Fault identification method and device for transformer on-load tap-changer, computer equipment, readable storage medium and program product
Through the combination of sound and vibration combined features and support vector machine learners, the problem of low diagnostic accuracy of on-load tap-changer signal in the prior art is solved, and high accuracy recognition and real-time performance improvement of transformer fault status are achieved.
Patent Information
- Application Number
- CN202411015030.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-26
- Publication Date
- 2024-11-01
AI Technical Summary
The prior art cannot effectively distinguish on-load tap-off signals caused by different mechanisms, and the accuracy of fault status diagnosis is low, so it is impossible to fully characterize the status information in non-stationary and strong time-varying vibration signals during switch switching.
By obtaining the sound signal and vibration signal data of the transformer on-load tap-off switch in different fault states, data cleaning and feature extraction are carried out, acoustic and vibrating joint features are constructed, and iterative training is performed using the target algorithm framework of the support vector machine as the basis learner. Generate a strong classifier model for fault identification.
It significantly improves the accuracy of fault identification, reduces data processing volume and improves real-time performance, and can more comprehensively and accurately describe the nonlinear characteristics of vibration signals, achieving accurate perception of the fault status of the on-load tap-off switch of the transformer.
Smart Images

Figure CN118885902A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of power equipment, and in particular to a method, device, computer equipment, computer-readable storage medium and computer program product for identifying a fault of an on-load tap changer of a transformer. Background Art
[0002] On-load tap changers are important components for voltage regulation in power systems. Reasonable regulation of on-load tap changers can effectively reduce system voltage fluctuations and improve the flexibility of power grid dispatching. Research and analysis show that unreasonable operation and failure of on-load tap changers are important inducing factors for voltage instability and other types of accidents in power systems.
[0003] In traditional technology, the monitoring methods for on-load tap changers are mainly concentrated on one of the vibration and acoustic methods, including envelope analysis, wavelet analysis and chaos characteristic analysis. However, since there are many types of faults in on-load tap changers, and different degrees of the same mechanical fault will cause changes in the signal morphology, most of the existing processing methods make judgments by comparing the differences in the signal time domain waveform, frequency domain and energy spectrum, but cannot fully characterize the state information contained in such non-stationary and highly time-varying vibration signals during the switching process. Therefore, the current technology cannot effectively distinguish the on-load tap changer signals caused by different mechanisms, and there is a problem of low accuracy in fault state diagnosis. Summary of the invention
[0004] Based on this, it is necessary to provide a transformer on-load tap changer fault identification method, device, computer equipment, computer readable storage medium and computer program product to address the above technical problems.
[0005] In a first aspect, the present application provides a fault identification method for an on-load tap changer of a transformer, comprising:
[0006] Obtain the sound signal data and vibration signal data of the transformer on-load tap changer under different fault conditions, and perform data cleaning;
[0007] Converting the sound signal data from a time domain signal into a time-frequency image, and constructing a voiceprint feature of the sound signal data based on the texture feature of the time-frequency image;
[0008] Performing multivariate variational modal decomposition on the vibration signal data to extract modal components of the vibration signal data, and using entropy to describe nonlinear characteristics of the vibration signal data based on the modal components;
[0009] Constructing a sound-vibration joint feature according to the voiceprint feature of the sound signal data and the nonlinear feature of the vibration signal data, and selecting a training sample set and a test sample set from the sound-vibration joint feature;
[0010] The training sample set is input into a target algorithm framework with a support vector machine as the base learner for iterative training, and the test sample set is used for testing and adjustment to obtain a strong classifier model including multiple support vector machines; the strong classifier model is used to identify faults of the transformer on-load tap changer.
[0011] In one embodiment, converting the sound signal data from a time domain signal into a time-frequency image, and constructing a voiceprint feature of the sound signal data based on a texture feature of the time-frequency image, comprises:
[0012] Performing a generalized S transform on the sound signal data to obtain the time-frequency image; analyzing the time-frequency image to obtain the box dimension, directionality and contrast of the time-frequency image as the texture feature of the time-frequency image; constructing the voiceprint feature based on the box dimension, the directionality and the contrast.
[0013] In one embodiment, performing multivariate variational modal decomposition on the vibration signal data to extract modal components of the vibration signal data, and using entropy to describe nonlinear characteristics of the vibration signal data based on the modal components, includes:
[0014] The vibration signal data is subjected to multivariate variational modal decomposition to obtain a multivariate signal in matrix form, and a preset number of modal components are extracted from the multivariate signal; based on the modal components, the spectral entropy, energy entropy and permutation entropy of the vibration signal data are respectively calculated as the nonlinear characteristics of the vibration signal data.
[0015] In one embodiment, the step of inputting the training sample set into a target algorithm framework using a support vector machine as a base learner for iterative training includes:
[0016] In the target algorithm framework, the sample weights of the support vector machine are initialized; the support vector machine is iteratively trained according to the training sample set, and the sample weights of the support vector machine are updated according to the training results; the support vector machines are weighted combined according to the updated sample weights to obtain a preliminarily trained classifier model.
[0017] In one embodiment, the testing and adjusting using the test sample set includes:
[0018] The test sample set is input into the preliminarily trained classifier model for testing to obtain a fault classification result; if the classification accuracy of the fault classification result does not reach a preset value, the preliminarily trained classifier model is adjusted according to the test sample set until the classification accuracy reaches the preset value.
[0019] In one embodiment, the data cleaning includes:
[0020] Identify duplicate data in the sound signal data and the vibration signal data, clear the duplicate data, and filter the sound signal data and the vibration signal data through a low-pass filter to remove noise data, so as to obtain the sound signal data and the vibration signal data after data cleaning.
[0021] In a second aspect, the present application further provides a fault identification device for an on-load tap changer of a transformer, comprising:
[0022] A data acquisition module is used to acquire the sound signal data and vibration signal data of the transformer on-load tap changer under different fault conditions and perform data cleaning;
[0023] A voiceprint construction module, used to convert the sound signal data from a time domain signal into a time-frequency image, and construct a voiceprint feature of the sound signal data based on the texture features of the time-frequency image;
[0024] A feature description module, used for performing multivariate variational modal decomposition on the vibration signal data to extract modal components of the vibration signal data, and using entropy to describe nonlinear features of the vibration signal data based on the modal components;
[0025] A feature combination module, configured to construct a sound-vibration combination feature according to the voiceprint feature of the sound signal data and the nonlinear feature of the vibration signal data, and select a training sample set and a test sample set from the sound-vibration combination feature;
[0026] A model training module is used to input the training sample set into a target algorithm framework with a support vector machine as the base learner for iterative training, and use the test sample set for testing and adjustment to obtain a strong classifier model including multiple support vector machines; the strong classifier model is used to identify faults of the transformer on-load tap changer.
[0027] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:
[0028] Acquire the sound signal data and vibration signal data of the transformer on-load tap changer under different fault states, and perform data cleaning; convert the sound signal data from a time domain signal into a time-frequency image, and construct the soundprint feature of the sound signal data based on the texture feature of the time-frequency image; perform multivariate variational modal decomposition on the vibration signal data to extract the modal components of the vibration signal data, and use entropy to describe the nonlinear characteristics of the vibration signal data based on the modal components; construct an acoustic-vibration joint feature according to the soundprint feature of the sound signal data and the nonlinear feature of the vibration signal data, and select a training sample set and a test sample set from the acoustic-vibration joint feature; input the training sample set into a target algorithm framework with a support vector machine as the base learner for iterative training, and use the test sample set for testing and adjustment to obtain a strong classifier model including multiple support vector machines; the strong classifier model is used to identify faults of the transformer on-load tap changer.
[0029] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:
[0030] Acquire the sound signal data and vibration signal data of the transformer on-load tap changer under different fault states, and perform data cleaning; convert the sound signal data from a time domain signal into a time-frequency image, and construct the soundprint feature of the sound signal data based on the texture feature of the time-frequency image; perform multivariate variational modal decomposition on the vibration signal data to extract the modal components of the vibration signal data, and use entropy to describe the nonlinear characteristics of the vibration signal data based on the modal components; construct an acoustic-vibration joint feature according to the soundprint feature of the sound signal data and the nonlinear feature of the vibration signal data, and select a training sample set and a test sample set from the acoustic-vibration joint feature; input the training sample set into a target algorithm framework with a support vector machine as the base learner for iterative training, and use the test sample set for testing and adjustment to obtain a strong classifier model including multiple support vector machines; the strong classifier model is used to identify faults of the transformer on-load tap changer.
[0031] In a fifth aspect, the present application further provides a computer program product, including a computer program, which implements the following steps when executed by a processor:
[0032] Acquire the sound signal data and vibration signal data of the transformer on-load tap changer under different fault states, and perform data cleaning; convert the sound signal data from a time domain signal into a time-frequency image, and construct the soundprint feature of the sound signal data based on the texture feature of the time-frequency image; perform multivariate variational modal decomposition on the vibration signal data to extract the modal components of the vibration signal data, and use entropy to describe the nonlinear characteristics of the vibration signal data based on the modal components; construct an acoustic-vibration joint feature according to the soundprint feature of the sound signal data and the nonlinear feature of the vibration signal data, and select a training sample set and a test sample set from the acoustic-vibration joint feature; input the training sample set into a target algorithm framework with a support vector machine as the base learner for iterative training, and use the test sample set for testing and adjustment to obtain a strong classifier model including multiple support vector machines; the strong classifier model is used to identify faults of the transformer on-load tap changer.
[0033] The above-mentioned fault identification method, device, computer equipment, computer-readable storage medium and computer program product of the transformer on-load tap changer collect sound signal data and vibration signal data of the transformer on-load tap changer under different fault states for joint feature analysis. Compared with analyzing only one of the features, the accuracy of fault identification is significantly improved; the data length of the sound-vibration joint feature is also shorter than when the two features are used together, which reduces the data processing amount and improves the data processing speed and the real-time performance of fault identification; by performing multivariate variational mode decomposition on the vibration signal data, the modal components of the vibration signal data can be effectively extracted, and the features can be extracted by entropy calculation, which can more comprehensively and accurately describe the nonlinear characteristics of the vibration signal data, and has good advantages for signal processing and fault diagnosis; finally, the training sample set is iteratively trained through the target algorithm framework with support vector machine as the base learner, and the test sample set is used for testing and adjustment, so as to effectively obtain a strong classifier model integrating multiple support vector machines, which can realize accurate perception of the fault state of the transformer on-load tap changer, thereby improving the accuracy of the fault state diagnosis of the on-load tap changer. BRIEF DESCRIPTION OF THE DRAWINGS
[0034] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related technologies, the drawings required for use in the embodiments of the present application or the related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.
[0035] FIG1 is an application environment diagram of a fault identification method for an on-load tap changer of a transformer in one embodiment;
[0036] Figure 2 1 is a flow chart of a method for identifying a fault of an on-load tap changer of a transformer in one embodiment;
[0037] Figure 3 A schematic diagram of a flow chart of steps for constructing voiceprint features in one embodiment;
[0038] Figure 4 A line graph showing the difference between adjacent correlation coefficients under different K values in one embodiment;
[0039] Figure 5 It is a schematic flow chart of a fault identification method of a transformer on-load tap changer in a specific embodiment;
[0040] Figure 6 is a structural block diagram of a fault identification device for an on-load tap changer of a transformer in one embodiment;
[0041] FIG. 7 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION
[0042] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.
[0043] The transformer on-load tap changer fault identification method provided in the embodiment of the present application can be applied in the application environment shown in FIG1. The terminal communicates with the server through a network. The data storage system can store data that the server needs to process. The data storage system can be integrated on the server, or placed on the cloud or other network servers.
[0044] Specifically, the transformer on-load tap changer fault identification method provided in the embodiment of the present application can be executed by a server.
[0045] Exemplarily, the server obtains the sound signal data and vibration signal data of the transformer on-load tap changer under different fault states, and performs data cleaning; the server converts the sound signal data from a time domain signal into a time-frequency image, and constructs the soundprint features of the sound signal data based on the texture features of the time-frequency image; the server performs multivariate variational modal decomposition on the vibration signal data to extract the modal components of the vibration signal data, and uses entropy to describe the nonlinear characteristics of the vibration signal data based on the modal components; the server constructs a sound-vibration joint feature based on the soundprint features of the sound signal data and the nonlinear characteristics of the vibration signal data, and selects a training sample set and a test sample set from the sound-vibration joint feature; the server inputs the training sample set into a target algorithm framework with a support vector machine as the base learner for iterative training, and uses the test sample set for testing and adjustment to obtain a strong classifier model containing multiple support vector machines; the strong classifier model is used to identify faults of the transformer on-load tap changer.
[0046] In the application environment shown in FIG1 , the terminal may be, but is not limited to, various personal computers, notebook computers, smart phones, and tablet computers. The server may be implemented as an independent server or a server cluster consisting of multiple servers.
[0047] In one embodiment, Figure 2 As shown, a fault identification method for a transformer on-load tap changer is provided, and the method is applied to the server in FIG1 as an example for description, and includes the following steps:
[0048] Step S201, obtaining sound signal data and vibration signal data of the transformer on-load tap changer in different fault states, and performing data cleaning.
[0049] Among them, different fault states include but are not limited to closing, opening, refusing to close and refusing to open.
[0050] Specifically, the server obtains the sound signal data and vibration signal data corresponding to the transformer on-load tap changer in the closed, open, refused to close and refused to open states, and performs data cleaning on the sound signal data and the vibration signal data.
[0051] Step S202: convert the sound signal data from the time domain signal into a time-frequency image, and construct a voiceprint feature of the sound signal data based on the texture feature of the time-frequency image.
[0052] Among them, texture features can be box dimension, orientation and contrast, etc.
[0053] Specifically, the server performs a generalized S transform on the sound signal to convert the one-dimensional time domain signal into a two-dimensional time-frequency image, and constructs the voiceprint features of the sound signal data based on the changes in the image texture in the time-frequency image.
[0054] Step S203 , performing multivariate variational modal decomposition on the vibration signal data to extract modal components of the vibration signal data, and using entropy to describe the nonlinear characteristics of the vibration signal data based on the modal components.
[0055] Among them, Multivariate Variational Mode Decomposition (MVMD) is a method used for signal processing and data decomposition. It decomposes multivariate signals into multiple modal components to reveal the intrinsic components and structures of the signals.
[0056] Specifically, the server performs multivariate variational modal decomposition on the vibration signal to extract modal components, calculates spectrum entropy, energy entropy, and permutation entropy respectively, and uses entropy to describe the nonlinear characteristics of the vibration signal data.
[0057] Step S204: construct an acoustic-vibration joint feature according to the voiceprint feature of the sound signal data and the nonlinear feature of the vibration signal data, and select a training sample set and a test sample set from the acoustic-vibration joint feature.
[0058] The acoustic vibration joint feature may be joint feature data including both voiceprint features and nonlinear features.
[0059] Specifically, the server constructs a sound-vibration joint feature according to the voiceprint features of the sound signal data and the nonlinear features of the vibration signal data, and randomly selects a certain number of training sample sets and test sample sets from the sound-vibration joint feature.
[0060] Step S205, input the training sample set into the target algorithm framework with the support vector machine as the base learner for iterative training, and use the test sample set for testing and adjustment to obtain a strong classifier model including multiple support vector machines; the strong classifier model is used to identify faults of the transformer on-load tap changer.
[0061] Among them, the base learner can be the basic unit or component for building the target algorithm framework. In ensemble learning, multiple base learners are combined to improve the prediction performance. Compared with a single model, it can better handle complex data and improve generalization ability.
[0062] The target algorithm may be an AdaBoost (Adaptive Boosting) algorithm, and the strong classifier may be an AdaBoost-SVM.
[0063] Specifically, the server inputs the training sample set into the AdaBoost-SVM classification model based on the Gray Wolf Optimization Algorithm, uses the AdaBoost algorithm for continuous iteration, integrates multiple SVM trainers, and uses the Gray Wolf Optimization Algorithm to find the parameters and coefficients of the SVM model to improve the overall classification effect.
[0064] In the above-mentioned fault identification method of the transformer on-load tap changer, by collecting the sound signal data and vibration signal data of the transformer on-load tap changer under different fault states for joint feature analysis, the accuracy of fault identification is significantly improved compared with analyzing only one of the features; the data length of the sound-vibration joint feature is also shorter than when the two features are used together, which reduces the data processing amount and improves the data processing speed and the real-time performance of fault identification; by performing multivariate variational modal decomposition on the vibration signal data, the modal components of the vibration signal data can be effectively extracted, and by extracting features through entropy calculation, the nonlinear characteristics of the vibration signal data can be more comprehensively and accurately described, which has good advantages for signal processing and fault diagnosis; finally, the training sample set is iteratively trained through the target algorithm framework with support vector machine as the base learner, and the test sample set is used for testing and adjustment, so as to effectively obtain a strong classifier model integrating multiple support vector machines, which can realize accurate perception of the fault state of the transformer on-load tap changer, thereby improving the accuracy of the fault state diagnosis of the on-load tap changer.
[0065] In one embodiment, if Figure 3 As shown, in the above step S202, the sound signal data is converted from the time domain signal into a time-frequency image, and the voiceprint feature of the sound signal data is constructed based on the texture feature of the time-frequency image, which specifically includes the following steps:
[0066] Step S301, performing a generalized S transform on the sound signal data to obtain a time-frequency image.
[0067] Step S302: Analyze the time-frequency image to obtain the box dimension, orientation and contrast of the time-frequency image as texture features of the time-frequency image.
[0068] Step S303: constructing voiceprint features based on box dimension, directionality and contrast.
[0069] Among them, the generalized S transform enhances the performance of the classic S transform by introducing different window functions or adjusting parameters to better adapt to the characteristics of different types of signals.
[0070] (1) Formula for generalized S transform of sound signal
[0071]
[0072] (2)Calculate the box dimension, directionality, and contrast
[0073] 2.1) Box dimension
[0074] The box dimension is particularly sensitive to the texture roughness of the image spatial distribution and can quantitatively characterize the distribution law of the time-frequency diagram. The calculation process is as follows: Divide the spectrogram of size M×M into sub-blocks of L×L (1 < L ≤ M / 2, L is an integer), and let r = L / M. Then each grid is a column of L×L×L 1 boxes, and L = L×G / M, where G is the gray level. Assume that in the grid, the maximum and minimum values of the image gray level fall into the kth (maximum value) and lth (minimum value) boxes. Then the number of boxes required for the grid image is:
[0075]
[0076] The number of boxes required to cover the entire image is :
[0077]
[0078] The box dimension can be calculated by the following formula:
[0079]
[0080] Use the least squares method to fit and calculate and the slope, and the absolute value obtained is the box dimension.
[0081] 2.2) Directionality
[0082] Directionality describes the global characteristics of the texture image and characterizes the trend of the basic texture unit and its arrangement in various directions. Calculate the gradient vector for each pixel, and its modulus and direction are defined as follows:
[0083]
[0084] In the formula: is the modulus of the gradient vector; and are respectively the change amounts of in the horizontal and vertical directions;
[0085] Draw a histogram and calculate the peak value of the gradient statistical histogram to obtain the directionality. The formula is as follows:
[0086]
[0087] In the formula: is the amplitude of the Pth peak, is the number of histogram peaks, represents the Dth peak of the histogram, is the normalization factor.
[0088] 2.3) Contrast
[0089] Contrast is used to reflect the grayscale difference of the image. The difference in grayscale value reflects the strength of contrast. If the grayscale peak tends to 0 or 255, the kurtosis is used. To characterize its deviation, the formula is calculated as follows:
[0090]
[0091] In the formula , It is the fourth-order moment of the grayscale mean of the entire image.
[0092] Specifically, the server performs a generalized S transform on the sound signal data to obtain a time-frequency image; analyzes the time-frequency image to obtain the box dimension, directionality, and contrast of the time-frequency image as texture features of the time-frequency image; and constructs a voiceprint feature based on the box dimension, directionality, and contrast.
[0093] In this embodiment, the box dimension, directionality and contrast of the time-frequency image are obtained by analyzing the time-frequency image, and the voiceprint feature is scientifically and accurately constructed based on the box dimension, directionality and contrast.
[0094] In one of the embodiments, the vibration signal is subjected to MVMD decomposition to extract modal components, entropy is used to describe the nonlinear characteristics of the vibration signal, and spectrum entropy, energy entropy, and permutation entropy are calculated respectively.
[0095] (1) Decomposition principle of vibration signal MVMD
[0096] Original signal Decompose into k modal functions , use Hilbert transform to get the corresponding marginal spectrum, and use the corresponding center frequency of each component The exponential term modulates it to the corresponding baseband.
[0097]
[0098]
[0099]
[0100] in, is the instantaneous amplitude, is the instantaneous phase, is the unit pulse function;
[0101] Constructing constrained variational models based on component bandwidth values estimated by L2 norm
[0102]
[0103] Introducing Lagrange multipliers With the penalty factor α, the constrained model is changed to an unconstrained model.
[0104]
[0105] The multiplication operator alternating direction method is used to alternately update each modal component and its center frequency , and finally solve , and The update expression is:
[0106]
[0107]
[0108]
[0109] By alternating updates , and The vibration signal Decomposed into k eigenmode components.
[0110] (2) When the following conditions are met, the variational mode decomposition stops iterating:
[0111]
[0112] Here, ε is a constant.
[0113] (3) Number of modal decompositions K
[0114] The number of modal decompositions K is a key parameter in the MVMD decomposition process, which affects the entire MVMD decomposition process and the final result. The size of the K value is usually determined by the center frequency observation method in research, that is, to judge the center frequency distance of adjacent modal components. The correlation coefficient can be used to judge. When the signal is over-decomposed, the adjacent modal components will have similar signals and similar frequencies. At this time, the correlation coefficient between the two components and the original signal is close. The steps are as follows:
[0115] 3.1) Initialize the K value, set K=2;
[0116] 3.2) Perform MVMD decomposition on the signal to obtain K IMF components;
[0117] 3.3) Calculate the correlation coefficient V between each IMF component and the original signal according to the following formula:
[0118]
[0119] In the formula, are the Kth IMF components obtained by MVMD decomposition.
[0120] (4) Calculate the difference between adjacent correlation coefficients
[0121] Calculate the difference between adjacent correlation coefficients and compare with 0.1. If they are all greater than 0.1, increase the K value until a value less than 0.1 appears. At this time, take K=K-1. Select the normal closing signal and calculate the correlation coefficient values between each component and the original signal when K ranges from 2 to 6, as shown in Table 1 below.
[0122] Table 1 Correlation coefficients between components and original signals at different K values
[0123]
[0124] Calculate the difference of adjacent correlation coefficients under different K values in Table 1 as follows Figure 4 As shown, it can be found that when K=6, the difference is less than 0.1, so the best modal component is 5.
[0125] (5) Calculation of spectrum entropy, energy entropy, and permutation entropy
[0126] 5.1) Calculation of power spectrum
[0127] The power spectrum describes the energy change of the signal from the perspective of frequency. The calculation steps are as follows: First, perform Fourier transform on each modal component, and then calculate the power spectrum of each modal component. , the power spectrum of each component calculated accounts for the weight of all components .
[0128]
[0129] Calculate the spectrum entropy value
[0130]
[0131] 5.2) Calculation of energy entropy
[0132] Energy entropy mainly describes the energy change of the signal from the perspective of time domain. It can highlight the concentration and dispersion of energy distribution. The more dispersed the energy distribution is, the more obvious the energy melting value characteristics are. Energy entropy is calculated by the proportion of each modal energy in all modal components. The energy of each component :
[0133]
[0134] The second step is to calculate the energy proportion of each sub-band :
[0135]
[0136] Finally calculate the energy entropy value :
[0137]
[0138] 5.3) Calculation of permutation entropy
[0139] Permutation entropy is an effective method to characterize the complexity and randomness of time series, and it is highly targeted at non-stationary shock vibration signals.
[0140] For signal sequences , reconstruct its phase space and obtain the following matrix:
[0141]
[0142] Where: , the delay time m=8 and the embedding dimension are determined by the mutual information method and GP algorithm Arrange the elements of each component in ascending order, and obtain different symbol sequences according to the reconstructed component index. There are m kinds of symbol sequences in the m-dimensional phase space mapping, and the probabilities of s different symbol sequences appearing are , the calculation formula is as follows
[0143]
[0144] Will Perform normalization, that is:
[0145]
[0146] The permutation entropy of the five modes of the vibration signal can be calculated.
[0147] For example, the specific steps of the intelligent sensing method of the transformer on-load tap changer are as follows:
[0148] Taking SVM (Support Vector Machine) as the base classifier, the AdaBoost algorithm is used to adjust the sample weights in real time according to the classification error of the base classifier SVM. For the samples with incorrect classification, their training weight is increased in the next iteration; conversely, the training weight of the samples with correct classification is reduced. Finally, the trained SVM base classifiers are weightedly combined into a strong classifier AdaBoost-SVM to enhance the accuracy and stability of the classifier.
[0149] Based on the weight formula of AdaBoost weak classifier, the correct situation of the current base classifier is taken into account to determine the optimal range of weight. Its mathematical expression is:
[0150]
[0151]
[0152]
[0153] In the formula, Reflects the correct sample weight distribution state of the classification result; m is a constant in [0, 1]. It is monotonically increasing in (0, 1). When the independent variable is in the range of 0-0.8, it is monotonically increasing and the first-order derivative changes relatively slowly. When the independent variable tends to 1, The weights of AdaBoost’s weak classifiers are calculated by To improve, because The value of increases monotonically with the accuracy, and its value is within (0, 1). The smaller the value, the worse the classification result of the current weak classifier. If it is close to 1, the accuracy of the current result is close to 100%. Therefore, it is necessary to increase the weight of the weak classifier and the weight of the correctly classified samples at this time.
[0154] Then, the constructed acoustic vibration feature vector is input into the AdaBoost-SVM classification model based on the Grey Wolf optimization algorithm, and the AdaBoost algorithm is used for continuous iteration to integrate multiple SVM trainers. The test samples are input into the trained model for testing to obtain the classification results, thus realizing the intelligent perception of the fault status of OLTC (On Load Tap Changer).
[0155] In some embodiments, in a laboratory environment, a piezoelectric acceleration sensor, a charge amplifier, an audio collector, and an oscilloscope are used to perform a simulation experiment on a composite SYJZZ type OLTC.
[0156] The sampling frequency during the experiment was set to 50Hz. The vibration signals and sound signals of OLTC in normal, closing, opening, closing refusal and opening refusal were collected, and MVMD decomposition and S transform were performed respectively. It was found that there were 100 data in the vibration signal feature and voiceprint feature sample sets, including 21 normal closing, 24 closing refusal, 28 normal opening and 27 opening refusal.
[0157] The analysis results are shown in the following table:
[0158] Table 2 Vibration characteristics collected under OLTC fault conditions in the embodiment
[0159]
[0160] Table 3 Voiceprint features collected under OLTC fault conditions in the embodiment
[0161]
[0162] The feature vectors constructed by combining sound and vibration are shown in Tables 2 and 3. The test samples are input into the strong classifier AdaBoost-SVM, and the fault detection success rate is found to be 95%.
[0163] The test results were compared and analyzed with traditional SVM, BP neural network, and RF algorithms. The parameter values were set according to engineering experience. The diagnostic results are as follows:
[0164] Table 4 OLTC fault status monitoring success rate
[0165]
[0166] According to the comparison results, the intelligent evaluation model proposed in the present invention has the best recognition accuracy among the four machine learning models when identifying the OLTC fault status.
[0167] After testing the examples, the following conclusions were drawn:
[0168] ① When the joint features of sound signal data and vibration signal data are used for analysis, the diagnostic accuracy is higher than that of traditional vibration signals;
[0169] ②Using MVMD to perform multivariate decomposition of OLTC fault signals and extracting features through entropy calculation can describe the characteristics of the signal more comprehensively and accurately, which has good advantages for signal processing and fault diagnosis;
[0170] ③ Using the AdaBoost-SVM classification model based on the Grey Wolf Optimization Algorithm, it is proposed to use AdaBoost to integrate multiple SVM trainer test sample data, and the diagnostic signal results have better efficiency and accuracy.
[0171] In summary, this application proposes an intelligent perception method using MVMD and acoustic vibration signals to improve the accuracy of signal decomposition and feature extraction, enhance the state assessment performance of the classification model, and thus improve the accuracy and reliability of OLTC fault diagnosis results.
[0172] In one embodiment, Figure 5 As shown, a fault identification method for a transformer on-load tap changer in a specific embodiment is provided, which specifically includes the following steps:
[0173] Step S501, obtaining sound signal data and vibration signal data of the transformer on-load tap changer in different fault states.
[0174] Step S502, identifying repeated data in the sound signal data and the vibration signal data, clearing the repeated data, and filtering the sound signal data and the vibration signal data through a low-pass filter to remove noise data, thereby obtaining sound signal data and vibration signal data after data cleaning.
[0175] Step S503, performing a generalized S transform on the sound signal data to obtain a time-frequency image; analyzing the time-frequency image to obtain the box dimension, directionality and contrast of the time-frequency image as the texture feature of the time-frequency image; constructing a voiceprint feature based on the box dimension, directionality and contrast.
[0176] Step S504, performing multivariate variational modal decomposition on the vibration signal data to obtain a multivariate signal in matrix form, and extracting a preset number of modal components from the multivariate signal; based on the modal components, respectively calculating the spectral entropy, energy entropy and permutation entropy of the vibration signal data as nonlinear characteristics of the vibration signal data.
[0177] Step S505: construct an acoustic-vibration joint feature according to the voiceprint feature of the sound signal data and the nonlinear feature of the vibration signal data, and select a training sample set and a test sample set from the acoustic-vibration joint feature.
[0178] Step S506, in the target algorithm framework, initialize the sample weights of the support vector machine; iteratively train the support vector machine according to the training sample set, and update the sample weights of the support vector machine according to the training results; perform weighted combination of the support vector machine according to the updated sample weights to obtain a preliminarily trained classifier model.
[0179] Step S507, input the test sample set into the preliminarily trained classifier model for testing to obtain a fault classification result; when the classification accuracy of the fault classification result does not reach a preset value, adjust the preliminarily trained classifier model according to the test sample set until the classification accuracy reaches the preset value, and obtain a strong classifier model including multiple support vector machines.
[0180] The beneficial effects brought by the above embodiments are as follows:
[0181] The present application uses MVMD to decompose the OLTC fault status signal, which can achieve effective decomposition of the fault signal, extract more characteristic sound signal and vibration signal combined fault characteristics, and then achieve accurate perception of the OLTC fault status, effectively improving the accuracy of on-load tap changer fault status diagnosis.
[0182] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.
[0183] Based on the same inventive concept, the embodiment of the present application also provides a transformer on-load tap changer fault identification device for implementing the transformer on-load tap changer fault identification method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in the embodiments of one or more transformer on-load tap changer fault identification devices provided below can refer to the limitations of the transformer on-load tap changer fault identification method above, and will not be repeated here.
[0184] In an exemplary embodiment, Figure 6 As shown, a fault identification device for a transformer on-load tap changer is provided, comprising:
[0185] The data acquisition module 601 is used to acquire the sound signal data and vibration signal data of the transformer on-load tap changer under different fault states, and perform data cleaning;
[0186] A voiceprint construction module 602 is used to convert the sound signal data from a time domain signal into a time-frequency image, and construct a voiceprint feature of the sound signal data based on the texture feature of the time-frequency image;
[0187] A feature description module 603 is used to perform multivariate variational modal decomposition on the vibration signal data to extract modal components of the vibration signal data, and to describe nonlinear features of the vibration signal data using entropy based on the modal components;
[0188] A feature combination module 604 is used to construct a sound-vibration combination feature according to the voiceprint feature of the sound signal data and the nonlinear feature of the vibration signal data, and select a training sample set and a test sample set from the sound-vibration combination feature;
[0189] The model training module 605 is used to input the training sample set into the target algorithm framework with the support vector machine as the base learner for iterative training, and use the test sample set for testing and adjustment to obtain a strong classifier model including multiple support vector machines; the strong classifier model is used to identify faults of the transformer on-load tap changer.
[0190] In one embodiment, the voiceprint construction module 602 is further used to perform a generalized S transform on the sound signal data to obtain the time-frequency image; analyze the time-frequency image to obtain the box dimension, directionality and contrast of the time-frequency image as the texture feature of the time-frequency image; and construct the voiceprint feature based on the box dimension, the directionality and the contrast.
[0191] In one embodiment, the feature description module 603 is also used to perform multivariate variational modal decomposition on the vibration signal data to obtain a multivariate signal in matrix form, and extract a preset number of modal components from the multivariate signal; based on the modal components, the spectral entropy, energy entropy and permutation entropy of the vibration signal data are respectively calculated as the nonlinear features of the vibration signal data.
[0192] In one embodiment, the model training module 605 is also used to initialize the sample weights of the support vector machine in the target algorithm framework; iteratively train the support vector machine according to the training sample set, and update the sample weights of the support vector machine according to the training results; and weightedly combine the support vector machines according to the updated sample weights to obtain a preliminarily trained classifier model.
[0193] In one embodiment, the model training module 605 is also used to input the test sample set into the preliminarily trained classifier model for testing to obtain a fault classification result; when the classification accuracy of the fault classification result does not reach a preset value, the preliminarily trained classifier model is adjusted according to the test sample set until the classification accuracy reaches the preset value.
[0194] In one embodiment, the data acquisition module 601 is also used to identify duplicate data in the sound signal data and the vibration signal data, clear the duplicate data, and filter the sound signal data and the vibration signal data through a low-pass filter to remove noise data, thereby obtaining the sound signal data and the vibration signal data after data cleaning.
[0195] Each module in the above-mentioned transformer on-load tap changer fault identification device can be implemented in whole or in part by software, hardware and a combination thereof. Each of the above-mentioned modules can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each of the above modules.
[0196] In an exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be shown in FIG7. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a fault identification method for a transformer on-load tap changer is implemented.
[0197] Those skilled in the art will understand that the structure shown in FIG. 7 is merely a block diagram of a partial structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.
[0198] In one embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above method embodiments when executing the computer program.
[0199] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.
[0200] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.
[0201] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.
[0202] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and 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 embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application 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 memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic 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 and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.
[0203] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, 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, they should be considered to be within the scope of this application.
[0204] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations 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 shall be subject to the attached claims.
Claims
1. A fault identification method for a transformer on-load tap changer, It is characterized in that The method comprises: Obtain the sound signal data and vibration signal data of the transformer on-load tap changer under different fault conditions, and perform data cleaning; Converting the sound signal data from a time domain signal into a time-frequency image, and constructing a voiceprint feature of the sound signal data based on the texture feature of the time-frequency image; Performing multivariate variational modal decomposition on the vibration signal data to extract modal components of the vibration signal data, and using entropy to describe nonlinear characteristics of the vibration signal data based on the modal components; Constructing a sound-vibration joint feature according to the voiceprint feature of the sound signal data and the nonlinear feature of the vibration signal data, and selecting a training sample set and a test sample set from the sound-vibration joint feature; The training sample set is input into a target algorithm framework with a support vector machine as the base learner for iterative training, and the test sample set is used for testing and adjustment to obtain a strong classifier model including multiple support vector machines; the strong classifier model is used to identify faults of the transformer on-load tap changer.
2. The method according to claim 1, It is characterized in that The converting the sound signal data from a time domain signal into a time-frequency image, and constructing a voiceprint feature of the sound signal data based on the texture feature of the time-frequency image, comprises: Performing a generalized S transform on the sound signal data to obtain the time-frequency image; Analyzing the time-frequency image to obtain the box dimension, orientation and contrast of the time-frequency image as the texture features of the time-frequency image; The voiceprint feature is constructed based on the box dimension, the directionality and the contrast.
3. The method according to claim 1, It is characterized in that The performing multivariate variational modal decomposition on the vibration signal data to extract modal components of the vibration signal data, and using entropy to describe nonlinear characteristics of the vibration signal data based on the modal components, comprises: Performing multivariate variational modal decomposition on the vibration signal data to obtain a multivariate signal in matrix form, and extracting a preset number of modal components from the multivariate signal; According to the modal components, the spectrum entropy, energy entropy and permutation entropy of the vibration signal data are calculated respectively as the nonlinear characteristics of the vibration signal data.
4. The method according to claim 1, It is characterized in that The step of inputting the training sample set into a target algorithm framework using a support vector machine as a base learner for iterative training comprises: In the target algorithm framework, initializing the sample weights of the support vector machine; Iteratively training the support vector machine according to the training sample set, and updating the sample weights of the support vector machine according to the training results; The support vector machines are weighted combined according to the updated sample weights to obtain a preliminarily trained classifier model.
5. The method according to claim 4, It is characterized in that The testing and adjustment using the test sample set includes: Inputting the test sample set into the preliminarily trained classifier model for testing to obtain a fault classification result; When the classification accuracy of the fault classification result does not reach the preset value, the preliminarily trained classifier model is adjusted according to the test sample set until the classification accuracy reaches the preset value.
6. The method according to any one of claims 1 to 5, It is characterized in that The data cleaning comprises: Identify duplicate data in the sound signal data and the vibration signal data, clear the duplicate data, and filter the sound signal data and the vibration signal data through a low-pass filter to remove noise data, so as to obtain the sound signal data and the vibration signal data after data cleaning.
7. A fault identification device for a transformer on-load tap changer, It is characterized in that The device comprises: A data acquisition module is used to acquire the sound signal data and vibration signal data of the transformer on-load tap changer under different fault conditions and perform data cleaning; A voiceprint construction module, used to convert the sound signal data from a time domain signal into a time-frequency image, and construct a voiceprint feature of the sound signal data based on the texture features of the time-frequency image; A feature description module, used for performing multivariate variational modal decomposition on the vibration signal data to extract modal components of the vibration signal data, and using entropy to describe nonlinear features of the vibration signal data based on the modal components; A feature combination module, configured to construct a sound-vibration combination feature according to the voiceprint feature of the sound signal data and the nonlinear feature of the vibration signal data, and select a training sample set and a test sample set from the sound-vibration combination feature; A model training module is used to input the training sample set into a target algorithm framework with a support vector machine as the base learner for iterative training, and use the test sample set for testing and adjustment to obtain a strong classifier model including multiple support vector machines; the strong classifier model is used to identify faults of the transformer on-load tap changer.
8. A computer device comprising a memory and a processor, wherein the memory stores a computer program. It is characterized in that 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, It is 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, It is 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.
Citation Information
Cited By
Switch discharge detection method and system based on multi-source information fusion voiceprint signal
CN119249134A
Hub motor fault diagnosis method and system based on SBAO improved RLMD
CN119471371A
Electrical equipment electric leakage identification method and identification device based on voiceprint signal
CN119511144A
Electrical equipment leakage identification method and identification device based on voiceprint signal
CN119511144B
Method and system for recognizing voiceprint defect of power master equipment
CN119517084A
