Ferromagnetic Resonance Identification of Built-in Instrument Transformer for GIS Based on Improved Kurtosis Graph and GoogLeNet
In view of overheating,insulation cracking and even damage of instrument transformer due to ferromagnetic resonance of the dead tank electromagnetic potential transformer(PT)in the gas insulated metal enclosed switchgear(GIS),a method for identifying the ferromagnetic resonance of the instrument transformer in GIS is proposed.Firstly,aiming at the problem that ferroresonance fault is easily disturbed by uncertain noise and causes signal feature drowning,the frequency band of fast spectral kurtosis is refined by interpolation,the sub bands is used to enhance of decomposition of the high-frequency part,and keep the expression of the high frequency part of the signal in the kurtosis graph more sufficient,and the characteristics of the ferromagnetic resonance signal is preserved in the form of images.Then,on this basis,the structure of GoogLeNet convolutional neural network is adjusted in two aspects:on the one hand,the convolution operation is added to enhance the network's ability to extract potential features;on the other hand,reducing the number of sampling layers solves the problem of insufficient measured data.Finally,the method is verified by using simulation and laboratory monitoring data,and the results show its effectiveness in improving fault identification accuracy and saving training time.
GISelectromagnetic voltage transformerferromagnetic resonanceparameter matching
万方数据
维普
