Scholarlink

The reliability of power transmission and distribution heavily relies on the performance of outdoor insulators, which are constantly exposed to environmental stress, leading to degradation over time. This article presented an advanced approach for monitoring the condition of outdoor insulators, employing a Swin Transformer classifier. The study utilized the Chinese Power Line Insulator Dataset (CPLID) and addressed data insufficiency through image augmentation techniques. Additionally, the Swin2SR image super-resolution technique was used to mitigate motion blur in the images. Subsequently, pre-processed images were fed into the Swin Transformer V2 model for condition classification, utilizing a transfer learning strategy. Furthermore, a web-based application was developed for remote monitoring of outdoor insulators. The experimental results demonstrated that the proposed model, enhanced with image augmentation and super-resolution techniques, achieved a remarkable 100% detection accuracy. This approach facilitated proactive maintenance, reduced downtime, and ensured power systems’ security and reliability.

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

Insulator Condition Monitoring Using Swin Transformer Classifier Enhanced with Image Augmentation and Swin Super-Resolution Technique

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Satyajit Panigrahy,
  • Subrata Karmakar

摘要

The reliability of power transmission and distribution heavily relies on the performance of outdoor insulators, which are constantly exposed to environmental stress, leading to degradation over time. This article presented an advanced approach for monitoring the condition of outdoor insulators, employing a Swin Transformer classifier. The study utilized the Chinese Power Line Insulator Dataset (CPLID) and addressed data insufficiency through image augmentation techniques. Additionally, the Swin2SR image super-resolution technique was used to mitigate motion blur in the images. Subsequently, pre-processed images were fed into the Swin Transformer V2 model for condition classification, utilizing a transfer learning strategy. Furthermore, a web-based application was developed for remote monitoring of outdoor insulators. The experimental results demonstrated that the proposed model, enhanced with image augmentation and super-resolution techniques, achieved a remarkable 100% detection accuracy. This approach facilitated proactive maintenance, reduced downtime, and ensured power systems’ security and reliability.