基于深度学习的输电线路绝缘子故障检测方法  

Insulator Fault Detection Method for Transmission Line Based on Deep Learning

作　　者：杨桢[1] 刘易宸 李鑫[1] 许雪飞 YANG Zhen;LIU Yichen;LI Xin;XU Xuefei(Faculty of Electrical and Control Engineering,Liaoning Technical University,Huludao 125105)

机构地区：[1]辽宁工程技术大学电气与控制工程学院,葫芦岛125105

出　　处：《电气工程学报》2024年第2期325-334,共10页Journal of Electrical Engineering

摘　　要：针对航拍巡检高压输电线路上绝缘子目标易受复杂背景和部分遮挡影响,造成传统算法难以准确检测的问题,建立了一种基于改进YOLOv5的输电线路绝缘子检测模型。首先,利用GA+K的检测框优化算法对选择的模型进行改进,来提高识别精度;然后,在YOLOv5算法框架中融合CBAM模块来提升图像中故障目标区域的显著度;其次,采用高斯函数改进YOLOv5中的非极大值抑制方法,提高对遮挡目标的识别准确率;最后利用辽宁某电网公司提供的无人机巡检图像制作数据集,并将所提算法与4种经典目标检测算法进行比较。试验结果表明,相比于4种对比算法,该算法能够在保证较高检测精度的同时具有较好的实时性,并且平均检测精度可以达到95.1%,每张图片的检测时间为0.04 s,兼具目标检测的准确性和实时性。In view of the problem that the insulator target of high voltage transmission line is susceptible to complex background and partial occlusion,a model based on improved YOLOv5 is developed.First,the detection box optimization algorithm of GA+K is used to improve the selected model to improve the recognition accuracy;then,the CBAM module is integrated in the YOLOv5 algorithm framework to improve the salience of the fault target area in the image;second,the Gaussian function is used to improve the non-maximum suppression method in YOLOv5 to improve the recognition accuracy of the occlusion target;finally,the UAV inspection image provided by a power grid company in Liaoning Province is used,and the proposed algorithm is compared with four classical target detection algorithms.The experimental results show that compared with the four comparison algorithms,this algorithm can guarantee high detection accuracy with good real-time performance,and the average detection accuracy can reach 95.1%.The detection time of each image is 0.04 s,which has both the accuracy and real-time performance of target detection.

关 键 词：绝缘子缺陷 故障检测 深度学习 目标识别 YOLOv5 

分 类 号：TM755[电气工程—电力系统及自动化]