Cable joint partial discharge signal denoising method, device and medium

Abstract

The embodiment of the invention discloses a cable joint partial discharge signal denoising method and device and a medium, belongs to the technical field of discharge signal denoising, and solves the problem that during partial discharge signal denoising, comprehensive and effective signal extraction is difficult to perform in an environment in which white noise and periodic narrow-band interference exist at the same time. Variational mode decomposition is carried out on the local noisy signal, and the mode number is determined through the energy relative change rate between adjacent intrinsic mode components obtained through decomposition; constructing a multi-feature index set through iterative updating; performing white noise dominant mode rejection through weighted fusion to obtain a primarily denoised discharge signal; determining initial periodic narrowband interference frequency information corresponding to the primarily denoised discharge signal, and performing iterative correction to obtain frequency information of target narrowband interference; and constructing a guide signal according to the frequency information of the target narrowband interference, performing singular value decomposition, and performing signal reconstruction according to a decomposition result to obtain a denoised target discharge signal.

Description

Technical Field

[0001] The present application relates to the technical field of discharge signal denoising, and in particular to a method, device and medium for denoising partial discharge signals of cable joints. Background Art

[0002] Cable joints are critical connecting components in power systems, and their insulation condition is directly related to the safety and reliability of grid operation. Partial discharge (PD), a key early sign of insulation defects, is crucial for preventing power accidents and extending equipment life. In recent years, with the widespread deployment of ultra-high voltage (UHV) cable systems, online PD detection technology has been widely researched and applied.

[0003] However, in actual detection, partial discharge signals are often accompanied by complex interference signals, mainly including broadband white noise (such as thermal noise and internal noise of electronic devices) and periodic narrowband interference (such as power frequency interference and power electronics switching frequency interference). These interference signals seriously affect the accuracy of partial discharge signal extraction and recognition, resulting in misjudgments and missed detections, and reducing the reliability of the diagnostic system.

[0004] In existing technologies, to denoise partial discharge signals and preserve true discharge characteristics, commonly used signal filtering methods include bandpass filtering, empirical mode decomposition, and wavelet transforms. While these methods have achieved some success in certain applications, they still have limitations in terms of denoising accuracy, parameter adaptability, and targeted suppression of different types of interference. High-fidelity signal extraction is often difficult in scenarios with high noise backgrounds or where the signal and noise spectra overlap. This is particularly true in complex environments where both white noise and periodic narrowband interference are present, making comprehensive and effective signal extraction difficult. Summary of the Invention

[0005] The embodiments of the present application provide a method, device, and medium for denoising partial discharge signals from cable connectors, which are used to solve the following technical problem: existing methods for denoising partial discharge signals are difficult to achieve comprehensive and effective signal extraction when faced with complex environments where white noise and periodic narrowband interference coexist on site.

[0006] The embodiments of this application adopt the following technical solutions:

[0007] The embodiment of the present application provides a method for denoising a local discharge signal of a cable joint. The method comprises: performing variational modal decomposition on a local noise signal corresponding to the cable joint, determining a modal number by the relative rate of change of energy between adjacent intrinsic modal components obtained by decomposition; obtaining a target intrinsic modal function set and a frequency set corresponding to the local noise signal through iterative updating based on the modal number, and constructing a multi-feature index set based on the target intrinsic modal function set and the frequency set; performing weighted fusion on the multi-feature index set to eliminate the white noise dominant mode to obtain a preliminary denoised discharge signal; determining the initial periodic narrowband interference frequency information corresponding to the preliminary denoised discharge signal, and iteratively correcting the initial periodic narrowband interference frequency to obtain the frequency information of the target narrowband interference; constructing a guide signal based on the frequency information of the target narrowband interference, performing singular value decomposition based on the guide signal, and reconstructing the signal according to the decomposition result to obtain the target denoised discharge signal.

[0008] The embodiment of the present application proposes a method for adaptively determining the number of modes based on the relative rate of change of energy of the intrinsic mode component, which avoids the problem of the dependence of the number of modes on empirical selection in traditional variational mode decomposition and improves the accuracy and stability of the decomposition. Secondly, the embodiment of the present application effectively identifies and eliminates the dominant mode of white noise by integrating a weighted fusion strategy of multiple feature indicators such as kurtosis, energy proportion and center frequency, combined with normalization and weighted scoring functions, thereby enhancing the ability to distinguish between signals and noise. In addition, the embodiment of the present application introduces a singular value decomposition method assisted by a guide signal, which realizes active enhancement and truncation of periodic narrowband interference by constructing a frequency-matched and amplitude-enhanced sinusoidal reference signal, thereby improving the effectiveness and accuracy of singular value decomposition in removing periodic interference, while ensuring effective noise reduction, retaining the key waveform characteristics of the local discharge signal.

[0009] In one implementation of the present application, the modal number is determined by the relative rate of change of energy between adjacent eigenmodal components obtained by decomposition, specifically including:

[0010] Function-based:

[0011]

[0012]

[0013] The relative energy change rate between adjacent eigenmode components is obtained, and the K value corresponding to the preset change rate threshold condition is used as the mode number; where ER is the energy ratio between the reconstructed signal and the original signal; x(t) is the original signal; h′(t) is the reconstructed signal; u K (t),u K-1(t) is the two adjacent eigenmode components; γ is the relative rate of change of energy between adjacent eigenmode components. When γ≤0.1, ER is in a stable state.

[0014] In one implementation of the present application, white noise-dominated modes are eliminated by weighted fusion of multiple feature indicator sets, specifically including: classifying multiple feature indicators in the multiple feature indicator set; wherein the multiple feature indicators include a signal kurtosis indicator, a signal energy ratio indicator, and a signal center frequency indicator; based on the categories of multiple feature indicators, matching corresponding weighting coefficients, and constructing a comprehensive scoring function, so as to perform weighted fusion of the multiple feature indicator set corresponding to the Kth mode through the comprehensive scoring function to obtain an indicator score; comparing the indicator score with the preset scoring indicator, and eliminating the white noise-dominated mode based on the comparison result.

[0015] In one implementation of the present application, based on the categories of multiple feature indicators, corresponding weighting coefficients are matched to construct a comprehensive scoring function, specifically including:

[0016] Function-based:

[0017]

[0018] Construct a comprehensive scoring function;

[0019] Among them, S k is the comprehensive score corresponding to the Kth mode; ω1 is the weighting coefficient corresponding to the signal kurtosis index; ω2 is the weighting coefficient corresponding to the signal energy ratio index; ω3 is the weighting coefficient corresponding to the signal center frequency index; Ku k is the signal kurtosis corresponding to the Kth mode; Ku min is the minimum kurtosis; Ku max is the maximum kurtosis; E k is the energy proportion corresponding to the Kth mode; E min is the minimum energy proportion; E max is the maximum energy proportion; ω k is the center frequency corresponding to the Kth mode; ω min is the minimum center frequency; ω max is the maximum center frequency.

[0020] In one implementation of the present application, obtaining a preliminary denoised discharge signal specifically includes: calibrating the remaining discharge signal into a modal component containing effective partial discharge information; and superimposing and reconstructing the calibrated modal components to obtain a preliminary denoised discharge signal.

[0021] In one implementation of the present application, the initial periodic narrowband interference frequency information corresponding to the discharge signal after preliminary denoising is determined, and the initial periodic narrowband interference frequency is iteratively corrected to obtain the frequency information of the target narrowband interference, specifically including: performing a fast Fourier transform on the discharge signal after preliminary denoising to obtain a spectrum; using a threshold method to screen peak points in the spectrum, and determining the number of frequencies of periodic narrowband interference based on the skewness corresponding to each peak point; obtaining the initial periodic narrowband interference frequency based on the spectrum correction function, the sampling frequency and the number of sampling points; obtaining the frequency shift distance based on the adopted frequency, the number of sampling points and the initial periodic narrowband interference frequency; iteratively correcting the spectral line amplitude in the spectrum correction function based on the frequency shift distance to determine the target narrowband interference frequency through the iteratively corrected spectral line amplitude; obtaining the frequency information of the target narrowband interference based on the number of frequencies of periodic narrowband interference and the target narrowband interference frequency.

[0022] In one implementation of the present application, constructing a pilot signal according to the frequency information of the target narrowband interference specifically includes: the pilot signal is:

[0023]

[0024] Among them, g(t) is the guidance signal; y max is the maximum amplitude of the discharge signal after preliminary denoising; f in is the target narrowband interference frequency; m is the number of narrowband interference frequencies; φ i is the phase offset.

[0025] In one implementation of the present application, singular value decomposition is performed based on the guide signal to reconstruct the signal according to the decomposition result to obtain the target denoised discharge signal, specifically including: superimposing the constructed guide signal onto the preliminary denoised discharge signal to obtain a mixed signal; constructing a Hankel matrix based on the mixed signal, and performing singular value decomposition to obtain the corresponding singular value sequence and characteristic subspace; setting the singular values ​​of the preset sequence positions to zero, and reconstructing the signal based on the remaining singular values ​​to obtain the target denoised discharge signal.

[0026] An embodiment of the present application provides a device for denoising a local discharge signal of a cable joint, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so as to enable the at least one processor to: perform variational modal decomposition on a local noise signal corresponding to the cable joint, and determine a modal number by using the relative energy change rate between adjacent intrinsic modal components obtained by the decomposition; obtain a target intrinsic modal function set and a frequency set corresponding to the local noise signal through iterative updating based on the modal number, and construct a multi-feature index set based on the target intrinsic modal function set and the frequency set; perform weighted fusion on the multi-feature index set to eliminate white noise-dominated modes to obtain a preliminary denoised discharge signal; determine initial periodic narrowband interference frequency information corresponding to the preliminary denoised discharge signal, and iteratively correct the initial periodic narrowband interference frequency to obtain frequency information of the target narrowband interference; construct a guide signal based on the frequency information of the target narrowband interference, perform singular value decomposition based on the guide signal, and reconstruct the signal based on the decomposition result to obtain the target denoised discharge signal.

[0027] An embodiment of the present application provides a non-volatile computer storage medium storing computer-executable instructions, wherein the computer-executable instructions are configured to: perform variational modal decomposition on a local noise signal corresponding to a cable connector, and determine a modal number by the relative rate of change of energy between adjacent eigenmodal components obtained by the decomposition; obtain a target eigenmodal function set and a frequency set corresponding to the local noise signal through iterative updating based on the modal number, and construct a multi-feature index set based on the target eigenmodal function set and the frequency set; perform weighted fusion on the multi-feature index set to eliminate white noise-dominated modes to obtain a preliminary denoised discharge signal; determine initial periodic narrowband interference frequency information corresponding to the preliminary denoised discharge signal, and iteratively correct the initial periodic narrowband interference frequency to obtain frequency information of the target narrowband interference; construct a guide signal based on the frequency information of the target narrowband interference, and perform singular value decomposition based on the guide signal to reconstruct the signal according to the decomposition result to obtain the target denoised discharge signal.

[0028] At least one of the above technical solutions adopted in the embodiments of the present application can achieve the following beneficial effects: The embodiments of the present application propose an adaptive method for determining the modal number based on the relative rate of change of the energy of the intrinsic mode component, which avoids the problem of the modal number relying on experience in the traditional variational modal decomposition and improves the accuracy and stability of the decomposition. Secondly, the embodiments of the present application effectively identify and eliminate the dominant mode of white noise by integrating the multi-feature index weighted fusion strategy of kurtosis, energy proportion and center frequency, combined with normalization and weighted scoring functions, thereby enhancing the ability to distinguish between signals and noise. In addition, the embodiments of the present application introduce a singular value decomposition method assisted by a guide signal, which realizes the active enhancement and truncation of periodic narrowband interference by constructing a frequency-matched and amplitude-enhanced sinusoidal reference signal, thereby improving the effectiveness and accuracy of the singular value decomposition in removing periodic interference, while ensuring effective noise reduction, retaining the key waveform characteristics of the local discharge signal. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments described in the present application. For those skilled in the art, other drawings can be obtained based on these drawings without inventive work. In the drawings:

[0030] Figure 1 A flow chart of a cable joint partial discharge signal denoising method provided in an embodiment of the present application;

[0031] Figure 2 A schematic diagram of an original partial discharge signal provided in an embodiment of the present application;

[0032] Figure 3 A schematic diagram of a noise-stained signal provided in an embodiment of the present application;

[0033] Figure 4 A schematic diagram of a discharge signal after preliminary denoising provided in an embodiment of the present application;

[0034] Figure 5 A schematic diagram of a discharge signal after target denoising provided in an embodiment of the present application;

[0035] Figure 6 A schematic structural diagram of a cable joint partial discharge signal denoising device provided in an embodiment of the present application.

[0036] Reference numerals:

[0037] 200: Cable joint partial discharge signal denoising device, 201: Processor, 202: Memory. DETAILED DESCRIPTION

[0038] Embodiments of the present application provide a method, device, and medium for denoising partial discharge signals from cable connectors.

[0039] In order to enable those skilled in the art to better understand the technical solutions in this application, the following will clearly and completely describe the technical solutions in the embodiments of this application in conjunction with the drawings in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this application.

[0040] The technical solutions proposed in the embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0041] Figure 1 A flow chart of a cable joint partial discharge signal denoising method provided in an embodiment of the present application is shown in FIG. Figure 1 As shown in FIG, the cable joint partial discharge signal denoising method includes the following steps:

[0042] Step 101: Perform variational modal decomposition on the local noise signal corresponding to the cable joint, and determine the modal number through the relative energy change rate between adjacent eigenmodal components obtained by the decomposition.

[0043] In one implementation of the present application, white noise is first removed by using a VMD (Variational Mode Decomposition) algorithm based on adaptively determining the number of modes.

[0044] Specifically, VMD is a signal time-frequency analysis method based on variational theory. Its purpose is to decompose a complex signal f(t) into K intrinsic mode functions (IMFs) with finite bandwidth and different center frequencies, requiring the total bandwidth of the mode to be minimized, that is,

[0045]

[0046] Among them, u k represents the kth mode function, ω k represents the center frequency of each IMF, represents the time derivative operator, and δ(t) represents the Dirac function.

[0047] Introduce the penalty factor α and Lagrangian operator λ(t):

[0048]

[0049] Update through iteration un+1 , the optimal solution obtained is the target IMF set {u k} and its corresponding frequency set {ω k}. Introduce a discrimination mechanism to identify the target IMF set {u k By filtering out the dominant white noise mode and superimposing the remaining modal components that are determined to contain valid PD information, the denoised signal can be obtained, namely:

[0050] x(t)=h′(t)+n(t) (3)

[0051] Where x(t) is the original signal, h′(t) is the reconstructed signal, and n(t) is the noise signal.

[0052] In one implementation of the present application, the method for adaptively determining the modal number K is as follows:

[0053] As the number of decomposition layers increases, the energy ratio ER between the reconstructed signal and the original signal tends to be stable. The expression of ER is as follows:

[0054]

[0055] When the adjacent IMF components u K (t),u K-1 When the relative energy change rate γ of (t) is ≤ 0.1, the ER is considered to be stable, and the K at this time is taken as the optimal K value.

[0056]

[0057] Step 102: Based on the modal number, a target intrinsic mode function set and a frequency set corresponding to the local noise signal are obtained through iterative updating, so as to construct a multi-feature index set based on the target intrinsic mode function set and the frequency set.

[0058] After obtaining the optimal modal number K, update iteratively u n+1 , we get the target IMF set {u k} and its corresponding frequency set {ω k Traditional VMD calculates the signal's kurtosis Ku and uses a threshold of 3 to filter out h′(t). The expression for kurtosis Ku is shown below.

[0059]

[0060] Where μ and η represent the mean and standard deviation of the signal, respectively, and E(x-μ) 4 is the fourth-order mathematical expectation of the signal.

[0061] However, in a strong noise environment, this method is difficult to accurately filter IMFs, resulting in a decrease in denoising performance. To this end, the embodiment of the present application constructs a multi-feature index set based on the signal's kurtosis (partial discharge signals usually have high kurtosis characteristics, while the kurtosis of white noise is generally close to or lower than 3), energy proportion (partial discharge signals exhibit pulse aggregation characteristics, with concentrated energy, and a modal energy proportion significantly higher than that of white noise modes, while white noise modes have dispersed energy and a small proportion), and center frequency (white noise modes are usually concentrated in the medium and high frequency regions, while partial discharge signals are concentrated in the medium and low frequency bands), and proposes a modal discrimination scheme based on weighted fusion of multiple feature indicators for automatically identifying and eliminating the dominant mode of white noise.

[0062] Step 103 : performing weighted fusion on multiple feature index sets to remove the white noise dominant mode, so as to obtain a preliminary denoised discharge signal.

[0063] In one implementation of the present application, multiple feature indicators in a multi-feature indicator set are classified; the multiple feature indicators include a signal kurtosis indicator, a signal energy ratio indicator, and a signal center frequency indicator. Based on the categories of the multiple feature indicators, corresponding weighting coefficients are matched to construct a comprehensive scoring function. The comprehensive scoring function is used to perform a weighted fusion of the multi-feature indicator set corresponding to the Kth mode to obtain an indicator score. The indicator score is compared with a preset scoring indicator, and the white noise-dominated mode is eliminated based on the comparison result.

[0064] Specifically, the signal kurtosis index, signal energy ratio index and signal center frequency index are normalized respectively, and the weighted coefficient is introduced to construct the comprehensive scoring function S. The comprehensive score S of each mode k As shown below:

[0065]

[0066] Among them, ω1 is the weighting coefficient corresponding to the signal kurtosis index; ω2 is the weighting coefficient corresponding to the signal energy ratio index; ω3 is the weighting coefficient corresponding to the signal center frequency index; Kuk is the signal kurtosis corresponding to the Kth mode; Ku min is the minimum kurtosis; Ku max is the maximum kurtosis; E k is the energy proportion corresponding to the Kth mode; E min is the minimum energy proportion; E max is the maximum energy proportion; ω k is the center frequency corresponding to the Kth mode; ω min is the minimum center frequency; ω max is the maximum center frequency. The empirical weights are usually set to ω1 = 0.4, ω2 = 0.3, and ω3 = 0.3; it should be noted that the empirical weights can be fine-tuned according to the distribution of the data set or the application scenario.k Compare with the set threshold T to determine whether the mode is dominated by white noise: If S k >T, the kth mode is considered to be a white noise mode and is removed. The recommended detection threshold range is T∈[0.65,0.75], and T=0.7 is usually taken.

[0067] In one implementation of the present application, the residual discharge signal is calibrated into a modal component containing effective partial discharge information, and the calibrated modal components are superimposed and reconstructed to obtain a preliminary denoised discharge signal.

[0068] Specifically, the white noise-dominated mode is screened out, and the remaining modal components that are determined to contain valid partial discharge information are superimposed and reconstructed to obtain a preliminary denoised discharge signal.

[0069] Step 104 : determining the initial periodic narrowband interference frequency information corresponding to the discharge signal after preliminary denoising, and iteratively correcting the initial periodic narrowband interference frequency to obtain the frequency information of the target narrowband interference.

[0070] In one implementation of the present application, a fast Fourier transform is performed on the discharge signal after preliminary denoising to obtain a spectrum. Peak points are screened in the spectrum by a threshold method, and the number of frequencies of periodic narrowband interference is determined based on the skewness corresponding to each peak point. The initial periodic narrowband interference frequency is obtained based on the spectrum correction function, the sampling frequency and the number of sampling points. The frequency shift distance is obtained based on the adopted frequency, the number of sampling points and the initial periodic narrowband interference frequency. The spectral line amplitude in the spectrum correction function is iteratively corrected based on the frequency shift distance to determine the target narrowband interference frequency through the iteratively corrected spectral line amplitude. Based on the number of frequencies of periodic narrowband interference and the target narrowband interference frequency, the frequency information of the target narrowband interference is obtained.

[0071] Specifically, the process of removing periodic narrowband interference by the SVD (Singular Value Decomposition) algorithm with a guidance signal includes obtaining the number of frequencies of periodic narrowband interference in the discharge signal after preliminary denoising, obtaining an accurate frequency estimate of the narrowband interference signal, and constructing a guidance signal to remove the periodic narrowband interference.

[0072] Furthermore, the frequency number of periodic narrowband interference in the discharge signal after preliminary denoising is obtained, including:

[0073] Perform fast Fourier transform on the preliminary denoised signal to obtain the spectrum. According to the classical threshold method, a smaller threshold T is introduced. m , according to the number of peak points in the spectrum, the frequency number of periodic narrowband interference is preliminarily obtained.

[0074]

[0075] Wherein, N is the sampling data point, σ is the standard deviation of the signal, β is the coefficient, and the value range is 0.1 to 0.5. In the embodiment of the present application, β is 0.4.

[0076] In order to accurately obtain the frequency number of periodic narrowband interference, the skewness SKE of each peak point in the spectrum is calculated:

[0077]

[0078] Among them, x i is the signal sample point, μ is the sample mean, σ is the sample standard deviation, and N is the total number of samples.

[0079] Furthermore, periodic narrowband interference signals follow a normal distribution, with a theoretical skewness of 0. However, due to spectral leakage, their skewness is generally less than 0, and the temporal spectrum distribution exhibits a negative deviation. However, partial discharge signals deviate significantly from a normal distribution, have a certain time range, and their skewness is often greater than 0. Therefore, if the calculated skewness of a peak point is less than 0, the center frequency of the periodic narrowband interference corresponding to the peak point can be determined, thereby determining the number m of frequencies of the periodic narrowband interference.

[0080] Furthermore, obtaining an accurate frequency estimate of the narrowband interference signal includes:

[0081] Identify the main frequency components of the narrowband interference, obtain the serial number k of the main peak spectrum line and the serial number k+r of the nearby sub-peak spectrum line, and record the corresponding amplitudes R(k) and R(k+r). Based on the amplitude relationship between the spectrum lines, calculate the spectrum correction value σ for subsequent fine-tuning of the frequency estimation. The correction formula is as follows:

[0082]

[0083] Using the sampling frequency f s And the number of sampling points N, preliminarily estimate the frequency f of narrowband interference r , which is calculated as follows:

[0084]

[0085] Perform spectral line fine-tuning in the frequency domain and calculate the frequency shift distance f d , and correct the spectrum lines k and k+r. The spectrum line amplitudes after frequency shift are again recorded as R(k) and R(k+r). The frequency shift distance f d The calculation formula is as follows:

[0086]

[0087] Repeat the above steps and use the updated spectrum line amplitude to perform fine frequency estimation to obtain the frequency f of narrowband interference with higher accuracy.in (i=1, 2,…m).

[0088] Step 105 : construct a pilot signal according to the frequency information of the target narrowband interference, perform singular value decomposition based on the pilot signal, and reconstruct the signal according to the decomposition result to obtain the target denoised discharge signal.

[0089] In one implementation of the present application, based on the obtained number of narrowband interference frequencies m and the accurately estimated frequency f in (i=1, 2, ...m), construct the guidance signal g(t), which is expressed as follows:

[0090]

[0091] Among them, g(t) is the guidance signal; y max is the maximum amplitude of the discharge signal after preliminary denoising; f in is the target narrowband interference frequency; m is the number of narrowband interference frequencies; φ i is the phase offset.

[0092] Furthermore, the constructed pilot signal g(t) is superimposed on the preliminary denoised signal to obtain the mixed signal m(t). A Hankel matrix is ​​constructed based on m(t), and SVD decomposition is performed to obtain the corresponding singular value sequence and characteristic subspace. Since the narrowband interference and the pilot signal highly overlap in the characteristic space, their principal components are concentrated in the first 2m singular values. Therefore, the first 2m singular values ​​are set to zero, while the remaining values ​​are retained. Signal reconstruction is then performed to obtain the final partial discharge signal.

[0093] In one implementation of the present application, when conducting a cable partial discharge test in a laboratory, a high-frequency current sensor is used to collect high-frequency current signals at the cable joint, white noise and periodic narrowband interference are added, and denoising is performed using the method proposed in the embodiment of the present application. The method flow is as follows:

[0094] (1) Adaptively determine the modal number K: perform VMD decomposition on the noisy signal and calculate the energy of each IMF component. When the relative energy change rate of adjacent IMF components γ≤0.1, K is the optimal value.

[0095] (2) IMF component screening based on weighted fusion of multiple feature indicators: the three feature indicators of signal kurtosis, energy proportion and center frequency are normalized respectively, and the weighted coefficient is introduced to construct a comprehensive scoring function. k The threshold T is set to distinguish between the white noise-dominated mode and the partial discharge-dominated mode;

[0096] (3) Obtaining a preliminary denoised signal: superimposing and reconstructing the modal components determined to contain valid PD information to obtain a preliminary denoised signal;

[0097] (4) Determine the number and frequency of narrowband interference: Use the classical threshold method to screen the peak points, then calculate the skewness of each peak point to accurately determine the number m of periodic narrowband interference frequencies. Calculate the preliminary estimate of the narrowband interference frequency, calculate the frequency shift distance based on the correction value, and iterate multiple times to obtain a higher-precision narrowband interference frequency.

[0098] (5) SVD decomposition with pilot signal: According to the number and frequency of narrowband interference, a pilot signal is constructed and SVD decomposition is performed. The first 2m singular values ​​are set to zero to remove periodic narrowband interference.

[0099] (6) Obtaining the final denoised signal: Perform signal reconstruction based on the remaining singular values ​​to obtain the final denoised partial discharge signal.

[0100] Figure 2 A schematic diagram of an original partial discharge signal provided in an embodiment of the present application is shown as follows: Figure 2 As shown, the horizontal axis is time and the vertical axis is amplitude. Figure 2 The PD signal in is the original signal before the noise signal is added.

[0101] Figure 3 A schematic diagram of a noise signal provided in an embodiment of the present application is shown in FIG. Figure 2 Based on the original partial discharge signal, the high-frequency current signal is collected at the cable joint through a high-frequency current through-core sensor, and white noise and periodic narrowband interference are added to obtain Figure 3 The noisy signal shown. Figure 4 A schematic diagram of a discharge signal after preliminary denoising provided in an embodiment of the present application is provided. Figure 3 On this basis, the three characteristic indicators of signal kurtosis, energy proportion and center frequency are normalized respectively, and the weighted coefficient is introduced to construct a comprehensive scoring function. k The threshold T is set to distinguish the white noise dominant mode from the partial discharge dominant mode, and the modal components containing effective partial discharge information are superimposed and reconstructed to obtain Figure 4 The discharge signal after preliminary denoising is shown. Figure 5 A schematic diagram of a discharge signal after target denoising provided in an embodiment of the present application is provided. Figure 4 On this basis, the classical threshold method is used to screen the peak points, calculate the skewness of each peak point, and accurately determine the number m of periodic narrowband interference frequencies. The frequency of the narrowband interference is preliminarily estimated, and the frequency shift distance is calculated based on the correction value. After multiple iterations, the frequency of the narrowband interference with higher accuracy is obtained. According to the number and frequency of the narrowband interference, the guidance signal is constructed and SVD decomposition is performed. The first 2m singular values ​​are set to zero to remove the periodic narrowband interference. The signal is reconstructed based on the remaining singular values ​​to obtain Figure 5 The target discharge signal after denoising is shown.

[0102] Figure 6 This is a schematic diagram of the structure of a cable joint partial discharge signal denoising device provided in an embodiment of the present application. Figure 6 As shown, a cable joint partial discharge signal denoising device 200 includes: at least one processor 201; and a memory 202 in communication with the at least one processor 201; wherein the memory 202 stores instructions that can be executed by the at least one processor 201, and the instructions are executed by the at least one processor 201 to enable the at least one processor 201 to: perform variational modal decomposition on the local noise signal corresponding to the cable joint, and determine the modal number by the relative energy change rate between adjacent eigenmodal components obtained by the decomposition; based on the modal number, obtain the corresponding local noise signal through iterative updating. The target intrinsic mode function set and frequency set are used to construct a multi-feature indicator set based on the target intrinsic mode function set and frequency set; the white noise dominant mode is eliminated by weighted fusion of the multi-feature indicator set to obtain a preliminary denoised discharge signal; the initial periodic narrowband interference frequency information corresponding to the preliminary denoised discharge signal is determined, and the initial periodic narrowband interference frequency is iteratively corrected to obtain the frequency information of the target narrowband interference; a guidance signal is constructed according to the frequency information of the target narrowband interference, and singular value decomposition is performed based on the guidance signal to reconstruct the signal according to the decomposition result to obtain the target denoised discharge signal.

[0103] An embodiment of the present application provides a non-volatile computer storage medium storing computer-executable instructions, wherein the computer-executable instructions are configured to: perform variational modal decomposition on a local noise signal corresponding to a cable connector, and determine a modal number by the relative rate of change of energy between adjacent eigenmodal components obtained by the decomposition; obtain a target eigenmodal function set and a frequency set corresponding to the local noise signal through iterative updating based on the modal number, and construct a multi-feature index set based on the target eigenmodal function set and the frequency set; perform weighted fusion on the multi-feature index set to eliminate white noise-dominated modes to obtain a preliminary denoised discharge signal; determine initial periodic narrowband interference frequency information corresponding to the preliminary denoised discharge signal, and iteratively correct the initial periodic narrowband interference frequency to obtain frequency information of the target narrowband interference; construct a guide signal based on the frequency information of the target narrowband interference, and perform singular value decomposition based on the guide signal to reconstruct the signal according to the decomposition result to obtain the target denoised discharge signal.

[0104] The various embodiments in this application are described in a progressive manner. Similar portions between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from the other embodiments. In particular, the device, apparatus, and non-volatile computer storage medium embodiments are generally similar to the method embodiments, so their descriptions are relatively simple. For relevant portions, refer to the descriptions of the method embodiments.

[0105] The foregoing is merely an embodiment of the present application and is not intended to limit the present application. It will be apparent to those skilled in the art that various modifications and variations may be made to the embodiments of the present application. However, such modifications or substitutions do not deviate from the spirit and scope of the technical solutions of the various embodiments of the present application.

Claims

1. A method for denoising partial discharge signals of cable joints, characterized in that: The method comprises: Perform variational modal decomposition on the local noise signal corresponding to the cable joint, and determine the mode number by the relative energy change rate between adjacent eigenmodal components obtained by decomposition; Based on the modal number, a target intrinsic mode function set and a frequency set corresponding to the local noise signal are obtained through iterative updating, so as to construct a multi-feature index set based on the target intrinsic mode function set and the frequency set; By performing weighted fusion on the multiple feature index sets, white noise dominant mode is eliminated to obtain a preliminary denoised discharge signal; Determining the initial periodic narrowband interference frequency information corresponding to the discharge signal after preliminary denoising, and iteratively correcting the initial periodic narrowband interference frequency to obtain the frequency information of the target narrowband interference; A pilot signal is constructed according to the frequency information of the target narrowband interference, and a singular value decomposition is performed based on the pilot signal to reconstruct the signal according to the decomposition result to obtain a target denoised discharge signal.

2. A cable joint partial discharge signal denoising method according to claim 1, characterized in that: The relative rate of change of energy between adjacent eigenmode components obtained by decomposition is used to determine the mode number, which specifically includes: Function-based: Obtaining the relative energy change rate between the adjacent eigenmode components, and taking the K value corresponding to when the relative energy change rate meets a preset change rate threshold condition as the mode number; Where ER is the energy ratio between the reconstructed signal and the original signal; x(t) is the original signal; h′(t) is the reconstructed signal; u K (t),u K-1 (t) is the two adjacent eigenmode components; γ is the relative rate of change of energy between adjacent eigenmode components. When γ≤0.1, ER is in a stable state.

3. A cable joint partial discharge signal denoising method according to claim 1, characterized in that: The white noise dominant mode is eliminated by weighted fusion of the multiple feature index sets, specifically including: Classifying multiple feature indicators in the multi-feature indicator set; wherein the multiple feature indicators include a signal kurtosis indicator, a signal energy ratio indicator, and a signal center frequency indicator; Based on the categories of the plurality of feature indicators, corresponding weighting coefficients are matched to construct a comprehensive scoring function, so as to perform weighted fusion on the set of multiple feature indicators corresponding to the Kth modality through the comprehensive scoring function to obtain an indicator score; The index score is compared with the preset score index, and the white noise dominant mode is eliminated based on the comparison result.

4. A cable joint partial discharge signal denoising method according to claim 3, characterized in that: The categories based on the plurality of characteristic indicators are matched with corresponding weighting coefficients to construct a comprehensive scoring function, which specifically includes: Function-based: Constructing the comprehensive scoring function; Among them, S k is the comprehensive score corresponding to the Kth mode; ω1 is the weighting coefficient corresponding to the signal kurtosis index; ω2 is the weighting coefficient corresponding to the signal energy ratio index; ω3 is the weighting coefficient corresponding to the signal center frequency index; Ku k is the signal kurtosis corresponding to the Kth mode; Ku min is the minimum kurtosis; Ku max is the maximum kurtosis; E k is the energy proportion corresponding to the Kth mode; E min is the minimum energy proportion; E max is the maximum energy proportion; ω k is the center frequency corresponding to the Kth mode; ω min is the minimum center frequency; ω max is the maximum center frequency.

5. A cable joint partial discharge signal denoising method according to claim 3, characterized in that: The obtaining of the preliminary denoised discharge signal specifically includes: The residual discharge signal is calibrated as the modal component containing effective partial discharge information; The calibrated modal components are superimposed and reconstructed to obtain the preliminary denoised discharge signal.

6. A cable joint partial discharge signal denoising method according to claim 1, characterized in that: The determining of the initial periodic narrowband interference frequency information corresponding to the preliminary denoised discharge signal and iteratively correcting the initial periodic narrowband interference frequency to obtain the frequency information of the target narrowband interference specifically includes: Performing fast Fourier transform on the discharge signal after preliminary denoising to obtain a frequency spectrum; Using a threshold method, peak points are screened in the spectrum, and the number of frequencies of periodic narrowband interference is determined based on the skewness corresponding to each peak point. Based on the spectrum correction function, sampling frequency and number of sampling points, the initial periodic narrowband interference frequency is obtained; Obtaining a frequency shift distance based on the adopted frequency, the number of sampling points, and the initial periodic narrowband interference frequency; Iteratively correcting the spectrum line amplitude in the spectrum correction function based on the frequency shift distance, so as to determine the target narrowband interference frequency through the iteratively corrected spectrum line amplitude; Frequency information of the target narrowband interference is obtained based on the number of frequencies of the periodic narrowband interference and the target narrowband interference frequency.

7. A cable joint partial discharge signal denoising method according to claim 1, characterized in that: The constructing of the guidance signal according to the frequency information of the target narrowband interference specifically includes: The guiding signal is: Among them, g(t) is the guidance signal; y max is the maximum amplitude of the discharge signal after preliminary denoising; f in is the target narrowband interference frequency; m is the number of narrowband interference frequencies; φ i is the phase offset.

8. The method for denoising a cable joint partial discharge signal according to claim 1, characterized in that: The performing of singular value decomposition based on the guide signal to reconstruct the signal according to the decomposition result to obtain the target denoised discharge signal specifically includes: Superimposing the constructed guiding signal onto the discharge signal after preliminary denoising to obtain a mixed signal; Constructing a Hankel matrix based on the mixed signal and performing singular value decomposition to obtain a corresponding singular value sequence and characteristic subspace; The singular values ​​at the preset sequence positions are set to zero, and the signal is reconstructed based on the remaining singular values ​​to obtain the target denoised discharge signal.

9. A cable joint partial discharge signal denoising device, characterized in that: The device comprises a memory for storing computer program instructions and a processor for executing the program instructions, wherein when the computer program instructions are executed by the processor, the device is triggered to execute the method according to any one of claims 1 to 8.

10. A non-volatile computer storage medium storing computer executable instructions, characterized in that: The computer executable instructions can execute the method according to any one of claims 1 to 8.

Images