基于深度学习的配电线路瓷绝缘子缺陷识别  

Defect Identification of Porcelain Insulators in Distribution Lines Based on Deep Learning

作　　者：严宇平 吴广财 王泽涌 YAN Yuping;WU Guangcai;WANG Zeyong(Guangdong Power Grid Co.Ltd.,Guangzhou 510699,China;Guangdong Electric Power Information Technology Co.Ltd.,Guangzhou 510630,China)

机构地区：[1]广东电网有限公司,广东广州510699 [2]广东电力信息科技有限公司,广东广州510630

出　　处：《工业加热》2023年第1期70-74,共5页Industrial Heating

基　　金：广东省科技研发项目(201091325)。

摘　　要：考虑当前方法在识别瓷绝缘子缺陷时,受到缺陷类别的影响,导致识别精度较低的问题,提出了基于深度学习的配电线路瓷绝缘子缺陷识别方法。根据瓷绝缘子可靠性测试原理,采用兆欧表测试配电线路电源电压,通过计算配电线路的电源负载特性,引入测试电压波纹系数,对配电线路瓷绝缘子缺陷进行电子信号检测。根据配电线路瓷绝缘子表面的缺陷区域分布规律,计算瓷绝缘子缺陷图像的深度学习中心和隶属度函数。利用深度学习特征的最佳数量,定位检索配电线路瓷绝缘子缺陷特征,通过对瓷绝缘子缺陷图像的二值化处理,实现配电线路瓷绝缘子缺陷识别。实验结果表明,所提方法在识别瓷绝缘子脱落、灼蚀时缺陷识别指标都达到了9,破损时缺陷识别指标均控制在8以上,具有较高的识别精度。Considering that the current method is affected by the defect category when identifying the defects of porcelain insulators, resulting in low identification accuracy, a defect identification method of porcelain insulators of distribution lines based on deep learning is proposed. According to the reliability test principle of porcelain insulator, the megameter is used to test the power supply voltage of distribution line. By calculating the power load characteristics of distribution line, the ripple coefficient of test voltage is introduced to detect the defects of porcelain insulator in distribution line. According to the distribution law of defect area on the surface of porcelain insulator of distribution line, the depth learning center and membership function of porcelain insulator defect image are calculated. The optimal number of deep learning features is used to locate and retrieve the defect features of porcelain insulators in distribution lines. Through the binary processing of porcelain insulator defect images, the defect recognition of porcelain insulators in distribution lines is realized. The experimental results show that the proposed method can reach 9 when the porcelain insulator falls off and burnt, and the defect recognition index is controlled above 8 when the porcelain insulator is broken, which has high recognition accuracy.

关 键 词：深度学习 瓷绝缘子 配电线路 定位检索 缺陷识别 二值化处理 

分 类 号：TM726[电气工程—电力系统及自动化]