Enhancing Condition Monitoring of Outdoor Insulator through EfficientNet Classifier

Abstract

Outdoor insulators play a pivotal role in ensuring reliable power transmission and distribution. However, these insulators are exposed to various environmental stresses, which can lead to degradation over time and compromise their performance. This article proposes an enhanced condition monitoring approach utilizing an EfficientNet classifier to address this challenge. This work used an open repository named the Chinese Power Line Insulator Dataset (CPLID). To overcome the data insufficiency problem, an image augmentation technique was also used. After proper preprocessing, the captured images were input to EfficientNet models for condition classification using a transfer learning strategy. Finally, A web-based app was developed using Gradio to monitor outdoor insulators remotely. The experimental results show that the EfficientNet B4 outperforms all the other versions of EfficientNet with 99.41% detection accuracy, thereby enabling proactive maintenance, reducing downtime, and ensuring the security and reliability of power systems.