Title: Power transmission line anomaly detection scheme based on CNN-transformer model

Authors: Ming Gao; Wenfei Zhang

Addresses: School of Electric Power Engineering, South China University of Technology, Guangzhou, Guangdong, China ' Guangzhou Power Supply Bureau of Guangdong Power Grid Co., Ltd., Guangzhou, Guangdong, China

Abstract: The anomaly of power transmission lines has resulted in the failure of power delivery system, which brings about tremendous loss for the economy and industry. The wider distribution of power delivery system has imposed huge challenges on monitoring and making response to the anomaly cases in a short time. In this work, we introduce an autonomous anomaly detection system by exploiting the Computer Vision (CV) and Internet of Thing (IoT) techniques. At the first step, we design and develop an IoT sensor that can detect and feedback physical conditions around the power tower. Once the anomaly situation occurs, the on-site image acquisition is carried out by drones. To simplify the construction of image analysis pipelines while maintaining high accuracy, we adopt the State-of-The-Art (SOTA) cascaded Convolutional Neural Network (CNN)-transformer model. According to our experiment results, the CNN-transformer model is able to provide promising performance for anomaly detection of power lines, achieving higher average precision while consuming almost the same computing resources. The proposed anomaly detection scheme is of importance for realising large-scale and autonomous anomaly detection for power lines.

Keywords: IoT sensor; power automation; anomaly detection; computer vision; neural network.

DOI: 10.1504/IJGUC.2021.119565

International Journal of Grid and Utility Computing, 2021 Vol.12 No.4, pp.388 - 395

Received: 25 Jul 2020
Accepted: 27 Aug 2020
Published online: 09 Dec 2021 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article