基于红外图像的电力设备识别及发热故障诊断方法研究进展  

Research Progress of Power Equipment Identification and Thermal Fault Diagnosis Based on Infrared Images

作　　者：刘传洋 吴一全[1] LIU Chuanyang;WU Yiquan(College of Electronic and Information Engineering,Nanjing University of Aeronautics and Astronautics,Nanjing 211106,Jiangsu Province,China)

机构地区：[1]南京航空航天大学电子信息工程学院,江苏省南京市211106

出　　处：《中国电机工程学报》2025年第6期2171-2195,I0011,共26页PROCEEDINGS OF THE CHINESE SOCIETY FOR ELECTRICAL ENGINEERING

基　　金：国家自然科学基金项目(61573183);高校自然科学研究项目(2023AH052358,CZ2022ZRZ07)。

摘　　要：在电力大数据背景下,依托机器视觉和深度学习技术,从海量的红外图像数据中实现电力设备识别及发热故障诊断,已经成为该领域运维工作亟待解决的问题。该文以红外图像为研究对象,综述了基于传统图像处理和基于深度学习两类方法的红外图像中电力设备识别及发热故障诊断研究进展。首先,概述电力设备红外图像识别及发热故障诊断的发展历程及技术流程;然后,阐明基于传统图像处理的电力设备识别及发热故障诊断方法,从图像预处理、图像配准、图像分割、特征提取与分类、发热故障诊断5个方面进行归纳总结;阐述基于深度学习的变电站设备和输电线路设备识别及发热故障诊断方法,与传统图像处理方法相比,深度学习方法能够快速准确地识别电力设备发热故障;最后,指出基于深度学习的视觉技术在电力设备识别及发热故障诊断应用中存在的问题,基于现有的深度学习技术和最近的研究思路,对未来研究工作进行展望。In the context of power big data,relying on deep learning and machine vision technology to realize power equipment identification and thermal fault diagnosis from massive infrared image data has become an urgent problem for operation and maintenance work in this field.In this paper,infrared image is taken as the research object,and the research progress of power equipment identification and thermal fault diagnosis in infrared images based on the methods of traditional image processing and deep learning is reviewed.First,the development and technical process of infrared image recognition and thermal fault diagnosis of power equipment are described briefly.Then,the traditional image processing methods for power equipment identification and thermal fault diagnosis are introduced,which are summarized from five aspects:image preprocessing,image registration,image segmentation,feature extraction and classification,and thermal fault diagnosis.The deep learning-based identification and thermal fault diagnosis methods for substation equipment and transmission lines equipment are emphasized.Compared with the traditional image processing method,the deep learning method can identify the thermal fault of power equipment quickly and accurately.Finally,the problems existing in the application of vision technology based on deep learning in power equipment identification and thermal fault diagnosis are pointed out.Based on the existing deep learning technology and recent research ideas,the future research work is prospected.

关 键 词：电力设备识别 发热故障诊断 红外图像 传统图像处理 深度学习 视觉检测 

分 类 号：TM76[电气工程—电力系统及自动化] TP277[自动化与计算机技术—检测技术与自动化装置]