Converter station partial discharge pulse interference suppression method based on feature extraction

Abstract

The invention relates to the field of converter station power equipment partial discharge detection, in particular to a converter station partial discharge pulse interference suppression method based on feature extraction. Comprising the following steps: numbering partial discharge pulse signal data to obtain numbered sequence data; a Morlet mother wavelet function is adopted to calculate a first characteristic parameter and a second characteristic parameter based on the serial number sequence data, normalization processing is performed on the first characteristic parameter and the second characteristic parameter, and a scatter diagram is drawn; determining the average value of the maximum amplitude values of the pulse signals in the subset corresponding to each data cluster in the scatter diagram; and if the amplitude of the pulse signal data in the subset corresponding to the data cluster does not meet the preset threshold range set based on the average value, rejecting the subset from the original signal data set. According to the method, the partial discharge signal is subjected to feature extraction, so that the partial discharge signal and the pulse interference signal are effectively separated, the interference signal is effectively suppressed, and the accuracy and the anti-interference capability of partial discharge detection of power equipment are improved.

Description

Technical Field

[0001] The present application relates to the field of partial discharge detection of power equipment in converter stations, and in particular to a method for suppressing partial discharge pulse interference in converter stations based on feature extraction. Background Art

[0002] In the power system, the operating status of power equipment will directly affect the safety of the entire system. When a major failure occurs in the power equipment, it may cause the power system to be paralyzed. Therefore, controlling the insulation status inside the power equipment is very important to maintain the safe and reliable operation of the entire power system.

[0003] Partial discharge is a localized discharge phenomenon that often occurs inside power equipment. Under the long-term effect of partial discharge, the insulation materials of power equipment will age, which may cause equipment breakdown and cause power outages. Therefore, it is of great practical significance to conduct partial discharge tests on power equipment. However, during the partial discharge test, power electronic equipment in the power system often generates pulse interference. Due to the similarity between the pulse signal from the power electronic device and the partial discharge signal waveform, this will interfere with the insulation diagnosis of the power equipment, thereby affecting the subsequent operation status assessment work.

[0004] Therefore, it is necessary to suppress the pulse interference signal from the outside during the partial discharge test. Summary of the invention

[0005] In view of this, it is necessary to provide a method for suppressing partial discharge pulse interference in a converter station based on feature extraction, so as to suppress the pulse interference signal from the outside during the partial discharge test.

[0006] A method for suppressing partial discharge pulse interference in a converter station based on feature extraction, comprising: numbering partial discharge pulse signal data to obtain number sequence data;

[0007] Using a Morlet mother wavelet function to calculate a first characteristic parameter and a second characteristic parameter based on the numbering sequence data, normalizing the first characteristic parameter and the second characteristic parameter and drawing a scatter plot;

[0008] Determine the average value of the maximum amplitude of the pulse signal in the subset corresponding to each data cluster in the scatter plot;

[0009] If the amplitude of the pulse signal data in the subset corresponding to the data cluster does not meet the preset threshold range set based on the average value, the subset corresponding to the data cluster is removed from the original signal data set.

[0010] Preferably, the first characteristic parameter and the second characteristic parameter are calculated based on the numbering sequence data using a Morlet mother wavelet function, including:

[0011] Using a Morlet mother wavelet function to calculate a first characteristic parameter and a second characteristic parameter based on the numbering sequence data according to a preset scale parameter and a time shift parameter;

[0012] The lth scale parameter a l ,satisfy:

[0013] The kth time shift parameter τ k ,satisfy:

[0014] The total number of scale parameters a is m, and the total number of time shift parameters τ is n; min and f max are the lower and upper frequency limits, respectively, f s is the sampling frequency.

[0015] Preferably, the first characteristic parameter satisfies:

[0016]

[0017] The second characteristic parameter satisfies:

[0018]

[0019] Wherein, fp1 is the first characteristic parameter, fp2 is the second characteristic parameter; the total number of the scale parameters a is m, the total number of the time shift parameters τ is n; f min and f max are the lower and upper frequency limits, respectively, f s is the sampling frequency; * represents the complex conjugate operation.

[0020] Preferably, normalizing the first characteristic parameter and the second characteristic parameter and drawing a scatter plot includes:

[0021] Extract the maximum value fp of the first characteristic parameter 1max and the maximum value of the second characteristic parameter fp 2max ;

[0022] The maximum value fp of the first characteristic parameter 1max and the maximum value of the second characteristic parameter fp 2max Divide them by their corresponding maximum values ​​respectively for normalization;

[0023] The normalized data are plotted in the form of a scatter plot in a two-dimensional coordinate system.

[0024] Preferably, determining the average value of the maximum amplitude of the pulse signal in the subset corresponding to each data cluster in the scatter plot includes:

[0025] The characteristic parameter grid traversal search method is used to determine the four sets of extreme value coordinates corresponding to each data cluster;

[0026] Extracting the serial numbers of the characteristic parameter data points in the quadrilateral region determined by the four sets of extreme value coordinates;

[0027] Based on the numbers of the extracted characteristic parameter data points, determining the pulse signal in the subset corresponding to each data cluster;

[0028] Determine the average value of the maximum amplitude of the pulse signal in the subset corresponding to each data cluster.

[0029] Preferably, the preset threshold range set based on the average value satisfies: 0.9Fave~1.1Fave; wherein Fave represents the average value of the maximum amplitude.

[0030] Preferably, after removing the pulse signal number subset original signal data that does not meet the preset threshold range set based on the average value, the method further includes:

[0031] The phase distribution analysis is performed on the partial discharge data after the noise signal is removed.

[0032] Compared with the prior art, the beneficial effect of the present application is that by performing feature extraction on local discharge signals and pulse interference signals in the scale time shift domain, the local discharge signals and pulse interference signals can be effectively separated, and the pulse interference signals from the outside world can be effectively suppressed during the local discharge test, thereby improving the accuracy and anti-interference ability of local discharge detection of power equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is a flow chart of a method for suppressing local discharge pulse interference in a converter station based on feature extraction provided in an embodiment of the present application.

[0034] Figure 2 This is an example diagram of the time domain waveform of the pulse signal provided in the embodiment of the present application.

[0035] Figure 3 This is an example diagram of a scatter plot drawn based on the calculated characteristic parameters provided in an embodiment of the present application.

[0036] Figure 4 This is an example diagram of two data clusters determined by using a feature parameter grid traversal search method provided in an embodiment of the present application.

[0037] Figure 5This is an example diagram of a phase distribution spectrum of a local discharge data set containing impulse noise provided in an embodiment of the present application.

[0038] Figure 6 This is an example diagram of the phase distribution of a partial discharge data set after the noise signal is removed according to an embodiment of the present application. DETAILED DESCRIPTION

[0039] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0040] See also Figure 1-5 ,like Figure 1 As shown, the method for suppressing partial discharge pulse interference in a converter station based on feature extraction provided in an embodiment of the present application includes the following steps:

[0041] S101: Numbering partial discharge pulse signal data to obtain numbering sequence data.

[0042] like Figure 2 As shown in FIG. 1 , for a partial discharge data set of a dry-type epoxy resin insulated transformer with power electronic pulse noise interference, for each pulse signal p(t), it is numbered in sequence, namely p1(t), p2(t), p3(t), ..., pi(t), and a number sequence data of a total of i pulse signals is obtained. Assume that in this embodiment, there are 3365 pulse signals in total, and the time domain waveform example of the first pulse signal is shown in FIG. Figure 2 as shown.

[0043] S102: using a Morlet mother wavelet function to calculate a first characteristic parameter and a second characteristic parameter based on the numbering sequence data, normalizing the first characteristic parameter and the second characteristic parameter and drawing a scatter plot.

[0044] The Morlet mother wavelet is used to process the pulse signal in the numbered sequence data, where the Morlet mother wavelet expression is:

[0045]

[0046] Among them, f b Refers to the bandwidth factor of the waveform, which affects the spectrum width of the wavelet. crefers to the central frequency, which is the main frequency component of the wavelet. These two parameters together determine the shape of the Morlet wavelet and its distribution characteristics in the time and frequency domains. In the process of feature extraction of partial discharge pulse signals, choosing the appropriate "f b ” and “f c "The value is crucial to ensure the accuracy of signal analysis. As a preferred method, the value f b =f c =2.

[0047] Further, a Morlet mother wavelet function is used to calculate a first characteristic parameter and a second characteristic parameter based on the numbering sequence data according to a preset scale parameter and a time shift parameter;

[0048] The lth scale parameter a l ,satisfy: The kth time shift parameter τ k ,satisfy: Among them, the total number of scale parameters a is m, the total number of time shift parameters τ is n; f min and f max They are the lower and upper frequency limits (which the detection system can detect during the partial discharge test), f s It is the sampling frequency (of the detection system during partial discharge testing).

[0049] The calculated first characteristic parameter fp1 and second characteristic parameter fp2 satisfy the following formulas respectively:

[0050]

[0051] Here, * represents the complex conjugate operation.

[0052] After calculating the first characteristic parameter fp1 and the second characteristic parameter fp2, for a total of i pulse signals p in the partial discharge data set i (t), the characteristic parameter calculated for each pulse signal is (fp i1 ,fp i2 ), extract the maximum value of the characteristic parameter fp1 and record it as fp 1max , extract the maximum value of the characteristic parameter fp2 and record it as fp 2max , for these i groups of feature parameters, they are divided by their corresponding maximum values, and then the i groups of feature parameters are normalized, that is, (fp i1 / fp 1max ,fp i2 / fp 2max ). For example, in the present embodiment, there are 3365 sets of characteristic parameters, and its fp 1max=3.05168875941025×10 -5 ;fp 2max =0.424837972209585.

[0053] Finally, the normalized i groups of feature parameters are plotted in the form of a scatter plot in a two-dimensional coordinate system, as shown in Figure 3 Example of a scatter plot of the feature parameters shown.

[0054] S103: Determine the average value of the maximum amplitude of the pulse signal in the subset corresponding to each data cluster in the scatter plot.

[0055] Observe and determine the number of data clusters in the scatter plot, and record it as z. For each data cluster, use four sets of extreme value coordinates, namely (a1, b1), (a2, b2), (a3, b3), (a4, b4), to determine a quadrilateral area, and extract the pulse signal number corresponding to each data cluster. Among them, the specific coordinate values ​​of the four sets of extreme value coordinates are determined by the characteristic parameter grid traversal search method, starting from 0 in the range of 0-1, and changing the horizontal and vertical coordinates numerically with a preset step size, such as 0.001, until it is observed that the quadrilateral area determined by the four sets of extreme value coordinates completely covers the scattered points in each data cluster.

[0056] For example, after the traversal search method, for the first data cluster, the four sets of extreme value coordinates are (0,0), (0.13,0), (0.13,0.1), (0,0.1); for the second data cluster, the four sets of extreme value coordinates are (0.14,0), (0.14,1), (1,0), (1,1), and the quadrilateral area corresponding to each set of extreme value coordinates is as follows: Figure 4 shown.

[0057] After the numerical values ​​of the extreme value coordinates corresponding to each data cluster are determined, the numbers of the characteristic parameter data points within the quadrilateral area determined by the extreme value coordinates are extracted.

[0058] The signal subsets of the partial discharge data are determined by using the numbers corresponding to the pulse signals of each extracted data cluster, and a total of z signal subsets are obtained. For each subset, the maximum amplitude information of the pulse signal waveform data corresponding to the subset is extracted, and the average value of the maximum amplitude of the pulse signal data contained in the subset is calculated. The calculation result is F ave .

[0059] S104: If the amplitude of the pulse signal data in the subset corresponding to the data cluster does not meet the preset threshold range set based on the average value, the original signal data of the pulse signal number subset is eliminated.

[0060] Based on the average Fave A preset threshold range is set, and noise signals and effective discharge signals are distinguished according to the threshold range. Specifically, by checking how many proportions of the maximum amplitudes of the pulse signals in the subset fall within the preset threshold range. If this proportion exceeds a certain level, it is considered that the subset is mainly composed of noise; otherwise, it is considered that the subset is mainly composed of effective discharge signals. The proportion of the maximum amplitude of the pulse signal falling within the preset threshold range can be set according to actual needs, experience, etc.

[0061] Furthermore, after removing the subsets corresponding to the data clusters that do not meet the preset threshold range set based on the average value from the original signal data set, the method further includes: performing phase distribution analysis on the partial discharge data after the noise signal is removed.

[0062] The purpose of performing phase distribution analysis on the partial discharge data set after noise signal removal is to verify whether the noise signal has been successfully suppressed and to ensure that the remaining signal is a valid partial discharge signal. By comparing the phase distribution spectra of the data set before and after noise removal, it can be intuitively seen whether the external pulse noise interference has been effectively removed. Such analysis helps to confirm the effectiveness of signal processing, thereby ensuring the accuracy of subsequent partial discharge detection of power equipment. In addition, this also provides a reliable data basis for further fault diagnosis and status assessment.

[0063] Further, in this application, based on the average value F ave The preset threshold range is set to 0.9F ave ~1.1F ave , if this ratio exceeds 90%, the subset is considered to be mainly composed of noise. Figure 5 Figure 2 shows an example of a phase distribution spectrum of a partial discharge data set containing impulse noise.

[0064] Furthermore, the subset mainly composed of noise determined in the previous step is removed from the original signal data set, thus achieving noise suppression. An example of the phase distribution of the partial discharge data set after removing the noise signal.

[0065] In summary, the present application creatively performs feature extraction on local discharge signals and pulse interference signals in the scale time shift domain, and the feature extraction algorithm provided by the present application can effectively realize the separation of local discharge signals and pulse interference signals, and obtain a pure local discharge signal data set, which can effectively suppress the pulse interference signals from the outside world during the local discharge test process, and improve the accuracy and anti-interference ability of local discharge detection of power equipment.

Claims

1. A method for suppressing partial discharge pulse interference in a converter station based on feature extraction, characterized in that: include: Numbering the partial discharge pulse signal data to obtain numbered sequence data; Using a Morlet mother wavelet function to calculate a first characteristic parameter and a second characteristic parameter based on the numbering sequence data, normalizing the first characteristic parameter and the second characteristic parameter and drawing a scatter plot; Determine the average value of the maximum amplitude of the pulse signal in the subset corresponding to each data cluster in the scatter plot; If the amplitude of the pulse signal in the subset corresponding to the data cluster does not meet the preset threshold range set based on the average value, the subset corresponding to the data cluster is removed from the original signal data set.

2. The method according to claim 1, characterized in that The first characteristic parameter and the second characteristic parameter are calculated based on the numbering sequence data using the Morlet mother wavelet function, including: Using a Morlet mother wavelet function to calculate a first characteristic parameter and a second characteristic parameter based on the numbering sequence data according to a preset scale parameter and a time shift parameter; The lth scale parameter a l ,satisfy: The kth time shift parameter τ k ,satisfy: The total number of scale parameters a is m, and the total number of time shift parameters τ is n; min and f max are the lower and upper frequency limits, respectively, f s is the sampling frequency.

3. The method according to claim 2, characterized in that include: The first characteristic parameter satisfies: The second characteristic parameter satisfies: Wherein, fp1 is the first characteristic parameter, fp2 is the second characteristic parameter; the total number of the scale parameters a is m, the total number of the time shift parameters τ is n; f min and f max are the lower and upper frequency limits, respectively, f s is the sampling frequency; * represents the complex conjugate operation.

4. The method according to claim 3, characterized in that: Normalizing the first characteristic parameter and the second characteristic parameter and drawing a scatter plot, including: Extract the maximum value fp of the first characteristic parameter 1max and the maximum value fp of the second characteristic parameter 2max ; The maximum value fp of the first characteristic parameter 1max and the maximum value fp of the second characteristic parameter 2max Divide them by their corresponding maximum values ​​respectively for normalization; The normalized data are plotted in the form of a scatter plot in a two-dimensional coordinate system.

5. The method according to claim 4, characterized in that Determining the average value of the maximum amplitude of the pulse signal in the subset corresponding to each data cluster in the scatter plot includes: The characteristic parameter grid traversal search method is used to determine the four sets of extreme value coordinates corresponding to each data cluster; The numbers of the characteristic parameter data points in the quadrilateral area determined by the four sets of extreme value coordinates are extracted; based on the numbers of the extracted characteristic parameter data points, the pulse signals in the subset corresponding to each data cluster are determined; and the average value of the maximum amplitude of the pulse signals in the subset corresponding to each data cluster is determined.

6. The method according to claim 1, characterized in that The preset threshold range set based on the average value satisfies: 0.9Fave~1.1Fave; wherein Fave represents the average value of the maximum amplitude.

7. The method according to claim 1, characterized in that After removing the subsets corresponding to the data clusters that do not meet the preset threshold range set based on the average value from the original signal data set, the method further includes: performing phase distribution analysis on the partial discharge data after the noise signal is removed.

Images