基于增强现实技术的电力设备故障识别方法研究  

Research on power equipment fault identification method based on augmented reality technology

作　　者：徐兴国 林楚伟 陈伟武 江永 XU Xingguo;LIN Chuwei;CHEN Weiwu;JIANG Yong(Haimen Power Plant,Huaneng Power International Inc.,Shantou 515132,China)

机构地区：[1]华能国际电力股份有限公司海门电厂,广东汕头515132

出　　处：《电子设计工程》2020年第23期149-152,157,共5页Electronic Design Engineering

基　　金：广东省工业攻关项目(2016ZR3398)。

摘　　要：针对电力设备巡检智能化水平较低的现状,文中将增强现实(Augmented Reality,AR)技术应用于电力设备巡检过程。文中从智能巡检终端、服务器与数据库3个层面构建了基于AR技术的电力设备智能巡检系统架构。提出基于AR技术和深度神经网络(Deep Neural Networks,DNN)算法的电力设备故障识别方法,将智能巡检终端采集的图像作为输入,在线识别电力设备可能存在的故障类型。通过仿真测试表明,所提方法故障识别时间与支持向量机(Support Vector Machine,SVM)与BP神经网络(Back Propagation-Neural network,BP-NN)算法相近。但是各类故障识别准确率均大于98%,大于SVM与BP-NN算法,所提方法能够快速准确地识别电力设备故障类型。In view of the low level of intelligent inspection of power equipment,Augmented Reality(AR)technology is applied to the inspection process of power equipment.In this paper,the architecture of intelligent inspection system for power equipment based on AR technology is constructed from three levels:intelligent inspection terminal,server and database.A fault identification method of power equipment based on AR technology and Deep Neural Networks(DNN)algorithm is proposed.The image collected by intelligent inspection terminal is used as input to identify the possible fault types of power equipment online.The simulation test shows that the fault identification time of the proposed method is similar to Support Vector Machine(SVM)and BP Neural network(BP-NN)algorithm,but the accuracy of all kinds of fault identification is greater than 98%,greater than SVM and BP-NN.The proposed method can identify the fault types of power equipment quickly and accurately.

关 键 词：增强现实 故障识别 深度学习 智能巡检 

分 类 号：TN99[电子电信—信号与信息处理] TP277[电子电信—信息与通信工程]