Study On The Characteristics Information Of Partial Discharge In Transformer Oil-paper Insulation And Its Development Characteristics

Power transformers are critical to safe operation of power grid equipment, Oil-paper insulation is the most common insulation method of power transformer., And the event of partial discharge can lead to aging and disruption of the insulation, therefore carry Partial Discharge characteristics, to understand the development of operation Transformer's Partial Discharge and the aging conditions of materials, is a very good response to internal insulation of latent defects to determine the extent of transformer inside of insulation deterioration, And the failure nipped in the bud, safe and reliable operation for transformer has very important significance..This paper analyzes the transformer insulation structure and the common failure of partial discharge, designs oil-paper partial discharge of transformer test platform, and makes the surface discharge model and the air-gap discharge model in oil, with step-up method and the constant pressure method to study the partial discharge characteristics. By analyzing the changes of time-domain signal, and the variation of two-dimensional map,surface discharge in oil along with the increase of applied voltage or discharge time of growth, maximum discharge magnitude and discharge frequency increasing trend emerged; air-gap discharge in oil along with the increase of applied voltage or discharge time of growth, maximum discharge magnitude and discharge times emerged first increase and then decrease and finally increased. The characteristic parameters of phase spectra H qmax (?) of the surface partial discharge in oil: Sk showed a growth trend, Ku first increases and then decreases. The characteristic parameters of phase spectra H n(?) of the surface partial discharge: Sk and Ku are both greater than zero, and showing a gradually increasing trend. The characteristic parameters of phase spectrum H qmax (?) of the oilpaper air-gap discharge: Sk and Ku increase with the growth discharge time. Sk of the phase spectra H n(?)presents first grows and then increases tendency.In the analysis of kernel principal component analysis (KPCA) , the 29 statistical operators are reduced, According to the information generated at different periods of the process combined with the clustering results, the partial discharge development process can be divided into initial discharge stage, discharge development stage, discharge stabilization stage and pre-breakdown stage. Using wavelet neural network, combined with dynamic clustering of results, determine the network input and output structure, partial discharge at different stages of the development process are identified, the results in good agreement with the cluster analysis.