电力系统及其自动化学报2019，Vol.31Issue(10)：24-30,7.DOI:10.19635/j.cnki.csu-epsa.000074
GIS局部放电时域波形图像的模式识别方法
Pattern Recognition Method for Time-domain Waveform Images of GIS Partial Discharge
刘创华 1何金 2张春晖 2曹梦 2宋晓博 2朱旭亮3
作者信息
- 1. 国网天津市电力公司,天津 300232
- 2. 国网天津市电力公司电力科学研究院,天津 300384
- 3. 国网天津市电力公司电力科学研究院,天津 300384
- 折叠
摘要
Abstract
A large amount of data obtained from the on-site detection of gas insulated switchgear(GIS)partial dis-charge are time-domain waveform images. However,the traditional partial discharge pattern recognition method cannot be directly applied to the defect classification of these images. Aiming at the partial discharge time-domain waveform im-ages,different image recognition techniques,such as image segmentation,image gray processing,image binarization, image enhancement and image compression,are used to preprocess these images. Through on-site detection in the sub-station,image data sets of types of partial discharge defect are established. Afterwards,a support vector machine (SVM)model,which adopts radial basis function(RBF)as the kernel function,is used for the preprocessed images that only contain pulsed voltage of partial discharge. A directed acyclic graph(DAG)classifier is obtained using the SMO optimization method,which is further directly used to perform pattern recognition. Experimental results show that the accuracy of six types of partial discharge defect using the SVM model is above 85%,which is superior to that using the back propagation neural network(BPNN)model;the proposed method does not need to extract features manually, and it has higher recognition accuracy.关键词
模式识别/图像/支持向量机/局部放电Key words
pattern recognition/image/support vector machine(SVM)/partial discharge分类
信息技术与安全科学引用本文复制引用
刘创华,何金,张春晖,曹梦,宋晓博,朱旭亮..GIS局部放电时域波形图像的模式识别方法[J].电力系统及其自动化学报,2019,31(10):24-30,7.
