基于机器学习的输电线路设备细微缺陷智能检测方法  

Intelligent detection method for minor defects of transmission line equipment based on machine learning

作　　者：宁柏锋 董召杰 NING Baifeng;DONG Zhaojie(Shenzhen Power Supply Co.Ltd,Shenzhen Guangdong 518000,China;Dingxin Information Technology Co.,Ltd.,Guangzhou 518000,China)

机构地区：[1]深圳供电局有限公司,广东深圳518000 [2]鼎信信息科技有限责任公司,广州518000

出　　处：《自动化与仪器仪表》2020年第5期161-165,共5页Automation & Instrumentation

基　　金：中国南方电网有限责任公司科技项目(No.090000KK52170124)。

摘　　要：针对现有输电线路设备细微缺陷智能检测方法存在的GZ-复刻率较高的问题,设计了基于机器学习的输电线路设备细微缺陷智能检测方法。根据设备实时监测、运行、历史故障维修次数等数据,构建基于机器学习的NFA模型,区分不同类型的输电线路设备的细微缺陷;对图像的亮度、对比度、拼接方式等方面进行处理,引入Adaboost算法完成图像识别,最后通过对相位编组梯度的计算,实现基于机器学习的输电线路设备细微缺陷的智能检测。实验结果表明,设计方法的GZ-复刻率平均值比传统方法的GZ-复刻率平均值降低了47.3%,故障信号检测性能更好,可以证明基于机器机器学习的输电线路设备细微缺陷智能检测方法的综合有效性。Aiming at the problem of high GZ-rewriting rate in the existing intelligent detection methods for transmission line equipment,an intelligent detection method for transmission line equipment micro-defects based on machine learning is designed.According to the data of real-time monitoring,operation and maintenance times of historical faults,the NFA model based on machine learning is constructed to distinguish the minor defects of different types of transmission line equipment.The brightness,contrast and splicing methods of images are processed,and the image recognition is completed by introducing the Adaboost algorithm.Finally,the phase is identified by using the Adaboost algorithm.The calculation of marshalling gradient realizes the intelligent detection of transmission line equipment’s fine defects based on machine learning.The experimental results show that the average gz-reproduction rate of the design method is 47.3% lower than that of the traditional method,and the fault signal detection performance is better,which can prove the comprehensive effectiveness of the transmission line equipment subtle defect intelligent detection method based on machine learning.

关 键 词：机器学习 输电线路设备 ADABOOST算法 细微缺陷 智能检测方法 

分 类 号：TP23[自动化与计算机技术—检测技术与自动化装置]