Features of vibration signal of power transformer using local wave method

Abstract

The mechanical faults of a transformer are predicted with monitoring its vibration, which is a new technology put forward recently for condition based monitoring of transformers. The key technology of vibration method is how to identify the effective features of the vibration signal. Here, a new self-adapting time-frequency analysis method - local wave method which is developed from the empirical mode decomposition is applied, and it is combined with the Hilbert transform to get the time-frequency distribution of local wave, which can qualitatively describe the relationship between time and frequency, and achieves a effective recognition of the non-linear, non-stationary signals. Experimental results in this paper demonstrate that the features of the vibration signal can be effectively found out with local wave method, and the energy distribution of the vibration signal to frequency bands can obtained. The mechanical status of the core and winding of a transformer can be effectively diagnosed by the frequency bands-energy distribution. The local wave method is a effective method for diagnosing the mechanical faults and the severity of the faults of power transformers.