Study On Intelligent Recognition Technology About GIS Partial Discharge Faults

GIS (Gas Insulated Substation) is widely used in power systems for its advantages ofgood insulation, compact structure and easy maintenance. In case of malfunction of the GIS,power system will break down, causing great loss. Therefore, it is important to achieve faultdiagnosis of the GIS equipments.In this paper, a comparison between various detection methods of GIS equipment ismade, and the application of BP neural network in malfunction recognition is introduced.For the identification of partial discharge fault, in this paper a signal recognition system forGIS partial discharge is designed based on BP neural network. And the structure of systemand process for data analysis are introduced. The formation about spectrum of dischargesignals and extraction of characteristic quantities for spectrum are introduced. A Labviewprogram is designed to extract and solve the characteristic quantities, as well as theconstruction and training of the BP neural network are realized based on mixedprogramming using Labview and Matlab. The normalization of training samples in neuralnetwork are discussed and achieved. In order to determine the parameter of the number ofneural network hidden layer units, an experiment about the effect of test is made bychanging thus parameter, then the parameter with which the experiment get the best effectis chosen as the neural network hidden layer unit. Then, the BP neural network using thenetwork parameters obtained from the training process is rebuilt and simulated in Labview,realizing the function of classification and identification. Finally, a design of identificationscheme based on embedded processor is introduced briefly. In the end, the paper drawsconclusions as follows: BP neural network do achieve the intelligent recognition of GISpartial discharge, and the recognition rate and accuracy are high enough to meet therequirements of equipment fault detection.