基于随机森林的复合绝缘子憎水性等级检测  

Hydrophobicity Level Detection of Composite Insulators Based on Random Forest

作　　者：乔逸卓 张红旗[1] 马昕宇 QIAO Yizhuo;ZHANG Hongqi;MA Xinyu(College of Mechanical and Electrical Engineering,Inner Mongolia Agricultural University,Hohhot 010010,China;Wuhan Long'an Group Co.,Ltd.,Wuhan 430074,China)

机构地区：[1]内蒙古农业大学机电工程学院,呼和浩特010010 [2]武汉龙安集团有限责任公司,武汉430074

出　　处：《内蒙古电力技术》2024年第5期94-100,共7页Inner Mongolia Electric Power

基　　金：风能太阳能利用技术教育部重点实验室开放基金项目“含高比例风电集群接入的蒙西电网经济调度理论研究”(2021ZD01)。

摘　　要：针对传统复合绝缘子憎水性检测费时费力,因受环境等因素影响,检测结果误差较大的问题,提出一种基于随机森林的复合绝缘子憎水性等级检测方法。该方法是一种机器学习算法,通过构建多个决策树并投票来建立一个分类模型,采用图像处理方法中的增强、平滑、去噪、分割等操作对复合绝缘子伞裙表面进行预处理,从中提取憎水性相关特征数据,最后运用随机森林完成复合绝缘子憎水性等级检测。检测分析时,当决策树数量在100左右时,准确率稳定在92.88%,与传统复合绝缘子憎水性检测方法相比,提高了憎水性检测效率及准确率。Aiming at the issues that the traditional composite insulator hydrophobicity detection is time⁃consuming and labor⁃intensive,and there are large errors of detection results affected by the environment and other factors,the author proposes a composite insulator hydrophobicity level detection method based on Random Forest.This method is a machine learning algorithm,through building a classification model by constructing multiple decision trees and voting,preprocesses the surface of the composite insulator umbrella skirt using image processing methods such as enhancement,smoothing,denoising,and segmentation,to extract hydrophobic related feature data,and finally completes the detection of composite insulators hydrophobic level by using the Random Forest.During the process of detecting and analyzing,when the number of decision trees is around 100,the accuracy rate is stable at 92.88%,which improves the efficiency and accuracy of hydrophobicity detection comparing with the traditional composit insulator hydrophobicity detecting method.

关 键 词：复合绝缘子 憎水性 随机森林 图像处理 决策树 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术] TM216[自动化与计算机技术—计算机科学与技术]