Research On Fault Diagnosis Of Power Transformer Based On Acoustic Features

As one of the most important equipment in the power system, the transformerundertakes the voltage transformation from primary side to secondary side, the distributionand transmission of power and some other critical tasks. Therefore, the technology ofelectrical equipment’s failure detection is a practical subject combined with engineeringpractice which plays an important role in the practical application of electrical productions.It is of great theoretical and practical significance to study on an ideal diagnostic methodfor power transformer failures’ prediction and early detection.This article is based on the acoustic signal as the characteristic signal which is emittedfrom the operation transformer to carry out the fault diagnosis. The electric and magneticfields do not interfere with the detection and dissemination of the acoustic signals, and themonitoring process does not affect the normal operation of the transformers, so it is a moreeffective means to detect the failure of the transformers by the acoustic signals. This papermainly studies the audible sounds with frequency band of20Hz-20kHz. This paper hasmainly studied the theoretical mechanism of acoustic signal emitted by different types oftransformer failure and its physical characteristic. Capacitive acoustic sensor or stethoscopehave been used to get the acoustic signal on the site. The paper has mainly studied thewavelet algorithm and valve algorithm to filter out the noises and compared with the resultsof traditional methods such as frequency spectrum analysis, wavelet analysis andHilbert-Huang transformation. Farther more, field test data have been obtained from on-sitetransformer sound signals, and analyze the sound signals by using software to get thephysical characteristics. In the initial stage of the research, laboratory has set up a platformby studying the discharge characteristics of the physical volume of the sound signals to setup the foundation for transformer fault detection.Research shows that by using the acoustic signal’s characteristic to detect thetransformer failure is fast and accurate. By using the data processing method in this paper,the characteristic of the signal can be used to decide the failure types of transformer. Thesound signals obtained from laboratory simulation discharge test have a fundamentalfrequency of250Hz is one of the most important characteristic. The next step of this study will focus on collecting a large amount of sound signals and improving the intelligentdatabase design of the transformer fault diagnosis.