Uhf Envelop Analysis And Defects Identification For Partial Discharge In GIS

Enclosed gas-insulated switchgear (GIS) is widely used in cities such as high voltageelectricity power transmission system. Partial Discharge serves as both a warning of and arepresentation of the degraded insulation of the GIS, and partial discharge may cause furtherdegradation of the insulation. In order to improve the safety and reliability of GIS, A largeamount of research has been carried out both at home and abroad. UHF signals analysis isone of the main methods of the study in GIS. So far, however, pattern recognition for UHFsignals is still a pending issue which is the forefront. The difficulties of the researchincluding: Transmission mechanism and modeling of UHF signals of Partial discharge,stability and reliability of detection system, selection and acquisition of characteristicparameters of UHF signal, Classifier design for pattern recognition, pattern-based databasesand long-term data accumulation.In this paper, the theory of electromagnetic is used for analysis of UHF signals generated bypartial discharge in GIS. The principle of resonant and transmission characteristics isanalyzed too.A GIS partial discharge testing system is established in the laboratory. Five kinds of typicaldefects was made in GIS and partial discharge testing were carried out. The time-domainsignal and its envelope signal of partial discharge was recorded. A large number ofexperiments showed that the time-domain characteristics of the partial discharge signals andthe envelopes generated by the same defect model are about the same. Both the time-domaincharacteristics of the partial discharge signals and the envelope shapes generated by thedifferent defect models are different. Therefore, the partial discharge detection and defectspattern recognition method based on the envelope detection is proposed in this paper.Compared with the PRPD methods (Phase Resolved Partial Discharge), it can remove thepulse interference by the characteristic parameters of the envelope and reduce the wrongreports by interference. This method is independent on the phase of AC, so it is suitable forGIS of DC applications. The use of envelope detector avoids ultra-wideband real-timesampling requirement and a huge amount of data on the treatment.Noise interference is one of the key factors affecting the accurate measurement of partialdischarge. Thesis analyzes various kinds of interference signal and their characteristics which affect the measurement in GIS. Hardware band-pass filter, band-stop filters andwavelet-based signal processing algorithm are used to remove interference. Hardware filtercan effectively remove the narrowband interference, but can not eliminate the white-noise.The mathematical method of Hilbert transform was used to study the signal changes ofwhite-noise before and after envelope detector. The existence of white-noise affects theextraction accuracy of signal envelope in time domain feature. Thesis compare and studyseveral types of wavelet methods for de-noise and an improved threshold method wasproposed which obtain good de-noising effect with testing data. Whether GIS is in goodcondition or not and finding where defects are the ultimate goal of signal extraction andanalysis of characteristic parameters. This work is used to guide the repair and maintenance.Data preprocessing, feature extraction, classifier design, principle and algorithm of trainingof network were studied in the paper. According to the feature of the envelop signal, thesisused the time-domain parameters for the characteristic parameters and the BP neural networkclassifier for defect types of pattern recognition. Time-domain parameters can be a gooddescription of the characteristics of the envelope signal. BP neural network has simplestructure, high stability, fault tolerance, and anti-interference ability; and it’s self-organized.As a classifier of pattern recognition, its hit rates are as much as more than96％withmeasurement data for five different defects.Partial discharge detection system based on envelope detector was developed and realized inthis paper, which including the UHF sensor, UHF amplifier and data acquisition circuit.