摘要
输电线路悬挂异物会引发输电线路单相接地、相间短路等停电事故,因此本文提出一种基于卷积神经网络与ECOC-SVM的输电线路异物检测方法。首先,本文构建气球、风筝、塑料和鸟巢4种输电线路异物图像数据集;然后采用Otsu自适应阈值分割、形态学处理等方法提取感兴趣区域;再利用DenseNet201提取感兴趣区域的特征;最后对ECOC-SVM模型进行训练、测试与结果分析。所用方法在4类异物上的平均识别准确率可达93.3%,有助于提高输电线路运维的效率。
Foreign bodies suspended on transmission lines will cause power outages such as single-phase grounding and phase-to-phase short-circuit.Therefore,this paper proposes a foreign body detection method for transmission lines based on convolutional neural network and ECOC-SVM.Firstly,four foreign body image datasets of the transmission line including balloon,kite,plastic and nest are established.Then the Otsu adaptive threshold segmentation,morphological processing and other methods are used to extract the region of interest,and Dense Net201 is used to extract the features of the region of interest.Finally,the ECOC-SVM model is trained,tested and analyzed.The average recognition accuracy of this method for the four foreign bodies can reach 93.3%,improving the efficiency of the transmission line operation and maintenance.
作者
余沿臻
邱志斌
周银彪
朱轩
王青
YU Yanzhen;QIU Zhibin;ZHOU Yinbiao;ZHU Xuan;WANG Qing(Department of Energy and Electrical Engineering,Nanchang University,Nanchang 330031,China;State Grid Jiangxi Electric Power Company,Nanchang 330000,China)
出处
《智慧电力》
北大核心
2022年第3期87-92,107,共7页
Smart Power
基金
国家自然科学基金资助项目(51967013)
江西省大学生创新创业训练计划项目(202110403086)。
作者简介
余沿臻(2001),男,福建漳州人,在读本科生,主要研究方向为输电线路异物检测和深度学习。
