基于Faster R-CNN的环网柜电缆相序检测  

Phase sequence detection of cable in ring main unit based on faster R-CNN

作　　者：李继东 赵锴 黄佳 郑静媛 张淞杰 LI Jidong;ZHAO Yan;HUANG Jia;ZHENG Jingyuan;ZHANG Songjie(State Grid Beijing Electric Power Corporation Fengtai Power Supply Bureau,Beijing 100073,China;College of Electrical Engineering,Zhejiang University,Hangzhou Zhejiang 310027,China)

机构地区：[1]国家电网北京丰台供电公司,北京100073 [2]浙江大学电气工程学院,浙江杭州310027

出　　处：《杭州电子科技大学学报（自然科学版）》2021年第4期61-66,共6页Journal of Hangzhou Dianzi University：Natural Sciences

基　　金：国家电网公司科技资助项目(52020518005F)。

摘　　要：随着环形配电网的普及,环网柜的运维变得越来越重要。传统的环网柜施工验收方式以人工巡视为主,耗时耗力,已不能满足城市配电网施工验收可靠性和实时性的要求。文章结合卷积神经网络技术,分析环网柜电缆特征,提出一种基于Faster R-CNN的环网柜电缆相序检测方法,通过区域候选网络对区域特征进行检测识别。实验结果表明,检测器交并比阈值为大于0.5时,Faster R-CNN方法在数据集上的平均检测精度最高为0.93,与同类主流目标检测算法相比,其精度均有不小的提升。With the popularization of ring power distribution network,the operation and maintenance of ring main unit has become more and more important.The traditional construction acceptance method is to inspect the installation situation by manual inspection,which is time-consuming and labor-intensive,and can no longer meet the reliability and real-time requirements of urban distribution network construction acceptance.Thus,combined with the convolutional neural network technology to analyze the characteristics of the ring main unit,the paper proposes a Faster R-CNN-based ring main unit s cable phase sequence detection method,which can detect region features by region proposal network.Experimental results show that the highest average detection accuracy of Faster R-CNN method on the data set is 0.93(when the Intersection-over-Union is greater than 0.5),which is a great improvement compared with the state-of-art object detection algorithms.

关 键 词：环网柜 卷积神经网络 目标检测 Faster R-CNN 

分 类 号：TM752[电气工程—电力系统及自动化]