Search

CN-122017716-A - Noise-dyeing signal denoising method and system for field verification of transformer partial discharge high-frequency current sensor

CN122017716ACN 122017716 ACN122017716 ACN 122017716ACN-122017716-A

Abstract

Aiming at the problem that a field calibration signal of a transformer partial discharge high-frequency current sensor is seriously polluted by field electromagnetic noise interference, the invention provides a self-adaptive denoising method based on Meyer wavelet and an improved smoothing threshold function, which comprises the steps of firstly, arranging a standard sensor according to an optimal penetration distance to obtain an original noise-dyeing signal, then, adaptively determining a decomposition level by data length, selecting Meyer wavelet for multi-scale decomposition, smoothly shrinking wavelet coefficients by adopting a continuously-conductive and progressive unbiased improved threshold function, and finally, reconstructing a high signal-to-noise ratio calibration signal by wavelet inverse transformation; the method does not need manual intervention on a threshold value, maintains orthogonality and high-order conductivity in the whole process, and is beneficial to improving the accuracy of the checking result of the high-frequency sensor to be tested.

Inventors

  • DING GUOCHENG
  • XIE JIA
  • YANG HAITAO
  • WU JIE
  • XIE YIMING
  • WU XINGWANG
  • HU XIAOYU
  • LIU WEI
  • LI JIANLIN
  • HUANG WEIMIN

Assignees

  • 国网安徽省电力有限公司电力科学研究院

Dates

Publication Date
20260512
Application Date
20251215

Claims (10)

  1. 1. The method for denoising the dyeing noise signal of the transformer partial discharge high-frequency current sensor field verification is characterized by comprising the following steps of: The method comprises the steps that firstly, an upper computer collects an original dyeing noise check signal S (t) output by a sensor to be tested; Step two, self-adaptively determining a wavelet decomposition level J according to the signal length N, selecting a proper wavelet basis function, and performing discrete wavelet decomposition on a dyeing noise signal S (t) to obtain wavelet coefficients of all scales; Step three, performing layer-by-layer shrinkage processing on the wavelet coefficient by adopting a continuously-conductive progressive unbiased improved threshold function, wherein the improved threshold function is as follows: where lambda is the common threshold, sigma is the noise standard deviation estimate, Representing the target wavelet coefficient processed by the threshold function; And step four, carrying out signal reconstruction by adopting wavelet inverse transformation on the denoised signal to obtain a denoised check signal.
  2. 2. The method of claim 1, wherein the wavelet basis function of step two selects a Meyer wavelet suitable for the denoising problem, and the frequency domain expression is as follows: Wherein beta (x) is an auxiliary tool, satisfying 。
  3. 3. The method of claim 1, wherein the wavelet decomposition level J of step two is calculated by the following formula: , where N is the length of the data signal S (t), Representing a downward integer.
  4. 4. The method according to claim 1, characterized in that the noise standard deviation σ employs a robust median estimate: Wherein the method comprises the steps of And (5) processing the layer 1 target wavelet coefficient for the threshold function.
  5. 5. The utility model provides a high frequency current sensor field verification's noise-dyed signal denoising system is put in transformer, its characterized in that includes: The signal acquisition module is used for acquiring an original dyeing noise check signal S (t) output by the sensor to be detected by the upper computer; the wavelet decomposition module is used for adaptively determining the wavelet decomposition series J according to the signal length N, selecting a proper wavelet basis function, and performing discrete wavelet decomposition on the dyeing noise signal S (t) to obtain wavelet coefficients of all scales; The denoising module is used for performing layer-by-layer shrinkage processing on the wavelet coefficient by adopting a continuously-conductive progressive unbiased improved threshold function, wherein the improved threshold function is as follows: where lambda is the common threshold, sigma is the noise standard deviation estimate, Representing the target wavelet coefficient processed by the threshold function; and the signal reconstruction module is used for carrying out signal reconstruction by adopting wavelet inverse transformation on the denoised signal to obtain a denoised check signal.
  6. 6. The system of claim 5, wherein the wavelet basis function of the wavelet decomposition module selects a Meyer wavelet suitable for the denoising problem, and wherein the frequency domain expression is as follows: Wherein beta (x) is an auxiliary tool, satisfying 。
  7. 7. The system of claim 5, wherein the wavelet decomposition level J of the wavelet decomposition module is calculated by the formula: , where N is the length of the data signal S (t), Representing a downward integer.
  8. 8. The method of claim 5, wherein the noise standard deviation σ employs a robust median estimate: Wherein the method comprises the steps of Is a layer 1 high frequency detail coefficient.
  9. 9. A processing device comprising at least one processor and at least one memory communicatively coupled to the processor, wherein the memory stores program instructions executable by the processor, the processor invoking the program instructions to perform the method of any of claims 1-4.
  10. 10. A computer readable storage medium storing computer instructions for causing a computer to perform the method of any one of claims 1 to 4.

Description

Noise-dyeing signal denoising method and system for field verification of transformer partial discharge high-frequency current sensor Technical Field The invention relates to the field of verification of partial discharge sensors, in particular to a method and a system for denoising a field verification signal of a transformer partial discharge high-frequency sensor. Background Partial discharge (PARTIAL DISCHARGE, PD) is an early typical characterization of transformer insulation degradation, and on-line monitoring of the partial discharge signal can be performed to provide early warning prior to equipment failure. The high-frequency current method (High Frequency Current Transformer, HFCT) has become one of the main means of on-line monitoring of the partial discharge of the transformer because of its convenient installation, wide frequency band and high sensitivity. The high-frequency current sensor in operation needs to be calibrated regularly to ensure that the amplitude-frequency response, the linearity and the sensitivity index meet the field test precision requirement. However, a strong electromagnetic environment exists in the transformer station site, and standard pulse signals injected in the calibration process are easily overlapped by interference such as spatial coupling, potential fluctuation of a grounding network, carrier communication and the like, so that wide-spectrum white noise and multi-frequency narrow-band noise in the range of 500 kHz-2.5 MHz are formed. The traditional oscilloscope or peak comparison method directly processes the noisy signal as a true value, and errors exist. The existing denoising technology mainly comprises a hard/soft threshold wavelet method, an empirical mode decomposition method, an adaptive filtering and spectral subtraction method, a channel or priori noise model reference method and a site difficulty in acquisition, wherein a fixed wavelet base (db 4 and sym 8) is adopted, the number of decomposition layers is artificially set, narrow-band interference is not sufficiently suppressed, a soft threshold has constant deviation, the empirical mode decomposition method is good in self-adaption and easy to generate mode aliasing, overshoot is generated on a pulse calibration signal, and the adaptive filtering and spectral subtraction method is difficult to acquire on site. Therefore, an online denoising method which does not need a reference channel, does not need manual intervention, and can simultaneously inhibit white noise and narrow-band interference is needed, so that the reliability and efficiency of on-site calibration of the high-frequency current sensor are improved. Disclosure of Invention The invention aims to overcome the defects of low denoising precision and strong parameter dependence in the existing high-frequency current sensor field calibration technology, and provides a self-adaptive denoising method based on Meyer wavelets and improved smoothing threshold values, which realizes synchronous suppression of electromagnetic interference of calibration pulse signals in a strong electromagnetic environment of a transformer substation. The invention solves the technical problems by the following technical means: A method for denoising a noise-dyed signal of a transformer partial discharge high-frequency current sensor field verification comprises the following steps: The method comprises the steps that firstly, an upper computer collects an original dyeing noise check signal S (t) output by a sensor to be tested; Step two, self-adaptively determining a wavelet decomposition level J according to the signal length N, selecting a proper wavelet basis function, and performing discrete wavelet decomposition on a dyeing noise signal S (t) to obtain wavelet coefficients of all scales; Step three, performing layer-by-layer shrinkage processing on the wavelet coefficient by adopting a continuously-conductive progressive unbiased improved threshold function, wherein the improved threshold function is as follows: Where lambda is the common threshold, sigma is the noise standard deviation estimate, Representing the target wavelet coefficient processed by the threshold function; And step four, carrying out signal reconstruction by adopting wavelet inverse transformation on the denoised signal to obtain a denoised check signal. Further, the wavelet basis function of the second step selects a Meyer wavelet suitable for the denoising problem, and the frequency domain expression is as follows: Wherein beta (x) is an auxiliary tool, satisfying 。 Further, the wavelet decomposition level J of the second step is calculated by the following formula: , where N is the length of the data signal S (t), Representing a downward integer. Further, the noise standard deviation σ uses a robust median estimate: Wherein the method comprises the steps of Is a layer 1 high frequency detail coefficient. The invention also provides a noise-dyeing signal denoising system for the on-sit