms other methods in classification accuracy, especially for long-sequence data.