Ultrasonic signal characteristic parameter extraction method and related device

Abstract

The invention belongs to the technical field of characteristic parameter extraction, and discloses an ultrasonic signal characteristic parameter extraction method and a related device, and the method comprises the steps: obtaining an ultrasonic signal caused by partial discharge, and obtaining a phase-distinguished partial discharge PRPD spectrogram and a time domain diagram of the partial discharge signal; obtaining a two-dimensional spectrogram of the discharge parameters according to the partial discharge PRPD spectrogram of the phase discrimination of the partial discharge signal; performing wavelet transform processing on the time-domain graph of the partial discharge signal to obtain different types of partial discharge time-frequency spectrograms; graying the time-frequency spectrogram to obtain a grey-scale image, and extracting statistical characteristics of the two-dimensional spectrogram, moment characteristics and texture characteristics of the grey-scale image and time-frequency characteristics of the discharge pulse; according to the method, multi-dimensional complementary and relatively complete characteristic parameters can be provided for input of partial discharge mode recognition, high-dimensional errors introduced in the training process are reduced, and the training efficiency is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of characteristic parameter extraction of ultrasonic signals, and in particular relates to a characteristic parameter extraction method of ultrasonic signals and related devices. Background Art

[0002] Gas insulated switchgear (GIS) and power transformers are important electrical equipment in the power system. They are in a pivotal position in the power system. Once a fault occurs, a large-scale power outage may occur, causing significant losses to the power system and the national economy. Therefore, the power system attaches great importance to the status of power transformers, especially the health of their insulating medium. With the continuous improvement of the voltage level of the power grid, high-voltage insulation problems have become an important factor affecting the stable operation of the power grid. Partial discharge has also become an important cause of insulation degradation of power transformers. Therefore, the detection and evaluation of partial discharge has become an important means of monitoring its insulation condition.

[0003] There is no universally recognized concept of partial discharge. It is generally believed that partial discharge is an electrical phenomenon in which discharge occurs only in some areas of the insulation system under the action of an electric field, but does not penetrate the conductors to which voltage is applied, that is, it has not yet completely broken down. Although partial discharge generally does not cause penetrating breakdown of insulation, it can cause local damage to dielectrics (especially organic dielectrics). If partial discharge exists for a long time, it may cause a reduction in the electrical strength of the insulation device under certain conditions. This slow development process poses a great hidden danger to high-voltage electrical equipment. Therefore, real-time detection of partial discharge is an important issue in the daily operation and maintenance of power systems.

[0004] The detection method of partial discharge is usually based on various phenomena caused by partial discharge, and the state of partial discharge is characterized by physical quantities that can describe the phenomenon. Partial discharge usually produces electric pulses, gas products, ultrasound, electromagnetic radiation, light, local overheating and other phenomena. Correspondingly, there are many detection methods such as electric pulse detection method, gas chromatography detection method, ultrasonic detection method, electromagnetic wave detection method, light detection method, etc.

[0005] The electric pulse detection method detects the pulse current caused by partial discharge in the grounding wire of the end screen of the casing of the electrical equipment, the grounding wire of the shell, the grounding wire of the core, and the winding by detecting impedance, and obtains some basic quantities of partial discharge; due to the low test frequency and narrow frequency band, this detection method obtains less information and is susceptible to electromagnetic interference. The gas chromatography detection method determines the degree and mode of partial discharge based on the decomposition gas produced by partial discharge; but the gas composition analysis is time-consuming and cannot achieve real-time detection. The ultra-high frequency detection method detects the ultra-high frequency electromagnetic wave signal of partial discharge of power equipment to obtain partial discharge information; but this detection method requires the pre-embedded special ultra-high frequency sensor in the power equipment. The optical detection method detects partial discharge by detecting the fluorescence caused by partial discharge through optical fiber; but this detection method requires complex optical fiber arrangement, and the resolution of the current optical fiber sensor cannot meet the actual engineering needs, and quantitative analysis and partial discharge pattern recognition cannot be performed.

[0006] The ultrasonic detection method uses an ultrasonic sensor fixed on the wall of the transformer box to receive the ultrasonic waves generated by the partial discharge inside the transformer, thereby detecting the size and location of the partial discharge. This detection method can perform real-time monitoring and offline detection without going deep into the electrical equipment, and is easy to locate and not susceptible to electromagnetic interference.

[0007] In the existing schemes, most of them only consider one of the phase characteristics, time domain characteristics and frequency domain characteristics of partial discharge signals, but do not consider them comprehensively. However, since partial discharge signals are transient non-stationary signals, and the signals may also contain interference signals in addition to discharge pulse signals, the signal volume is huge, and it is difficult to express all the characteristic information with a single characteristic analysis and to efficiently classify discharge defects, so it is impossible to obtain satisfactory recognition results.

[0008] Furthermore, high-dimensional input parameters are not always beneficial to pattern recognition. First, high-dimensional input parameters may lead to the "curse of dimensionality", that is, problems such as increased data sparsity, decreased computational efficiency, and overfitting, which greatly increase the training cost and inference complexity of the recognition model. Secondly, too high an input dimension reduces the ability to visualize data, making it difficult to build, analyze, and interpret the recognition model. In order to prevent high-dimensional data from introducing the "curse of dimensionality" in pattern recognition or dramatically increasing the cost of model construction and training, the feature parameter dimensions provided by the feature parameter extraction method need to adapt to the actual model of pattern recognition. Summary of the invention

[0009] The purpose of the present invention is to provide a method for extracting characteristic parameters of ultrasonic signals and related devices to solve the problems of inaccurate recognition results, high model construction and training costs.

[0010] To achieve the above object, the present invention adopts the following technical solutions:

[0011] In a first aspect, the present invention provides a method for extracting characteristic parameters of an ultrasonic signal, comprising:

[0012] Acquire the ultrasonic signal caused by partial discharge, obtain the phase-resolved partial discharge PRPD spectrum and time domain diagram of the partial discharge signal, and obtain the two-dimensional spectrum of the discharge parameters according to the phase-resolved partial discharge PRPD spectrum of the partial discharge signal;

[0013] The time domain diagram of partial discharge signal is processed by wavelet transform to obtain different types of partial discharge time-frequency spectrum diagrams;

[0014] The time-frequency spectrum is grayed to obtain a grayscale image, and the statistical features of the two-dimensional spectrum, the moment features and texture features of the grayscale image, and the time-frequency features of the discharge pulse are extracted.

[0015] The extracted features are subjected to dimensionality reduction processing and are used for partial discharge detection of high-voltage power equipment.

[0016] Furthermore, a two-dimensional spectrum of discharge parameters is obtained based on the phase-resolved partial discharge PRPD spectrum of the partial discharge signal, including:

[0017] The discharge amount of the phase-resolved partial discharge PRPD spectrum was corrected and the phase was divided. The 0-360° phase was continuously and evenly divided into 100 intervals. The number of discharges, average discharge amount and maximum discharge amount in each phase interval were counted, and the results were obtained respectively. spectrum; at the same time, the interval where the discharge amount is located is evenly divided to obtain the number of discharges in each discharge amount interval and obtain the qn spectrum;

[0018] in: is the phase, n is the number of discharges, q av is the average discharge capacity, q is the discharge capacity, q max is the maximum discharge capacity.

[0019] Furthermore, the time domain diagram of the partial discharge signal is processed by wavelet transform to obtain different types of partial discharge time-frequency spectrum diagrams, including:

[0020] The expression of wavelet transform is:

[0021]

[0022] Among them, f(t) is the original signal; is the basis function of wavelet transform, α and τ are the scale factor and displacement respectively.

[0023] Furthermore, the statistical features of the two-dimensional spectrum are extracted, including:

[0024] First, using the two-dimensional spectrum, calculate the skewness SK, steepness Ku, discharge factor Q0, mutual correlation coefficient cc and modified mutual correlation coefficient mcc:

[0025] The skewness SK is used to describe the degree of skewness of the spectrum shape compared to the shape of the Gaussian curve and is defined as:

[0026]

[0027] Steepness is used to describe the degree to which the shape of the spectrum is protruding compared to the shape of the Gaussian curve and is defined as:

[0028]

[0029] In the above two formulas: Δx is the phase window width; W is the number of phase windows in the power frequency half-cycle spectrum; x i is the phase of the ith phase window in the power frequency half-cycle spectrum; p i , μ, σ are the ordinate probability, mean and standard deviation in the i-th phase window in the spectrum, respectively;

[0030] The discharge factor Q0 is used to characterize the difference in discharge between the positive and negative half cycles of the spectrum and is defined as:

[0031]

[0032] The mutual correlation coefficient cc reflects the correlation between the positive and negative half-cycle patterns and is defined as:

[0033]

[0034] The corrected correlation coefficient describes The difference between the positive and negative half-cycle discharge modes of the spectrum is defined as:

[0035] mcc=Q0·cc

[0036] In the above three formulas: are the discharge amounts in the i-th phase window in the positive and negative half-cycle spectra, respectively. are the number of discharges in the i-th phase window in the positive and negative half-cycle spectrum, respectively.

[0037] Furthermore, the moment feature extraction of grayscale images includes:

[0038] Grayscale moment feature parameters come from grayscale images, which are composed of three-dimensional The grayscale image is obtained by spectral transformation. The spectrum is a coordinate plane, and the phase The phase window of the discharge amount q is divided into equal grids, and the number of discharges n of each grid is counted to form a three-dimensional The spectrogram is then converted into a grayscale image; the moment feature is used to describe the shape characteristics of the image, including the grayscale center of gravity and the main axis direction v of the grayscale image:

[0039] The grayscale centroid (x0, y0) is defined as the grayscale Figure 1 The ratio of the first-order origin moment to the zero-order origin moment is given by:

[0040]

[0041] Among them, m pq Represents the p+q-order origin moment of the grayscale image f(x,y), defined as:

[0042]

[0043] The principal axis direction v is defined as the ratio of the two second-order central moments, and the formula is:

[0044]

[0045] Among them, u pq The p+q central moment of the grayscale image f(x,y) is defined as:

[0046] u pq =∑ x ∑ y (x-x0) p (y-y0) q f(x,y);

[0047] Extract texture features of grayscale images, including:

[0048] The texture feature of the grayscale image describes the arrangement information of the grayscale values ​​in the grayscale image; the grayscale co-occurrence matrix GLCM reflects the grayscale information of the image based on direction, adjacent intervals and change amplitude;

[0049] Specifically, select four commonly used spatial angles The average and standard deviation of the energy ASM, entropy value ENT, moment of inertia CON, and correlation COR on the grayscale texture features are used as the extracted grayscale texture features:

[0050] Energy describes the uniformity of grayscale distribution. The smaller the energy value of the grayscale image, the more uneven the grayscale distribution. Conversely, the more uniform the grayscale distribution is. It is defined as:

[0051]

[0052] Where K is the grayscale level of the grayscale image, is the value of the element with coordinate (i, j) in the co-occurrence matrix;

[0053] The entropy value describes the complexity of the texture in the grayscale image. The smaller the entropy value, the simpler the texture of the grayscale image; on the contrary, the more complex it is. It is defined as:

[0054]

[0055] The moment of inertia describes the difference between adjacent textures of a grayscale image. The larger the moment of inertia in the image, the greater the difference between adjacent textures of the image, and the clearer the image will be. It is defined as:

[0056]

[0057] Correlation reflects the similarity of elements in the co-occurrence matrix in the horizontal and vertical directions and is defined as:

[0058]

[0059] In the formula,

[0060] The average level and difference of texture features at each angle are reflected by calculating the average value and standard deviation of various texture features at each angle, which is defined as:

[0061]

[0062] In the formula, are various texture features obtained at four angles; m = 4, are the four selected angles; Mean and SD are the mean and standard deviation of the texture feature parameters at the above angles respectively.

[0063] Furthermore, the time-frequency characteristics of the discharge pulse are extracted, including:

[0064] The time-frequency characteristics include the rise time t of the time domain pulse r , Fall time t d , 50% amplitude duration t 50％ , 10% amplitude duration t 10％ , total pulse duration t total , maximum pulse amplitude U max , equivalent time T j , Secondary Equivalent Time Width The skewness SK of the frequency domain waveform f , steepness Ku f , equivalent bandwidth F j and quadratic equivalent bandwidth

[0065] First, the acquired time domain pulses are processed as follows:

[0066]

[0067] In the formula, j represents the jth pulse; m represents that this pulse is composed of m points; b i represents the time domain pulse value of the i-th point; Δt is the sampling interval; Δt(i-1) represents the time corresponding to the i-th point; Fourier transform of the above signal is:

[0068]

[0069] In the formula, B i represents the amplitude of the i-th point in the frequency domain waveform; Δf(i-1) is the frequency value of the i-th point;

[0070] Secondly, the expected time value of a single pulse signal and the expected frequency For calculation, the first-order central moment is usually used to represent the frequency expectation and time expectation of the signal, which is defined as follows:

[0071]

[0072] Finally, the equivalent duration of a single pulse signal in the time domain and the equivalent bandwidth in the frequency domain are calculated as follows:

[0073]

[0074] When characterizing the characteristics of the discharge pulse waveform, a high-order statistic is introduced to calculate its K-order cumulant, which is defined as follows:

[0075]

[0076] Get the second equivalent time width and quadratic equivalent bandwidth

[0077] Furthermore, after feature extraction, the principal component analysis algorithm is used to reduce the dimension of the features:

[0078] Organize the extracted features into a sample dataset Where n represents the number of samples in the data set, that is, the number of samples of ultrasonic signals caused by partial discharge; is the i-th sample in the data, and the feature dimension of the sample is m; the data set is centralized so that the sample mean is 0, that is,

[0079] Use the principal component analysis algorithm to reduce the dimensionality of the sample data set:

[0080] First, the sample data is centralized, that is, let the sample x i Subtract the empirical mean of the sample According to the above steps, all samples in the sample data set are centralized to obtain the centralized sample data set X c ;

[0081] Next, calculate the covariance matrix The covariance matrix S t Perform eigenvalue decomposition to obtain the n eigenvalues ​​of the covariance matrix: λ1,λ2,...,λ n (λ1>λ2>...>λ n ), and the n eigenvectors q1,q2,...,q corresponding to the eigenvalues n ;

[0082] Again, according to the target dimension d of the dimensionality reduction algorithm, take the first d largest eigenvalues ​​λ1,λ2,...,λ d (λ1>λ2>...>λ d ) corresponding to the eigenvectors q1,q2,...,q d , and normalize the eigenvectors to get c1,c2,...,c d ;

[0083] Finally, the projection matrix W = [c1, c2, ..., c d ], the original data set is processed by dimension reduction to obtain the data Y=WTX after dimension reduction, and a recognition model is constructed based on the data after dimension reduction, which is used for the detection of partial discharge phenomenon of high-voltage power equipment.

[0084] In a second aspect, the present invention provides a system for extracting characteristic parameters of an ultrasonic signal, comprising:

[0085] A data acquisition module is used to acquire the ultrasonic signal caused by partial discharge, obtain the phase-resolved partial discharge PRPD spectrum and time domain diagram of the partial discharge signal, and obtain the two-dimensional spectrum of the discharge parameters according to the phase-resolved partial discharge PRPD spectrum of the partial discharge signal;

[0086] The time-frequency spectrum acquisition module is used to process the time-domain spectrum of the partial discharge signal by wavelet transform to obtain different types of partial discharge time-frequency spectrum;

[0087] The feature extraction module is used to grayscale the time-frequency spectrum to obtain a grayscale image, extract the statistical features of the two-dimensional spectrum, the moment features and texture features of the grayscale image, and the time-frequency features of the discharge pulse, and perform dimensionality reduction processing on the extracted features for the detection of partial discharge phenomena in high-voltage power equipment.

[0088] In a third aspect, the present invention provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of a method for extracting characteristic parameters of an ultrasonic signal when executing the computer program.

[0089] In a fourth aspect, the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of a method for extracting characteristic parameters of an ultrasonic signal are implemented.

[0090] Compared with the prior art, the present invention has the following technical effects:

[0091] The present invention can provide multi-dimensional complementary and relatively complete characteristic parameters for input of partial discharge pattern recognition, reduce high-dimensional errors introduced in the training process, and improve training efficiency.

[0092] The present invention introduces wavelet transform as a method for processing time domain signals, inherits and develops the localization idea of ​​short-time Fourier transform, and effectively overcomes the window problem brought by short-time Fourier transform, and can provide a "time-frequency" window that changes with frequency. At the same time, for time-varying non-stationary partial discharge pulse signals, wavelet transform can better meet the frequency requirements of signal changes and can better reflect the time-frequency characteristics of partial discharge signals.

[0093] Furthermore, the processing of partial discharge signals by wavelet transform can better reduce the impact of environmental noise and other interference signals on the final identification, enhance the algorithm's ability to characterize the characteristic parameters of partial discharge signals, and facilitate the identification of partial discharge patterns.

[0094] Furthermore, the present invention integrates the characteristics of various partial discharge signals and can provide multi-dimensional complementary and relatively complete characteristic parameters for input of partial discharge pattern recognition, which can effectively solve the problem of reduced recognition rate due to signal distortion.

[0095] Furthermore, the present invention introduces a principal component analysis algorithm to reduce the dimensionality of high-dimensional information extracted from ultrasonic information, effectively reducing the "dimensionality disaster" that may be caused by high-dimensional input parameters in the partial discharge recognition model. The principal component analysis algorithm maps the originally complex data set to a lower-dimensional subspace by retaining the most important information in the data, while maintaining most of the variation of the data set. In this way, not only can the computational efficiency of the recognition model be improved and the risk of overfitting be reduced, but the cost of model construction and training can also be significantly reduced, making the recognition of partial discharge patterns more accurate and reliable.

[0096] Furthermore, the dimensionality reduction processing of the principal component analysis algorithm introduced in the present invention can play an important role in improving data visualization capabilities. By reducing the high-dimensional input parameters to a subspace that is easier to understand and interpret, the characteristics and patterns of the data can be more intuitively displayed, helping analysts to understand and evaluate the characteristics of partial discharge more quickly, and providing more powerful support for fault diagnosis and prevention. In addition, dimensionality reduction processing can also help discover potential characteristic variables and associations, and further optimize the design and performance of pattern recognition algorithms.

[0097] Furthermore, the principal component analysis algorithm introduced in the present invention can bring significant improvements in data processing and transmission. The reduction of data dimensions not only reduces the storage and transmission costs, but also improves the efficiency and speed of data processing, providing a better solution for large-scale data processing. This is particularly important for real-time partial discharge detection systems, which can help the system to more quickly realize data collection, processing and identification, and improve the practicality and reliability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0098] Figure 1 It is a flow chart of the present invention.

[0099] Figure 2 This is a system structure diagram of the present invention. DETAILED DESCRIPTION

[0100] The present invention is further described below in conjunction with the accompanying drawings:

[0101] Example 1, please refer to Figure 1 , a method for extracting characteristic parameters of an ultrasonic signal, comprising:

[0102] Acquire the ultrasonic signal caused by partial discharge, obtain the phase-resolved partial discharge PRPD spectrum and time domain diagram of the partial discharge signal, and obtain the two-dimensional spectrum of the discharge parameters according to the phase-resolved partial discharge PRPD spectrum of the partial discharge signal;

[0103] The time domain diagram of partial discharge signal is processed by wavelet transform to obtain different types of partial discharge time-frequency spectrum diagrams;

[0104] The time-frequency spectrum is grayed to obtain a grayscale image, and the statistical features of the two-dimensional spectrum, the moment features and texture features of the grayscale image, and the time-frequency features of the discharge pulse are extracted.

[0105] The present invention can provide multi-dimensional complementary and relatively complete characteristic parameters for input of partial discharge pattern recognition, reduce high-dimensional errors introduced in the training process, and improve training efficiency.

[0106] Embodiment 2, the present invention provides a method for extracting characteristic parameters of an ultrasonic signal, comprising:

[0107] Step 1: Calculate the PRPD spectrum of the partial discharge signal Positive and negative half-week two-dimensional spectrum;

[0108] Specifically, the discharge amount of the PRPD spectrum was corrected, and the phase was divided. The 0-360° phase was continuously and evenly divided into 100 intervals. The number of discharges, average discharge amount and maximum discharge amount in each phase interval were counted, and the results were obtained respectively. At the same time, the interval where the discharge amount is located is evenly divided to obtain the number of discharges in each discharge amount interval and obtain the qn spectrum.

[0109] Step 2, applying wavelet transform to the time domain diagram of the partial discharge signal to obtain different types of partial discharge time-frequency spectrum diagrams;

[0110] Specifically, the expression of the wavelet transform is:

[0111]

[0112] Among them, f(t) is the original signal; is the basis function of wavelet transform, α and τ are the scale factor and displacement respectively.

[0113] Step 3, graying the time-frequency spectrum of the discharge model obtained by wavelet transform to obtain a grayscale image, and extracting the statistical features of the two-dimensional spectrum, the moment features and texture features of the grayscale image, and the time-frequency features of the discharge pulse.

[0114] Specifically, firstly, the skewness SK, steepness Ku, discharge factor Q0, cross-correlation coefficient cc and modified cross-correlation coefficient mcc are calculated using the two-dimensional spectrum. The skewness SK is used to describe the degree of deviation of the spectrum shape compared with the shape of the Gaussian curve, and is defined as:

[0115]

[0116] The steepness is used to describe the degree of protrusion of the spectrum shape compared to the shape of the Gaussian curve and is defined as:

[0117]

[0118] Where: Δx is the phase window width; W is the number of phase windows in the power frequency half-cycle spectrum; x i is the phase of the ith phase window in the power frequency half-cycle spectrum; p i , μ, and σ are the ordinate probability, mean, and standard deviation in the i-th phase window in the spectrum, respectively.

[0119] The discharge factor Q0 is used to characterize the difference in discharge between the positive and negative half cycles of the spectrum and is defined as:

[0120]

[0121] The mutual correlation coefficient cc reflects the correlation degree of the positive and negative half-cycle spectrum shapes and is defined as:

[0122]

[0123] The corrected correlation coefficient describes The difference between the positive and negative half-cycle discharge modes of the spectrum is defined as:

[0124] mcc=Q0·cc

[0125] Where: are the discharge amounts in the i-th phase window in the positive and negative half-cycle spectra, respectively. are the number of discharges in the i-th phase window in the positive and negative half-cycle spectrum, respectively.

[0126] Preferably, the local discharge signal is extracted 22-dimensional statistical parameter feature quantities such as the skewness of the positive and negative half-cycle spectra

[0127] Furthermore, the time-frequency spectrum grayscale image of the partial discharge signal is obtained by gray-scaling the time-frequency spectrum. This conversion method is a well-known technology and the specific process will not be described in detail.

[0128] Specifically, the grayscale centroid (x0, y0) is defined as Figure 1 The ratio of the first-order origin moment to the zero-order origin moment is given by:

[0129]

[0130] Among them, m pq Represents the p+q-order origin moment of the grayscale image f(x,y), defined as:

[0131]

[0132] Furthermore, the above main axis direction v is defined as the ratio of two second-order central moments, and the formula is:

[0133]

[0134] Among them, u pq The p+q central moment of the grayscale image f(x,y) is defined as:

[0135]

[0136] Preferably, the above moment feature formula is used to extract a total of 9 moment feature parameters of the positive and negative half-cycle grayscale images.

[0137] Furthermore, four common linear uncorrelated features in the gray-level co-occurrence matrix are selected, namely energy (ASM), entropy (ENT), moment of inertia (CON), and correlation (COR).

[0138] The energy describes the uniformity of grayscale distribution. The smaller the energy value of the grayscale image, the more uneven the grayscale distribution, and vice versa, the more uniform the grayscale distribution. This parameter is defined as:

[0139]

[0140] Where K is the grayscale level of the grayscale image, is the value of the element with coordinate (i,j) in the co-occurrence matrix.

[0141] The entropy value describes the complexity of the texture in the grayscale image. The smaller the entropy value, the simpler the texture of the grayscale image; conversely, the more complex it is. This parameter is defined as:

[0142]

[0143] The moment of inertia describes the difference between adjacent textures in a grayscale image. The larger the moment of inertia in an image, the greater the difference between adjacent textures in the image, and the clearer the image will be. This parameter is defined as:

[0144]

[0145] The correlation reflects the similarity of the elements in the co-occurrence matrix in the horizontal and vertical directions, which is defined as:

[0146]

[0147] In the formula,

[0148] The average level and difference of texture features at each angle are reflected by calculating the average value and standard deviation of the above-mentioned texture features at each angle, which is defined as:

[0149]

[0150] In the formula, are various texture features obtained at four angles; m = 4, Mean and SD are the average and standard deviation of the texture feature parameters at the above angles, respectively.

[0151] Preferably, the average values ​​and standard deviations of the energy, entropy, moment of inertia and correlation of the grayscale images of the four types of discharges in the above four directions are extracted as texture features, and the texture features have a total of 8 dimensions.

[0152] Furthermore, the rise time t of the time domain pulse r , Fall time t d , 50% amplitude duration t 50％ , 10% amplitude duration t 10％ , total pulse duration t total , maximum pulse amplitude U max , equivalent time T j , Secondary Equivalent Time Width The skewness SK of the frequency domain waveform f , steepness Ku f , equivalent bandwidth F j and quadratic equivalent bandwidth Extract.

[0153] For the rise time of the time domain pulse, t r , Fall time t d , 50% amplitude duration t 50％ , 10% amplitude duration t 10％ , total pulse duration t total , maximum pulse amplitude U max The definitions of these parameters are well-known techniques, and the remaining parameters are described in detail.

[0154] Because when processing pulse signal technology, the time center, width and frequency center, width of the signal are particularly important. They can not only characterize the central position of the signal in the time domain and frequency domain, but also represent the extended information of the signal in the time and frequency domain. Therefore, in order to facilitate the extraction of features, the present invention processes the signal as follows:

[0155] First, the acquired time domain pulses are processed as follows:

[0156]

[0157] In the formula, j represents the jth pulse; m represents that this pulse is composed of m points; b i represents the time domain pulse value of the i-th point; Δt is the sampling interval; Δt(i-1) represents the time corresponding to the i-th point. Perform Fourier transform on the above signal to obtain:

[0158]

[0159] In the formula, B i Represents the amplitude of the i-th point in the frequency domain waveform; Δf(i-1) is the frequency value of the i-th point

[0160] Secondly, the expected time value of a single pulse signal and the expected frequency For calculation, the first-order central moment is usually used to represent the frequency expectation and time expectation of the signal, which is defined as follows:

[0161]

[0162] Finally, the time standard deviation (equivalent duration) of a single pulse signal in the time domain and the frequency standard deviation (equivalent bandwidth) in the frequency domain are calculated as follows:

[0163]

[0164] When characterizing the characteristics of the discharge pulse waveform, high-order statistics can be introduced to calculate its K-order cumulative amount. This idea can be used to calculate high-order statistics of equivalent duration and equivalent bandwidth, which are defined as follows:

[0165]

[0166] Preferably, a total of 12-dimensional feature sequences are extracted through time-frequency domain pulse analysis.

[0167] Furthermore, the principal component analysis algorithm used in the present invention is shown below, which is mainly divided into two parts: normalization and dimensionality reduction.

[0168] First, the sample data is centralized, that is, let the sample x i Subtract the empirical mean of the sample According to the above steps, all samples in the sample data set are centralized to obtain the centralized sample data set X c ;

[0169] Next, calculate the covariance matrix The covariance matrix S t Perform eigenvalue decomposition to obtain the n eigenvalues ​​of the covariance matrix: λ1,λ2,...,λ n (λ1>λ2>...>λ n ), and the n eigenvectors q1,q2,...,q corresponding to the eigenvalues n ;

[0170] Again, according to the target dimension d of the dimensionality reduction algorithm, take the first d largest eigenvalues ​​λ1,λ2,...,λ d (λ1>λ2>...>λ d ) corresponding to the eigenvectors q1,q2,...,q d , and normalize the eigenvectors to get c1,c2,...,c d ;

[0171] Finally, the projection matrix W = [c1, c2, ..., cd ], the original data set is processed by dimension reduction, and the data after dimension reduction is Y=WTX.

[0172] The present invention organizes various information extracted from the ultrasonic signal caused by partial discharge into a sample data set Where n represents the number of samples in the data set, that is, the number of samples of ultrasonic signals caused by partial discharge; is the i-th sample in the data, and the feature dimension of the sample is m. In order to maintain generality, the data set is centered so that the sample mean is 0, that is,

[0173] Furthermore, the present invention uses the principal component analysis algorithm to perform dimensionality reduction processing on the sample data set. The principal component analysis projects the original high-dimensional data into a low-dimensional space, and makes the projected data best represent the key information of the original data. The data features obtained after the principal component analysis algorithm is used to reduce the data dimension is a linear combination of the original data features, and its coefficients come from the projection matrix.

[0174] Specifically, the principal component analysis algorithm finds the optimal projection matrix by solving the following objective function:

[0175]

[0176] Among them, Ι represents a d-dimensional identity matrix, represents the square of the l2 norm, and also represents the distance measure of the reconstruction error. The above formula defines the optimal projection matrix by minimizing the angle of the reconstruction error. In fact, W T x i Represents the sample after projection, WW T x i To reconstruct the sample, when the sum of the reconstruction errors of all samples is the smallest, it means that the low-dimensional projection generated by this method can better restore the original sample, and the projected information can greatly represent the information contained in the original high-dimensional sample, thus achieving the goal of data dimensionality reduction.

[0177] Furthermore, the optimal projection matrix required by the principal component analysis algorithm can also be obtained by maximizing the contrast angle between the samples after projection, that is, it is composed of the eigenvectors of the covariance matrix of the sample data, as shown in the following formula.

[0178]

[0179] in, is the sample x i The covariance matrix of , tr(·) represents the trace operation of the matrix. The above formula shows that the principal component analysis algorithm makes the samples more separable in the projection space by maximizing the contrast angle between the sample data after projection.

[0180] In fact, the above two equations are equivalent, because and is a constant, so the first term on the left side of the minimization equation is equivalent to the second term on the left side of the maximization equation, that is, the two equations are equivalent.

[0181] Preferably, for the problem raised by the above two equivalent equations, the present invention uses the sample covariance matrix S t The projection matrix is ​​constructed by using the eigenvectors corresponding to the first d largest eigenvalues.

[0182] At this point, the dimensionality-reduced data set of the multi-dimensional information extracted by the present invention is obtained, and the overall information is guaranteed to be complete as much as possible under the premise of ensuring the multi-dimensional complementarity of the information. Researchers in partial discharge pattern recognition can build a recognition model based on this data set for partial discharge phenomenon detection of high-voltage power equipment to ensure the safe operation of high-voltage power equipment.

[0183] In yet another embodiment of the present invention, a system for extracting characteristic parameters of an ultrasonic signal is provided, which can be used to implement the above-mentioned method for extracting characteristic parameters of an ultrasonic signal. Specifically, the system includes:

[0184] A data acquisition module is used to acquire the ultrasonic signal caused by partial discharge, obtain the phase-resolved partial discharge PRPD spectrum and time domain diagram of the partial discharge signal, and obtain the two-dimensional spectrum of the discharge parameters according to the phase-resolved partial discharge PRPD spectrum of the partial discharge signal;

[0185] The time-frequency spectrum acquisition module is used to process the time-domain spectrum of the partial discharge signal by wavelet transform to obtain different types of partial discharge time-frequency spectrum;

[0186] The feature extraction module is used to grayscale the time-frequency spectrum to obtain a grayscale image, and to extract the statistical features of the two-dimensional spectrum, the moment features and texture features of the grayscale image, and the time-frequency features of the discharge pulse.

[0187] The division of modules in the embodiments of the present invention is schematic and is only a logical function division. There may be other division methods in actual implementation. In addition, each functional module in each embodiment of the present invention may be integrated into one processor, or may exist physically separately, or two or more modules may be integrated into one module. The above-mentioned integrated modules may be implemented in the form of hardware or in the form of software functional modules.

[0188] In another embodiment of the present invention, a computer device is provided, the computer device includes a processor and a memory, the memory is used to store a computer program, the computer program includes program instructions, and the processor is used to execute the program instructions stored in the computer storage medium. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, which is suitable for implementing one or more instructions, and is specifically suitable for loading and executing one or more instructions in a computer storage medium to implement a corresponding method flow or a corresponding function; the processor described in the embodiment of the present invention can be used for the operation of a method for extracting characteristic parameters of an ultrasonic signal.

[0189] In another embodiment of the present invention, the present invention also provides a storage medium, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device for storing programs and data. It is understandable that the computer-readable storage medium here can include both built-in storage media in the computer device and, of course, extended storage media supported by the computer device. The computer-readable storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space, and these instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk memory. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the characteristic parameter extraction method of an ultrasonic signal in the above embodiment.

[0190] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0191] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0192] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0193] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0194] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for extracting characteristic parameters of an ultrasonic signal, characterized in that: include: Acquire the ultrasonic signal caused by partial discharge, obtain the phase-resolved partial discharge PRPD spectrum and time domain diagram of the partial discharge signal, and obtain the two-dimensional spectrum of the discharge parameters according to the phase-resolved partial discharge PRPD spectrum of the partial discharge signal; The time domain diagram of partial discharge signal is processed by wavelet transform to obtain different types of partial discharge time-frequency spectrum diagrams; The time-frequency spectrum is grayed out to obtain a grayscale image, and the statistical features of the two-dimensional spectrum, the moment features and texture features of the grayscale image, and the time-frequency features of the discharge pulse are extracted; The extracted features are subjected to dimensionality reduction processing and are used for partial discharge detection of high-voltage power equipment.

2. The method for extracting characteristic parameters of an ultrasonic signal according to claim 1, characterized in that: The two-dimensional spectrum of discharge parameters is obtained based on the phase-resolved partial discharge PRPD spectrum of the partial discharge signal, including: The discharge amount of the phase-resolved partial discharge PRPD spectrum was corrected and the phase was divided. The 0-360° phase was continuously and evenly divided into 100 intervals. The number of discharges, average discharge amount and maximum discharge amount in each phase interval were counted, and the results were obtained respectively. spectrum; at the same time, the interval where the discharge amount is located is evenly divided to obtain the number of discharges in each discharge amount interval and obtain the qn spectrum; in: is the phase, n is the number of discharges, q av is the average discharge capacity, q is the discharge capacity, q max is the maximum discharge capacity.

3. The method for extracting characteristic parameters of an ultrasonic signal according to claim 1, characterized in that: The time domain diagram of the partial discharge signal is processed by wavelet transform to obtain different types of partial discharge time-frequency spectrum diagrams, including: The expression of wavelet transform is: Among them, f(t) is the original signal; is the basis function of wavelet transform, α and τ are the scale factor and displacement respectively.

4. The method for extracting characteristic parameters of an ultrasonic signal according to claim 1, characterized in that: Extract the statistical features of the two-dimensional spectrum, including: First, using the two-dimensional spectrum, calculate the skewness SK, steepness Ku, discharge factor Q0, mutual correlation coefficient cc and modified mutual correlation coefficient mcc: The skewness SK is used to describe the degree of skewness of the spectrum shape compared to the shape of the Gaussian curve and is defined as: Steepness is used to describe the degree to which the shape of the spectrum is protruding compared to the shape of the Gaussian curve and is defined as: In the above two formulas: Δx is the phase window width; W is the number of phase windows in the power frequency half-cycle spectrum; x i is the phase of the ith phase window in the power frequency half-cycle spectrum; p i , μ, σ are the ordinate probability, mean and standard deviation in the i-th phase window in the spectrum, respectively; The discharge factor Q0 is used to characterize the difference in discharge between the positive and negative half cycles of the spectrum and is defined as: The mutual correlation coefficient cc reflects the correlation between the positive and negative half-cycle patterns and is defined as: The corrected correlation coefficient describes The difference between the positive and negative half-cycle discharge modes of the spectrum is defined as: mcc=Q0·cc In the above three formulas: are the discharge amounts in the i-th phase window in the positive and negative half-cycle spectra, respectively. are the number of discharges in the i-th phase window in the positive and negative half-cycle spectrum, respectively.

5. The method for extracting characteristic parameters of an ultrasonic signal according to claim 1, characterized in that: Moment feature extraction of grayscale images, including: Grayscale moment feature parameters come from grayscale images, which are composed of three-dimensional The grayscale image is obtained by spectral transformation. The spectrum is a coordinate plane, and the phase The phase window of the discharge amount q is divided into equal grids, and the number of discharges n of each grid is counted to form a three-dimensional The spectrogram is then converted into a grayscale image; the moment feature is used to describe the shape characteristics of the image, including the grayscale center of gravity and the main axis direction v of the grayscale image: The grayscale centroid (x0, y0) is defined as the ratio of the first-order origin moment of the grayscale image to the zero-order origin moment. The formula is: Among them, m pq Represents the p+q-order origin moment of the grayscale image f(x,y), defined as: The principal axis direction v is defined as the ratio of the two second-order central moments, and the formula is: Among them, u pq The p+q central moment of the grayscale image f(x,y) is defined as: u pq =∑ x ∑ y (x-x0) p (y-y0) q f(x,y); Extract texture features of grayscale images, including: The texture feature of the grayscale image describes the arrangement information of the grayscale values ​​in the grayscale image; the grayscale co-occurrence matrix GLCM reflects the grayscale information of the image based on direction, adjacent intervals and change amplitude; Specifically, select four commonly used spatial angles The average and standard deviation of the energy ASM, entropy value ENT, moment of inertia CON, and correlation COR on the grayscale texture features are used as the extracted grayscale texture features: Energy describes the uniformity of grayscale distribution. The smaller the energy value of the grayscale image, the more uneven the grayscale distribution. Conversely, the more uniform the grayscale distribution is. It is defined as: Where K is the grayscale level of the grayscale image, is the value of the element with coordinate (i, j) in the co-occurrence matrix; The entropy value describes the complexity of the texture in the grayscale image. The smaller the entropy value, the simpler the texture of the grayscale image; on the contrary, the more complex it is. It is defined as: The moment of inertia describes the difference between adjacent textures of a grayscale image. The larger the moment of inertia in the image, the greater the difference between adjacent textures of the image, and the clearer the image will be. It is defined as: Correlation reflects the similarity of elements in the co-occurrence matrix in the horizontal and vertical directions and is defined as: In the formula, The average level and difference of texture features at each angle are reflected by calculating the average value and standard deviation of various texture features at each angle, which is defined as: In the formula, are various texture features obtained at four angles; m = 4, are the four selected angles; Mean and SD are the mean and standard deviation of the texture feature parameters at the above angles respectively.

6. The method for extracting characteristic parameters of an ultrasonic signal according to claim 1, characterized in that: Extract the time-frequency characteristics of the discharge pulse, including: The time-frequency characteristics include the rise time t of the time domain pulse r , Fall time t d , 50% amplitude duration t 50％ , 10% amplitude duration t 10％ , total pulse duration t total , maximum pulse amplitude U max , equivalent time T j , Secondary Equivalent Time Width The skewness SK of the frequency domain waveform f , steepness Ku f , equivalent bandwidth F j and quadratic equivalent bandwidth First, the acquired time domain pulses are processed as follows: In the formula, j represents the jth pulse; m represents that this pulse is composed of m points; b i represents the time domain pulse value of the i-th point; Δt is the sampling interval; Δt(i-1) represents the time corresponding to the i-th point; Fourier transform of the above signal is: In the formula, B i represents the amplitude of the i-th point in the frequency domain waveform; Δf(i-1) is the frequency value of the i-th point; Secondly, the expected time value of a single pulse signal and the expected frequency For calculation, the first-order central moment is usually used to represent the frequency expectation and time expectation of the signal, which is defined as follows: Finally, the equivalent duration of a single pulse signal in the time domain and the equivalent bandwidth in the frequency domain are calculated as follows: When characterizing the characteristics of the discharge pulse waveform, a high-order statistic is introduced to calculate its K-order cumulant, which is defined as follows: Get the second equivalent time width and quadratic equivalent bandwidth 7. The method for extracting characteristic parameters of an ultrasonic signal according to claim 1, characterized in that: After feature extraction, the principal component analysis algorithm is used to reduce the dimension of the features: Organize the extracted features into a sample dataset Where n represents the number of samples in the data set, that is, the number of samples of ultrasonic signals caused by partial discharge; is the i-th sample in the data, and the feature dimension of the sample is m; the data set is centralized so that the sample mean is 0, that is, Use the principal component analysis algorithm to reduce the dimensionality of the sample data set: First, the sample data is centralized, that is, let the sample x i Subtract the empirical mean of the sample According to the above steps, all samples in the sample data set are centralized to obtain the centralized sample data set X c ; Next, calculate the covariance matrix The covariance matrix S t Perform eigenvalue decomposition to obtain the n eigenvalues ​​of the covariance matrix: λ1,λ2,...,λ n (λ1>λ2>...>λ n ), and the n eigenvectors q1,q2,...,q corresponding to the eigenvalues n ; Again, according to the target dimension d of the dimensionality reduction algorithm, take the first d largest eigenvalues ​​λ1,λ2,...,λ d (λ1>λ2>...>λ d ) corresponding to the eigenvectors q1,q2,...,q d , and normalize the eigenvectors to get c1,c2,...,c d ; Finally, the projection matrix W = [c1, c2, ..., c d ], the original data set is processed by dimension reduction, and the data after dimension reduction is Y=W T X, builds a recognition model based on the reduced-dimensional data for the detection of partial discharge phenomena in high-voltage power equipment.

8. A characteristic parameter extraction system for ultrasonic signals, characterized in that: include: A data acquisition module is used to acquire the ultrasonic signal caused by partial discharge, obtain the phase-resolved partial discharge PRPD spectrum and time domain diagram of the partial discharge signal, and obtain the two-dimensional spectrum of the discharge parameters according to the phase-resolved partial discharge PRPD spectrum of the partial discharge signal; The time-frequency spectrum acquisition module is used to process the time-domain spectrum of the partial discharge signal by wavelet transform to obtain different types of partial discharge time-frequency spectrum; The feature extraction module is used to grayscale the time-frequency spectrum to obtain a grayscale image, extract the statistical features of the two-dimensional spectrum, the moment features and texture features of the grayscale image, and the time-frequency features of the discharge pulse, and perform dimensionality reduction processing on the extracted features for the detection of partial discharge phenomena in high-voltage power equipment.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method for extracting characteristic parameters of an ultrasonic signal as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of a method for extracting characteristic parameters of an ultrasonic signal as claimed in any one of claims 1 to 7 are implemented.

Images