Blind Separation of Transformer Vibration Signal Based on Sensitive IMFs

Abstract

The vibration signal measured from the wall of the transformer oil tank is a mixture of various vibration source signals, and it is difficult to extract the characteristics of the transformer vibration signal and fault diagnosis. In this paper, a blind separation method of transformer vibration signal based on sensitive IMF is proposed aiming at the problem that the single-channel transformer vibration signal is difficult to separate. First, this paper analyzes the transformer vibration characteristics, and then the Intrinsic Mode Function (IMF) decomposed by the Variational Mode Decomposition (VMD) algorithm is screened and reconstructed based on the sensitive factor. The VMD algorithm is combined with the Fast Independent Component Analysis (FastICA) algorithm to separate the vibration signal. The method is verified by constructing a simulation signal according to the transformer vibration characteristics, and the method is applied to separate the transformer vibration signal to prove the effectiveness of the method.