Data enhancement method for vibration signals of converter transformer

Abstract

The invention relates to a data enhancement method for vibration signals of a converter transformer, and belongs to the field of electric power. The method comprises the following steps: firstly, a Markov transition field (MTF) matrix of a vibration signal sequence is calculated, and the matrix is converted into a two-dimensional feature map to retain the correlation of a time sequence; and on this basis, a self-attention module is introduced to optimize a generative network, which can accept information between long-distance pixel points and realize a global receptive field between the pixel points. Meanwhile, the identification network adopts convolution kernels of different sizes for parallel feature extraction so as to increase the range of a receptive field and obtain deeper features. According to the model, a two-dimensional image data set is established according to actually measured converter station vibration signals, effective training data can be further generated, and the problem that the converter transformer state recognition accuracy is low due to the fact that the number of data sets is unbalanced is solved.

Description

technical field [0001] The invention belongs to the field of electric power and relates to a data enhancement method for vibration signals of converter transformers. Background technique [0002] With the development of computer technology and the further upgrading of hardware, more and more fault classification models of power transmission and transformation equipment have emerged in the field of deep learning. In the EHVDC transmission system, the converter transformer is an important part, and its safe operation is directly related to the stability of the entire transmission system. At present, the state assessment of transformers based on vibration signals has attracted widespread attention, but there are few studies on its state assessment methods at home and abroad. The main reasons are as follows: [0003] (1) The converter transformer is the key equipment for UHV DC transmission, and its cost is very expensive. It is difficult to obtain vibration data under specia...

AI Technical Summary

Problems solved by technology

[0003] (1) The converter transformer is the key equipment for UHV DC transmission, and the cost is very expensive

It is difficult to obtain vibration data under special working conditions, which will also lead to a serious imbalance in the number of samples in its related data sets

[0004] (2) The vibration signal of the converter transformer is highly complex, and its deep learning model training based on time series is difficult and inefficient

[0008] (1) Due to the insufficient feature extraction ability of the fully connected layer or the convolutional layer of the one-dimensional sequence for the one-dimensional time series, the quality of the generated data is very poor

[0009] (2) There are a large number of parameters in the one-dimensional convolutional layer and the fully connected layer that need to be trained to make it difficult for the network to converge

Images