Research On Fault Diagnosis Method For Distribution Transformer Based On Acoustic Signal

Distribution transformer is an important power equipment direct user-oriented and its safe and reliable operation is the key to ensure the power supply for the normal production and life of users.The distribution transformer has the characteristics of wide distribution,large number and low value,so it is not suitable to monitor and diagnose the faults in the way of on-line monitoring.Therefore,it is significant to study the distribution transformer inspection method and develope the inspection equipment for improving the reliability of distribution transformer.Based on the characteristics that the acoustic signal under the different operating conditions of the distribution transformer is not the same,it is proposed to use.the noise diagnosis method to diagnose the distribution transformer,so as to provide a new inspection method and means for the daily inspection of the distribution transformer.In this paper,the ontology noise and fault sound signal generation mechanism of distribution transformer are studied.The mechanism and influencing factors of the noise signal of distribution transformer core and winding are studied.The mechanism of the fault acoustic signal of the distribution transformer is analyzed,and the mechanism of the discharge acoustic signal of the distribution transformer is analyzed emphatically.It provides the theoretical support for the fault diagnosis of the distribution transformer based on the acoustic signal.The method of analysing acoustic signal is studied in this paper.In view of the complicated environment of the distribution transformer and the problem of excessive interference,we propose a fast independent component analysis(FastICA)based on negative entropy to separate the target sound source.The complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN)which is the improve method of the ensemble empirical mode decomposition(EEMD)algorithm is used to analyze the non-stationary fault signal,solving the problem that it is difficult to completely eliminate the additional Gauss white noise in the method of EEMD.In this paper,the fault diagnosis method based on acoustic signal is proposed.Firstly,the target sound source is separated by FastICA algorithm.Then,The signals is decomposed separately by CEEMDAN to extract the singular spectral entropy which reflects the complexity of the signal and the degree of irregularity?the energy entropy of the CEEMDAN band energy entropy which reflects the energy characteristics of the signal and the marginal spectral entropy of the time-frequency characteristic of the signal,and the center-of-gravity frequency as the feature quantity.Aiming at the problem that the interference noise around the distribution transformer is too much and it is difficult to extract the feature of all the disturbing noise,a single support vector data description(SVDD)algorithm is proposed.The training sample only needs the characteristics of the target sample.The multi-classification of SVDD model is achieved by establishing SVDD models of different faults,and the important parameter rejection rate and kernel function in SVDD method are optimized by particle swarm optimization algorithm.In this paper,the acoustic signals of distribution transformer under different working conditions are collected and analyzed.Through the experiment,the typical needle discharge,the discharge of the surface discharge and the floating discharge sound signal in the distribution transformer are simulated,and the noise signal of the discharge fault is mixed with the noise of the distribution transformer in the laboratory.The proposed method is used to classify and identify the discharge fault.In addition,the noise signal of the distribution transformer in different working conditions is mixed with the acoustic signal of the discharge fault,and the method is used to classify and identify the discharge fault.The experimental results show that the proposed method can identify three different types of distribution transformers Discharge failure,in considering the different conditions of the sound signal mixture,the recognition rate can reach more than 91.5%.The proposed method can be used for distribution transformer fault diagnosis.