The Wavelet De-moise And Pattern-recognition Research In GIS PD

GIS is one of the important equipment of power system, its main trouble is inside insulating degradation. The measurement of the partial discharged signal offered important information for insulating whole situation within GIS. The partial discharges monitoring system divided into the measurement of the PD signal,the suppression of external disturbance,atlas analyse,extract eigencector and indentify&classify.This paper uses the ultrasonic method to measure the PD sigal, utilize the external sensor to measure the ultra acoustical signal let out from GIS and deal with in order to judge whether it's inside discharges to happen. The test prescent the method of sensor access, the design of passive wave filter, DSP providing choosing and the setup of bunches of communicationThe noise limits accuracy that trouble diagnoses of noise of the signal. This limitation will cause and can't take the proper remedy in time, thus cause the insulating state of GIS to worsen and even puncture further. For the background noise question appearing while measuring to the small signal, this text uses the wavelet to analyse the law of threshold value to deal with the signal. Compared with traditional method, the analytical method when small wave analysis is one kind of time-frequently method, distinguish more characteristics. This kind of characteristic reflects the non-steady signal sensitivly, apt to measure the unusual different signal. This method is practical and effective, and it is in common use to compare, even under the strong environment that interferes, can be made the gread filtering effectiveness when it can be choose the suitable threshold.This text has discussed two kinds of method that choose characteristic parameters which are theφ-q-n parameter and wavelet vary of characteristic parameter, use the latter one to carry on the research of the pattern-recognition. Through to carrying on the modeling, sample and is trained and tests actually because of the multi-layer perceiving machine of BP neural network, and then identify to the creedping discharge,corona and airspace discharge sigal. The experimental result show that utilizing the swavelet to vary and draw the pulse of discharging individually and discharge, the pattern-recognition method that the characteristic parameter carry on is superior to the traditional method, it have reduced the dimension of the signal with vectorial characteristic remarkably, and not only simplified the design of the neural network, also improved the precision of discerning.