AI识别技术在输配电网络故障诊断中的应用.pdfVIP

信息技术XINXIJISHU 2023年第6期 AI识别技术在输配电网络故障诊断中的应用 张峰，郭圣² ，李国强*，周华敏 摘要：常规方法分割图像中输配电网络像素点时，选择的分割点梯度不一致，使导线故障和器件故 障定位范围较大，为此提出AI识别技术在输配电网络故障诊断中的应用方法。获取输配电网络机巡 数据，转换红外图像像素点颜色，去除噪音像素点，根据灰度值选择种子像素点，将梯度幅值和方向一 致的种子点作为分割点，分割输配电网络像素区域，输入AI识别网络，提取分割图像特征进行分类学 习，输出故障诊断结果。实验结果表明，设计方法减小了导线故障点标记距离、器件元件故障区域标 记面积，提高了故障位置定位精度。 关键词：AI识别技术；输配电网络；故障诊断；红外图像 中图分类号：TM930.9 文献标识码：A D0I: 10.13274/j. cnki. hdzj. 2023. 06.031 Application of AI recognition technology in fault diagnosis of transmission and distribution network ZHANG Feng’, CUO Sheng, LI Guo-qiang, ZHOU Hua-min (1. Machine Patrol Management Center of Guangdong Power Grid Co. , Ltd. , Guangzhou 510000, China; 2. Guangdong Power Grid Co. , Ltd. , Guangzhou 510000, China) Abstract: When conventional methods segment the pixels of power transmission and distribution network in the image, the gradient of the selected segmentation points is inconsistent, which might make the location range of conductor fault and device fault larger. Therefore, the application of AI recognition technology in fault diagnosis of transmission and distribution network is proposed. The machine patrol data of power trans- mission and distribution network is obtained, the pixel color of infrared image is converted , and the noise pixel is removed. According to the gray value, the seed pixel is selected, and take the seed point with the same gradient amplitude and direction as the segmentation point to segment the pixel area of power transmis- sion and distribution network. Finally, AI recognition network is input to extract the segmentation image features for classification learning, and the fault diagnosis results are output. The experiment results show that the method reduces the marking distance of conductor fault point and the marking area of device fault area, and improves the fault location accuracy. Key wor