Transformer vibration signal de-noising method based on EEMD kurtosis threshold value

Abstract

The invention discloses a transformer vibration signal de-noising method based on an EEMD kurtosis threshold value. Many methods have been put forward by domestic and foreign scholars for de-noising of non-stationary signals and are roughly divided into three major types of time-frequency analysis methods based on time domain statistics, Fourier transform and wavelet transform. All the methods have respective advantages and disadvantages, for instance, the theory of the frequency domain method is mature, but signals with overlapped frequency are difficult to separate; although the wavelet transform method has the multi-resolution performance, the de-noising effect of the method usually depends on selection of a wavelet base and a threshold value. The method includes the steps that first, random signals are subjected to EEMD, then, each intrinsic mode function is subjected to autocorrelation function calculation, the kurtosis coefficient of each IMF is solved, next, the IMFs are differentiated through the threshold value, and therefore noisy signals can be removed. The method is used for de-noising of transformer vibration signals.

Description

Technical field: [0001] The invention relates to a transformer vibration signal noise reduction method based on EEMD kurtosis threshold. Background technique: [0002] As an important hub equipment in the power system, the transformer is not only very expensive, but also plays a very important role. Ensuring its safe and reliable operation is the key to the safe operation of the entire power grid. Therefore, real-time detection of its operating status is very necessary. At present, the transformer fault diagnosis method based on vibration signal analysis is a new type of diagnosis method, which has the characteristics of real-time and online operation, and has been widely used in power systems. However, the actual collected signals in the project contain a lot of noise. For example, when the transformer equipment fails in the early stage, the collected vibration signal contains random noise, which makes the effective information that can reflect the fault characteristics bec...

AI Technical Summary

Problems solved by technology

At present, scholars are also trying to use the EEMD method to denoise the signal. For example, using the characteristic that the energy density and average period of each component of white noise decomposed by EEMD are constant, the noise components that meet the conditions are eliminated, but the eliminated components There may be useful signals, and the solution is not given in the paper; there are also methods that use the difference between white noise and ordinary signal autocorrelation functions, first perform EMD decomposition on the noisy signal, and then artificially judge which IMF is the noise component, and then Perform soft threshold processing on the noise component. Although this method can effectively suppress the noise, the distinction between the noise component and the useful component can only be judged by humans, which has a certain degree of subjectivity.

Images