The Method Of Recognition Research For The Pattern Of Partial Discharge In GIS Based On The Fingerprint Of Acoustic Emission

Gas insulated station(GIS)is widely used in high voltage power transmission field for its small area covered,less maintenance workload,excellent insulation performance and high reliability.With the increase of the grid voltage level and system capacity,the fault inside the GIS equipment also increased.So it is very important to find the method for evaluating the GIS's internal state effectively.As the commonly used method for diagnosising and assessmenting the GIS internal insulation,the PD testing of acoustic emission is currently only depending on the amplitude and the phase to realize rough pattern recognition.Abundant information contained in the acoustic emission signals has not been applicated effectively.It is unable to provide guidance and support powerfully for assementing and diagnosising the condition of the GIS equiment insulation and the severity of the discharge.So the corresponding research is urgently needed to carry out.This paper obtain a large amout of AE PD signals original data of the high voltage spike,low voltage spike,suspension,free metal particles and the surface defects in the GIS through the measured,based on the research simulated in the laboratory.According to the difference of the spectra of PD type,extract the statistical characteristic parameters and fractal characteristic parameter and structure feature vector.Obtain the vector of the statistical characteristics,the fractal characteristic and the combined characteristics from the acoustic emission,that can be used for the recognition of the PD pattern.And realize the automatic identification based on the fingerprint of the acoustic emission by using BP artificial neural network.Upon examination,using integrated eigenvector to research the BP artificial neural network pattern recognition works best.The accuracy of Creeping discharge is the highest recognition as 94.1%.And the accuracy of high voltage spike discharge is the lowest recognition as 75.5%.The proposed method of PD pattern recognition based on acoustic emission signals can be applied to evaluate the state of GIS equipment insulation.