The Characteristic Information Extraction Of Air-Gap Discharge In Transformer Oil-Paper Insulation And Research On Its Process Dividing

With the gradual forming of national grid pattern, which includes transmission ofelectricity from the western to the eastern region, cross feed between north and south,and network around the country, the EHV/UHV transmission has become the trend ofthe grid development. Therefore, higher requirement for the safe and reliable operationof the electric instruments is put forward. As one of the crucial equipment in the powersystem, stable and reliable electricity supply of power transformer is meaningful for thehealthy development of our national economy. Due to the combined effect of electric,hot, and mechanical factors, the transformer often fails when it is operating under thenormal voltage. Air-gap discharge is the main fault form of partial discharge, which isone of the main transformer fault, and that is why the research on the developmentalcharacteristics of air-gap discharge in oil-paper insulation is significant for the diagnosisand monitor of power transformer.In order to study the varied rule of air-gap discharge in the oil-paper insulationform emerging, growing and break-down, this essay does the simulative experiment ofair-gap discharge by constant-voltage method based on the oil-paper discharge testplatform. The essay begins with the equivalent circuit of single air gap discharge model,and emphasizes the varied rule of the power during the discharge process, and alsocombined the numbers of discharge with the varied power to raise a newcharacteristic—the average discharge of per second which better describes the processof the characteristics of the discharge in different discharge stages.In order to divide the development of air-gap discharge into different dischargestages, this paper learn from the idea of PRPD, draws φ-W-n3D images and gets grayimages of discharge energy-phases. Summarizing the feature extraction methods formthe gray image, the wavelet moments in different scale parameters and translationparameters are extracted from discharge energy-phases gray images using waveletmoments algorithm.62discharge samples are extracted in different times of air-gapdischarge experiment and18-dimensional wavelet moments are calculated from positiveand negative power frequency of the energy-phase gray image of every sample.According to the “hold together” characteristic of these wavelet moments in the featurespace,62samples are divided as start discharge and shock develop stage, weakdischarge stage, outbreak discharge stage and pre-breakdown stage using fuzzy C-means clustering method. Compared with the result using the traditional systemclustering method, the clustering result using fuzzy C-means clustering method is better.