摘要
为实现对变电站GIS盆式绝缘子开裂的准确识别,并确定具体裂纹形式,引入卷积神经网络,开展变电站GIS盆式绝缘子开裂识别方法研究。引入X-DR检测技术,获取变电站GIS盆式绝缘子射线图像;应用软阈值法,对盆式绝缘子射线图像进行去噪处理;通过卷积神经网络的应用,构建开裂检测模型,并完成对模型的训练,实现对绝缘子是否开裂以及具体裂纹形式的识别。通过对比实验证明,新的识别方法可以检测到绝缘子开裂,且可以对裂纹形式给出准确的识别结果。
In order to accurately identify the crack of GIS basin insulator in substation and determine the specific crack form,convolutional neural network was introduced to design and study the crack identification method of GIS basin insulator in substation.The X-ray image of GIS basin insulator in substation was obtained by introducing X-DR detection technology.The soft threshold method is applied to denoise the radiographic image of the basin insulator.Through the application of convolutional neural network,the crack detection model is constructed,and the model is trained to recognize whether the insulator is cracked or not and the specific crack form.The experimental results show that the new identification method can detect insulator cracking and give accurate identification results of crack forms.
作者
张强
郭斌
刁文正
ZHANG Qiang;GUO Bin;DIAO Wenzheng(SDEE Hitachi High-Voltage Swichgear Co.,Ltd.,Jinan 250103,China)
出处
《电工技术》
2025年第9期252-254,258,共4页
Electric Engineering
关键词
卷积神经网络
GIS
开裂识别
盆式绝缘子
变电站
convolutional neural network
GIS
cracking identification
basin insulator
substation
作者简介
张强(1978-),研究方向为气体绝缘金属封闭开关的设计、研发、运行及维护;郭斌(1974-),研究方向为高压开关设备智能化的研发与应用;刁文正(1985-),研究方向为气体绝缘金属封闭开关的设计、开发及高压开关设备智能化的研发及应用。
