Research On Fault Diagnosis Technology Of Transformer By Vibration Method

Power transformer is a static device, used to change a numerical AC voltage into several different voltages of the same frequency. It is the important equipment in power system, the safe and reliable operation of the power transformer is the key of entire power grid safe operation. In recent years, the diagnosis method to test the transformer fault based on the vibration signal is taking shape, the specific system of transformer fault diagnosis based on vibration signal is to extract features further by collected and analysis vibration signal, according to the final feature extraction to evaluate the working condition of transformer core and windings. A large number of experiments show that when the transformer faults, the vibration signal is not stationary, but non-stationary. How to analyze the non-stationary signals, this paper uses the method of wavelet analysis to analysis signals, but the vibration signal of transformer fault diagnosis method based on wavelet analysis has a defect, it is that the wavelet decomposition coefficient has to meet the Gauss distribution, this is not the case in reality, this traditional Gauss distribution assumption is conflict with the compressed characteristics of wavelet transform.This paper first analyzes the generation mechanism of transformer fault signal, put forward the analysis method of wavelet transform, summarizes the knowledge of wavelet transform theory. Then, build a transformer vibration model and the signal is divided into several fault types of appropriate. After the wavelet transform of signals, using the energy spectrum method to analyze fault analysis, we get the generalized Gauss model. According to the probability density function of the generalized Gauss distribution is used to fit the signal, extracting the parameters of generalized Gauss distribution of wavelet coefficients as the feature of fault diagnosis signal. Extracting the parameters of generalized Gauss distribution of wavelet coefficients as the feature of signals for fault diagnosis and combined with support vector machine for signal classification. When considering the relativity of the wavelet coefficient, fault diagnoseis to signal with Wavelet Domain Hidden Markov model are established combine with the energy spectrum. Finally, to realize wavelet transform in the VC platform, count the energy and to realize the detection of fault signal.