Using support vector machine algorithm to achieve insulation defect pattern recognition of GIS

Summary

Gas insulated switchgear (GIS) has been widely used all over the world, the high reliability, low maintenance and compact size made it an attractive choice in many aspects. Though GIS has high reliability, the return of practical experience indicated that most of the failures were related to insulation problems. Partial discharge (PD) is one of the main reasons for causing internal insulation deterioration of GIS. Monitoring of PD can promptly and accurately determine the status of the internal insulation to prevent GIS accidents and ensure power system security and stability. To evaluate the condition of GIS accurately and provide appropriate guidance on their maintenance, a fundamental ultra-high frequency (UHF) database of partial discharges corresponding to different types of defects is presented for the observation of insulation state of GIS. In order to investigate the features of PD excited by different insulation defect models in GIS, the paper designed four kinds of typical PD models to simulate the actual insulation defects in GIS. And then set up the UHF-based PD testing system for these four typical defects and collect the UHF signals which present different morphological features of every type of PD. By investigating the PD characteristics of various types of defects, eight feature parameters are extracted from the UHF signals of PD to describe the typical features of each type of PD. A four-class support vector machine classifier is constructed based on support vector machine (SVM) algorithm which is a new learning method developed based on the statistical learning theory, and realizes the structural risk minimization theory. Meanwhile, the paper set up the UHF-based PD testing system for GIS and the experimental results indicate that the proposed method can extract entropy-sequence features of PD samples, possesses a high recognition rate above 88% and can effectively identify these four types of PDs in GIS.

Additional informations

Authors

G. LU, J. XIONG, S. YANG, Y. LIU, L. GAN, K. Zhou

Keywords

GIS, Insulation Defect Pattern Recognition, Support Vector Machine Algorithm