Feature extraction method and device of GIS ultrahigh frequency partial discharge signal

Abstract

The invention discloses a GIS ultrahigh frequency partial discharge signal feature extraction method and device, and relates to the field of electrical data processing, and the method comprises the steps: employing a center frequency method to optimize the number K of modal components of a VMD algorithm, and obtaining an optimal K value; decomposing the ultrahigh-frequency partial discharge signal by using a VMD algorithm of which the number of mode components is the optimal K value to obtain K intrinsic mode function components; performing different types of entropy calculation on each intrinsic mode function component to obtain values of different types of characteristic variables in each intrinsic mode function component, and constructing a characteristic matrix; and inputting the feature matrix into a dynamic network marker model based on a sample covariance matrix to extract dominant feature variables and key network markers corresponding to the dominant feature variables, calculating correlation quantitative indexes among the dominant feature variables, and constructing the extracted key features based on the key network markers and the correlation quantitative indexes. The problem that the robustness of feature expression is insufficient is solved.

Description

Technical Field

[0001] The present invention relates to the field of electrical data processing, and particularly relates to a method and device for extracting characteristics of UHF partial discharge signals of GIS. Background Art

[0002] In recent years, China has been actively promoting the construction of a new power system with new energy as the main body. Compared with the traditional power system, the power source structure of the new power system is mainly composed of new energy, which has obvious characteristics of randomness, volatility and intermittency, and the requirements for flexibility, controllability, safety and stability are also getting higher and higher. This brings huge challenges to the operation and maintenance of high-voltage power equipment that constitutes the main framework of the system. The risk of large-scale power outages caused by faults or damages of key power equipment is increasing day by day. Gas Insulated Switchgear (GIS), as a high-voltage power equipment in the power system, its condition is crucial for the safe operation of the power system. During the production, transportation, installation and operation of GIS, some defects that affect the insulation of the equipment will inevitably occur. After a period of operation under high voltage and large current, these defects will cause partial discharge inside GIS. Various insulation defects caused by partial discharge (PD) are the main culprits of equipment accidents and warning information for insulation hidden dangers. Such tiny discharges will corrode the insulation structure. If not discovered or treated in time, the deteriorated area will continue to expand, eventually leading to the overall breakdown of the insulation, affecting the normal operation of power equipment, and even possibly causing the paralysis of the entire power system in severe cases. At the same time, different insulation defects have different degrees of harm to the insulation state of the equipment. Therefore, accurately judging the type of insulation defect that causes the PD phenomenon is of great significance for guiding the maintenance work and reducing its potential risks in a timely manner.

[0003] Feature variable extraction is the key to the PD type recognition method, and its advantages and disadvantages directly affect the performance of the classification algorithm. At present, in the aspect of feature extraction of UHF partial discharge signals in GIS, due to the complex, non-stationary and non-linear characteristics of the UHF signals generated by GIS partial discharges, there are relatively high requirements for the analysis, processing and feature extraction of UHF partial discharge signals in GIS. Traditional time-domain or frequency-domain analysis is difficult to accurately extract the fault features of signals. Time-frequency analysis methods have the advantages of time and frequency localization, multi-scale analysis, etc., can characterize the mutual relationship between the signal frequency sequence and the time sequence, and are suitable for the analysis and processing of non-stationary signals. Among them, the wavelet transform method has the advantages of low entropy, multi-resolution, decorrelation, and flexible basis selection, but it has the limitations of weak adaptability and energy leakage. The empirical mode decomposition (EMD) method can effectively extract information from noisy non-linear and non-stationary processes, but this method has the defects of mode mixing and end-point effects. The ensemble empirical mode decomposition (EEMD) overcomes the defects of the EMD method, but this method has a large amount of calculation, and it is difficult for the edge computing gateway of the current on-line monitoring device to complete. Compared with the above methods, the variational mode decomposition (VMD) is an adaptive multi-resolution technology, which can reduce the computational complexity while ensuring the signal decomposition quality, has good performance in processing non-stationary and non-linear signals, can improve the reliability of signal decomposition and avoid redundant features. However, the mode decomposition parameters of this method have a great impact on the performance and there are certain limitations.

[0004] Meanwhile, most feature extraction methods mainly focus on the isolated extraction of single features or the parallel processing of independent multiple features. Taking the typical variational mode decomposition - entropy analysis method as an example, the traditional processing flow usually only extracts the independent entropy value features of each modal component. Such single entropy features expose significant limitations in complex noise environments: single metrics such as power spectral entropy are highly sensitive to noise perturbations, and there are significant differences in the stability performance of different entropy metrics, resulting in insufficient robustness of feature expression. Although integrating multiple entropy features can, to a certain extent, compensate for the defects of single metrics, the parallel processing mechanism of independent multiple features will generate a redundant feature space with high dimensions and high correlations, which is prone to cause the curse of dimensionality problem when directly input into the classification model. More critically, this processing mode ignores the internal correlation mechanism of different entropy features in the modal components, that is, various entropy metrics essentially reflect the feature attributes of different dimensions of the same signal, and their co-variation laws contain rich complementary information. It can be seen that there are still certain limitations in the mining of the non-linear coupling relationship between multi-dimensional features by the currently used related methods, which to some extent affects the integrity and accuracy of complex signal feature expression. Summary of the Invention

[0005] The purpose of this application is to propose a feature extraction method and device for GIS ultra-high frequency partial discharge signals in view of the above-mentioned technical problems.

[0006] In the first aspect, the present invention provides a feature extraction method for GIS ultra-high frequency partial discharge signals, including the following steps:

[0007] Obtain the GIS ultra-high frequency signal and perform preprocessing to obtain the ultra-high frequency partial discharge signal;

[0008] Obtain the training data and optimize the number of modal components K of the VMD algorithm using the center frequency method to obtain the optimal K value;

[0009] Use the VMD algorithm with the number of modal components being the optimal K value to decompose the ultra-high frequency partial discharge signal to obtain K intrinsic mode function components; calculate different types of entropy for each intrinsic mode function component respectively to obtain the values of different types of feature variables in each intrinsic mode function component; construct a feature matrix according to the values of different types of feature variables in each intrinsic mode function component;

[0010] Input the feature matrix into the dynamic network marker model based on the sample covariance matrix to extract the set of dominant feature variables and their corresponding index markers, use the set of index markers as the key network markers, calculate the correlation quantification index between the dominant feature variables, and construct the extracted key features based on the key network markers and the correlation quantification index.

[0011] Preferably, training data is obtained and the number of modal components K of the VMD algorithm is optimized using the central frequency method to obtain the optimal K value, which specifically includes:

[0012] Collect the ultra-high frequency partial discharge signals corresponding to different partial discharge types of the GIS device and form training data;

[0013] For the ultra-high frequency partial discharge signals of each partial discharge type, calculate the central frequency values of each intrinsic mode function component obtained by the VMD algorithm under different numbers of modal components K;

[0014] For each partial discharge type, calculate the relative change rate of the central frequency value of the last intrinsic mode function component under adjacent K values, as shown in the following formula:

[0015] ;

[0016] Where represents the central frequency value of the (K + 1)-th intrinsic mode function component when the number of modal components is K + 1, represents the central frequency value of the K-th intrinsic mode function component when the number of modal components is K, represents the relative change rate when the number of modal components is K;

[0017] Statistically count the number of modal components K corresponding to each partial discharge type when the relative change rate is less than or equal to the threshold, and use the number of modal components K corresponding to the most occurrences among all partial discharge types when the relative change rate is less than or equal to the threshold as the optimal K value.

[0018] Preferably, different types of entropy calculations include power spectral entropy calculation, sample entropy calculation, fuzzy entropy calculation, permutation entropy calculation, and distribution entropy calculation, and the result of each type of entropy calculation corresponds to a characteristic variable.

[0019] Preferably, the feature matrix is input into the dynamic network biomarker model based on the sample covariance matrix to extract the dominant characteristic variables and their corresponding index marker sets, which specifically includes:

[0020] Perform standardized processing on the feature matrix to obtain the feature matrix after standardized processing;

[0021] Construct the sample covariance matrix of the feature matrix after standardized processing, perform eigenvalue decomposition on the sample covariance matrix, and screen out the dominant characteristic variables according to the eigenvector corresponding to the largest eigenvalue of the sample covariance matrix and the ranking of its elements, and extract the set of index markers corresponding to the dominant characteristic variables.

[0022] Preferably, the eigenvalue decomposition process of the sample covariance matrix is as shown in the following formula:

[0023] ;

[0024] Among them, C represents the sample covariance matrix, is a diagonal matrix with the th eigenvalue of the sample covariance matrix C as the diagonal elements. The eigenvalues of the sample covariance matrix C are sorted from largest to smallest. The first eigenvalue of the sample covariance matrix C is the largest eigenvalue of the sample covariance matrix C. is an orthogonal matrix with the unit eigenvector as columns, , and n is the number of characteristic variables;

[0025] The element in the i-th row and j-th column of the sample covariance matrix C is expressed as:

[0026] ;

[0027] Among them, represents the covariance between the eigenvector in the th column and the eigenvector in the th column in the standardized characteristic matrix, , represents the largest eigenvalue of the sample covariance matrix, and are respectively the th element and the th element of the unit eigenvector corresponding to the largest eigenvalue of the sample covariance matrix, corresponding to the eigenvector in the th column and the eigenvector in the th column in the standardized characteristic matrix; is the th eigenvalue of the sample covariance matrix, and are respectively the th element and the th element of the unit eigenvector corresponding to the th eigenvalue of the sample covariance matrix;

[0028] Compare the eigenvector corresponding to the largest eigenvalue Among the n elements, select the characteristic variables corresponding to the columns in the standardized characteristic matrix of the top m elements in terms of numerical ranking as the dominant characteristic variables, and use the set of index marks of the dominant characteristic variables in the standardized characteristic matrix as the key network marks .

[0029] Preferably, for the dominant characteristic variables, an association quantification index is used to quantify the association between the dominant characteristic variables, as shown in the following formula:

[0030] ;

[0031] where m is the number of elements in the key network marks , the value of m is set to satisfy and , the P value is a positive real number in [0, 1], n is the number of characteristic variables, represents the characteristic vector of the th column in the standardized characteristic matrix and the th column of the characteristic vector The covariance between them, represents the absolute value, represents the L2 norm;

[0032] The key feature is the vector formed by splicing the key network marks and the association quantification index.

[0033] In a second aspect, the present invention provides a feature extraction device for GIS ultra-high frequency partial discharge signals, including:

[0034] A preprocessing module configured to obtain GIS ultra-high frequency signals and perform preprocessing to obtain ultra-high frequency partial discharge signals;

[0035] A K value optimization module configured to obtain training data and optimize the number of modal components K of the VMD algorithm by the center frequency method to obtain the optimal K value;

[0036] A feature decomposition module configured to decompose the ultra-high frequency partial discharge signal by using the VMD algorithm with the number of modal components being the optimal K value to obtain K intrinsic mode function components; calculate different types of entropy for each intrinsic mode function component respectively to obtain the values of different types of characteristic variables in each intrinsic mode function component; construct a characteristic matrix according to the values of different types of characteristic variables in each intrinsic mode function component;

[0037] The key feature extraction module is configured to input the feature matrix into a dynamic network biomarker model based on the sample covariance matrix to extract a set of dominant feature variables and their corresponding index markers, use the set of index markers as key network markers, calculate the correlation quantification index between the dominant feature variables, and construct the extracted key features based on the key network markers and the correlation quantification index.

[0038] In a third aspect, the present invention provides an electronic device, including one or more processors; a storage device for storing one or more programs, which when executed by the one or more processors cause the one or more processors to implement the method described in any implementation manner of the first aspect.

[0039] In a fourth aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the method described in any implementation manner of the first aspect.

[0040] In a fifth aspect, the present invention provides a computer program product, including a computer program, which when executed by a processor implements the method described in any implementation manner of the first aspect.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] (1) In order to effectively mine the internal features of different types of GIS partial discharge UHF signals, the feature extraction method of GIS UHF partial discharge signals proposed by the present invention first optimizes the number of modal decompositions K of the VMD algorithm by using the center frequency method, and uses the VMD algorithm to decompose the UHF partial discharge signal into intrinsic mode function components with different frequency scale features. Then, various entropy values are calculated for each of the decomposed intrinsic mode function components to further mine the feature attributes of different dimensions of the same signal, and then the corresponding feature matrix is obtained and standardized. And considering the non-linear coupling relationship between different entropy features, the feature matrix is input into a dynamic network biomarker model based on the sample covariance matrix to extract a set of index markers of the dominant feature variables and their corresponding correlation quantification indexes, enhancing the discrimination ability and robustness of the feature extraction of GIS UHF partial discharge signals.

[0043] (2) The feature extraction method of GIS UHF partial discharge signals proposed by the present invention constructs key features based on the set of index markers of the extracted dominant feature variables and their corresponding correlation quantification indexes, reveals the dynamic evolution of different GIS partial discharge types, considers the non-linear coupling relationship between feature variables, and does not cause the problem of dimensionality disaster, enhancing the distinguishability of different GIS partial discharge types.

[0044] (3) The key features extracted by the feature extraction method of GIS UHF partial discharge signals proposed by the present invention can be further input into the classification devices corresponding to different partial discharge types for identification to determine the corresponding partial discharge type. Using these key features for classification can significantly improve the identification accuracy of partial discharge types. Description of the Drawings

[0045] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0046] Figure 1 It is a flowchart of the feature extraction method of GIS UHF partial discharge signals in the embodiments of the present application;

[0047] Figure 2 It is a flowchart of the feature extraction method of GIS UHF partial discharge signals in the embodiments of the present application;

[0048] Figure 3 It is a time-domain diagram and a frequency-spectrum diagram of the UHF signal of tip discharge and its VMD decomposition result in the feature extraction method of GIS UHF partial discharge signals in the embodiments of the present application;

[0049] Figure 4 It is a time-domain diagram and a frequency-spectrum diagram of the UHF signal of air-gap discharge and its VMD decomposition result in the feature extraction method of GIS UHF partial discharge signals in the embodiments of the present application;

[0050] Figure 5 It is a time-domain diagram and a frequency-spectrum diagram of the UHF signal of floating discharge and its VMD decomposition result in the feature extraction method of GIS UHF partial discharge signals in the embodiments of the present application;

[0051] Figure 6 It is a schematic diagram of the key network markers of three partial discharge types in the feature extraction method of GIS UHF partial discharge signals in the embodiments of the present application;

[0052] Figure 7 It is a diagram of the correlation quantification index of three partial discharge types in the feature extraction method of GIS UHF partial discharge signals in the embodiments of the present application;

[0053] Figure 8 It is a schematic diagram of the feature extraction device of GIS UHF partial discharge signals in the embodiments of the present application;

[0054] Figure 9Schematic diagram of the hardware structure of the electronic device provided by the embodiment of the present invention. Detailed implementation manners

[0055] In order to make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings. Apparently, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0056] Figure 1 The figure shows a method for extracting characteristics of GIS ultra-high frequency partial discharge signals provided by an embodiment of the present application, including the following steps:

[0057] S1. Obtain GIS ultra-high frequency signals and perform preprocessing to obtain ultra-high frequency partial discharge signals.

[0058] Specifically, referring to Figure 2 , the preprocessing process includes interception and filtering. In one embodiment, the relevant signals of partial discharge in the ultra-high frequency signals are intercepted, and db4 wavelet five-layer decomposition is used for filtering to obtain ultra-high frequency partial discharge signals.

[0059] S2. Obtain training data and optimize the number of modal components K of the VMD algorithm by using the center frequency method to obtain the optimal K value.

[0060] In a specific embodiment, obtaining training data and optimizing the number of modal components K of the VMD algorithm by using the center frequency method to obtain the optimal K value specifically includes:

[0061] Collect the ultra-high frequency partial discharge signals obtained after preprocessing corresponding to different partial discharge types of the GIS device to form training data;

[0062] Calculate the center frequency values of the respective intrinsic mode function components obtained by the VMD algorithm under different numbers of modal components K for the ultra-high frequency partial discharge signals of each partial discharge type;

[0063] Calculate the relative change rate of the center frequency values of the last intrinsic mode function component under adjacent K values for each partial discharge type, as shown in the following formula:

[0064] ;

[0065] Wherein, represents the center frequency value of the (K + 1)th intrinsic mode function component when the number of modal components is K + 1, It represents the central frequency value of the Kth intrinsic mode function component when the number of modal components is K. It represents the relative change rate when the number of modal components is K.

[0066] Count the number of modal components K corresponding to each partial discharge type when the relative change rate is less than or equal to the threshold, and take the number of modal components K corresponding to the most frequently occurring one among all partial discharge types when the relative change rate is less than or equal to the threshold as the optimal K value.

[0067] Specifically, for the extraction of key features of GIS UHF partial discharge signals, first, the selection of the number of modal components K needs to be considered. If the value of K is too large, it will lead to modal repetition, and if it is too small, it will lead to under-decomposition of the mode. The embodiments of this application use the central frequency method to determine the value of K and try it in ascending order. In one of the embodiments, the threshold can be set to 10%. When the relative change rate (i.e., the judgment accuracy ) of the central frequency value of the last intrinsic mode function (IMF) component under adjacent K values is less than or equal to 10%, it can be considered that the change in the central frequency value is not significant, and the value of K at this time is the optimal K value. Take the relative change rate as the judgment accuracy.

[0068] The embodiments of this application have built a UHF partial discharge signal acquisition test platform for the typical defect electrode model of the GIS device, collected the UHF partial discharge signals, and randomly selected 150 UHF partial discharge signals of three different discharge types from the large number of collected UHF partial discharge signals as the verification samples of the feature extraction method in the embodiments of this application. The programming environment is Matlab2022b.

[0069] Considering that the GIS UHF partial discharge signal is a non-stationary and complex time series, the original UHF signal is decomposed into K different intrinsic mode function components by using the VMD algorithm. The determination of the value of K has a great influence on the decomposition result. The embodiments of this application use the central frequency method to determine the optimal K value. In one of the embodiments, let the penalty factor of the VMD algorithm , and the convergence criterion tolerance . The relationship between the value of K and the central frequency value obtained through multiple groups of experiments is shown in Tables 1 to 3.

[0070] Table 1 Central frequency values of IMF components under different K values for tip discharge:

[0071]

[0072] Table 2 Central frequency values of IMF components under different K values for air gap discharge:

[0073]

[0074] Table 3 Central frequency values of IMF components under different K values of suspension discharge:

[0075]

[0076] When the judgment accuracy is concerned, it can be considered that the change of the central frequency value is not significant, and the K value at this time is the optimal K value. The UHF partial discharge signals of three types of partial discharge types are respectively at K = 6, K = 7, K = 6 , and there is no obvious change in the central frequency value after that. Therefore, the best K value is selected as 6.

[0077] Based on the training data, the optimal K value is determined, and the variational mode decomposition (VMD) algorithm with the number of modal components being the optimal K value can also be used for feature decomposition of the UHF partial discharge signals of subsequent unknown partial discharge types.

[0078] S3. Use the VMD algorithm with the number of modal components being the optimal K value to decompose the UHF partial discharge signal, and obtain K intrinsic mode function components; calculate different types of entropy for each intrinsic mode function component respectively, and obtain the values of different types of characteristic variables in each intrinsic mode function component; construct a feature matrix according to the values of different types of characteristic variables in each intrinsic mode function component.

[0079] In a specific embodiment, different types of entropy calculations include power spectrum entropy calculation, sample entropy calculation, fuzzy entropy calculation, permutation entropy calculation and distribution entropy calculation, and the result of each type of entropy calculation corresponds to a characteristic variable.

[0080] Specifically, use the VMD algorithm with the number of modal components being 6 to decompose the UHF partial discharge signals of three typical partial discharge types, and the UHF partial discharge signals of three typical partial discharge types and the VMD decomposition results are as Figures 3 - 5 shown. It can be seen from the decomposition results of the given UHF partial discharge signals that each IMF component obtained after the UHF partial discharge signals are decomposed by the VMD algorithm has certain frequency and amplitude characteristics, effectively solving the modal aliasing problem existing in the empirical mode decomposition (EMD) algorithm. In addition, there are obvious differences in the detailed characteristics between the components obtained by decomposing the UHF partial discharge signals of different partial discharge types. It can be seen from this that after the UHF partial discharge signals are decomposed by the VMD algorithm, their internal characteristics are effectively mined, enhancing the distinguishability of the UHF partial discharge signals. For the UHF partial discharge signals of unknown partial discharge types, their corresponding partial discharge types can also be determined through the frequency and amplitude characteristics of each IMF component obtained after the UHF partial discharge signals are decomposed by the VMD algorithm.

[0081] In order to conduct a quantitative analysis of the IMF components after VMD decomposition, the power spectral entropy (SE), sample entropy (SampEn), fuzzy entropy (FE), permutation entropy (PE), and distribution entropy (DistEn) of the IMF components after VMD decomposition are calculated respectively to obtain the corresponding entropy features. These entropy features are mapped to the nodes of the dynamic network to construct a feature matrix containing multi-scale complexity information. Each column of entropy features in this feature matrix constitutes a feature variable.

[0082] The feature matrix is shown as follows:

[0083] ;

[0084] where and represent the values of the th feature variable and the th feature variable in the kth intrinsic mode function component respectively, k = 1, 2,..., K, , and n is the number of feature variables;

[0085] Specifically, the UHF partial discharge signals of three types of partial discharges are randomly selected and the corresponding feature matrices are calculated as shown in Table 4-6.

[0086] Table 4 Characteristic matrix table of tip discharge:

[0087]

[0088] Table 5 Characteristic matrix table of air gap discharge:

[0089]

[0090] Table 6 Characteristic matrix table of floating discharge:

[0091]

[0092] S4, input the feature matrix into the dynamic network marker model based on the sample covariance matrix to extract the set of dominant feature variables and their corresponding index marks, use the set of index marks as the key network marks, calculate the correlation quantification index between the dominant feature variables, and construct the extracted key features based on the key network marks and the correlation quantification index.

[0093] In a specific embodiment, the feature matrix is input into a dynamic network biomarker model based on the sample covariance matrix to extract the dominant feature variables and their corresponding index marker sets, which specifically includes:

[0094] Perform standardization processing on the feature matrix to obtain the feature matrix after standardization processing;

[0095] Construct the sample covariance matrix of the feature matrix after standardization processing, perform eigenvalue decomposition on the sample covariance matrix, and screen out the dominant feature variables according to the eigenvector corresponding to the largest eigenvalue of the sample covariance matrix and the ranking of its elements, and extract the set of index markers corresponding to the dominant feature variables.

[0096] In a specific embodiment, the eigenvalue decomposition process of the sample covariance matrix is shown as follows:

[0097] ;

[0098] Among them, C represents the sample covariance matrix, is a diagonal matrix with the th eigenvalue of the sample covariance matrix C as the diagonal elements. The eigenvalues of the sample covariance matrix C are sorted from largest to smallest. The first eigenvalue of the sample covariance matrix C is the largest eigenvalue of the sample covariance matrix C, is an orthogonal matrix with the unit eigenvectors as columns, , and n is the number of feature variables;

[0099] The element in the i-th row and j-th column of the sample covariance matrix C is expressed as:

[0100] ;

[0101] Among them, represents the covariance between the eigenvector in the th column and the eigenvector in the th column of the feature matrix after standardization processing, , represents the largest eigenvalue of the sample covariance matrix, and are respectively the th element and the th element of the unit eigenvector corresponding to the largest eigenvalue of the sample covariance matrix, corresponding to the The eigenvector of the column and the eigenvector of the column ; is the -th eigenvalue of the sample covariance matrix, and are respectively the -th eigenvalue of the corresponding unit eigenvector of the -th element and the -th element;

[0102] Compare the n elements in the eigenvector corresponding to the largest eigenvalue Select the eigenvariables of the corresponding columns in the standardized feature matrix corresponding to the top m elements in terms of numerical ranking as the dominant eigenvariables, and the set of index labels of the dominant eigenvariables in the standardized feature matrix as the key network labels .

[0103] In a specific embodiment, for the dominant eigenvariables, the correlation quantification index is used to quantify the correlation between the dominant eigenvariables, as shown in the following formula:

[0104] ;

[0105] where m is the number of elements in the key network label , the value of m is set to satisfy and , P is a positive real number in [0,1], n is the number of eigenvariables, represents the eigenvector of the -th column in the standardized feature matrix and the -th column of the eigenvector covariance between, represents the absolute value, represents the L2 norm;

[0106] The key feature is the vector formed by splicing the key network label and the correlation quantification index.

[0107] Specifically, each feature of the feature matrix is subjected to Z-Score standardization to make the mean of each feature 0 and the standard deviation 1, eliminating the influence of the dimension differences of each feature on the construction of the covariance matrix. Then, it is input into the Dynamic network markers based on sample covariancematrices (SDNM) model. Set the P-value to 0.7, and select the set of feature variables corresponding to the top m elements in the numerical ranking of the eigenvector corresponding to the largest eigenvalue in the standardized feature matrix as the Dynamic network marker (DNM), and obtain the key network markers and correlation indexes , and the following gives the and extracted from 20 samples randomly selected from each category as Figure 6 and Figure 7 shown. It can be seen that the key network markers of tip discharge , and the correlation indexes are mainly concentrated between 0.91 and 0.94. The key network markers of air-gap discharge , and the correlation indexes are mainly concentrated between 0.85 and 0.9. The key network markers of suspended discharge , and the correlation indexes are mainly concentrated between 0.94 and 0.97. Therefore, the key network markers and the correlation indexes have good cohesion and separability. Thus, the embodiments of the present application can select the key network markers and the correlation indexes as the key features of the UHF partial discharge signal. Specifically, the vector formed by splicing the key network markers and the correlation indexes can be used as the key feature.

[0108] To verify the advantages of the method proposed in the embodiments of the present application in the feature extraction of GIS UHF partial discharge signals, the key features extracted in the embodiments of the present application are input into a Support Vector Machine (SVM) for classification, together with the features extracted by using EMD, EEMD decomposition and the SDNM model, the features extracted by using VMD decomposition and separately extracting SE or PE features, and the features extracted by using VMD decomposition and calculating Multiple entropy (ME). All models divide the first 80% of the training data into the training set and the last 20% of the training data into the test set. The recognition results of each model are shown in Table 7.

[0109] Table 7 Accuracy of different feature extraction methods for partial discharge type recognition:

[0110]

[0111] The following conclusions can be drawn from the test results:

[0112] 1) From the recognition results of the three partial discharge type recognition methods of VMD-SDNM-SVM, EMD-SDNM-SVM and EEMD-SDNM-SVM, it can be seen that compared with the two signal modal decomposition algorithms of EMD and EEMD, the recognition accuracy of using the VMD algorithm to decompose the UHF partial discharge signal is increased by 16.4% and 10.95% respectively. It can be seen from this that the decomposition of the VMD algorithm can better realize the excavation of the internal characteristics of the UHF partial discharge signal, and the obtained components can better represent the original signal.

[0113] 2) By comparing the recognition results of the three partial discharge type recognition methods of VMD-SDNM-SVM, VMD-PE-SVM and VMD-SE-SVM, it can be seen that through the SDNM model, the complementary information of various entropy indexes is combined, and the non-linear coupling relationship between various entropy indexes is considered. Compared with only extracting the independent entropy value features of each modal component, it has stronger robustness and distinguishability.

[0114] 3) By comparing the recognition results of the VMD-SDNM-SVM and VMD-ME-SVM diagnostic models, it can be seen that through the SDNM model, not only the complementary information of various entropy indexes is fully integrated, and the non-linear coupling relationship between various entropy indexes is considered, but also the dimension disaster problem will not be caused, and the distinguishability of different partial discharge types of GIS is enhanced. Therefore, the partial discharge defects of GIS can be well classified.

[0115] Further referring to Figure 8 , as an implementation of the methods shown in the above figures, the present application provides an embodiment of a feature extraction device for UHF partial discharge signals of GIS. This device embodiment corresponds to Figure 1 the method embodiment shown, and this device can be specifically applied to various electronic devices.

[0116] The embodiment of the present application provides a feature extraction device for UHF partial discharge signals of GIS, including:

[0117] A preprocessing module 1, configured to obtain the UHF signal of GIS and perform preprocessing to obtain UHF partial discharge signals;

[0118] The K-value optimization module 2 is configured to obtain training data and optimize the number of modal components K of the VMD algorithm using the central frequency method to obtain the optimal K-value;

[0119] The feature decomposition module 3 is configured to decompose the UHF partial discharge signal using the VMD algorithm with the number of modal components being the optimal K-value to obtain K intrinsic mode function components; calculate different types of entropy for each intrinsic mode function component respectively to obtain the values of different types of feature variables in each intrinsic mode function component; construct a feature matrix based on the values of different types of feature variables in each intrinsic mode function component;

[0120] The key feature extraction module 4 is configured to input the feature matrix into the dynamic network biomarker model based on the sample covariance matrix to extract the set of dominant feature variables and their corresponding index labels, use the set of index labels as the key network labels, calculate the correlation quantification index between the dominant feature variables, and construct the extracted key features based on the key network labels and the correlation quantification index.

[0121] Furthermore, construct a GIS partial discharge type recognition model and train it to obtain a trained partial discharge type recognition model. The partial discharge type recognition model includes the above feature extraction device and classification device; input the key features into the trained partial discharge type recognition model, first extract the corresponding key features through the feature extraction device, and then input the key features into the classification device to identify the corresponding partial discharge type.

[0122] Specifically, the key features of different partial discharge types can be used to construct training data, and a classification device is trained for each different partial discharge type using the training data to obtain a trained classification device. The classification device can select classifiers such as SVM. Input the key features corresponding to the UHF partial discharge signal to be predicted into the trained classification device corresponding to each partial discharge type to identify the corresponding partial discharge type.

[0123] Figure 9 It is a schematic diagram of the hardware structure of the electronic device provided by the embodiment of the present invention. As Figure 9 shown, the electronic device of this embodiment includes: a processor 901 and a memory 902; wherein the memory 902 is used to store computer execution instructions; the processor 901 is used to execute the computer execution instructions stored in the memory to implement each step executed by the electronic device in the above embodiment. Specifically, reference can be made to the relevant descriptions in the foregoing method embodiments.

[0124] Optionally, the memory 902 can be either independent or integrated with the processor 901.

[0125] When the memory 902 is independently provided, the electronic device further includes a bus 903 for connecting the memory 902 and the processor 901.

[0126] An embodiment of the present invention further provides a computer storage medium, in which computer-executable instructions are stored. When the processor 901 executes the computer-executable instructions, the above method is implemented.

[0127] An embodiment of the present invention further provides a computer program product, including a computer program. When the computer program is executed by the processor 901, the above method is implemented.

[0128] In the embodiments provided by the present invention, it should be understood that the disclosed device and method can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of modules is only a logical function division. In actual implementation, there may be other division methods. For example, multiple modules can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed mutual coupling or direct coupling or communication connection can be through some interfaces, and the indirect coupling or communication connection of devices or modules can be in electrical, mechanical or other forms.

[0129] The modules described as separate components may or may not be physically separated. The components displayed as modules may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the modules can be selected according to actual needs to implement the solution of this embodiment.

[0130] In addition, in each embodiment of the present invention, the functional modules can be integrated in a processing unit, or each module can exist physically alone, or two or more modules can be integrated in one unit. The unit formed by the above modules can be implemented in the form of hardware, or in the form of a hardware plus a software functional unit.

[0131] The above integrated modules implemented in the form of software functional modules can be stored in a computer-readable storage medium. The above software functional modules are stored in a storage medium, including several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) or the processor 901 to execute some steps of the methods in various embodiments of the present application.

[0132] It should be understood that the above-mentioned processor 901 may be a Central Processing Unit (CPU), or may also be other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), etc. The general-purpose processor may be a microprocessor or the processor 901 may also be any conventional processor 901, etc. The steps of the method disclosed in combination with the invention can be directly embodied as being executed and completed by the hardware processor 901, or by a combination of the hardware and software modules in the processor 901.

[0133] The memory 902 may include high-speed RAM memory, and may also include non-volatile storage NVM, such as at least one disk memory, and may also be a USB flash drive, a mobile hard disk, a read-only memory, a magnetic disk, or an optical disc, etc.

[0134] The bus 903 may be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, an Extended Industry Standard Architecture (EISA) bus, etc. The bus 903 can be divided into an address bus, a data bus, a control bus, etc. For the sake of convenience in representation, the bus 903 in the attached drawings of this application is not limited to only one bus 903 or one type of bus 903.

[0135] The above storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic memory, flash memory, magnetic disk or optical disc. The storage medium can be any available medium that can be accessed by a general-purpose or special-purpose computer.

[0136] An exemplary storage medium is coupled to the processor 901, enabling the processor 901 to read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be a component of the processor 901. The processor 901 and the storage medium can be located in an Application Specific Integrated Circuits (ASIC). Of course, the processor 901 and the storage medium can also exist as discrete components in an electronic device or a master device.

[0137] Those of ordinary skill in the art can understand that all or part of the steps for implementing the above method embodiments can be completed by hardware related to program instructions. The foregoing program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps including the above method embodiments; and the foregoing storage medium includes: various media such as ROM, RAM, magnetic disks, or optical discs that can store program codes.

[0138] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for extracting features of UHF partial discharge signals of GIS, characterized in that, Including the following steps: Obtain GIS UHF signals and perform preprocessing to obtain UHF partial discharge signals; Obtain training data and optimize the number of modal components K of the VMD algorithm using the central frequency method to obtain the optimal K value; Decompose the UHF partial discharge signal using the VMD algorithm with the optimal K value of the number of modal components to obtain K intrinsic mode function components; calculate different types of entropy for each intrinsic mode function component respectively to obtain the values of different types of characteristic variables in each intrinsic mode function component; construct a feature matrix according to the values of different types of characteristic variables in each intrinsic mode function component; Input the feature matrix into the dynamic network biomarker model based on the sample covariance matrix to extract the set of dominant characteristic variables and their corresponding index labels, use the set of index labels as the key network labels, calculate the correlation quantification index between the dominant characteristic variables, and construct the extracted key features based on the key network labels and the correlation quantification index.

2. The method for extracting the characteristics of the UHF partial discharge signal of GIS according to claim 1, wherein Obtain training data and optimize the number of modal components K of the VMD algorithm using the central frequency method to obtain the optimal K value, specifically including: Collect UHF partial discharge signals corresponding to different partial discharge types of GIS devices to form training data; Calculate the central frequency values of each intrinsic mode function component obtained by the VMD algorithm under different numbers of modal components K for the UHF partial discharge signal of each partial discharge type; Calculate the relative change rate of the central frequency value of the last intrinsic mode function component under adjacent K values for each partial discharge type, as shown in the following formula: ； Among them, represents the central frequency value of the (K + 1)-th intrinsic mode function component when the number of modal components is K + 1, represents the central frequency value of the K-th intrinsic mode function component when the number of modal components is K, represents the relative change rate when the number of modal components is K; Count the number of modal components K corresponding to each partial discharge type when the relative change rate is less than or equal to the threshold, and use the number of modal components K corresponding to the most occurrences when the relative change rate is less than or equal to the threshold among all partial discharge types as the optimal K value.

3. The method for extracting features of GIS ultra-high frequency partial discharge signals according to claim 1, wherein, The different types of entropy calculations include power spectral entropy calculation, sample entropy calculation, fuzzy entropy calculation, permutation entropy calculation, and distribution entropy calculation, and the result of each type of entropy calculation corresponds to a characteristic variable.

4. The feature extraction method of GIS ultra-high frequency partial discharge signal according to claim 1, characterized in that, Input the feature matrix into the dynamic network biomarker model based on the sample covariance matrix to extract the set of dominant characteristic variables and their corresponding index label sets, specifically including: Perform standardization processing on the feature matrix to obtain the standardized feature matrix; Construct the sample covariance matrix of the standardized feature matrix, perform eigenvalue decomposition on the sample covariance matrix, and screen out the dominant characteristic variables according to the eigenvector corresponding to the largest eigenvalue of the sample covariance matrix and the ranking of its elements, and extract the set of index labels corresponding to the dominant characteristic variables.

5. The method for extracting the characteristics of the GIS ultra-high frequency partial discharge signal according to claim 4, wherein The eigenvalue decomposition process of the sample covariance matrix is shown in the following formula: ； where C represents the sample covariance matrix, is a diagonal matrix with the -th eigenvalue of the sample covariance matrix C as the diagonal element, and the eigenvalues of the sample covariance matrix C are sorted from largest to smallest, and the first eigenvalue of the sample covariance matrix C is the largest eigenvalue of the sample covariance matrix C, is an orthogonal matrix with the unit eigenvector as columns, , and n is the number of characteristic variables; The element in the \(i\)-th row and \(j\)-th column of the sample covariance matrix \(C\) is expressed as: ； Among them, represents the eigenvector of the th column and the eigenvector of the th column in the standardized feature matrix, and is the maximum eigenvalue of the sample covariance matrix, and are respectively the th element and the th element of the unit eigenvector corresponding to the maximum eigenvalue of the sample covariance matrix, corresponding to the eigenvector of the th column and the eigenvector of the th column in the standardized feature matrix respectively; is the th eigenvalue of the sample covariance matrix, and are respectively the th element and the th element of the unit eigenvector corresponding to the th eigenvalue of the sample covariance matrix; Compare the eigenvectors corresponding to the maximum eigenvalue among the n elements, select the eigenvariables of the corresponding columns in the standardized feature matrix corresponding to the top m elements in terms of numerical value as the dominant eigenvariables, and the set of index marks of the dominant eigenvariables in the standardized feature matrix as the key network marks .

6. The method for extracting features of GIS ultra-high frequency partial discharge signals according to claim 5, characterized in that For the dominant feature variables, a correlation quantification index is adopted to quantify the correlation between the dominant feature variables, as shown in the following formula: ； where m is the key network label is the number of elements in, and the value of m is set to satisfy and , where P is a positive real number in [0, 1], n is the number of feature variables, represents the th column eigenvector and the th column eigenvector in the covariance between, represents the absolute value, represents the L2 norm; The key feature is a vector formed by concatenating the key network label and the correlation quantification index.

7. A feature extraction device for UHF partial discharge signals of GIS, characterized in that, Including: A preprocessing module configured to obtain GIS UHF signals and perform preprocessing to obtain UHF partial discharge signals; The K-value optimization module is configured to obtain training data and optimize the number of modal components K of the VMD algorithm by using the center frequency method to obtain the optimal K-value; The feature decomposition module is configured to decompose the ultra-high frequency partial discharge signal by using the VMD algorithm with the number of modal components being the optimal K-value to obtain K intrinsic mode function components; calculate different types of entropy for each intrinsic mode function component respectively to obtain the values of different types of feature variables in each intrinsic mode function component; construct a feature matrix according to the values of different types of feature variables in each intrinsic mode function component; The key feature extraction module is configured to input the feature matrix into the dynamic network marker model based on the sample covariance matrix to extract the set of dominant feature variables and their corresponding index markers, use the set of index markers as the key network markers, calculate the correlation quantification index between the dominant feature variables, and construct the extracted key features based on the key network markers and the correlation quantification index.

8. An electronic device, comprising: One or more processors; A storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method according to any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the method according to any one of claims 1-6.

Images