Gas insulation equipment partial discharge detection method and device, electronic equipment, readable storage medium and program product

Abstract

The invention relates to a gas insulation equipment partial discharge detection method and device, electronic equipment, a readable storage medium and a program product, and the method comprises the steps: obtaining a pulse current PRPD atlas and an ultrahigh frequency PRPD atlas through signal samples collected under different excitation voltages, carrying out the feature extraction of the pulse current PRPD atlas and the ultrahigh frequency PRPD atlas, and carrying out the detection of partial discharge of gas insulation equipment. The method comprises the following steps: acquiring a characteristic signal matrix under different excitation voltages, performing dimension reduction processing on the characteristic signal matrix by adopting a UMAP algorithm to generate a corresponding low-dimensional characteristic representation, performing normalization and visualization processing on the low-dimensional characteristic representation, and acquiring a discharge mode distribution model under different excitation voltages; and detecting the discharge mode of the partial discharge signal of the gas insulation equipment based on the discharge mode distribution model. According to the method, an ultrahigh frequency signal and a pulse current signal are fused, and UMAP is used for carrying out dimension reduction processing, so that accurate detection and classification of discharge of metal particles in the gas insulation equipment are realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of discharge identification, in particular to a partial discharge detection method and device for gas insulated equipment, an electronic device, a readable storage medium and a program product. BACKGROUND

[0002] The gas insulated equipment refers to a metal-enclosed switchgear which uses gas as an insulating medium. The most common gas insulated equipment is a gas insulated switchgear (GIS), which is composed of circuit breakers, disconnectors, grounding switches, current transformers, voltage transformers, arresters, busbars, connecting pieces and outgoing terminal ends. All these devices or components are completely enclosed in a metal ground shell and filled with SF6 (sulfur hexafluoride) gas as an insulating and arc extinguishing medium. Due to its excellent insulation and arc extinguishing performance, the gas insulated equipment is widely used in power systems and undertakes key tasks.

[0003] However, metal particles generated during the installation and operation of the gas insulated equipment cannot be avoided, which may reduce the flashover voltage and cause discharge accidents. In the conventional technology, in order to achieve early detection, partial discharge signals caused by metal particles are generally detected by identification. However, the conventional detection method has low accuracy. SUMMARY

[0004] Therefore, it is necessary to provide a partial discharge detection method and device for gas insulated equipment, an electronic device, a computer readable storage medium and a computer program product, which can improve the detection accuracy.

[0005] In a first aspect, the present application provides a partial discharge detection method for gas insulated equipment, which comprises:

[0006] obtaining a pulse current PRPD spectrum based on pulse current signal samples collected under different excitation voltages, and obtaining an ultra-high frequency PRPD spectrum based on ultra-high frequency signal samples collected under different excitation voltages;

[0007] performing feature extraction on the pulse current PRPD spectrum and the ultra-high frequency PRPD spectrum to obtain a feature signal matrix under different excitation voltages;

[0008] performing dimensionality reduction processing on the feature signal matrix under different excitation voltages by using a UMAP algorithm to generate a corresponding low-dimensional feature representation;

[0009] performing normalization and visualization processing on the low-dimensional feature representation to obtain a discharge mode distribution model based on different excitation voltages;

[0010] Detect a discharge mode of the partial discharge signal of the gas insulated equipment based on the discharge mode distribution model.

[0011] In one of the embodiments, the feature extraction on the pulse current PRPD pattern and the ultra-high frequency PRPD pattern obtains a feature signal matrix under different excitation voltages, which includes:

[0012] The first feature extraction on the pulse current PRPD pattern obtains a first feature signal sequence of apparent discharge magnitude under different excitation voltages;

[0013] The second feature extraction on the ultra-high frequency PRPD pattern obtains a second feature signal sequence representing pattern morphology under different excitation voltages;

[0014] The first feature signal in the first feature signal sequence and the second feature signal in the second feature signal sequence under the same excitation voltage are obtained to generate a feature signal matrix corresponding to the excitation voltage.

[0015] In one of the embodiments,

[0016] The first feature extraction includes the extraction of maximum partial discharge amplitude, total partial discharge, average positive partial discharge, total positive partial discharge, average negative partial discharge, and total negative partial discharge based on the pulse current PRPD pattern;

[0017] The second feature extraction includes the extraction of ultra-high frequency total pulse number, standard deviation, skewness, and kurtosis based on the ultra-high frequency PRPD pattern.

[0018] In one of the embodiments, the UMAP algorithm is used to perform dimension reduction processing on the feature signal matrix under different excitation voltages to generate corresponding low-dimensional feature representations, which includes:

[0019] Based on the similarity between the feature signal matrices under different excitation voltages, the near-neighbor relationship of the feature signal matrix under each excitation voltage is determined;

[0020] The feature signal matrix under different excitation voltages is mapped to a low-dimensional space to obtain corresponding initial low-dimensional features;

[0021] The cross-entropy loss between the near-neighbor relationship of the feature signal matrix and the corresponding initial low-dimensional features is obtained;

[0022] The initial low-dimensional features are iteratively updated based on the cross-entropy loss until the cross-entropy loss converges or reaches a maximum number of iterations, and the updated initial low-dimensional features are determined as the low-dimensional feature representations corresponding to the feature signal matrix.

[0023] In one of the embodiments, the normalization and visualization processing of the low-dimensional feature representation is performed to obtain a discharge mode distribution model based on discharge mode distribution under different excitation voltages, which includes:

[0024] The low-dimensional feature representation is normalized to obtain a normalized low-dimensional feature representation;

[0025] The normalized low-dimensional feature representation is mapped to a three-dimensional space to obtain a distribution result of the low-dimensional feature representation in the three-dimensional space under different excitation voltages, wherein the distribution result includes the aggregation state of different clusters, and different clusters represent different discharge modes.

[0026] A corresponding discharge mode distribution model is generated based on the distribution result of the three-dimensional space.

[0027] In one of the embodiments, the discharge mode distribution model is used to detect the discharge mode of the partial discharge signal of the gas insulated equipment, which includes:

[0028] Based on the pulse current method and the ultra-high frequency method, the characteristic signal of the partial discharge signal of the gas insulated equipment is collected;

[0029] The UMAP algorithm is used to perform dimensionality reduction processing on the characteristic signal to generate a low-dimensional feature representation of the partial discharge signal of the gas insulated equipment;

[0030] The low-dimensional feature representation is mapped to the discharge mode distribution model to determine a target cluster in which the low-dimensional feature representation is located in the discharge mode distribution model;

[0031] The discharge mode corresponding to the target cluster is obtained, and the discharge mode corresponding to the target cluster is determined as the discharge mode of the partial discharge signal of the gas insulated equipment.

[0032] In a second aspect, the present application provides a partial discharge detection device for a gas insulated equipment, which includes:

[0033] A graph obtaining module is configured to obtain a pulse current PRPD graph based on pulse current signal samples collected under different excitation voltages, and obtain an ultra-high frequency PRPD graph based on ultra-high frequency signal samples collected under different excitation voltages;

[0034] A feature extraction module is configured to perform feature extraction on the pulse current PRPD graph and the ultra-high frequency PRPD graph to obtain a feature signal matrix under different excitation voltages;

[0035] A dimensionality reduction processing module is configured to perform dimensionality reduction processing on the feature signal matrix under different excitation voltages by using the UMAP algorithm to generate a corresponding low-dimensional feature representation;

[0036] The model obtaining module is configured to normalize and visualize the low-dimensional feature representation, and obtain a discharge mode distribution model based on the discharge under different excitation voltages.

[0037] The discharge detection module is configured to detect the discharge mode of the partial discharge signal of the gas insulated equipment based on the discharge mode distribution model.

[0038] In a third aspect, the present application provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the method in the first aspect when executing the computer program.

[0039] In a fourth aspect, the present application provides a computer readable storage medium, which stores a computer program, and the computer program implements the steps of the method in the first aspect when executed by a processor.

[0040] In a fifth aspect, the present application provides a computer program product, comprising a computer program, and the computer program implements the steps of the method in the first aspect when executed by a processor.

[0041] The gas insulated equipment partial discharge detection method, device, electronic device, computer readable storage medium and computer program product, by acquiring the pulse current PRPD spectrum and the ultra-high frequency PRPD spectrum based on the signal samples collected under different excitation voltages, extracting features from the pulse current PRPD spectrum and the ultra-high frequency PRPD spectrum, acquiring the feature signal matrix under different excitation voltages, performing dimension reduction processing on the feature signal matrix under different excitation voltages by using the UMAP algorithm, generating the corresponding low-dimensional feature representation, normalizing and visualizing the low-dimensional feature representation, and obtaining the discharge mode distribution model based on the discharge under different excitation voltages, thereby detecting the discharge mode of the partial discharge signal of the gas insulated equipment based on the discharge mode distribution model. By fusing the ultra-high frequency signal and the pulse current signal, and using the UMAP for dimension reduction processing, the internal metal particle discharge of the gas insulated equipment is accurately detected and classified, the processing efficiency is improved, and the accuracy and robustness of the discharge mode recognition are improved. BRIEF DESCRIPTION OF DRAWINGS

[0042] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related art, the drawings needed to be used in the description of the embodiments of the present application or the related art will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other related drawings can also be obtained without creative labor.

[0043] Figure 1 It is a flowchart of the gas insulated equipment partial discharge detection method in an embodiment.

[0044] Figure 2 Structure diagram of a measuring device in an embodiment;

[0045] Figure 3 Enlarged view of a test unit in an embodiment;

[0046] Figure 4 PRPD pattern diagram of pulse current in an embodiment;

[0047] Figure 5 PRPD pattern diagram of ultra-high frequency in an embodiment;

[0048] Figure 6 Flowchart of feature extraction steps in an embodiment;

[0049] Figure 7 Diagram of a first feature signal sequence extracted in an embodiment;

[0050] Figure 8 Diagram of a second feature signal sequence extracted in an embodiment;

[0051] Figure 9 Flowchart of UMAP algorithm steps in an embodiment;

[0052] Figure 10A Visualization diagram of a low-dimensional space in an embodiment;

[0053] Figure 10B Visualization diagram of a discharge mode distribution model in an embodiment;

[0054] Figure 11 Structure block diagram of a partial discharge detection device for gas insulated equipment in an embodiment;

[0055] Figure 12 Internal structure diagram of an electronic device in an embodiment. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical scheme and advantages of the present application clearer, the present application will be further described in detail below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0057] In an embodiment, as shown in Figure 1 , a partial discharge detection method for gas insulated equipment is provided, which comprises the following steps:

[0058] Step 102, obtaining pulse current PRPD pattern and ultra-high frequency PRPD pattern based on signal samples collected under different excitation voltages.

[0059] Specifically, the pulse current PRPD (Phase-resolved partial discharge) pattern can be obtained based on the pulse current signal samples collected under different excitation voltages, and the ultra-high frequency PRPD pattern can be obtained based on the ultra-high frequency signal samples collected under different excitation voltages.

[0060] The pulse current signal samples and the ultra-high frequency signal samples can be measured by a measuring device as shown in FIG. 1. Figure 2 The measuring device includes, in sequence, an UHF (Ultra-high frequency) sensor, a Pico oscilloscope, a computer, an MPD800, a coupling capacitor, a resistor, a transformer, and an AC voltage source. The computer is used for receiving and processing data. The Pico oscilloscope is used for displaying and analyzing signals. The MPD800 is connected to the UHF sensor and used for collecting pulse current and ultra-high frequency signals. The UHF sensor is used for detecting the partial discharge signals generated by the metal particles. The coupling capacitor is used for signal coupling. The resistor is used for resistance adjustment in the circuit. The transformer is used for voltage conversion. The AC voltage source provides power for the entire circuit. The test unit as the research object is connected to the measuring device, and its enlarged view is shown in FIG. 2, in which a metal particle with a diameter of 1 mm can be placed as a sample. Figure 3

[0061] In this embodiment, the partial discharge signal samples generated by the metal particles in the test unit under different excitation voltages can be collected based on the measuring device as shown in FIG. 1, and PRPD patterns can be constructed. Figure 2 For example, the pulse current PRPD pattern as shown in FIG. 3 (wherein the abscissa is the phase and the ordinate is the partial discharge amplitude) can be constructed based on the pulse current signal samples collected under different excitation voltages, and the ultra-high frequency PRPD pattern as shown in FIG. 4 can be constructed based on the ultra-high frequency signal samples collected under different excitation voltages. Figure 4 Figure 5

[0062] Step 104, feature extraction is performed on the pulse current PRPD pattern and the ultra-high frequency PRPD pattern to obtain a feature signal matrix under different excitation voltages.

[0063] The feature extraction is a process of extracting feature parameters having corresponding change rules for defect evolution caused by the metal particles from the pulse current PRPD pattern and the ultra-high frequency PRPD pattern. The feature signal matrix is a matrix composed of the extracted feature parameters.

[0064] ​​​Since the discharge amplitude increases with the increase of voltage, the characteristic parameters can be extracted from the PRPD spectrum to quantify the internal correlation. In pulse current measurement, the apparent discharge magnitude is the most direct indicator to characterize the defect-induced discharge process. Unlike pulse current PRPD, UHF PRPD focuses on the variation of discharge frequency with phase, which is more focused on the spectrum morphology. Therefore, the indicators of apparent discharge magnitude under different excitation voltages can be extracted as characteristic parameters based on the pulse current PRPD spectrum, and the indicators representing the spectrum morphology under different excitation voltages can be extracted as characteristic parameters based on the UHF PRPD spectrum, and the characteristic signal matrix corresponding to the excitation voltage is formed based on the extracted characteristic parameters. By fusing the characteristic parameters of UHF signals and pulse current signals, and then performing discharge analysis based on subsequent steps, the accuracy of analysis can be improved.

[0065] In step 106, the UMAP algorithm is used to perform dimensionality reduction processing on the characteristic signal matrix under different excitation voltages to generate the corresponding low-dimensional feature representation.

[0066] The UMAP (Uniform Manifold Approximation and Projection) algorithm is a nonlinear dimensionality reduction algorithm that maps high-dimensional data to low-dimensional space while preserving the global structure and local structure of the data, facilitating data visualization and clustering analysis.

[0067] The low-dimensional feature representation is a low-dimensional representation obtained by performing dimensionality reduction processing on the characteristic signal matrix based on the UMAP algorithm. In this embodiment, the UMAP algorithm is used to perform dimensionality reduction processing on the characteristic signal matrix under different excitation voltages to generate the corresponding low-dimensional feature representation, thereby facilitating subsequent analysis, which not only improves the calculation efficiency, but also accurately captures the potential geometric structure of the data set.

[0068] In step 108, the low-dimensional feature representation is normalized and visualized to obtain a discharge mode distribution model based on different excitation voltages.

[0069] The discharge mode distribution model includes the clustering state of the low-dimensional feature representation under different excitation voltages, and different clusters represent different discharge modes.

[0070] In this embodiment, the low-dimensional feature representation obtained above is normalized and visualized to obtain a discharge mode distribution model based on different excitation voltages for subsequent detection of actual partial discharge signals of gas insulated equipment.

[0071] In step 110, the discharge mode of the partial discharge signal of the gas insulated equipment is detected based on the discharge mode distribution model.

[0072] The discharge mode is used to represent the evolution of the defect caused by the metal particles, including a discharge initial stage without discharge risk, a discharge middle stage with discharge risk, and a discharge final stage with serious discharge risk.

[0073] In the embodiment, the discharge mode of the partial discharge signal of the gas insulated equipment can be detected based on the discharge mode distribution model obtained in the above steps. Since the clustering states of the low-dimensional feature representation under different excitation voltages are included in the discharge mode distribution model, and different clustering clusters represent different discharge modes. Therefore, the partial discharge signal of the gas insulated equipment can be mapped into the discharge mode distribution model to determine the clustering cluster after mapping. The discharge mode of the corresponding clustering cluster is determined as the discharge mode of the partial discharge signal, and the discharge detection of the partial discharge signal of the gas insulated equipment is realized.

[0074] In the above gas insulated equipment partial discharge detection method, the pulse current PRPD spectrum and the ultra-high frequency PRPD spectrum are obtained based on the signal samples collected under different excitation voltages, the feature extraction is performed on the pulse current PRPD spectrum and the ultra-high frequency PRPD spectrum, the feature signal matrix under different excitation voltages is obtained, the UMAP algorithm is used to perform dimension reduction processing on the feature signal matrix under different excitation voltages, the corresponding low-dimensional feature representation is generated, and the low-dimensional feature representation is normalized and visualized to obtain the discharge mode distribution model under different excitation voltages. The discharge mode of the partial discharge signal of the gas insulated equipment is detected based on the discharge mode distribution model. By fusing the ultra-high frequency signal and the pulse current signal and using UMAP for dimension reduction processing, accurate detection and classification of the discharge of the metal particles inside the gas insulated equipment are realized, the processing efficiency is improved, and the accuracy and robustness of the discharge mode recognition are improved.

[0075] In one exemplary embodiment, as shown in FIG. 1, the method includes the following steps: Figure 6 In step 104, the feature extraction is performed on the pulse current PRPD spectrum and the ultra-high frequency PRPD spectrum to obtain the feature signal matrix under different excitation voltages. Specifically, the feature extraction can include:

[0076] In step 602, the first feature extraction is performed on the pulse current PRPD spectrum to obtain the first feature signal sequence of the apparent discharge level under different excitation voltages.

[0077] The first feature extraction includes the extraction of the maximum partial discharge amplitude, the total partial discharge amount, the average value of the positive partial discharge, the total value of the positive partial discharge, the average value of the negative partial discharge, and the total value of the negative partial discharge based on the pulse current PRPD spectrum.

[0078] For example, the maximum partial discharge amplitude can be based on the average value of the maximum apparent discharge level The determination can be specifically calculated based on the following formula (1):

[0079] , (1)

[0080] wherein, represents the discharge magnitude of the jth sampling point in the ith measurement, and N is the total number of measurements. is the average value of the absolute value of the discharge magnitude of the sampling point with the largest discharge magnitude in each sampling of N samplings, that is, the maximum local discharge amplitude.

[0081] The total local discharge amount can be based on the average value of the average apparent discharge magnitude The determination can be specifically calculated based on the following formula (2):

[0082] , (2)

[0083] wherein, represents the discharge magnitude of the jth sampling point in the ith measurement. is the average value of the absolute value of the total amount of each sampling of N samplings, that is, the total local discharge amount.

[0084] The average value of the positive local discharge The average value of the negative local discharge , for representing the discharge characteristics of the positive and negative periods, which can be calculated based on the following formulas (3) and (4):

[0085] , , (3)

[0086] , , (4)

[0087] wherein, is the discharge magnitude of the s th sampling point of the positive half cycle in the ith measurement, is the discharge magnitude of the t th sampling point of the negative half cycle in the ith measurement.

[0088] The total value of the positive local discharge The total value of the negative local discharge , which can be calculated based on the following formulas (5) and (6):

[0089] , , (5)

[0090] , , (6)

[0091] Wherein, O is the sampling point number of positive half cycle, K is the sampling point number of negative half cycle.

[0092] In the embodiment, according to the above formulas (1) to (6), the pulse current PRPD spectrum under each excitation voltage is obtained based on the above pulse current characteristic parameters under each excitation voltage, and the first characteristic signal sequence of the apparent discharge magnitude under different excitation voltages is obtained based on the above pulse current characteristic parameters under each excitation voltage, and the result is as shown in Figure 4 Figure 7 Wherein, the abscissa is the excitation voltage, and each ordinate is the above corresponding pulse current characteristic parameter.

[0093] Step 604, the second characteristic extraction of the ultra-high frequency PRPD spectrum is carried out, and the second characteristic signal sequence of the spectrum form under different excitation voltages is obtained.

[0094] Wherein, the second characteristic extraction includes PRPD total pulse number, PRPD standard deviation, PRPD skewness and PRPD kurtosis extraction based on the ultra-high frequency PRPD spectrum.

[0095] For example, the PRPD total pulse number Specifically, it can be calculated based on the following formula (7):

[0096] , (7)

[0097] Wherein, i represents the sampling times, s represents the number of sampling points, The PD (Pulse Doppler, Pulse Doppler) pulse of the s-th sampling point with amplitude exceeding the threshold coefficient in the i-th sampling.

[0098] PRPD standard deviation For measuring the fluctuation range of discharge amplitude and the dispersion degree of data distribution, it can be calculated based on the following formula (8):

[0099] , (8)

[0100] Wherein, The average pulse number of sampling points in the i-th measurement, The pulse of the s-th sampling point in the i-th sampling, O is the number of sampling points, and N is the sampling times.

[0101] PRPD skewness And PRPD kurtosis , used for classifying discharge mode in PRPD. The greater the PRPD skewness indicates that the distribution is more asymmetric, and the higher the PRPD kurtosis indicates that the distribution is more sharp than the normal distribution, which can be determined by the following formulas (9) and (10):​

[0102] (9)

[0103] (10)

[0104] wherein, represents the standard deviation of the data.

[0105] In this embodiment, according to the above formulas (7) to (10), the above ultra-high frequency characteristic parameters under each excitation voltage can be obtained by calculation based on the ultra-high frequency PRPD spectrum shown in FIG. 6, and the second characteristic signal sequence representing the spectrum form under different excitation voltages can be obtained based on the above ultra-high frequency characteristic parameters under each excitation voltage, and the results are shown in FIG. 7, wherein the abscissa is the excitation voltage, and each ordinate is the above corresponding ultra-high frequency characteristic parameter. Figure 5 Figure 8

[0106] Step 606, obtaining the first characteristic signal in the first characteristic signal sequence and the second characteristic signal in the second characteristic signal sequence under the same excitation voltage, and generating a characteristic signal matrix under the corresponding excitation voltage.

[0107] Specifically, based on the first characteristic signal in the first characteristic signal sequence and the second characteristic signal in the second characteristic signal sequence under the same excitation voltage, a characteristic signal matrix under the corresponding excitation voltage is generated.

[0108] Since the first characteristic signal sequence is the pulse current characteristic parameter of the partial discharge signal measured based on the pulse current method, and the second characteristic signal sequence is the ultra-high frequency characteristic parameter of the partial discharge signal measured based on the ultra-high frequency method, therefore, by fusing the characteristic parameters obtained by the two kinds of measurement methods, the corresponding characteristic signal matrix :

[0109]

[0110] Further, the defect evolution caused by metal particles can be checked by the corresponding change rule of the 10 characteristic parameters.

[0111] However, if all parameters are used without selection to quantify the defect evolution process, not only the calculation complexity will be increased, but also the randomness of the results may be introduced due to the different orders of magnitude of the parameters. Although the typical correlation analysis can eliminate the parameters with lower contribution to the discharge analysis, the feature matrix processed by the typical correlation analysis is still in a high-dimensional space.

[0112] ​​Based on this, this embodiment performs dimensionality reduction processing on the above-mentioned feature signal matrix based on the Riemannian manifold hypothesis and uniform manifold approximation and projection (i.e., UMAP algorithm) of approximate nearest neighbor search, which not only improves the computational efficiency but also accurately captures the potential geometric structure of the data set.

[0113] In an exemplary embodiment, in step 106, the UMAP algorithm is used to perform dimensionality reduction processing on the characteristic signal matrix under different excitation voltages to generate corresponding low-dimensional feature representations, including:

[0114] Based on the similarity between the characteristic signal matrices under different excitation voltages, the neighbor relationship of the characteristic signal matrix under each excitation voltage is determined; the characteristic signal matrix under different excitation voltages is mapped to a low-dimensional space to obtain the corresponding initial low-dimensional features; the cross-entropy loss between the neighbor relationship of the characteristic signal matrix and the corresponding initial low-dimensional features is obtained; the initial low-dimensional features are iteratively updated based on the cross-entropy loss until the cross-entropy loss converges or reaches the maximum number of iterations, and the updated initial low-dimensional features are determined as the low-dimensional feature representation corresponding to the characteristic signal matrix.

[0115] For example, the UMAP algorithm is Figure 9 Shown, including:

[0116] Local fuzzy simplex set: By Construct a fuzzy probability model of the local neighborhood, determine the neighbors of each point and their similarity through fuzzy logic, and deal with noise and uncertainty. Specifically, we can use formula (11) to calculate the probability of each point High-dimensional similarity of nearest neighbors :

[0117] , (11)

[0118] in, is the vector composed of the i-th data point (the vector length is 10, which contains the 10 characteristic parameters in the above characteristic signal matrix). Similarly, is the vector composed of the jth data point. pij is the sample and The high-dimensional similarity between For samples The similarity coefficient can be obtained based on the initialization assumption or through iterative calculation. For samples The standard deviation of . and Used to ensure connectivity in local neighborhoods.

[0119] Assigning distribution weights: Equation (12) assigns weights to different neighborhoods or data points by using a T-distribution with parameters, optimizing the representation after dimensionality reduction while preserving key structures.

[0120] , (12)

[0121] where, and are equation coefficients used to control the decay rate of the distribution, which can be obtained by program iteration. is the corresponding coordinate result after dimensionality reduction, is the corresponding coordinate result after dimensionality reduction. is the weight, representing the weight of each data point , in the dimensionality reduction calculation.

[0122] Symmetrization of T-distribution: Equation (13) adjusts the symmetry of the probability distribution in the low-dimensional space to eliminate directional bias and ensure the symmetry of the adjacency relationship.

[0123] , (13)

[0124] where, represents the result of the weight matrix in the i-th row and j-th column, represents the result of the j-th row and i-th column.

[0125] Embedding target space: Map high-dimensional data to low-dimensional space to generate visual low-dimensional representation and preserve local and global structures. For example, reduce high-dimensional (including 10 features) to three dimensions, so that low-dimensional features can be represented by three-dimensional plots.

[0126] Calculate cross-entropy: Cross-entropy as a loss function drives the optimization process by minimizing the inconsistency between the two space distributions, measuring the difference between the probability distributions in high-dimensional and low-dimensional spaces. Equation (14) minimizes the cross-entropy between high-dimensional and low-dimensional probability distributions to optimize low-dimensional embedding coordinates . Equation (15) iteratively updates low-dimensional coordinates to make the structure of the low-dimensional space reflect the topology of the high-dimensional data as much as possible. The specific formula is as follows:

[0127] , (14)

[0128] , (15)

[0129] where, is the data point result after dimensionality reduction; and is an adjustment coefficient, and the optimal parameter value is obtained by iterative calculation; t is the number of iterations. represents the (t+1)th iteration result and the tth iteration result and the adjustment coefficient , and the cross-entropy loss between them.

[0130] Termination judgment: by checking the convergence of the loss function or whether the maximum number of iterations is reached, it is determined when the algorithm stops iteration optimization to avoid overfitting or redundant calculation. If the loss function converges or the maximum number of iterations is reached, the final dimensionality reduction processing result is output, that is, the final is output. If the loss function does not converge or the maximum number of iterations is not reached, the embedding target space is returned to continue iteration optimization.

[0131] In this embodiment, UMAP optimizes the low-dimensional embedding to minimize the distribution difference by constructing a fuzzy graph structure of high-dimensional data, and finally generates a low-dimensional feature representation that preserves the topological characteristics of the original data. Then, after normalizing the scale of the low-dimensional feature representation, it is mapped to a visualization result.

[0132] For example, in step 108, the low-dimensional feature representation is normalized and visualized to obtain a discharge mode distribution model based on different excitation voltages, which can specifically include:

[0133] The low-dimensional feature representation is normalized to obtain a normalized low-dimensional feature representation. For example, the low-dimensional feature representation is normalized to a value between 0 and 1, so as to unify the parameter dimension.

[0134] The normalized low-dimensional feature representation is mapped to a three-dimensional space as shown in Figure 10A , to obtain the distribution result of the low-dimensional feature representation in the three-dimensional space under different excitation voltages, and then a corresponding discharge mode distribution model can be generated based on the above distribution result in the three-dimensional space. Specifically Figure 10B , wherein the distribution result includes the aggregation state of different clusters, and different clusters represent different discharge modes. As Figure 10B can be seen, there are three clusters in total, so the categories of the discharge process of the test sample under different excitation voltages include approximately 3 categories (such as the initial stage of discharge, the middle stage of discharge, and the final stage of discharge). Each scatter point represents a , so the discharge category of each can be classified according to the calculation result in the figure, so as to obtain the discharge mode distribution model. Then, the discharge category corresponding to the current can be evaluated based on the discharge mode distribution model, that is, the current discharge mode is evaluated.

[0135] In an example embodiment, in step 110, detecting the discharge mode of the partial discharge signal of the gas insulated device based on the discharge mode distribution model can specifically include:

[0136] Based on the pulse current method and the ultra-high frequency method, the characteristic signal of the partial discharge signal of the gas insulated device is collected, similar to the above , containing 10 characteristic parameters in the characteristic signal matrix. Then the UMAP algorithm is used to perform dimension reduction processing on the characteristic signal to generate a low-dimensional feature representation of the partial discharge signal of the gas insulated device. The dimension reduction processing can refer to the above embodiments, and this embodiment will not be described here.

[0137] The low-dimensional feature representation can be further mapped to a low-dimensional space as shown in Figure 10A to determine the target cluster in which the low-dimensional feature representation is located in the discharge mode distribution model. Since each cluster represents a discharge category, i.e., a discharge mode, the discharge mode corresponding to the target cluster can be determined as the discharge mode of the partial discharge signal of the gas insulated device based on the acquired discharge mode corresponding to the target cluster, thereby achieving accurate detection of the discharge mode of the partial discharge signal of the gas insulated device.

[0138] In the above embodiments, by combining the pulse current method and the ultra-high frequency method, the characteristic signal of the partial discharge signal of the gas insulated device is collected, and the high-dimensional characteristic signal is efficiently processed by the UMAP algorithm for dimension reduction to be mapped into the discharge mode distribution model, thereby realizing visualization and recognition of different discharge states, and improving the accuracy and reliability of the partial discharge diagnosis.

[0139] It should be understood that although each step in the flowchart involved in each of the above embodiments is displayed in sequence according to the arrow, these steps are not necessarily executed in sequence according to the arrow. Unless otherwise specified herein, the execution of these steps is not strictly limited in sequence, and these steps can be executed in other sequences. Moreover, at least part of the steps in the flowchart involved in each of the above embodiments can include multiple steps or stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution sequence of these steps or stages is not necessarily sequential, but can be executed in rotation or alternation with at least part of other steps or steps or stages in other steps.

[0140] Based on the same inventive concept, the embodiments of the present application also provide a gas insulated equipment partial discharge detection device for implementing the above-mentioned gas insulated equipment partial discharge detection method. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme described in the above-mentioned method, so the specific limitations in one or more gas insulated equipment partial discharge detection device embodiments provided below can refer to the limitations of the gas insulated equipment partial discharge detection method described above, which will not be repeated here.

[0141] In one exemplary embodiment, as shown in Figure 11 A gas insulated equipment partial discharge detection device is provided, comprising: a graph acquisition module 1102, a feature extraction module 1104, a dimension reduction processing module 1106, a model acquisition module 1108, and a discharge detection module 1110, wherein:

[0142] The graph acquisition module 1102 is configured to acquire a pulse current PRPD graph based on pulse current signal samples collected under different excitation voltages, and acquire an ultra-high frequency PRPD graph based on ultra-high frequency signal samples collected under different excitation voltages.

[0143] The feature extraction module 1104 is configured to perform feature extraction on the pulse current PRPD graph and the ultra-high frequency PRPD graph to acquire a feature signal matrix under different excitation voltages.

[0144] The dimension reduction processing module 1106 is configured to perform dimension reduction processing on the feature signal matrix under different excitation voltages using a UMAP algorithm to generate a corresponding low-dimensional feature representation.

[0145] The model acquisition module 1108 is configured to perform normalization and visualization processing on the low-dimensional feature representation to acquire a discharge mode distribution model based on different excitation voltages.

[0146] The discharge detection module 1110 is configured to detect a discharge mode of the gas insulated equipment partial discharge signal based on the discharge mode distribution model.

[0147] In one exemplary embodiment, the feature extraction module is specifically configured to: perform first feature extraction on the pulse current PRPD graph to acquire a first feature signal sequence of apparent discharge magnitude under different excitation voltages; perform second feature extraction on the ultra-high frequency PRPD graph to acquire a second feature signal sequence representing graph morphology under different excitation voltages; acquire a first feature signal in the first feature signal sequence and a second feature signal in the second feature signal sequence under the same excitation voltage to generate a feature signal matrix under the corresponding excitation voltage.

[0148] In an example embodiment, the first feature extraction includes extraction of maximum partial discharge amplitude, total partial discharge amount, average value of positive partial discharge, total value of positive partial discharge, average value of negative partial discharge, and total value of negative partial discharge based on the pulse current PRPD spectrum; and the second feature extraction includes extraction of ultra-high frequency total pulse number, standard deviation, skewness, and kurtosis based on the ultra-high frequency PRPD spectrum.

[0149] In an example embodiment, the dimension reduction processing module is specifically configured to: determine a near-neighbor relationship of the feature signal matrix under each excitation voltage based on similarity between the feature signal matrices under different excitation voltages; map the feature signal matrices under different excitation voltages to a low-dimensional space to obtain corresponding initial low-dimensional features; obtain a cross-entropy loss between the near-neighbor relationship of the feature signal matrix and the corresponding initial low-dimensional features; and iteratively update the initial low-dimensional features based on the cross-entropy loss until the cross-entropy loss converges or a maximum number of iterations is reached, and determine the updated initial low-dimensional features as the low-dimensional feature representation corresponding to the feature signal matrix.

[0150] In an example embodiment, the model obtaining module is specifically configured to: perform normalization processing on the low-dimensional feature representation to obtain a normalized low-dimensional feature representation; map the normalized low-dimensional feature representation to a three-dimensional space to obtain a distribution result of the low-dimensional feature representation under different excitation voltages in the three-dimensional space, the distribution result including aggregation states of different clusters, and different clusters representing different discharge modes; and generate a corresponding discharge mode distribution model based on the distribution result of the three-dimensional space.

[0151] In an example embodiment, the discharge detection module is specifically configured to: collect feature signals of the partial discharge signals of the gas insulated equipment based on the pulse current method and the ultra-high frequency method; perform dimension reduction processing on the feature signals by using a UMAP algorithm to generate a low-dimensional feature representation of the partial discharge signals of the gas insulated equipment; map the low-dimensional feature representation into the discharge mode distribution model to determine a target cluster in which the low-dimensional feature representation is located in the discharge mode distribution model; obtain a discharge mode corresponding to the target cluster, and determine the discharge mode corresponding to the target cluster as the discharge mode of the partial discharge signals of the gas insulated equipment.

[0152] The above-mentioned various modules in the gas insulated equipment partial discharge detection device can be all or partially realized by software, hardware, and combinations thereof. The above-mentioned various modules can be embedded in or independent of a processor in a computer device in a hardware form, or can be stored in a memory in a computer device in a software form, so as to be called and executed by a processor to perform operations corresponding to the above-mentioned various modules.

[0153] In an example embodiment, an electronic device is provided, an internal structure diagram of which can be as shown in FIG. 25. Figure 12 The electronic device includes a processor, a memory, an input / output interface, a communication interface, a display unit, and an input device. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface, the display unit, and the input device are connected to the system bus through the input / output interface. Among them, the processor of the electronic device is used to provide computing and control capabilities. The memory of the electronic device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The input / output interface of the electronic device is used to exchange information between the processor and external devices. The communication interface of the electronic device is used to communicate with external terminals in a wired or wireless manner. The wireless manner can be achieved through WIFI, mobile cellular network, Near Field Communication (NFC), or other technologies. The computer program is executed by the processor to implement a partial discharge detection method of a gas insulated device. The display unit of the electronic device is used to form a visually visible picture, which can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen. The input device of the electronic device can be a touch layer overlaid on the display screen, or a key, trackball, or touchpad arranged on the shell of the electronic device, or an external keyboard, touchpad, or mouse, etc.

[0154] Those skilled in the art can understand that Figure 12 The structure shown in the above embodiment is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the electronic device to which the scheme of the present application is applied. The specific electronic device can include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0155] In an example embodiment, an electronic device is provided, including a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps in the above method embodiments.

[0156] In an embodiment, a computer readable storage medium is provided, storing a computer program, and the computer program is executed by a processor to implement the steps in the above method embodiments.

[0157] In an embodiment, a computer program product is provided, including a computer program, and the computer program is executed by a processor to implement the steps in the above method embodiments.

[0158] It should be noted that the user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties, and the collection, use and processing of related data need to comply with relevant regulations.

[0159] It can be understood by those skilled in the art that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing related hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium. When the computer program is executed, it can include the processes of the above-mentioned embodiments of each method. In the embodiments provided in the present application, any reference to memory, database or other medium can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical storage, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration but not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The database involved in the embodiments provided in the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a block chain, etc., without being limited thereto. The processor involved in the embodiments provided in the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., without being limited thereto.

[0160] Any technical features in the above embodiments can be combined, and for the sake of brevity, not all possible combinations are described above, however, any combination of these technical features is deemed to be within the scope of the present application.

[0161] The above embodiments only express several implementation manners of the present application, and the description is relatively specific and detailed, but it should not be understood as a limitation on the patent scope of the present application. It should be pointed out that, for ordinary skilled persons in the art, several modifications and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. A method for detecting partial discharge of gas-insulated equipment, characterized in that: The method comprises: Obtaining a pulse current PRPD spectrum based on pulse current signal samples collected under different excitation voltages, and obtaining an ultra-high frequency PRPD spectrum based on ultra-high frequency signal samples collected under different excitation voltages; Performing feature extraction on the pulse current PRPD spectrum and the ultra-high frequency PRPD spectrum to obtain a characteristic signal matrix under different excitation voltages; Using the UMAP algorithm to perform dimensionality reduction processing on the characteristic signal matrix under different excitation voltages to generate corresponding low-dimensional feature representations; Normalizing and visualizing the low-dimensional feature representation to obtain a discharge pattern distribution model under different excitation voltages; The discharge mode of the partial discharge signal of the gas-insulated equipment is detected based on the discharge mode distribution model.

2. The method according to claim 1, characterized in that The feature extraction of the pulse current PRPD spectrum and the ultra-high frequency PRPD spectrum to obtain a feature signal matrix under different excitation voltages includes: Performing a first feature extraction on the pulse current PRPD spectrum to obtain a first feature signal sequence of the apparent discharge magnitude under different excitation voltages; Performing a second feature extraction on the UHF PRPD spectrum to obtain a second feature signal sequence representing the spectrum morphology under different excitation voltages; A first characteristic signal in the first characteristic signal sequence and a second characteristic signal in the second characteristic signal sequence under the same excitation voltage are acquired to generate a characteristic signal matrix under the corresponding excitation voltage.

3. The method according to claim 2, characterized in that The first feature extraction includes extracting the maximum partial discharge amplitude, the total partial discharge amount, the average value of positive partial discharge, the total value of positive partial discharge, the average value of negative partial discharge, and the total value of negative partial discharge based on the pulse current PRPD spectrum; The second feature extraction includes extracting the total number of ultra-high frequency pulses, standard deviation, skewness and kurtosis based on the ultra-high frequency PRPD spectrum.

4. The method according to claim 1, wherein The UMAP algorithm is used to perform dimensionality reduction processing on the characteristic signal matrix under different excitation voltages to generate corresponding low-dimensional feature representations, including: Determining the neighbor relationship of the characteristic signal matrix under each excitation voltage based on the similarity between the characteristic signal matrices under different excitation voltages; Mapping the characteristic signal matrix under different excitation voltages to a low-dimensional space to obtain corresponding initial low-dimensional features; Obtaining the cross entropy loss between the neighbor relationship of the feature signal matrix and the corresponding initial low-dimensional features; The initial low-dimensional features are iteratively updated based on the cross-entropy loss until the cross-entropy loss converges or reaches a maximum number of iterations, and the updated initial low-dimensional features are determined as the low-dimensional feature representation corresponding to the feature signal matrix.

5. The method according to claim 1, wherein Normalizing and visualizing the low-dimensional feature representation to obtain a discharge pattern distribution model based on different excitation voltages includes: Normalizing the low-dimensional feature representation to obtain a normalized low-dimensional feature representation; Mapping the normalized low-dimensional feature representation to a three-dimensional space to obtain distribution results of the low-dimensional feature representation in the three-dimensional space under different excitation voltages, wherein the distribution results include the aggregation states of different clusters, and different clusters represent different discharge modes; A corresponding discharge mode distribution model is generated based on the distribution result in the three-dimensional space.

6. The method according to any one of claims 1 to 5, characterized in that The detecting the discharge mode of the partial discharge signal of the gas-insulated equipment based on the discharge mode distribution model includes: Based on the pulse current method and the ultra-high frequency method, a characteristic signal of the partial discharge signal of the gas-insulated equipment is collected; Performing dimensionality reduction processing on the characteristic signal using the UMAP algorithm to generate a low-dimensional feature representation of the partial discharge signal of the gas-insulated equipment; Mapping the low-dimensional feature representation to the discharge pattern distribution model, and determining a target cluster in which the low-dimensional feature representation is located in the discharge pattern distribution model; A discharge mode corresponding to the target cluster is acquired, and the discharge mode corresponding to the target cluster is determined as the discharge mode of the partial discharge signal of the gas-insulated equipment.

7. A gas-insulated equipment partial discharge detection device, characterized in that: The device comprises: A spectrum acquisition module, used to acquire a pulse current PRPD spectrum based on pulse current signal samples collected under different excitation voltages, and to acquire an ultra-high frequency PRPD spectrum based on ultra-high frequency signal samples collected under different excitation voltages; A feature extraction module is used to extract features from the pulse current PRPD spectrum and the ultra-high frequency PRPD spectrum to obtain a feature signal matrix under different excitation voltages; A dimensionality reduction processing module is used to perform dimensionality reduction processing on the characteristic signal matrix under different excitation voltages using the UMAP algorithm to generate corresponding low-dimensional feature representations; A model acquisition module, configured to normalize and visualize the low-dimensional feature representation to obtain a discharge pattern distribution model based on different excitation voltages; A discharge detection module is configured to detect a discharge mode of a partial discharge signal of the gas-insulated equipment based on the discharge mode distribution model.

8. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

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

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

Images