Winding fault diagnosis of power transformer based on vibration distribution features

摘要

Vibration analysis method is an important means to realize live monitoring and fault diagnosis of power transformers. The key to the fault diagnosis method based on vibration analysis is to extract state features (values or vectors) from complex vibration signals on oil tank wall. Most of the traditional state feature extraction methods choose a single measured point's vibration signal to extract its features of time domain or frequency one, and ignore vibration distribution features among various measured points. Here, amplitude barycenter and barycenter trajectory of vibration distribution were studied and analyzed from the view point of vibration barycenter to propose 4 quantized parameters. Based on these 4 quantized parameters and the support vector machine (SVM) classification algorithm, the winding fault diagnosis model for power transformer based on vibration distribution features was proposed. The measurement results of actual power transformer winding fault tests and analysis ones of measured data samples acquired from more than 10 power transformers on site showed that the proposed vibration distribution features and quantized parameters can effectively reflect changes in transformer winding deformation and compression force relaxation, etc.;the winding fault diagnosis model based on vibration distribution features can also correctly detect and diagnose mechanical structure states of power transformer winding.

关键词