Analysis on transformer vibration signal recognition based on convolutional neural network

Abstract

Listen

Translate article

In order to study the relationship between the transformer vibration and the operation state, the wavelet analysis method and the convolutional neural network method were used to analyze the transformer vibration signal. This paper proposes a transformer based on convolution neural network-based surface vibration signal feature extraction method. The result show that the convolution of neural network in different station transformer surface vibration signal classification has a lot of advantage, as the integration of feature extraction and classification recognition process together can effectively classify vibration signal recognition processing. This method is feasible for classification and identification by providing an accuracy value of 92.74 %. The future perspective of this research will focus on a generalized network model and parameters through experimentation for further investigation of accuracy and efficiency of this method.