Font Size: a A A

Signal Processing And Localization For PD Source In Substation Based On Electromagnetic Antenna Array

Posted on:2015-03-19Degree:DoctorType:Dissertation
Country:ChinaCandidate:H J HouFull Text:PDF
GTID:1222330452466614Subject:High Voltage and Insulation Technology
Abstract/Summary:PDF Full Text Request
Partial discharge (PD) signals contain abundant information on the insulating state.Meanwhile, PD is not only the sign and manifestation of insulation degradation, but also thereason for its further degradation. Thus, PD test has been one of the most effective methods toinsulation testing and diagnosis and avoid insulation breakdown failures. Accuracy of theexiting method to estimate PD characteristics based on PD maps is not high. However, thelocalization for PD source is useful for finding the fault quickly and effectively. Ultra-highfrequency (UHF) electromagnetic waves have the advmtages of strong anti-interference, highsensitivity and stable transmitting speed. Therefore, it’s a current hot issue to apply UHFmethod to the localization and fault diagnosis of PD.The PD signals detected in substation always include the interference of white noises,random pulses, and because of the limitation of sampling rate of the signal acquisition system,the collected signals also include periodical narrowband interference and other interferencesignals. The interference can bring great difficulties to the signal delay estimation, PD sourcelocalization, separation and other issues. Firstly, the composition and structure of PDdetection system in substation based on antenna arrays has been introduced. The localizationalgorithm for PD souce in substation based on time delay sequence has been discussed deeply,and the origin of its localization error has been analyzed. Then the relationship betweenlocalization error and the placement of antenna array under the same time difference error hasbeen discussed theoretically, and optimal placement of the antenna may has been provided.The development of modem signal processing and higher-order statistics theory providesnew ideas for key problems of PD signal processing. In this paper, based on high orderstatistical theory, time delay estimation algorithms based on the method of fourth-ordercumulant and bispectmm estimation have been investigated. The problem of Gaussian noiseof unknown correlation and periodic narrowband interference mixed in PD signals is solved.PD UHF signals collected from substation are analyzed using this algorithm. Then the timedelay sequence is used for locating PD source, and the localization error is within the Im,which meets the accuracy of localization for PD source in substation. The attenuation characteristics of electromagnetic fields radiated by PD and antennasensitivity coeiffcient of receiving antenna has been analyzed, and a new localizationalgorithm based on the attenuation model of the electromagnetic fields has been proposed. Anovel localization algorithm for partial discharge in substation, based on the L-shaped antennaarray signal processing, and the thought of estimation signal parameter via rotationalinvariance techniques(ESPRIT) for the signals’ direction of arrival(DOA) and high orderstatistics has been proposed. Planar coordinates of PD sources can be obtained by solvingthe intersection of two straight in DO A, so solving nonlinear equations is avoided. Signalscollected in substation is analyzed by this algorithm, and the PD localization error is withinIm.PD UHF signals are modeled by autoregressive moving average (ARMA) model, and thetheoretical basis and realization method of the model order and parameters estimation usinghigher order cumulants are proposed. The amplitude of signal’s Fourier transformation withconstant amplitude factor can be reconstructed by parameters of ARM A model, and the phasewith linear phase difference can reconstructed by bispectrum estimation. Then thetime-domain signal with constant factor and time shitf is constructed. Based on Fisher-likeclass separation degree between signals characteristic parameters are selected from thereconstructed spectrum or the bispectrum of signal. The characteristic parameters are utilizedto train the radial basis function neural network (RBFNN) for separating the observed signals.Finally, the prototype of substation PD monitoring and fault warning system has beendevelopped, and the feasibility of PD source localization in substation is verified by theresults of tests in laboratory and substation...
Keywords/Search Tags:Subsation, Partial discharge, UHF, Localization algorithm, Timedelay estimation, High-order statistics, Array signal processing, Signalreconstruction, Signal separation
PDF Full Text Request
Related items