摘要
为了实现连续识别局部放电故障检测,利用数据采集与传输单元(DCTU)、高频电流互感器(HFCT)传感器和操作服务器构建了局部放电在线检测系统。运用深度残差学习(ResNet)和深度级可分离卷积(MobileNet)两类迁移学习模型建立了配电网局部放电在线检测系统及持续运行模式。通过对局部放电的连续监测获取真实局部放电数据,并与卷积神经网络(CNN)和支持向量机(SVM)两个基准模型进行对比研究。测试结果表明,所设计的局部放电在线检测系统对真实局部放电数据利用率高,所提出的ResNet50 V2和MobileNet50 V2迁移学习模型的检测精度可分别达到96.2%和97.4%。
In order to realize continuous identification of PD fault detection,this paper uses data acquisition and transmission unit(DCTU),high frequency current transformer(HFCT)sensor and operation server to establish an online PD detection system.Two kinds of transfer learning models,deep resistive learning(RESNET)and deep level separable convolution(MobileNet),are used to establish an online partial discharge detection system and continuous operation mode of distribution network.Through continuous monitoring of partial discharge,real partial discharge data are obtained.And compared with the convolutional neural network(CNN)and support vector machine(SVM)two benchmark models,the test results show that the online PD detection system designed in this paper has high utilization rate of real PD data,and the detection accuracy of the proposed ResNet50 V2 and MobileNet50 V2 transfer learning models can reach 96.2%and 97.4%,respectively.
作者
郭晓川
姜明
陈力
GUO Xiaochuan;JIANG Ming;CHEN Li(Jiangsu Electric Power Information Technology Co.,Ltd.,Nanjing 210000,China;Nanjing Ningzhong Human Resources Consulting Service Co.,Ltd.,Nanjing 210000,China)
出处
《微型电脑应用》
2023年第6期178-181,共4页
Microcomputer Applications
关键词
局部放电
人工智能
在线检测
配电网
迁移学习
partial discharge
artificial intelligence
online detection
distribution network
transfer learning
