Denoising method and system for noisy signal of transformer partial discharge high-frequency current sensor field calibration

Abstract

The invention provides a self-adaptive denoising method based on Meyer wavelets and an improved smooth threshold function in order to solve the problem that a transformer partial discharge high-frequency current sensor field calibration signal is seriously polluted by field electromagnetic noise interference. The method comprises the following steps: firstly, arranging a standard sensor according to an optimal cross-core distance to obtain an original noisy signal; then adaptively determining a decomposition series according to a data length, selecting Meyer wavelets to perform multi-scale decomposition, performing smooth contraction on wavelet coefficients by adopting a continuously derivable, progressive and unbiased improved threshold function, and finally reconstructing a high signal-to-noise ratio verification signal through wavelet inverse transformation; the method does not need manual intervention of the threshold value, keeps orthogonality and high-order derivability in the whole process, and is beneficial to improving the accuracy of the verification result of the high-frequency sensor to be tested.

Description

Technical Field

[0001] This invention relates to the field of partial discharge sensor calibration, specifically to a method and system for denoising the on-site calibration signal of a transformer partial discharge high-frequency sensor. Background Technology

[0002] Partial discharge (PD) is a typical early sign of transformer insulation degradation. Online monitoring of PD signals can provide early warning before equipment failure. The High Frequency Current Transformer (HFCT) method has become one of the mainstream methods for online monitoring of transformer PD due to its convenient installation, wide bandwidth, and high sensitivity. High-frequency current sensors in operation must be calibrated regularly to ensure that their amplitude-frequency response, linearity, and sensitivity meet the accuracy requirements of on-site testing.

[0003] However, substation sites have strong electromagnetic environments, and the standard pulse signal injected during calibration is easily superimposed by interference from spatial coupling, grounding grid potential fluctuations, and carrier communication, forming broadband white noise and multi-frequency narrowband noise in the range of 500 kHz to 2.5 MHz. Traditional oscilloscopes or peak comparison methods directly treat the noisy signal as the true value, which introduces errors.

[0004] Existing denoising techniques mainly include: hard / soft threshold wavelet methods, which use fixed wavelet bases (such as db4, sym8) and manually set decomposition levels, resulting in insufficient suppression of narrowband interference, and soft thresholds exhibit constant deviations; empirical mode decomposition methods, which have good adaptability but are prone to mode aliasing, causing overshoot in pulse calibration signals; and adaptive filtering and spectral subtraction, which require reference channels or prior noise models, making them difficult to obtain on-site. Therefore, there is an urgent need for an online denoising method that requires no reference channel, no manual intervention, and can simultaneously suppress white noise and narrowband interference to improve the reliability and efficiency of on-site calibration of high-frequency current sensors. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of low denoising accuracy and strong parameter dependence in existing high-frequency current sensor field calibration technology, and to provide an adaptive denoising method based on Meyer wavelet and improved smoothing threshold, so as to realize synchronous suppression of electromagnetic interference of calibration pulse signal in strong electromagnetic environment of substation.

[0006] The present invention solves the above-mentioned technical problems through the following technical means:

[0007] A method for denoising the noise-polluted signal during on-site calibration of a transformer partial discharge high-frequency current sensor includes the following steps: Step 1: The host computer acquires the original noise-impaired verification signal S(t) output by the sensor under test; Step 2: Adaptively determine the wavelet decomposition level J based on the signal length N, and select appropriate wavelet basis functions to perform discrete wavelet decomposition on the noisy signal S(t) to obtain wavelet coefficients at each scale; Step 3: Apply a continuously differentiable and asymptotically unbiased improved threshold function to perform a layer-by-layer shrinkage process on the wavelet coefficients; the improved threshold function is:

[0008] in l For general thresholds, s This is an estimate of the noise standard deviation. This represents the target wavelet coefficients after thresholding. Step 4: Reconstruct the denoised signal by performing inverse wavelet transform on the denoised signal to obtain the denoised verification signal.

[0009] Furthermore, in step two, the wavelet basis function is selected from the Meyer wavelet, which is suitable for this denoising problem. The frequency domain expression is as follows:

[0010] in β ( x ) is an auxiliary tool that satisfies .

[0011] Furthermore, the wavelet decomposition series in step two... J Calculated using the following formula: , in N For data signal S( t ) length, This indicates rounding down to the nearest integer.

[0012] Furthermore, the noise standard deviation σ is estimated using robust median estimation:

[0013] in This represents the high-frequency detail coefficients for the first layer.

[0014] This invention also provides a noise reduction system for on-site calibration of high-frequency current sensors for transformer partial discharge, comprising: Signal acquisition module: The host computer acquires the raw noise-impaired verification signal S(t) output by the sensor under test; Wavelet decomposition module: Based on the signal length N, the wavelet decomposition level J is adaptively determined, and a suitable wavelet basis function is selected to perform discrete wavelet decomposition on the noisy signal S(t) to obtain wavelet coefficients at each scale; Denoising module: Employs a continuously differentiable and asymptotically unbiased improved threshold function to perform layer-by-layer shrinkage processing on the wavelet coefficients; the improved threshold function is:

[0015] in l For general thresholds, s This is an estimate of the noise standard deviation. This represents the target wavelet coefficients after thresholding.

[0016] Furthermore, the wavelet basis function selected by the wavelet decomposition module is the Meyer wavelet, which is suitable for this denoising problem. Its frequency domain expression is as follows:

[0017] in β ( x ) is an auxiliary tool that satisfies .

[0018] Furthermore, the wavelet decomposition level of the wavelet decomposition module J Calculated using the following formula: , in N For data signal S( t ) length, This indicates rounding down to the nearest integer.

[0019] Furthermore, the noise standard deviation σ is estimated using robust median estimation:

[0020] in This represents the high-frequency detail coefficients for the first layer.

[0021] The present invention also provides a processing device, including at least one processor and at least one memory communicatively connected to the processor, wherein: the memory stores program instructions executable by the processor, and the processor can execute the above-described method by calling the program instructions.

[0022] The present invention also provides a computer-readable storage medium storing computer instructions that cause the computer to perform the above-described method.

[0023] The advantages of this invention are: The method of this invention is applicable to field environments, takes into account strong electromagnetic interference, and has adaptive noise reduction. Its decomposition level, threshold and noise variance are estimated from the data itself without manual intervention. It maintains orthogonality and high-order differentiability throughout the process, which helps to improve the accuracy of the calibration results of the high-frequency sensor under test.

[0024] This method relies on single-channel sampling only and can be extended to the field verification of high-frequency sensors such as GIS and cable terminals. Attached Figure Description

[0025] Figure 1 This is the original image of the noisy signal; Figure 2 This is a diagram of the denoised signal; Figure 3 A comparison was made between the classic soft and hard functions and the improved threshold function of this invention. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] This embodiment provides a method for denoising the noise-enhanced signal during on-site verification of a transformer partial discharge high-frequency current sensor, including the following steps: Step 1: The host computer acquires the original noise-impaired verification signal S(t) output by the sensor under test; Step 2: Adaptively determine the wavelet decomposition level J based on the signal length N, and select appropriate wavelet basis functions to perform discrete wavelet decomposition on the noisy signal S(t) to obtain wavelet coefficients at each scale; Step 3: Apply a continuously differentiable and asymptotically unbiased improved threshold function to perform a layer-by-layer shrinkage process on the wavelet coefficients; the improved threshold function is:

[0028] in l For general thresholds, s This is an estimate of the noise standard deviation. This represents the target wavelet coefficients after thresholding. Step 4: Reconstruct the denoised signal by performing inverse wavelet transform on the denoised signal to obtain the denoised verification signal.

[0029] Specifically, in step two, the wavelet basis function is selected from the Meyer wavelet, which is suitable for this denoising problem. The frequency domain expression is as follows:

[0030] in β ( x ) is an auxiliary tool that satisfies .

[0031] Wavelet decomposition series J Calculated using the following formula: , in N For data signal S( t ) length, This indicates rounding down to the nearest integer.

[0032] The noise standard deviation σ is estimated using a robust median estimate.

[0033] in This represents the high-frequency detail coefficients for the first layer.

[0034] To better understand the advantages of the threshold function, this embodiment compares the classic soft and hard threshold functions with the improved threshold function of this patent, such as... Figure 3 As shown, it is clear that the threshold function transition in this embodiment is significantly smoother, the denoising effect is better, and the corresponding signal-to-noise ratio is higher.

[0035] The derivation procedure for the threshold function mentioned above is as follows: For the observed wavelet coefficients y (Including noise), its true coefficient x The estimated value We obtain the following by solving the equation: (1) Where P λ ( t ) is about The penalty function, Introduce a non-convex penalty function P λ ( t ), so that its derivative is at t≥ l It has a contraction term of the following form: (2) The optimal condition for optimizing formula (1) is: (3) Equivalent to: (4) This is an implicit equation. To obtain the explicit threshold function, an approximation is used, for larger values... ,have Substitute into formula (2) and define oh = y This yields the explicit threshold rule: (5) For | oh |< l ,set up or ( oh ) = 0, yielding the complete piecewise function: .

[0036] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for denoising the polluted signal during on-site verification of a transformer partial discharge high-frequency current sensor, characterized in that, Includes the following steps: Step 1: The host computer acquires the original noise-impaired verification signal S(t) output by the sensor under test; Step 2: Adaptively determine the wavelet decomposition level J based on the signal length N, and select appropriate wavelet basis functions to perform discrete wavelet decomposition on the noisy signal S(t) to obtain wavelet coefficients at each scale; Step 3: Apply a continuously differentiable and asymptotically unbiased improved threshold function to perform a layer-by-layer shrinkage process on the wavelet coefficients; the improved threshold function is: in λ For general thresholds, σ This is an estimate of the noise standard deviation. This represents the target wavelet coefficients after thresholding. Step 4: Reconstruct the denoised signal by performing inverse wavelet transform on the denoised signal to obtain the denoised verification signal.

2. The method according to claim 1, characterized in that, In step two, the wavelet basis function selection is a suitable Meyer wavelet for this denoising problem, and its frequency domain expression is as follows: in β ( x ) is an auxiliary tool that satisfies .

3. The method according to claim 1, characterized in that, The wavelet decomposition series in step two J Calculated using the following formula: ， in N For data signal S( t ) length, This indicates rounding down to the nearest integer.

4. The method according to claim 1, characterized in that, The noise standard deviation σ is estimated using a robust median estimate. in These are the target wavelet coefficients of the first layer after threshold function processing.

5. A noise reduction system for on-site calibration of a transformer partial discharge high-frequency current sensor, characterized in that, include: Signal acquisition module: The host computer acquires the raw noise-impaired verification signal S(t) output by the sensor under test; Wavelet decomposition module: Based on the signal length N, the wavelet decomposition level J is adaptively determined, and a suitable wavelet basis function is selected to perform discrete wavelet decomposition on the noisy signal S(t) to obtain wavelet coefficients at each scale; Denoising module: Employs a continuously differentiable and asymptotically unbiased improved threshold function to perform layer-by-layer shrinkage processing on the wavelet coefficients; the improved threshold function is: in λ For general thresholds, σ This is an estimate of the noise standard deviation. This represents the target wavelet coefficients after thresholding. Signal reconstruction module: The signal is reconstructed by using inverse wavelet transform on the denoised signal to obtain the denoised verification signal.

6. The system according to claim 5, characterized in that, The wavelet basis function selected by the wavelet decomposition module is the Meyer wavelet, which is suitable for this denoising problem. Its frequency domain expression is as follows: in β ( x ) is an auxiliary tool that satisfies .

7. The system according to claim 5, characterized in that, The wavelet decomposition level of the wavelet decomposition module J Calculated using the following formula: ， in N For data signal S( t ) length, This indicates rounding down to the nearest integer.

8. The method according to claim 5, characterized in that, The noise standard deviation σ is estimated using a robust median estimate. in This represents the high-frequency detail coefficients for the first layer.

9. A processing device, characterized in that, It includes at least one processor and at least one memory communicatively connected to the processor, wherein: the memory stores program instructions executable by the processor, and the processor can execute the method as described in any one of claims 1 to 4 by invoking the program instructions.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause the computer to perform the method as described in any one of claims 1 to 4.

Images