Research Of PD On-line Monitoring And Pattern Recognition On The Dry-type Transformer

As one type of electrical transformers, dry-type transformers are often used in the special occasions,so they are very important in electric power system.Partial discharge is not only the main factors causing insulation deterioration and destruction of dry-type transformer,but also is an important sign of the insulation defects;There are different types of partial discharge,and the different discharges are caused by different insulation defects. It will contribute to find the insulation defect nature and location of the transformer more quickly if we can judge the type of partial discharges,and it is convenient for people to examine and repair.So it has important significance to do the research on PD on-line monitoring and pattern recognition technology of dry type transformer.Firstly, this paper analyzes the occurrence mechanism of the partial discharge and the reason why partial discharge occurs the ultrasonic wave,as well as studies the insulation defect types of the dry-type transformer.It is discovered that the factors that causes the insulation defects are mainly the selection of raw material, design of insulation structure and casting process of winding,etc,besides,they correspondingly occurs the tip discharge, creeping discharge and air discharge. For these three discharge types, the author produces three discharge models, and simulates the three types of discharge occurrence in dry-type transformers under the laboratory conditions.The raw data of PD ultrasonic signals is obtained.Secondly, the transform of wavelet packet analysis theory is very suitable for analyzing non-stationary signal. In this paper,the wavelet packet theory is used to decompose the partial discharge ultrasonic signal,the statistics of coefficients are extracted as pattern recognition input features which contains the most information energy.It identifies partial discharge type by means of Probabilistic neural network(PNN)classifier at last.Finally, The partial discharge online monitoring and fault early warning system of dry-type transformer based on ultrasonic signal is developed on the platform of virtual instrument.The system is writted by LabVIEW graphical programminglanguag and Matlab software tool.It can realize many functions,such as preservate,display the discharge waveform on-time,set discharge threshold and fault warning and so on.It also can analyze the saved datas,which provides the function such as history waveform reproduction,trend analysis, partial discharge PD source location and PD type of pattern recognition etc.This system can provide many informations of PD to operation personnel for the dry-type transformer’s safe operation.It has been successfully used on the main transformer of the offshore drilling platform.