CN-121997132-A - Signal processing method and system combining singular value decomposition and wavelet threshold denoising

CN121997132ACN 121997132 ACN121997132 ACN 121997132ACN-121997132-A

Abstract

The invention discloses a signal processing method and a system for combining singular value decomposition and wavelet threshold denoising. The method comprises the steps of obtaining a fault signal containing noise, carrying out matrix conversion and singular value decomposition on the fault signal, calculating energy contribution values of the singular values, processing wavelet decomposition coefficients by adopting improved thresholds and threshold functions on the singular values with the energy contribution values higher than a set threshold, and reconstructing an original signal through inverse wavelet transformation. By implementing the method of the invention, more flexible and effective denoising can be realized, and the precision and reliability of fault detection are improved.

Inventors

LI CHENYING
ZHOU LI
CAO JINGXING
ZHANG WEI
TAN XIAO
WANG QI
WU SHUQUN
ZHANG YIMING

Assignees

国网江苏省电力有限公司电力科学研究院
国网江苏省电力有限公司
江苏省电力试验研究院有限公司
东南大学溧阳研究院

Dates

Publication Date: 20260508
Application Date: 20260121

Claims (10)

1. The signal processing method combining singular value decomposition and wavelet threshold denoising is characterized by comprising the following steps: Acquiring a fault signal containing noise; performing matrix conversion and singular value decomposition on the fault signals, and calculating energy contribution values of the singular values; the singular values with energy contribution values higher than the set threshold are processed with wavelet decomposition coefficients using improved thresholds and threshold functions and the original signal is reconstructed by inverse wavelet transform.
2. The signal processing method combining singular value decomposition and wavelet thresholding according to claim 1, wherein said performing matrix conversion, singular value decomposition on the fault signal, and calculating energy contribution values for each singular value comprises: SVD decomposition is carried out on the fault signals, and the fault signals are converted into matrixes according to a Hankel matrix construction principle; And carrying out singular value decomposition on the matrix to obtain a group of singular values describing different components of the fault signal, and calculating an energy contribution value corresponding to each singular value.
3. The signal processing method combining singular value decomposition and wavelet thresholding according to claim 2, wherein the energy contribution value is calculated by the formula , wherein, ; ; Is a singular value.
4. The signal processing method combining singular value decomposition and wavelet thresholding according to claim 3, wherein after said performing matrix conversion, singular value decomposition on said fault signal and calculating the energy contribution value of each singular value, further comprising: and setting the singular value with the energy contribution value not higher than the set threshold value to zero.
5. The signal processing method combining singular value decomposition and wavelet thresholding according to claim 1, wherein said processing the wavelet decomposition coefficients with improved thresholding and thresholding functions for singular values with energy contribution values above a set threshold and reconstructing the original signal by inverse wavelet transform comprises: Selecting a wavelet basis function and a decomposition layer number according to signal characteristics corresponding to singular values of which the energy contribution values are higher than a set threshold value, and executing wavelet transformation on the signals to obtain wavelet decomposition coefficients; processing the wavelet decomposition coefficients using an improved threshold and threshold function to obtain processed wavelet decomposition coefficients; and carrying out inverse wavelet transformation on the processed wavelet decomposition coefficients to reconstruct the original signal.
6. The signal processing method combining singular value decomposition and wavelet thresholding according to claim 5, wherein the improved thresholding is ; The number of the decomposition layers; is an improved threshold.
7. The signal processing method combining singular value decomposition and wavelet thresholding according to claim 6, wherein the thresholding function is expressed as A, b is a function adjusting factor, a is an adjustable constant; 。
8. The signal processing method combining singular value decomposition with wavelet thresholding according to claim 4, wherein the wavelet basis function is a cofi wavelet basis function.
9. The signal processing method combining singular value decomposition and wavelet thresholding according to claim 4, wherein the number of decomposition layers is 5.
10. A signal processing system combining singular value decomposition with wavelet threshold denoising, comprising: An acquisition unit configured to acquire a noise-containing fault signal; the computing unit is used for performing matrix conversion and singular value decomposition on the fault signals and computing energy contribution values of the singular values; And the reconstruction unit is used for processing the wavelet decomposition coefficients by adopting improved thresholds and threshold functions on the singular values with energy contribution values higher than the set threshold, and reconstructing the original signals through inverse wavelet transformation.

Description

Signal processing method and system combining singular value decomposition and wavelet threshold denoising Technical Field The invention relates to the technical field of fault diagnosis of a low-voltage cable, in particular to a signal processing method and system combining singular value decomposition and wavelet threshold denoising. Background In the field of fault detection of low-voltage power supply cables, signal quality plays a decisive role in the accuracy and reliability of fault feature extraction. However, in actual operation, accuracy of signal processing is greatly limited due to noise interference widely existing in impulse noise, periodic interference, non-stationary noise, and the like. Particularly under the requirement of high-precision fault detection, the traditional denoising method often has a great deal of trouble. Conventionally, SVD (singular value decomposition ) denoising methods rely on manual experience to determine the number of singular values and preset a fixed cutoff threshold based on these experiences. This approach is not ideal for cable fault signals with time-variability and non-stationarity because a fixed threshold cannot accommodate signal characteristic variations in different operating environments, which can easily lead to loss of key signal features or failure of effective removal of residual noise. In addition, although SVD can capture the main component of a signal through matrix decomposition, it has limited retention capability for local abrupt features such as discharge pulses, which makes it poor in handling cable partial discharge faults with obvious transient features. Wavelet transformation is used as a common denoising means, noise is suppressed by setting different thresholds, but a single wavelet threshold denoising algorithm has inherent defects, a hard threshold function can cause signal distortion, and a soft threshold function can cause detail information loss. In a mixed noise environment such as white noise superimposed power frequency interference, it becomes a challenge to simultaneously preserve global features and reconstruct local details. Conventional methods typically use a fixed threshold for processing, but this may lead to problems of excessive filtering of the useful signal in low frequency coefficients and residual noise in high frequency coefficients as the number of wavelet decomposition layers increases. In summary, the prior art faces many challenges in coping with cable fault detection in complex noise environments. Therefore, it is necessary to design a new method to implement more flexible and effective denoising, and improve the accuracy and reliability of fault detection. Disclosure of Invention The invention aims to overcome the defects of the prior art and provides a signal processing method and a system combining singular value decomposition and wavelet threshold denoising. In order to achieve the above purpose, the invention adopts the following technical scheme that the signal processing method combining singular value decomposition and wavelet threshold denoising comprises the following steps: Acquiring a fault signal containing noise; performing matrix conversion and singular value decomposition on the fault signals, and calculating energy contribution values of the singular values; the singular values with energy contribution values higher than the set threshold are processed with wavelet decomposition coefficients using improved thresholds and threshold functions and the original signal is reconstructed by inverse wavelet transform. The further technical scheme is that the method for performing matrix conversion and singular value decomposition on the fault signal and calculating the energy contribution value of each singular value comprises the following steps: SVD decomposition is carried out on the fault signals, and the fault signals are converted into matrixes according to a Hankel matrix construction principle; And carrying out singular value decomposition on the matrix to obtain a group of singular values describing different components of the fault signal, and calculating an energy contribution value corresponding to each singular value. The further technical proposal is that the calculation formula of the energy contribution value is as follows, wherein,;;Is a singular value. The method comprises the following steps of performing matrix conversion and singular value decomposition on the fault signal, and calculating energy contribution values of the singular values, wherein the method further comprises the following steps: and setting the singular value with the energy contribution value not higher than the set threshold value to zero. The method further comprises the technical scheme that the improved threshold and threshold function are adopted for the singular values with energy contribution values higher than a set threshold to process wavelet decomposition coefficients, and the original signals are re