Transformer fault acoustic signature diagnostic sample enhancement method, system and storage medium

Abstract

This invention provides a method, system, and storage medium for enhancing transformer fault acoustic signature diagnostic samples. The method includes: constructing the acoustic wave equation for the vibration of the core and windings propagating through the medium and the medium interface boundary conditions; performing sound field simulation on a fault model based on the acoustic wave equation and the medium interface boundary conditions to obtain a fault simulation sound field; determining a physical simulation dataset based on the fault simulation sound field; performing sparse representation on the physical simulation dataset to obtain a sparse dictionary and a sparse coefficient matrix; performing transfer learning on real fault samples to obtain a transfer dictionary; and performing sample enhancement based on the sparse dictionary, the sparse coefficient matrix, and the transfer dictionary to obtain enhanced transformer fault acoustic signature samples. In this embodiment, sample enhancement using the sparse dictionary, the sparse coefficient matrix, and the transfer dictionary effectively enhances the samples of transformer faults.

Description

Technical Field

[0001] This invention relates to the field of fault diagnosis technology, and in particular to a method, system and storage medium for enhancing transformer fault acoustic signature diagnostic samples. Background Technology

[0002] Fault detection technology for oil-immersed transformers has long relied on methods such as oil chromatography and partial discharge detection, which have limitations such as response lag and complex operation. In recent years, fault monitoring based on acoustic waveform analysis has become a research hotspot due to its non-invasive and real-time advantages. It identifies internal mechanical faults (such as winding deformation and core loosening) by analyzing the vibration acoustic waves of the transformer.

[0003] In the existing process of transformer vibration and acoustic fault identification, the scarcity of transformer fault samples leads to the data-driven fault identification model being prone to overfitting. Therefore, how to enhance the transformer fault samples has become an urgent problem to be solved. Summary of the Invention

[0004] The purpose of this invention is to provide a method, system, and storage medium for enhancing transformer fault acoustic signature diagnostic samples, in order to solve the problem of how to enhance transformer fault samples.

[0005] This invention is implemented as follows: a method for enhancing transformer fault acoustic signature diagnostic samples, the method comprising:

[0006] A vibration mechanism model of transformer core winding is established, and electromagnetic force correction is applied to the vibration mechanism model of transformer core winding to obtain a vibration model of transformer core winding.

[0007] Fault simulation was performed on the vibration model of the transformer core winding to obtain the fault model, and the acoustic wave equation and medium interface boundary conditions of the vibration of the core and winding in the medium propagation were constructed.

[0008] Based on the acoustic wave equation and the medium interface boundary conditions, the fault model is subjected to acoustic field simulation to obtain the fault simulation acoustic field, and the physical simulation dataset is determined based on the fault simulation acoustic field.

[0009] The physical simulation dataset is sparsely represented to obtain a sparse dictionary and a sparse coefficient matrix, and transfer learning is performed on real fault samples to obtain a transfer dictionary.

[0010] Sample enhancement is performed based on the sparse dictionary, the sparse coefficient matrix, and the transfer dictionary to obtain enhanced samples of transformer fault acoustic signatures.

[0011] Preferably, the magnetostriction of the transformer core winding in the vibration mechanism model is expressed as:

[0012]

[0013] in, This indicates the normal magnetostriction rate. Indicates the saturation magnetostriction. Indicates the saturation flux of the iron core. Indicates the number of turns in the high-voltage coil. Indicates the winding voltage amplitude. Indicates the effective area of ​​magnetic flux. Indicates the frequency of alternating current. t Represents a time variable;

[0014] The electromagnetic force on the winding coil in the transformer core winding vibration mechanism model is expressed as:

[0015]

[0016] in, Indicates the radius of the winding coil. This indicates the magnitude of the current applied in the winding. This represents the proportionality coefficient between leakage flux density and current. This indicates the electromagnetic force acting on the coil;

[0017] The electromagnetic force formula in the vibration model of the transformer core winding is expressed as follows:

[0018]

[0019] in, This represents the correction factor for winding deformation. This represents the electromagnetic force experienced by the coil in the vibration model of the transformer core winding.

[0020] Preferably, the fault models include a winding deformation fault model, a core loosening fault model, and a winding displacement fault model;

[0021] The winding deformation fault model is represented as follows:

[0022]

[0023] in, Indicates deformation sensitivity factor, This indicates the relative deformation of the winding radius;

[0024] The core loosening fault model is represented as follows:

[0025]

[0026] in, Indicates the magnetostriction rate after loosening. Indicates the coupling coefficient of the loose iron core. Indicates the gap between laminates. Indicates the thickness of the silicon steel sheet;

[0027] The winding displacement fault model is represented as follows:

[0028]

[0029] in, Indicates the cross-sectional area of ​​the iron core. This indicates the elastic modulus of silicon steel. Represents the strain-stress conversion factor. Indicates the quality of the iron core. Indicates the stiffness coefficient. It is the vibration displacement of the iron core.

[0030] Preferably, the acoustic wave equation is expressed as:

[0031]

[0032] in, Indicates the winding. Indicates insulating oil. Indicates the mechanical casing. Indicates the first Sound pressure in a medium Indicates the first The speed of sound in a medium Indicates the first The viscosity coefficient of the medium, Indicates the first The density of the medium, Represents the Laplace operator. To express differentiation;

[0033] The medium interface boundary conditions include solid-liquid interface acoustic pressure continuity condition, normal particle velocity continuity condition, and liquid-solid interface boundary condition as acoustic impedance constraint.

[0034] The solid-liquid interface acoustic pressure continuity condition is expressed as follows:

[0035]

[0036] in, The sound pressure level of the incident sound wave. This represents the sound pressure level of the reflected sound wave. Indicates the sound pressure level of the transmitted sound wave;

[0037] The continuity condition for the normal particle velocity is expressed as follows:

[0038]

[0039] in, This represents the total sound pressure of the first medium. This represents the total sound pressure of the second medium. This indicates the direction of the normal to the first medium.

[0040] The liquid-solid interface boundary condition, expressed as an acoustic impedance constraint, is as follows:

[0041]

[0042]

[0043] in, Represents the angular frequency of a sound wave. This represents the acoustic impedance of the outer surface of the mechanical housing. This represents the total sound pressure of the third medium.

[0044] Preferably, the physical simulation dataset is determined based on the fault simulation sound field, including:

[0045] A sound field reconstruction model is established based on the equivalent source method, and the surface of the transformer tank is discretized into an equivalent sound source region.

[0046] Acquire the measurement point signals in the fault simulation sound field, and calculate the correlation coefficient between the equivalent sound source signal and the measurement point signals;

[0047] The target acquisition point in the measurement point signal is determined based on the correlation coefficient, and the data of the target acquisition point in the fault simulation sound field is obtained to obtain the physical simulation dataset.

[0048] Preferably, the formula used to calculate the correlation coefficient between the equivalent sound source signal and the measurement point signal includes:

[0049]

[0050] in, Indicates the zero-prevention correction term. This represents the frequency amplitude of the equivalent sound source signal. This indicates the frequency amplitude of the signal at the measurement point. Indicates the first n The equivalent sound source signal and the first m The correlation coefficient between signals at each measurement point Indicates the first m The signal at each measurement point at all frequency points f The mean of the above, Indicates the first n An equivalent sound source signal at all frequency points f The mean of the above, This indicates the total number of frequency points.

[0051] Preferably, the physical simulation dataset is sparsely represented to obtain a sparse dictionary and a sparse coefficient matrix, and transfer learning is performed on real fault samples to obtain a transfer dictionary, including:

[0052] The physical simulation dataset is subjected to DC removal and bandpass filtering, and the bandpass-filtered physical simulation dataset is subjected to wavelet packet decomposition to obtain sub-band coefficients.

[0053] The sub-band coefficients are concatenated to obtain an initial dictionary, and the initial dictionary is iteratively optimized to obtain the sparse dictionary and the sparse coefficient matrix.

[0054] The real fault samples are sparsely reconstructed based on the sparse dictionary to obtain sample coefficients, and the reconstruction residuals are calculated based on the sample coefficients.

[0055] The real fault samples are labeled with credibility based on the reconstructed residuals to obtain credible labeled samples, and the sparse coefficient set is determined based on the credible labeled samples.

[0056] The sparse dictionary is used as the initial value of the dictionary, and the initial value of the dictionary is iteratively updated according to the trusted labeled samples and the sparse coefficient set to obtain the migration dictionary.

[0057] Preferably, sample enhancement is performed based on the sparse dictionary, the sparse coefficient matrix, and the transfer dictionary to obtain enhanced transformer fault acoustic signature samples, including:

[0058] Candidate augmentation samples are generated based on the sparse coefficient matrix and the transfer dictionary, and the residuals of the candidate augmentation samples under the sparse dictionary are calculated to obtain the augmentation residuals;

[0059] The candidate enhancement samples are screened based on the enhancement residual to obtain screened samples, and the screened samples are mixed with the trusted labeled samples to obtain the transformer fault acoustic text enhancement samples.

[0060] Another objective of this invention is to provide a transformer fault acoustic signature diagnostic sample enhancement system, the system comprising:

[0061] The model building module is used to establish a vibration mechanism model of the transformer core winding and to perform electromagnetic force correction on the vibration mechanism model of the transformer core winding to obtain a vibration model of the transformer core winding.

[0062] The fault simulation module is used to simulate the vibration model of the transformer core winding to obtain the fault model, and to construct the acoustic wave equation and medium interface boundary conditions for the vibration of the core and winding in the medium propagation.

[0063] The fault simulation module is used to perform sound field simulation on the fault model according to the acoustic wave equation and the medium interface boundary conditions, obtain the fault simulation sound field, and determine the physical simulation dataset according to the fault simulation sound field.

[0064] The sparse transfer module is used to perform sparse representation on the physical simulation dataset to obtain a sparse dictionary and a sparse coefficient matrix, and to perform transfer learning on real fault samples to obtain a transfer dictionary.

[0065] The sample enhancement module is used to perform sample enhancement based on the sparse dictionary, the sparse coefficient matrix, and the migration dictionary to obtain enhanced samples of transformer fault acoustic signatures.

[0066] In this embodiment of the invention, by simulating the vibration model of the transformer core winding, the fault state of the transformer can be effectively simulated. By constructing the acoustic wave equation and the boundary conditions of the medium interface for the vibration of the core and winding in the medium, the acoustic field simulation of the fault model can be effectively guaranteed. Based on the fault simulation acoustic field, the physical simulation dataset can be effectively determined. By performing sparse representation on the physical simulation dataset, the sparse dictionary and sparse coefficient matrix can be effectively obtained. By performing transfer learning on real fault samples, the transfer dictionary can be effectively obtained. By performing sample augmentation through the sparse dictionary, sparse coefficient matrix and transfer dictionary, the sample augmentation effect of the transformer fault samples can be effectively achieved. Attached Figure Description

[0067] Figure 1 This is a flowchart of the transformer fault acoustic signature diagnostic sample enhancement method provided in the first embodiment of the present invention;

[0068] Figure 2 This is a schematic diagram illustrating a specific implementation of the transformer fault acoustic signature diagnostic sample enhancement method provided in the first embodiment of the present invention;

[0069] Figure 3 This is a schematic diagram of the structure of the transformer fault acoustic signature diagnostic sample enhancement system provided in the second embodiment of the present invention;

[0070] Figure 4 This is a schematic diagram of the transformer fault acoustic signature diagnostic sample enhancement system provided in the third embodiment of the present invention;

[0071] Figure 5 This is a schematic diagram of the structure of the terminal device provided in the fifth embodiment of the present invention. Detailed Implementation

[0072] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0073] To illustrate the technical solution described in this invention, specific embodiments are described below.

[0074] Example 1

[0075] Please see Figures 1 to 2 This is a flowchart of a transformer fault acoustic signature enhancement method provided in the first embodiment of the present invention. This method can be applied to any device or system and includes the following steps:

[0076] Step S10: Establish a vibration mechanism model of the transformer core winding, and perform electromagnetic force correction on the vibration mechanism model of the transformer core winding to obtain a vibration model of the transformer core winding.

[0077] Among them, a vibration mechanism model of transformer core winding is established in multiphysics simulation software. The influence of winding deformation on leakage flux is considered. Electromagnetic force correction is performed on the vibration mechanism model of transformer core winding to obtain the vibration model of transformer core winding.

[0078] In this step, the transformer core is made of laminated silicon steel sheets. When alternating magnetic flux passes through the core, the silicon steel sheets undergo periodic expansion and contraction deformation due to the magnetostrictive effect. The electromagnetic forces between the core columns and the yoke, and between the windings and the core, change periodically with the alternating current. The magnetostriction rate of the core in the transformer core winding vibration mechanism model is expressed as:

[0079]

[0080] in, This indicates the normal magnetostriction rate. Indicates the saturation magnetostriction. Indicates the saturation flux of the iron core. Indicates the number of turns in the high-voltage coil. Indicates the winding voltage amplitude. Indicates the effective area of ​​magnetic flux. Indicates the frequency of alternating current. t Represents a time variable;

[0081] Under the influence of various factors such as electromagnetic force and mechanical force, the geometry, size, or position of the transformer winding changes. The electromagnetic force on the winding coil in the transformer core winding vibration mechanism model is represented as follows:

[0082]

[0083] in, Indicates the radius of the winding coil. This indicates the magnitude of the current applied in the winding. This represents the proportionality coefficient between leakage flux density and current. This indicates the electromagnetic force acting on the coil.

[0084] The electromagnetic force formula in the vibration model of the transformer core winding is expressed as follows:

[0085]

[0086] in, This represents the correction factor for winding deformation. This represents the electromagnetic force experienced by the coil in the vibration model of the transformer core winding.

[0087] Step S20: Perform fault simulation on the vibration model of the transformer core winding to obtain the fault model, and construct the acoustic wave equation and medium interface boundary conditions for the vibration of the core and winding in the medium propagation.

[0088] Based on the transformer core winding vibration model, winding deformation faults are introduced to construct a winding deformation fault model, core loosening faults are introduced to construct a core loosening fault model, and winding displacement faults are introduced to construct a winding displacement fault model. The acoustic wave equation and medium interface boundary conditions for the propagation of core and winding vibrations in a solid-liquid-solid multi-medium environment are also established.

[0089] Optionally, the fault models include winding deformation fault models, core loosening fault models, and winding displacement fault models;

[0090] When a transformer winding undergoes mechanical deformation, its geometric parameters (such as coil radius and turn spacing) change, leading to significant changes in the leakage magnetic field distribution characteristics and electromagnetic force amplitude-frequency characteristics. The winding deformation fault model is represented as follows:

[0091]

[0092] in, Indicates deformation sensitivity factor, This indicates the relative deformation of the winding radius;

[0093] Loosening between iron core laminations leads to a weakening of magnetostrictive strain constraint. The iron core loosening fault model is represented as follows:

[0094]

[0095] in, Indicates the magnetostriction rate after loosening. Indicates the coupling coefficient of the loose iron core. Indicates the gap between laminates. Indicates the thickness of the silicon steel sheet;

[0096] The winding displacement fault model is represented as follows:

[0097]

[0098] in, Indicates the cross-sectional area of ​​the iron core. This indicates the elastic modulus of silicon steel. Represents the strain-stress conversion factor. Indicates the quality of the iron core. Indicates the stiffness coefficient. It is the vibration displacement of the iron core.

[0099] In this step, the sound source of the transformer winding and core vibration is transmitted to the air through the oil and mechanical casing. Its propagation medium is "solid-liquid-solid" multi-medium propagation. The attenuation, impedance and refraction of the sound pattern are different when it propagates in different media. Establishing the acoustic wave method and boundary conditions at the medium interface is the key to accurately analyzing the fault sound pattern of oil-immersed transformer.

[0100] The acoustic wave equation is expressed as:

[0101]

[0102] in, Indicates the winding. Indicates insulating oil. Indicates the mechanical casing. Indicates the first Sound pressure in a medium Indicates the first The speed of sound in a medium Indicates the first The viscosity coefficient of the medium, Indicates the first The density of the medium, Represents the Laplace operator. To express differentiation;

[0103] The medium interface boundary conditions include solid-liquid interface acoustic pressure continuity condition, normal particle velocity continuity condition, and liquid-solid interface boundary condition as acoustic impedance constraint.

[0104] The solid-liquid interface acoustic pressure continuity condition is expressed as follows:

[0105]

[0106] in, The sound pressure level of the incident sound wave. This represents the sound pressure level of the reflected sound wave. Indicates the sound pressure level of the transmitted sound wave;

[0107] The continuity condition for the normal particle velocity is expressed as follows:

[0108]

[0109] in, This represents the total sound pressure of the first medium. This represents the total sound pressure of the second medium. This indicates the direction of the normal to the first medium.

[0110] The liquid-solid interface boundary condition, expressed as an acoustic impedance constraint, is as follows:

[0111]

[0112]

[0113] in, Represents the angular frequency of a sound wave. This represents the acoustic impedance of the outer surface of the mechanical housing. This represents the total sound pressure of the third medium.

[0114] Step S30: Based on the acoustic wave equation and the medium interface boundary conditions, perform acoustic field simulation on the fault model to obtain the fault simulation acoustic field, and determine the physical simulation dataset based on the fault simulation acoustic field.

[0115] Among them, finite element simulation software is used to simulate the winding deformation fault model, winding loosening fault model, and winding displacement fault model under the acoustic wave equation and medium interface boundary conditions, generating the sound field of winding deformation fault, winding loosening fault, and winding displacement fault, and determining the physical simulation dataset based on the sound field of winding deformation fault, winding loosening fault, and winding displacement fault.

[0116] Optionally, a physical simulation dataset is determined based on the fault simulation sound field, including:

[0117] A sound field reconstruction model is established based on the equivalent source method, and the surface of the transformer tank is discretized into an equivalent sound source region.

[0118] Acquire the measurement point signals in the fault simulation sound field, and calculate the correlation coefficient between the equivalent sound source signal and the measurement point signals;

[0119] The target acquisition point in the measurement point signal is determined based on the correlation coefficient, and the data of the target acquisition point in the fault simulation sound field is obtained to obtain the physical simulation dataset.

[0120] Specifically, the first sub-time sequence diagram is output by selecting the most correlated measurement point in the sound field of winding deformation fault, the second sub-time sequence diagram is output by selecting the most correlated measurement point in the sound field of winding loosening fault, and the third sub-time sequence diagram is output by selecting the most correlated measurement point in the sound field of winding displacement fault. The first, second, and third sub-time sequence diagrams are combined to obtain the physical simulation dataset.

[0121] Furthermore, the formula used to calculate the correlation coefficient between the equivalent sound source signal and the measurement point signal includes:

[0122]

[0123] in, Indicates the zero-prevention correction term. This represents the frequency amplitude of the equivalent sound source signal. This indicates the frequency amplitude of the signal at the measurement point. Indicates the first n The equivalent sound source signal and the first m The correlation coefficient between signals at each measurement point Indicates the first m The signal at each measurement point at all frequency points f The mean of the above, Indicates the first n An equivalent sound source signal at all frequency points f The mean of the above, This indicates the total number of frequency points.

[0124] Step S40: Perform sparse representation on the physical simulation dataset to obtain a sparse dictionary and a sparse coefficient matrix, and perform transfer learning on real fault samples to obtain a transfer dictionary;

[0125] Optionally, the physical simulation dataset is sparsely represented to obtain a sparse dictionary and a sparse coefficient matrix, and transfer learning is performed on real fault samples to obtain a transfer dictionary, including:

[0126] The physical simulation dataset is subjected to DC removal and bandpass filtering, and the bandpass-filtered physical simulation dataset is subjected to wavelet packet decomposition to obtain sub-band coefficients.

[0127] Among them, the physical simulation dataset is preprocessed by removing DC and bandpass filtering to ensure that the length of each sample vector in the physical simulation dataset is the same.

[0128]

[0129] in, Represents a physical simulation dataset. This represents the first sample vector in the physical simulation dataset. This indicates the length of the sample vector.

[0130] The sub-band coefficients are concatenated to obtain an initial dictionary, and the initial dictionary is iteratively optimized to obtain the sparse dictionary and the sparse coefficient matrix.

[0131] The K-Singular Value Decomposition (K-SVD) algorithm is used to iteratively optimize the initial dictionary and solve the sparse representation problem.

[0132]

[0133]

[0134]

[0135] in, Represents a sparse dictionary. Represents a sparse coefficient matrix. This indicates the number of non-zero elements in the vector. Indicates the constraint sparsity. Represents the th element in the sparse coefficient matrix A sparse coefficient, This represents the value of the variable that minimizes the objective function. The sparse coding vector represents the combined weights of the fault components. Indicates limiting conditions. This represents the square root of the sum of the squares of all elements of the Frobenius matrix.

[0136] The iterative process of K-SVD includes: first, fixing the sparse dictionary for each sample vector, solving for the sparse coefficients using Orthogonal Matching Pursuit (OMP), then optimizing each column of the sparse dictionary column by column, updating each column of atoms by performing SVD decomposition on the local residual matrix, until the iteration converges. After the iteration is complete, the final sparse dictionary and its sparse coefficient moments are output.

[0137] Preferably, to eliminate the energy differences between different sparse coefficients, each column vector of the sparse coefficient matrix is ​​normalized according to its norm:

[0138]

[0139] The real fault samples are sparsely reconstructed based on the sparse dictionary to obtain sample coefficients, and the reconstruction residuals are calculated based on the sample coefficients.

[0140] The process involves preprocessing real fault samples to ensure consistent sample lengths between the real fault samples and the physical simulation dataset, and then using a sparse dictionary to sparsely reconstruct each target domain sample from the real fault samples.

[0141]

[0142]

[0143] in, This represents a real fault sample. Indicates the length of the target domain sample. This represents the first target domain sample. Let the coefficient of the j-th sample be denoted as . Indicates the norm constraint threshold. Let j represent the j-th target domain sample.

[0144] Optionally, the formula used to calculate the reconstructed residuals based on the sample coefficients includes:

[0145]

[0146] in, Represents the reconstructed residual;

[0147] The real fault samples are labeled with credibility based on the reconstructed residuals to obtain credible labeled samples, and the sparse coefficient set is determined based on the credible labeled samples.

[0148] Specifically, the reconstructed residual is compared with the residual threshold, and the target domain samples whose reconstructed residual is less than or equal to the residual threshold are labeled with confidence to obtain the confident labeled samples. The sparse coefficients corresponding to the confident labeled samples are obtained to obtain the sparse coefficient set.

[0149] The sparse dictionary is used as the initial value of the dictionary, and the initial value of the dictionary is iteratively updated according to the trusted labeled samples and the sparse coefficient set to obtain the migration dictionary;

[0150] The approach employs incremental updates, using a sparse dictionary as the initial value for the transfer dictionary. Trusted labeled samples are gradually integrated through "online dictionary learning" to obtain the final transfer dictionary. Specifically, let:

[0151]

[0152] If the number of iterations is less than or equal to the number of trusted labeled samples, then select the current iteration number of trusted labeled samples and their sparsity coefficients. , , These represent the coefficient matrix, residual matrix, and transition dictionary at the initial state, respectively.

[0153] If the number of iterations is greater than or equal to the number of trusted labeled samples, then either use or randomly select any pair of real fault samples and sample coefficients.

[0154] In the In this iteration, the cumulative matrix is ​​updated first:

[0155]

[0156] in, and They represent the first i The accumulated coefficient matrix and accumulated residual matrix in each iteration. and They represent the first i The accumulated coefficient matrix and accumulated residual matrix in -1 iterations, Indicates the first i Sample coefficients, Indicates the first i One target domain sample, Indicates matrix transpose;

[0157] Then, the dictionary update subproblem is solved:

[0158]

[0159] in, This refers to the updated dictionary. Let D denote the trace of the matrix, D denote the dictionary matrix, and C denote the dictionary constraint set.

[0160] In the updated dictionary, the current trusted labeled samples Fine-tuning of the sparsity coefficient is then performed:

[0161]

[0162] If the new sparsity coefficient If it can significantly reduce the reconstruction error, then it replaces the original sparse coefficients. If the coefficients are not used, they will be used in the next iteration; otherwise, the original coefficients will be retained.

[0163] Repeat the iterations until the preset number of iterations is reached, then output the final transition dictionary. ,at this time, It retains the basic structure of the source domain simulation samples while incorporating features from a small number of real samples in the target domain, and can be used to generate a large number of candidate augmented samples.

[0164] Step S50: Perform sample enhancement based on the sparse dictionary, the sparse coefficient matrix, and the transfer dictionary to obtain enhanced transformer fault acoustic signature samples;

[0165] Optionally, sample enhancement is performed based on the sparse dictionary, the sparse coefficient matrix, and the transfer dictionary to obtain enhanced transformer fault acoustic signature samples, including:

[0166] Candidate augmentation samples are generated based on the sparse coefficient matrix and the transfer dictionary, and the residuals of the candidate augmentation samples under the sparse dictionary are calculated to obtain the augmentation residuals;

[0167] Wherein, the sparse coefficients corresponding to each source domain sample This can be obtained by directly reconstructing under the migration dictionary:

[0168]

[0169] in, This represents candidate augmentation samples; this process is called "interaction one," which is generated using only sparse coefficients and a transfer dictionary.

[0170] To further ensure the quality of the generated samples, the residual of each candidate augmented sample under the sparse dictionary can be calculated:

[0171]

[0172] in, This indicates that the residual has been enhanced.

[0173] The candidate enhancement samples are screened based on the enhancement residual to obtain screened samples, and the screened samples are mixed with the trusted labeled samples to obtain the transformer fault acoustic text enhancement samples;

[0174] The process involves comparing the enhanced residual with the preset residual, deleting candidate enhanced samples whose enhanced residual is greater than the preset residual, and determining the remaining candidate enhanced samples as screening samples.

[0175] In this embodiment, by simulating the vibration model of the transformer core winding, the fault state of the transformer can be effectively simulated. By constructing the acoustic wave equation and the boundary conditions of the medium interface for the vibration of the core and winding in the medium, the acoustic field simulation of the fault model can be effectively guaranteed. Based on the fault simulation acoustic field, the physical simulation dataset can be effectively determined. By performing sparse representation on the physical simulation dataset, the sparse dictionary and sparse coefficient matrix can be effectively obtained. By performing transfer learning on real fault samples, the transfer dictionary can be effectively obtained. By performing sample augmentation through the sparse dictionary, sparse coefficient matrix and transfer dictionary, the sample augmentation effect of the transformer fault samples can be effectively achieved.

[0176] Example 2

[0177] Please see Figure 3 This is a schematic diagram of the structure of the transformer fault acoustic signature diagnostic sample enhancement system 100 provided in the second embodiment of the present invention, including:

[0178] The model building module 10 is used to establish a vibration mechanism model of the transformer core winding and to perform electromagnetic force correction on the vibration mechanism model of the transformer core winding to obtain a vibration model of the transformer core winding.

[0179] The fault simulation module 11 is used to simulate the vibration model of the transformer core winding to obtain the fault model, and to construct the acoustic wave equation and medium interface boundary conditions for the vibration of the core and winding in the medium propagation.

[0180] The fault simulation module 12 is used to perform sound field simulation on the fault model according to the acoustic wave equation and the medium interface boundary conditions, to obtain the fault simulation sound field, and to determine the physical simulation dataset according to the fault simulation sound field.

[0181] The sparse transfer module 13 is used to perform sparse representation on the physical simulation dataset to obtain a sparse dictionary and a sparse coefficient matrix, and to perform transfer learning on real fault samples to obtain a transfer dictionary.

[0182] The sample enhancement module 14 is used to perform sample enhancement based on the sparse dictionary, the sparse coefficient matrix and the migration dictionary to obtain enhanced samples of transformer fault acoustic signatures.

[0183] In this embodiment, by simulating the vibration model of the transformer core winding, the fault state of the transformer can be effectively simulated. By constructing the acoustic wave equation and the boundary conditions of the medium interface for the vibration of the core and winding in the medium, the acoustic field simulation of the fault model can be effectively guaranteed. Based on the fault simulation acoustic field, the physical simulation dataset can be effectively determined. By performing sparse representation on the physical simulation dataset, the sparse dictionary and sparse coefficient matrix can be effectively obtained. By performing transfer learning on real fault samples, the transfer dictionary can be effectively obtained. By performing sample augmentation through the sparse dictionary, sparse coefficient matrix and transfer dictionary, the sample augmentation effect of the transformer fault samples can be effectively achieved.

[0184] Example 3

[0185] Please see Figure 4 This is a flowchart of a transformer fault acoustic signature enhancement method provided in the third embodiment of the present invention. This method can be applied to any device or system and includes the following steps:

[0186] Step S1: Establish a geometric model of the transformer core in a multiphysics simulation software. Based on the principles of electromagnetics and mechanical dynamics, establish a model to describe the mechanical vibration of the transformer core components, the core and windings, under the action of magnetostriction and electromagnetic force.

[0187] Optionally, in step S1:

[0188] The core of a transformer is made of laminated silicon steel sheets. When an alternating magnetic flux passes through the core, the silicon steel sheets will undergo periodic expansion and contraction deformation due to the magnetostrictive effect. The electromagnetic forces between the core columns and the yoke, and between the windings and the core, will change periodically with the alternating current.

[0189] Step S2: Based on the transformer model established in Step S1, introduce winding deformation faults to construct a winding deformation fault model; introduce core loosening faults to construct a core loosening fault model; and introduce winding displacement faults to construct a winding displacement fault model.

[0190] Step S3: Establish the acoustic wave equation and boundary conditions for the propagation of vibrations of the iron core and windings in a solid-liquid-solid multi-medium environment;

[0191] Step S4: Use finite element simulation software to perform sound field simulation of the fault and obtain the sound pattern timing diagram at the measurement point;

[0192] Optionally, in step S4:

[0193] Step S4.1: Simulate the faults. Using finite element simulation software, simulate the winding deformation fault model, winding loosening fault model, and winding displacement fault model under the acoustic wave equation and boundary conditions, and generate the sound field of winding deformation fault, winding loosening fault, and winding displacement fault.

[0194] Step S4.2: Obtain the timing diagram of the acoustic signature at the measurement point.

[0195] The surface of the transformer tank is discretized into N equivalent sound source regions;

[0196] A sound field reconstruction model is established based on the equivalent source method. The surface of the transformer tank is discretized into N equivalent sound source regions. The equivalent sound source signals and the candidate measurement point signals are subjected to windowed Fourier transforms and the spectrum matrix is ​​normalized. The measurement point with the maximum correlation is selected based on the Pearson correlation matrix.

[0197] In the acoustic field of winding deformation fault, the maximum correlation measurement point is selected, and the acoustic fingerprint time sequence diagram is output. The sound pressure time domain signal is extracted to generate the acoustic fingerprint time sequence diagram. In the acoustic field of winding loosening fault, the maximum correlation measurement point is selected, and the acoustic fingerprint time sequence diagram is output. The sound pressure time domain signal is extracted to generate the acoustic fingerprint time sequence diagram. In the acoustic field of winding displacement fault, the maximum correlation measurement point is selected, and the acoustic fingerprint time sequence diagram is output. The sound pressure time domain signal is extracted to generate the acoustic fingerprint time sequence diagram.

[0198] Step S5: Fault sound sample transfer enhancement based on sparse dictionaries of source and target domains: Construct a basic dictionary using source domain knowledge, carefully select reliable samples from the target domain as a bridge, dynamically adjust the feature representation to adapt it to the target domain, and finally use the adjusted feature representation to interactively synthesize a large number of transformer fault soundprint enhancement samples.

[0199] Optionally, in step S5, the fault sound sample transfer enhancement between the transformer multiphysics fault sample generation data (source domain data) and the actual transformer collected data (target domain data) specifically includes the following steps:

[0200] Step S5.1: Source Domain Sparse Dictionary Learning and Sparse Coefficient Acquisition

[0201] First, the simulated clean fault sound sample set in the source domain is preprocessed with DC removal and bandpass filtering to ensure that each sample vector has the same length. Then, wavelet packet decomposition is used to perform multi-scale, multi-band decomposition on each sample vector, generating sub-band coefficients, which are then concatenated to construct an initial dictionary. The K-SVD algorithm is then used for iterative optimization.

[0202] The sparse coefficients are obtained through Orthogonal Matching Pursuit (OMP). Then, each column of the dictionary is optimized column by column, and the atoms in each column are updated by performing SVD decomposition on the local residual matrix until the iteration converges. After the iteration is completed, the final dictionary of the source domain and its sparse coefficient matrix are output.

[0203] Step S5.2: Screening of Trusted Annotated Samples in the Target Domain

[0204] First, a small set of labeled fault sound samples from the actual target domain is collected and subjected to the same preprocessing operations, maintaining the sample length consistent with the source domain. Then, a sparse dictionary is used to sparsely reconstruct each target domain sample, and the corresponding reconstruction residual is calculated.

[0205] The real fault samples are labeled with credibility based on the reconstructed residuals to obtain credible labeled samples, and the sparse coefficient set is determined based on the credible labeled samples.

[0206] The sparse dictionary is used as the initial value of the dictionary, and the initial value of the dictionary is iteratively updated according to the trusted labeled samples and the sparse coefficient set to obtain the migration dictionary.

[0207] Step S5.3: Online iterative update of the migration dictionary

[0208] Using an incremental update approach, a sparse dictionary is used as the initial value for the transfer dictionary. A small number of trusted labeled samples from the target domain are gradually incorporated through "online dictionary learning" to obtain the final transfer dictionary.

[0209] Step S5.4: Interactive Sample Generation and Enhancement in Dual-Target Domains

[0210] Using the source domain sparse coefficient matrix obtained in step S5.1 and the transfer dictionary obtained in step S5.3, a set of pseudo-target domain samples can be generated. For the sparse coefficient vector corresponding to each source domain sample, candidate enhancement samples are reconstructed under the transfer dictionary.

[0211] This process is called "Interaction 1," which utilizes only source domain coefficients and a transfer dictionary for generation. Subsequently, the real, reliable labeled samples selected in step S5.2 are mixed with the candidate enhancement samples to construct the enhanced transformer fault acoustic signature samples:

[0212] This mixing process, known as "interaction two," balances the diversity and realism of the target domain samples by considering both a small number of real samples and a large number of generated samples. To further ensure the quality of the generated samples, the residual of each candidate augmented sample under a sparse dictionary can be calculated. Based on the calculated residual, the candidate augmented samples are screened. The remaining high-quality candidate augmented samples are then remixed with real and reliable samples to obtain transformer fault acoustic signature augmented samples.

[0213] Example 4

[0214] This is a flowchart of a transformer fault acoustic signature enhancement method according to the fourth embodiment of the present invention. This method can be applied to any device or system and includes the following steps:

[0215] Step 1. Modeling the vibration mechanism of the core and windings

[0216] Step 1.1 Calibration of magnetostrictive parameters of iron core

[0217] The saturation magnetostriction, saturation magnetic flux, and elastic modulus of silicon steel sheets were measured experimentally.

[0218] The corrected magnetostriction is calculated based on the gap between the iron core laminations and the thickness of the silicon steel sheet.

[0219] Step 1.2: Calibration of winding deformation parameters

[0220] Apply different current amplitudes and measure the deformation of the winding radius;

[0221] Fit the deformation sensitivity factor to determine the corrected leakage flux ratio coefficient;

[0222] Step 2. Construction and Simulation of Multiphysics Coupled Model

[0223] Step 2.1: Electromagnetic Field Simulation Settings

[0224] Import the 3D geometric model of the transformer into COMSOL and define the winding material properties (conductivity, permeability).

[0225] Set up winding current excitation (amplitude I = 100 A, frequency f = 50 Hz) and solve for the leakage magnetic field distribution;

[0226] Step 2.2 Solving the coupling of force field and sound field

[0227] The electromagnetic force data is mapped to the solid mechanics module to define the elastic modulus and damping coefficient of the core and windings.

[0228] Set the acoustic module media parameters (structural components = 7850 kg / m³, insulating oil = 850 kg / m³, outer shell = 7800 kg / m³).

[0229] Solve the vibration displacement and sound pressure distribution simultaneously, with a time step of 1 μs and a total duration of 0.1 s.

[0230] Step 3. Simulation Result Verification

[0231] Step 3.1: Extraction of sound pressure signal at the measuring point

[0232] Ten measuring points were set on the surface of the transformer casing to output the sound pressure time-domain signal;

[0233] A short-time Fourier transform is performed on the sound pressure time-domain signal to generate a soundprint time-frequency map.

[0234] Step 3.2: Error Analysis

[0235] Compare the measured sound pressure data with the simulation results and calculate the average relative error (required to be <5%). If the error exceeds the limit, adjust the mesh density or time step and solve again.

[0236] II. Cross-domain migration enhancement method for scarce fault voiceprint samples

[0237] Step 4.1: Fault Parameter Combination Design

[0238] Nine fault scenarios are generated by preset winding deformation degree and core loosening level; each scenario runs simulation and outputs sound pressure time domain signal and corresponding label.

[0239] Step 4.2: Equivalent sound source screening

[0240] The surface of the fuel tank is discretized into 50 equivalent sound source regions;

[0241] Calculate the Pearson correlation coefficient between each sound source and the candidate measurement points, and select the top 5 measurement points with the highest correlation as data acquisition points.

[0242] Step 5. Sparse dictionary transfer learning

[0243] Step 5.1: Initial dictionary training

[0244] The simulation dataset was processed by frame segmentation (frame length 512 points, overlap rate 50%).

[0245] The initial dictionary was trained using the K-SVD algorithm, with sparsity set to 10.

[0246] Extract the shared sparse coefficient matrix.

[0247] Step 5.2: Migration Dictionary Optimization

[0248] Load the actual collected samples, normalize them, and align them with the simulation data;

[0249] A fixed shared sparse coefficient matrix is ​​used to optimize the target dictionary through gradient descent, and the optimization is iterated until the loss function converges (usually requiring 100 to 200 rounds).

[0250] Step 5.3: Enhanced Sample Generation

[0251] Reconstructing the signal using a transfer dictionary and a shared sparse coefficient matrix;

[0252] Inject measured ambient noise (signal-to-noise ratio set to 20 dB) into the reconstructed signal to generate an enhanced sample set.

[0253] Step 6. Cross-domain sample validation

[0254] Step 6.1: Visualization of Feature Distribution

[0255] MFCC features were extracted from simulation data, augmented data, and measured data.

[0256] We used t-SNE to reduce the dimensionality to 2D space and verified the overlap of feature distributions (objective: cosine similarity between augmented data and measured data > 0.85).

[0257] Step 6.2: Cross-domain recognition test

[0258] Train a support vector machine (SVM) classifier using augmented data in the source domain and 50 real test samples in the target domain;

[0259] Calculate the cross-domain recognition accuracy (target: improve by 15%~20% compared to pure simulation data).

[0260] Example 5

[0261] Figure 5 This is a structural block diagram of a terminal device 2 provided in the fifth embodiment of this application. For example... Figure 4 As shown, the terminal device 2 in this embodiment includes a processor 20, a memory 21, and a computer program 22 stored in the memory 21 and executable on the processor 20, such as a program for a transformer fault acoustic signature diagnostic sample enhancement method. When the processor 20 executes the computer program 22, it implements the steps in the various embodiments of the transformer fault acoustic signature diagnostic sample enhancement method described above.

[0262] For example, the computer program 22 may be divided into one or more modules, which are stored in the memory 21 and executed by the processor 20 to complete this application. The one or more modules may be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program 22 in the terminal device 2. The terminal device may include, but is not limited to, the processor 20 and the memory 21.

[0263] The processor 20 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0264] The memory 21 can be an internal storage unit of the terminal device 2, such as a hard drive or memory of the terminal device 2. The memory 21 can also be an external storage device of the terminal device 2, such as a plug-in hard drive, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the terminal device 2. Furthermore, the memory 21 can include both internal and external storage units of the terminal device 2. The memory 21 is used to store the computer program and other programs and data required by the terminal device. The memory 21 can also be used to temporarily store data that has been output or will be output.

[0265] Furthermore, the functional modules in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0266] If an integrated module is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. This computer-readable storage medium can be non-volatile or volatile. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable storage medium can include: any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the contents of a computer-readable storage medium may be appropriately added to or subtracted from the contents as required by the legislation and patent practice in a jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, a computer-readable storage medium may not include electrical carrier signals and telecommunication signals.

[0267] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for enhancing transformer fault acoustic signature diagnostic samples, characterized in that, The method includes: A vibration mechanism model of transformer core winding is established, and electromagnetic force correction is applied to the vibration mechanism model of transformer core winding to obtain a vibration model of transformer core winding. Fault simulation was performed on the vibration model of the transformer core winding to obtain the fault model, and the acoustic wave equation and medium interface boundary conditions of the vibration of the core and winding in the medium propagation were constructed. Based on the acoustic wave equation and the medium interface boundary conditions, the fault model is subjected to acoustic field simulation to obtain the fault simulation acoustic field, and the physical simulation dataset is determined based on the fault simulation acoustic field. The physical simulation dataset is sparsely represented to obtain a sparse dictionary and a sparse coefficient matrix, and transfer learning is performed on real fault samples to obtain a transfer dictionary. Based on the sparse dictionary, the sparse coefficient matrix, and the transfer dictionary, sample enhancement is performed to obtain transformer fault acoustic signature enhancement samples. The physical simulation dataset is sparsely represented to obtain a sparse dictionary and a sparse coefficient matrix. Transfer learning is then performed on real fault samples to obtain a transfer dictionary, including: The physical simulation dataset is subjected to DC removal and bandpass filtering, and the bandpass-filtered physical simulation dataset is subjected to wavelet packet decomposition to obtain sub-band coefficients. The sub-band coefficients are concatenated to obtain an initial dictionary, and the initial dictionary is iteratively optimized to obtain the sparse dictionary and the sparse coefficient matrix. The real fault samples are sparsely reconstructed based on the sparse dictionary to obtain sample coefficients, and the reconstruction residuals are calculated based on the sample coefficients. The real fault samples are labeled with credibility based on the reconstructed residuals to obtain credible labeled samples, and the sparse coefficient set is determined based on the credible labeled samples. The sparse dictionary is used as the initial value of the dictionary, and the initial value of the dictionary is iteratively updated according to the trusted labeled samples and the sparse coefficient set to obtain the migration dictionary.

2. The transformer fault acoustic signature diagnostic sample enhancement method as described in claim 1, characterized in that, In the transformer core winding vibration mechanism model, the magnetostriction of the core is expressed as: in, This indicates the normal magnetostriction rate. Indicates the saturation magnetostriction. Indicates the saturation flux of the iron core. Indicates the number of turns in the high-voltage coil. Indicates the winding voltage amplitude. Indicates the effective area of ​​magnetic flux. Indicates the frequency of alternating current. t Represents a time variable; The electromagnetic force on the winding coil in the transformer core winding vibration mechanism model is expressed as: in, Indicates the radius of the winding coil. This indicates the magnitude of the current applied in the winding. This represents the proportionality coefficient between leakage flux density and current. This indicates the electromagnetic force acting on the coil; The electromagnetic force formula in the vibration model of the transformer core winding is expressed as follows: in, This represents the correction factor for winding deformation. This represents the electromagnetic force experienced by the coil in the vibration model of the transformer core winding.

3. The transformer fault acoustic signature diagnostic sample enhancement method as described in claim 2, characterized in that, The fault models include winding deformation fault models, core loosening fault models, and winding displacement fault models. The winding deformation fault model is represented as follows: in, Indicates deformation sensitivity factor, This indicates the relative deformation of the winding radius; The core loosening fault model is represented as follows: in, Indicates the magnetostriction rate after loosening. Indicates the coupling coefficient of the loose iron core. Indicates the gap between laminates. Indicates the thickness of the silicon steel sheet; The winding displacement fault model is represented as follows: in, Indicates the cross-sectional area of ​​the iron core. This indicates the elastic modulus of silicon steel. Represents the strain-stress conversion factor. Indicates the quality of the iron core. Indicates the stiffness coefficient. It is the vibration displacement of the iron core.

4. The transformer fault acoustic signature diagnostic sample enhancement method as described in claim 1, characterized in that, The acoustic wave equation is expressed as: in, Indicates the winding. Indicates insulating oil. Indicates the mechanical casing. Indicates the first Sound pressure in a medium Indicates the first The speed of sound in a medium Indicates the first The viscosity coefficient of the medium, Indicates the first The density of the medium, Represents the Laplace operator. To express differentiation; The medium interface boundary conditions include solid-liquid interface acoustic pressure continuity condition, normal particle velocity continuity condition, and liquid-solid interface boundary condition as acoustic impedance constraint. The solid-liquid interface acoustic pressure continuity condition is expressed as follows: in, The sound pressure level of the incident sound wave. This represents the sound pressure level of the reflected sound wave. Indicates the sound pressure level of the transmitted sound wave; The continuity condition for the normal particle velocity is expressed as follows: in, This represents the total sound pressure of the first medium. This represents the total sound pressure of the second medium. This indicates the direction of the normal to the first medium. The liquid-solid interface boundary condition, expressed as an acoustic impedance constraint, is as follows: in, Represents the angular frequency of a sound wave. This represents the acoustic impedance of the outer surface of the mechanical housing. This represents the total sound pressure of the third medium.

5. The transformer fault acoustic signature diagnostic sample enhancement method as described in claim 1, characterized in that, The physical simulation dataset is determined based on the fault simulation sound field, including: A sound field reconstruction model is established based on the equivalent source method, and the surface of the transformer tank is discretized into an equivalent sound source region. Acquire the measurement point signals in the fault simulation sound field, and calculate the correlation coefficient between the equivalent sound source signal and the measurement point signals; The target acquisition point in the measurement point signal is determined based on the correlation coefficient, and the data of the target acquisition point in the fault simulation sound field is obtained to obtain the physical simulation dataset.

6. The transformer fault acoustic signature diagnostic sample enhancement method as described in claim 5, characterized in that, The formula used to calculate the correlation coefficient between the equivalent sound source signal and the measurement point signal includes: in, Indicates the zero-prevention correction term. This represents the frequency amplitude of the equivalent sound source signal. This indicates the frequency amplitude of the signal at the measurement point. Indicates the first n The equivalent sound source signal and the first m The correlation coefficient between signals at each measurement point Indicates the first m The signal at each measurement point at all frequency points f The mean of the above, Indicates the first n An equivalent sound source signal at all frequency points f The mean of the above, This indicates the total number of frequency points.

7. The transformer fault acoustic signature diagnostic sample enhancement method as described in claim 1, characterized in that, Based on the sparse dictionary, the sparse coefficient matrix, and the transfer dictionary, sample enhancement is performed to obtain transformer fault acoustic signature enhancement samples, including: Candidate augmentation samples are generated based on the sparse coefficient matrix and the transfer dictionary, and the residuals of the candidate augmentation samples under the sparse dictionary are calculated to obtain the augmentation residuals; The candidate enhancement samples are screened based on the enhancement residual to obtain screened samples, and the screened samples are mixed with the trusted labeled samples to obtain the transformer fault acoustic text enhancement samples.

8. A transformer fault acoustic signature diagnostic sample enhancement system, characterized in that, The system includes: The model building module is used to establish a vibration mechanism model of the transformer core winding and to perform electromagnetic force correction on the vibration mechanism model of the transformer core winding to obtain a vibration model of the transformer core winding. The fault simulation module is used to simulate the vibration model of the transformer core winding to obtain the fault model, and to construct the acoustic wave equation and medium interface boundary conditions for the vibration of the core and winding in the medium propagation. The fault simulation module is used to perform sound field simulation on the fault model according to the acoustic wave equation and the medium interface boundary conditions, obtain the fault simulation sound field, and determine the physical simulation dataset according to the fault simulation sound field. The sparse transfer module is used to perform sparse representation on the physical simulation dataset to obtain a sparse dictionary and a sparse coefficient matrix, and to perform transfer learning on real fault samples to obtain a transfer dictionary. The sample enhancement module is used to perform sample enhancement based on the sparse dictionary, the sparse coefficient matrix and the transfer dictionary to obtain enhanced samples of transformer fault acoustic signatures. The sparse migration module is also used to: perform DC removal and bandpass filtering on the physical simulation dataset, and perform wavelet packet decomposition on the bandpass filtered physical simulation dataset to obtain sub-band coefficients; The sub-band coefficients are concatenated to obtain an initial dictionary, and the initial dictionary is iteratively optimized to obtain the sparse dictionary and the sparse coefficient matrix. The real fault samples are sparsely reconstructed based on the sparse dictionary to obtain sample coefficients, and the reconstruction residuals are calculated based on the sample coefficients. The real fault samples are labeled with credibility based on the reconstructed residuals to obtain credible labeled samples, and the sparse coefficient set is determined based on the credible labeled samples. The sparse dictionary is used as the initial value of the dictionary, and the initial value of the dictionary is iteratively updated according to the trusted labeled samples and the sparse coefficient set to obtain the migration dictionary.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 7.

Images