基于图像深度学习与三维电场仿真的电力用复合绝缘子缺陷挖掘分析

Defect Mining and Analysis of Electric Power Composite Insulator Based on Image Deep Learning and Three-dimensional Electric Field Simulation

鉴于复合绝缘子运行环境差异较大、表面电场分布特征各异,亟需合理评估现役运行复合绝缘子运行状态。为此,首先应用紫外、红外成像技术对典型复合绝缘子进行观测形成电-热影像图谱数据库;然后应用小波成像技术对原始图像进行去噪处理,并基于聚类算法对复合绝缘子进行图像区域分割,联合视觉深度学习实现了复合绝缘子类型自适应识别和外轮廓匹配。同时应用三维建模技术对典型复合绝缘子电场分布规律进行仿真模拟,实现复合绝缘子高场强区域电场分布可视化;最终联合深度学习图像后处理结果和电场分布特征形成现役复合绝缘子运行状态评估策略。研究结果对于电力系统用复合绝缘子运行维护具有较好指导意义。

The composite insulators have different operating environment and surface electric field distribution characteristics,so it is urgent to make reasonable evaluation of the operation status.Firstly,the typical composite insulators are observed by ultraviolet and infrared imaging technology to form the database of electrical-thermal image.The original images are denoised by the wavelet imaging technology,and the image region of composite insulators is segmented based on the clustering algorithm.The adaptive recognition of composite insulators and the matching of external contours are realized by visual deep learning.At the same time,three-dimensional modeling technology is applied to simulate the electric field distribution law of typical composite insulators.The visualization of electric field in the high field strength area of composite insulators is realized.Finally,the post-processing results of image and the characteristics of electric field distribution are combined to form evaluation strategy of composite insulators operation status.This study has guiding significance for operation and maintenance of composite insulators used in power systems.