Transformer vibration feature extraction method, device and equipment and storage medium

Abstract

The invention discloses a transformer vibration feature extraction method, device and equipment and a storage medium, and relates to the technical field of transformer vibration feature selection, and the transformer vibration feature extraction method comprises the steps: extracting m types of feature vectors from n paths of transformer vibration signals in the time domain, frequency domain and time-frequency domain directions; normalizing the n * m feature vectors to obtain a fusion feature matrix; determining a fusion feature factor of each feature vector in the fusion feature matrix; and screening the vibration characteristics of the transformer according to the fusion characteristic factor of each characteristic vector in the fusion characteristic matrix. According to the method, by calculating fusion feature factors of multi-type and multi-channel vibration feature quantities, the relevancy and redundancy of the features are synthesized, and selection of the vibration features of the transformer is effectively achieved.

Description

Technical Field

[0001] This application relates to the technical field of transformer vibration feature selection, and particularly to a method, device, equipment and storage medium for extracting transformer vibration features. Background Art

[0002] The vibration feature quantity data for on-line monitoring of transformers is huge, and there is a large amount of redundant and irrelevant data among the feature quantities. If the feature quantities are not selectively distinguished, it will ultimately affect the accuracy and efficiency of on-line monitoring of transformers. The existing vibration feature quantity dimensionality reduction methods only consider the correlation between features and the original data, while ignoring the redundancy between the features to be selected and the selected features. Therefore, the optimal selection of features cannot be achieved. Summary of the Invention

[0003] Based on this, in view of the above problems, it is necessary to propose a method, device, equipment and storage medium for extracting transformer vibration features.

[0004] This application provides a method for extracting transformer vibration features, and the method includes: Extracting m types of feature vectors from n channels of transformer vibration signals in the time domain, frequency domain, and time-frequency domain directions; Normalizing the n×m feature vectors to obtain a fusion feature matrix; Determining the fusion feature factors of each feature vector in the fusion feature matrix; Selecting the transformer vibration features according to the fusion feature factors of each feature vector in the fusion feature matrix.

[0005] Among them, the determining the fusion feature factors of each feature vector in the fusion feature matrix includes: Randomly and evenly dividing each feature vector in the fusion feature matrix into a positive sample set and a negative sample set; Obtaining the neighboring samples of the feature vector from the positive sample set and the negative sample set respectively according to the distances between the feature vectors; Determining the fusion feature factors of each feature vector in the fusion feature matrix according to the neighboring samples of each feature vector.

[0006] Among them, the neighboring samples of the feature vector include the same-class samples and different-class samples of the feature vector; the determining the fusion feature factors of each feature vector in the fusion feature matrix according to the neighboring samples of each feature vector specifically includes: According to Determining the fusion feature factors of each feature vector in the fusion feature matrix, where is the i-th feature vector in the fusion feature matrix E, is a similar sample of the i-th eigenvector in the fused feature matrix E, is a dissimilar sample of the i-th eigenvector in the fused feature matrix E.

[0007] Among them, the obtaining of the neighboring samples of the eigenvector from the positive sample set and the negative sample set respectively according to the distances between the eigenvectors includes: Obtaining the eigenvectors with the maximum distance from the eigenvector from the positive sample set and the negative sample set respectively to obtain neighboring samples, Among them, the neighboring samples of the eigenvector in the same sample set are similar samples, and the neighboring samples of the eigenvector in different sample sets are dissimilar samples.

[0008] Among them, the screening of the transformer vibration characteristics according to the fusion feature factors of the eigenvectors in the fused feature matrix includes: In the fused feature matrix, screening the eigenvectors with fusion feature factors greater than the threshold as the transformer vibration characteristics.

[0009] Among them, the threshold is 2.5.

[0010] Among them, it further includes: Using the extracted transformer vibration characteristics to identify the usage status of the transformer.

[0011] This application also provides a transformer vibration characteristic extraction device, and the device includes: An eigenvector extraction module, configured to extract m types of eigenvectors from n-channel transformer vibration signals in the time domain, frequency domain, and time-frequency domain directions; A feature matrix determination module, configured to normalize the n×m eigenvectors to obtain a fused feature matrix; A fusion feature factor determination module, configured to determine the fusion feature factors of the eigenvectors in the fused feature matrix; A vibration characteristic screening module, configured to screen the transformer vibration characteristics according to the fusion feature factors of the eigenvectors in the fused feature matrix.

[0012] This application also provides a computer device, including a memory and a processor, where the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the following steps: Extracting m types of eigenvectors from n-channel transformer vibration signals in the time domain, frequency domain, and time-frequency domain directions; Normalizing the n×m eigenvectors to obtain a fused feature matrix; Determining the fusion feature factors of the eigenvectors in the fused feature matrix; Screen the transformer vibration characteristics according to the fusion feature factors of each eigenvector in the fusion feature matrix.

[0013] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the following steps: Extract m types of eigenvectors from n-channel transformer vibration signals in the time domain, frequency domain, and time-frequency domain directions; Normalize the n×m eigenvectors to obtain a fusion feature matrix; Determine the fusion feature factors of each eigenvector in the fusion feature matrix; Screen the transformer vibration characteristics according to the fusion feature factors of each eigenvector in the fusion feature matrix.

[0014] The transformer vibration characteristic extraction method provided by this application has the following beneficial effects: In the transformer vibration characteristic extraction method provided by this application, m types of eigenvectors are extracted from n-channel transformer vibration signals in the time domain, frequency domain, and time-frequency domain directions, and the n×m eigenvectors are normalized to obtain a fusion feature matrix; determine the fusion feature factors of each eigenvector in the fusion feature matrix; the fusion feature factors of various types and multi-channel vibration eigenvectors are obtained, and the transformer vibration characteristics are screened according to the fusion feature factors of each eigenvector in the fusion feature matrix, which can effectively remove redundant data in the vibration characteristic quantity and retain key information, thereby improving the accuracy of transformer condition monitoring. The optimized characteristic data is more representative and reliable, and can provide a more accurate basis for transformer condition assessment and fault diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of this application, and those of ordinary skill in the art can obtain other drawings based on these drawings without creative efforts.

[0016] Among them: Figure 1 is a schematic flowchart of a transformer vibration characteristic extraction method in an embodiment; Figure 2 is a structural diagram of a transformer vibration characteristic extraction device in an embodiment; Figure 3 is a schematic structural diagram of a computer device in an embodiment; Figure 4 is a schematic structural diagram of a computer-readable storage medium in an embodiment. Specific embodiments

[0017] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present application without creative efforts belong to the scope of protection of the present application.

[0018] In an embodiment of the present application, a method for extracting transformer vibration characteristics is provided. Please refer to Figure 1 , Figure 1 which is a schematic flowchart of the method for extracting transformer vibration characteristics in an embodiment; the method for extracting transformer vibration characteristics includes steps S1 to S4.

[0019] Step S1, extracting m types of feature vectors from n-channel transformer vibration signals in the time domain, frequency domain, and time-frequency domain directions; Specifically, the time-domain feature vectors include mean, variance, standard deviation, peak value, waveform index, margin index, etc., the frequency-domain feature vectors include amplitude spectrum, power spectrum, frequency, frequency distribution, frequency moment, etc., and the time-frequency domain feature vectors include wavelet transform coefficients, short-time Fourier transform coefficients, Hilbert-Huang transform coefficients, etc.

[0020] Step S2, normalizing the n×m feature vectors to obtain a fused feature matrix; Specifically, m types of feature vectors are extracted from n-channel transformer vibration signals in the time domain, frequency domain, and time-frequency domain directions, and the n×m feature vectors are normalized to obtain a fused feature matrix E = [E1, E2,..., E m ; Step S3, determining the fusion feature factors of each feature vector in the fused feature matrix; In some embodiments, the determining the fusion feature factors of each feature vector in the fused feature matrix includes: Randomly and evenly dividing each feature vector in the fused feature matrix into a positive sample set and a negative sample set; According to the distances between the feature vectors, obtaining the neighboring samples of the feature vectors from the positive sample set and the negative sample set respectively; Determining the fusion feature factors of each feature vector in the fused feature matrix according to the neighboring samples of each feature vector.

[0021] In some embodiments, the neighboring samples of the feature vector include the same-class samples and different-class samples of the feature vector; determining the fusion feature factors of each feature vector in the fusion feature matrix according to the neighboring samples of each feature vector specifically includes: According to determine the fusion feature factors of each feature vector in the fusion feature matrix, where is the i-th feature vector in the fusion feature matrix E, is the same-class sample of the i-th feature vector in the fusion feature matrix E, is the different-class sample of the i-th feature vector in the fusion feature matrix E.

[0022] Specifically, the larger the value of the fusion feature factor F, the greater the contribution of the feature vector E i to sample classification.

[0023] In some embodiments, obtaining the neighboring samples of the feature vector from the positive sample set and the negative sample set respectively according to the distances between the feature vectors includes: Obtain the feature vectors with the maximum distance from the feature vector from the positive sample set and the negative sample set respectively to obtain neighboring samples, where the neighboring samples in the same sample set as the feature vector are the same-class samples, and the neighboring samples in different sample sets from the feature vector are the different-class samples.

[0024] Step S4, screen the transformer vibration features according to the fusion feature factors of each feature vector in the fusion feature matrix.

[0025] In some embodiments, screening the transformer vibration features according to the fusion feature factors of each feature vector in the fusion feature matrix includes: In the fusion feature matrix, screen the feature vectors with fusion feature factors greater than the threshold as the transformer vibration features.

[0026] In some embodiments, the threshold is 2.5.

[0027] In some embodiments, the threshold 2.5 of the fusion feature factor F can be adjusted according to the actual application scenario.

[0028] In some embodiments, the transformer vibration feature extraction method further includes: using the extracted transformer vibration features to identify the usage status of the transformer.

[0029] Specifically, the transformer vibration features can be used for transformer fault diagnosis, identifying the types of internal faults of the transformer, or can be used for on-line vibration monitoring of the transformer to improve the accuracy and sensitivity of monitoring.

[0030] In some embodiments, the transformer vibration feature extraction method includes: Extract m types of feature vectors from n-channel transformer vibration signals in the time domain, frequency domain, and time-frequency domain directions. After normalizing the feature vectors of the multi-channel signals of n×m, a fused feature matrix E = [E1, E2,..., E m ; Calculate the fused feature factor F of each vector in the fused feature matrix through the Relief algorithm. The basic steps are as follows: a. Divide the fused feature matrix E into a positive sample set E+ and a negative sample set E-. The feature vectors in E+ and E- are randomly selected and have equal numbers.

[0031] b. Starting from the feature vector E1, calculate the distance between E1 and each feature vector, and select the feature vectors with the maximum distance in E+ and E-, which are called neighboring samples Z+ and Z-. Z+ is a homogeneous sample, and Z- is a heterogeneous sample. If E1 ∈ E+, then hit = Z1+, miss = Z1-; if E1 ∈ E-, then hit = Z1-, miss = Z1+. Based on this, select the neighboring samples of each feature vector.

[0032] c. Calculate the fused feature factor F. The calculation formula is , where diff ( E i , hit i ) represents the i th feature vector and the diff value between it and the hit obtained in step b corresponding to it; Select the sensitive features with the fused feature factor greater than 2.5, that is, the transformer vibration features, to achieve the optimization of vibration features.

[0033] It should be noted that E1, E2,..., E m are all n*1 vectors. Calculating the distance of each feature vector is to calculate the Euclidean distance between them; the calculated fused feature factor should be F i , and each time F i is calculated, hit i and miss i should be reselected according to step b; in actual use, when F ≥ 2.5, the number or sensitivity of the selected vibration features is better. In some embodiments, a better range can also be selected through a machine learning algorithm in the follow-up, or a custom threshold can be set according to the actual situation.

[0034] Adopting the technical solution of this embodiment, through the fusion feature factor optimization method, redundant data in the vibration feature quantities can be effectively removed, key information can be retained, thereby improving the accuracy of transformer condition monitoring. The optimized sensitive features can better reflect the true state of the transformer, thus improving the sensitivity of the monitoring system, detecting potential faults in a timely manner. By reducing the feature dimension, the data storage and processing costs can be reduced, the monitoring system can be simplified, and the efficiency can be improved. The optimized feature data is more representative and reliable, and can provide a more accurate basis for transformer condition assessment and fault diagnosis. All in all, the technical solution of this embodiment effectively solves the problems of large vibration feature dimension and serious data redundancy of the transformer through the fusion feature factor optimization method, improves the accuracy and sensitivity of on-line vibration monitoring of the transformer, and has significant technical and economic benefits.

[0035] In an embodiment of the present application, a transformer vibration feature extraction device is provided. Please refer to Figure 2 , Figure 2 which is a structural diagram of the transformer vibration feature extraction device in an embodiment. The transformer vibration feature extraction device includes: a feature vector extraction module 201, a feature matrix determination module 202, a fusion feature factor determination module 203, and a vibration feature screening module 204.

[0036] Among them, the feature vector extraction module 201 is configured to extract m types of feature vectors from n-channel transformer vibration signals in the time domain, frequency domain, and time-frequency domain directions; The feature matrix determination module 202 is configured to normalize the n×m feature vectors to obtain a fusion feature matrix; The fusion feature factor determination module 203 is configured to determine the fusion feature factors of each feature vector in the fusion feature matrix; The vibration feature screening module 204 is configured to screen the transformer vibration features according to the fusion feature factors of each feature vector in the fusion feature matrix.

[0037] In some embodiments, the fusion feature factor determination module 203 is further configured to: randomly and evenly divide each feature vector in the fusion feature matrix into a positive sample set and a negative sample set; Obtain the neighboring samples of the feature vector from the positive sample set and the negative sample set respectively according to the distance between each feature vector; Determine the fusion feature factors of each feature vector in the fusion feature matrix according to the neighboring samples of each feature vector.

[0038] In some embodiments, the fusion feature factor determination module 203 is further configured to: according to Determine the fusion feature factors of each eigenvector in the fusion feature matrix, where is the i-th eigenvector in the fusion feature matrix E, is the same-class sample of the i-th eigenvector in the fusion feature matrix E, is the different-class sample of the i-th eigenvector in the fusion feature matrix E.

[0039] In some embodiments, the fusion feature factor determination module 203 is further configured to: respectively obtain the eigenvectors with the maximum distance from the positive sample set and the negative sample set to the eigenvector, so as to obtain adjacent samples, where the adjacent sample of the eigenvector in the same sample set is the same-class sample, and the adjacent sample of the eigenvector in different sample sets is the different-class sample.

[0040] In some embodiments, the vibration feature screening module 204 is further configured to: in the fusion feature matrix, screen the eigenvectors whose fusion feature factors are greater than the threshold as the transformer vibration features.

[0041] In some embodiments, the vibration feature screening module 204 is further configured to: determine the threshold to be 2.5.

[0042] In some embodiments, the transformer vibration feature extraction device further includes an application module 205, and the application module 205 is configured to: Use the extracted transformer vibration features to identify the usage status of the transformer.

[0043] For other details of each module of the transformer vibration feature extraction device to implement the above technical solutions, reference can be made to the descriptions in the above-provided transformer vibration feature extraction method, which will not be elaborated here.

[0044] In the embodiments of the present application, a computer device is provided. Please refer to Figure 3 , Figure 3 which is a schematic structural diagram of a computer device in an embodiment. The device includes a memory 301 and a processor 302. The memory 301 stores a computer program. When the computer program is executed by the processor 302, the processor 302 is caused to execute the following steps: Extract m types of eigenvectors from n-channel transformer vibration signals in the time domain, frequency domain, and time-frequency domain directions; Normalize the n×m eigenvectors to obtain a fusion feature matrix; Determine the fusion feature factors of each eigenvector in the fusion feature matrix; According to the fusion feature factors of each eigenvector in the fusion feature matrix, screen the transformer vibration features.

[0045] Among them, the processor 302 can also be referred to as a CPU (Central Processing Unit). The processor 302 may be an integrated circuit chip with signal processing capabilities. The processor 302 can also be a general-purpose processor, a DSP (Digital Signal Process), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gata Array), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. The general-purpose processor can be a microprocessor or the processor 302 can also be any conventional processor, etc.

[0046] In an embodiment of the present application, a computer-readable storage medium is provided. Please refer to Figure 4 , Figure 4 which is a schematic structural diagram of a computer-readable storage medium in an embodiment. A readable computer program 401 is stored on the storage medium. Among them, the computer program 401 can be stored on the above storage medium in the form of a software product, including several instructions for causing a computer device (which can be a personal computer, a service machine, or a network device, etc.) or a processor to perform the following steps: Extract m types of feature vectors from the vibration signals of n transformers in the time domain, frequency domain, and time-frequency domain directions; Normalize the feature vectors of n×m to obtain a fusion feature matrix; Determine the fusion feature factors of each feature vector in the fusion feature matrix; Screen the transformer vibration characteristics according to the fusion feature factors of each feature vector in the fusion feature matrix.

[0047] The foregoing storage medium includes: various media that can store program codes such as USB flash drives, mobile hard disks, magnetic disks or optical discs, ROM (Read-Only Memory), RAM (Random Access Memory), etc., or terminal devices such as computers, service machines, mobile phones, and tablets.

[0048] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the embodiments provided in the present application can include non-volatile and / or volatile memories. Non-volatile memories can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memories can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM), etc.

[0049] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0050] The above-described embodiments merely represent several implementation manners of the present application. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, 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 extracting vibration characteristics of a transformer, characterized in that The method includes: extracting m types of feature vectors of n-channel transformer vibration signals in the time domain, frequency domain, and time-frequency domain; normalizing the n×m feature vectors to obtain a fusion feature matrix; determining the fusion feature factors of each feature vector in the fusion feature matrix; screening the transformer vibration features according to the fusion feature factors of each feature vector in the fusion feature matrix.

2. The transformer vibration feature extraction method according to claim 1, wherein The determining the fusion feature factors of each feature vector in the fusion feature matrix includes: randomly and evenly dividing each feature vector in the fusion feature matrix into a positive sample set and a negative sample set; obtaining the neighboring samples of the feature vector from the positive sample set and the negative sample set respectively according to the distances between the feature vectors; determining the fusion feature factors of each feature vector in the fusion feature matrix according to the neighboring samples of each feature vector.

3. The transformer vibration feature extraction method according to claim 2, wherein, The neighboring samples of the feature vector include the same-class samples and different-class samples of the feature vector; the determining the fusion feature factors of each feature vector in the fusion feature matrix according to the neighboring samples of each feature vector specifically includes: According to determine the fusion feature factors of each eigenvector in the fusion feature matrix, where is the i-th eigenvector in the fusion feature matrix E, is the same-class sample of the i-th eigenvector in the fusion feature matrix E, is the different-class sample of the i-th eigenvector in the fusion feature matrix E.

4. The transformer vibration feature extraction method according to claim 2, wherein The obtaining the neighboring samples of the feature vector from the positive sample set and the negative sample set respectively according to the distances between the feature vectors includes: respectively obtaining the feature vectors with the maximum distance from the feature vector from the positive sample set and the negative sample set to obtain the neighboring samples, wherein, the neighboring samples in the same sample set as the feature vector are the same-class samples, and the neighboring samples in different sample sets from the feature vector are the different-class samples.

5. The transformer vibration feature extraction method according to claim 1, wherein The screening the transformer vibration features according to the fusion feature factors of each feature vector in the fusion feature matrix includes: screening the feature vectors with fusion feature factors greater than the threshold in the fusion feature matrix as the transformer vibration features.

6. The method for extracting transformer vibration characteristics according to claim 5, wherein The threshold is 2.

5.

7. The transformer vibration feature extraction method according to claim 6, wherein It further includes: identifying the usage status of the transformer by using the extracted transformer vibration features.

8. A transformer vibration feature extraction device, characterized in that, The device includes: a feature vector extraction module for extracting m types of feature vectors of n-channel transformer vibration signals in the time domain, frequency domain, and time-frequency domain; a feature matrix determination module for normalizing the n×m feature vectors to obtain a fusion feature matrix; a fusion feature factor determination module for determining the fusion feature factors of each feature vector in the fusion feature matrix; a vibration feature screening module for screening the transformer vibration features according to the fusion feature factors of each feature vector in the fusion feature matrix.

9. A computer device, characterized in that, It includes a memory and a processor, and the memory stores a computer program. When the computer program is executed by the processor, the processor executes the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, Stores a computer program. When the computer program is executed by the processor, the processor executes the steps of the method according to any one of claims 1 to 7.

Images