Current signal compression acquisition and reconstruction method considering faulty arc identification and related equipment

Abstract

The present invention relates to the technical field of current signal processing, and in particular to a method, device, storage medium and electronic device for compressing, collecting and reconstructing current signals. The method includes: performing sliding acquisition on the current signal using a flexible Blackman window to obtain a sliding acquisition signal; constructing a partial orthogonal Gaussian measurement matrix; performing compressed acquisition on the sliding acquisition signal using the partial orthogonal Gaussian measurement matrix to obtain a compressed signal; performing sparse transformation on the sliding acquisition signal to solve for a frequency-domain transformation matrix; calculating a sensing matrix using the frequency-domain transformation matrix and the partial orthogonal Gaussian measurement matrix, and then iteratively solving for a sparse signal according to the compressed signal and the sensing matrix using an algorithm; calculating reconstructed data according to the frequency-domain transformation matrix and the sparse signal. The compressed acquisition and reconstruction of the present invention realizes low-volume data acquisition and effectively retains the high-frequency fault arc characteristics in the data, solving the problem of missing fault arc characteristics after traditional current signal acquisition.

Description

Technical Field

[0001] The present invention relates to the technical field of current signal processing, and in particular, to a method, device, storage medium, and electronic device for compressed acquisition and reconstruction of current signals considering fault arc identification. Background Art

[0002] The statements in this part only provide background technical information related to the present invention, and do not necessarily constitute prior art.

[0003] In modern power systems, the arc phenomenon is a common and complex electrical fault form, widely existing in various power facilities such as high-voltage equipment, substations, and transmission lines. The arc can not only cause equipment damage and system power outages, but also trigger fires, resulting in casualties and significant economic losses. Therefore, real-time monitoring and analysis of multi-source current signals have important practical significance.

[0004] Traditional fault arc identification methods usually rely on the monitoring of non-electrical quantities such as light and gas, which require a large number of monitoring devices, are complex to install, and have high costs. Identifying fault arcs by monitoring electrical parameters such as voltage, current, and power does not require additional monitoring devices, reducing the hardware cost. However, since the current signal presents high-frequency characteristics when a fault arc appears, and traditional signal monitoring sampling needs to satisfy the Nyquist-Shannon theorem, it leads to huge costs for sampling, transmission, and processing of high-frequency signals, and places higher requirements on the storage space and computing power of on-site monitoring devices. Therefore, how to efficiently and accurately collect and process multi-source current signals to ensure the reliability of fault arc identification has become an urgent problem to be solved. Summary of the Invention

[0005] To overcome the deficiencies of the background art, the present invention provides a method, device, storage medium, and electronic device for compressed acquisition and reconstruction of current signals considering fault arc identification. By compressing, acquiring, and reconstructing high-frequency current signals that can characterize the characteristics of fault arcs, low-volume data acquisition is achieved, and high-frequency fault arc characteristics in the data are effectively retained, solving the problem of missing fault arc characteristics after traditional current signal acquisition.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] In a first aspect, a method for compressed acquisition and reconstruction of current signals is provided, and the method includes:

[0008] Using a flexible Blackman window to perform sliding acquisition on the current signal to obtain a sliding acquisition signal ;

[0009] Constructing a partial orthogonal Gaussian measurement matrix ;

[0010] Using the partial orthogonal Gaussian measurement matrix to perform compressive acquisition on the sliding acquisition signal to obtain a compressed signal after compressive acquisition ;

[0011] Performing sparse transformation on the sliding acquisition signal to solve and obtain a frequency-domain transformation matrix ;

[0012] Using the frequency-domain transformation matrix and the partial orthogonal Gaussian measurement matrix to calculate and obtain a sensing matrix D, and then iteratively solving for a sparse signal according to the compressed signal and the sensing matrix D using an algorithm ;

[0013] Calculating and obtaining reconstructed data according to the frequency-domain transformation matrix and the sparse signal .

[0014] Furthermore,

[0015] The functional expression of the flexible Blackman window is:

[0016] (1)

[0017] In formula (1), t is the acquisition time of the current signal, T is the adjustable window width.

[0018] Furthermore,

[0019] Using the flexible Blackman window to perform sliding acquisition on the current signal to obtain a sliding acquisition signal , specifically as follows:

[0020] (2)

[0021] (3)

[0022] where is the number of discrete data points acquired, is the sampling interval, is the impulse function, .

[0023] Furthermore,

[0024] Adjusting the window width according to the reconstructed signal deviation of adjacent windows of the current signal T ​, the deviation of the reconstructed signal of adjacent windows of the current signal is:

[0025] (4)

[0026] In formula (4), and are the (h - 1)-th and h-th windows respectively;

[0027] When an arc fault occurs, the current harmonic frequency increases, resulting in increasing, then the window width T is reduced;

[0028] When decreases, the window width T is enlarged.

[0029] Furthermore,

[0030] the construction of the partially orthogonal Gaussian measurement matrix includes:

[0031] Based on the standard Gaussian measurement matrix with independent and identically distributed elements, construct the partially orthogonal Gaussian measurement matrix ;

[0032] The random variables in the standard Gaussian measurement matrix

[0033] (5)

[0034] In formula (5), is a Gaussian distribution with a mean of 0 and a variance of .

[0035] Furthermore,

[0036] The standard Gaussian measurement matrix is represented as the following matrix A , with a size of :

[0037] (6)

[0038] Extract the first columns of the matrix to form the matrix :

[0039] (7)

[0040] Orthogonalize the matrix , select the first column as the orthogonal basis vector , the jThe column needs to subtract its projections on all previous orthogonal basis vectors and be normalized:

[0041] (8)

[0042] To obtain the orthogonalized matrix:

[0043] (9)

[0044] Replace the first columns of the matrix with the matrix to obtain the partial orthogonal Gaussian matrix :

[0045] (10).

[0046] Furthermore,

[0047] When extracting the first columns of the matrix to form the matrix , k the ՛ value reference is 0.2× , and the deviation of the reconstructed signal of the current signal adjacent window is dynamically adjusted according to Equation (4) ;

[0048] When increases, decrease the value;

[0049] When decreases, increase the value;

[0050] The value selection range is (0.1~0.3)× .

[0051] Furthermore,

[0052] According to compressive sensing, the calculation formula of the compressed signal is as follows:

[0053] (11).

[0054] Furthermore,

[0055] Use discrete wavelet transform to perform sparse transform on the sliding acquisition signal to obtain the frequency domain transform matrix .

[0056] Furthermore,

[0057] The use of discrete wavelet transform for the sliding acquisition signal Perform a sparse transform to obtain a frequency-domain transform matrix , including:

[0058] Use the Daubechies wavelet basis of the multi-Bessey wavelet to perform the sliding acquisition signal Perform N - 1 decompositions to obtain 1 low-frequency component and N - 1 high-frequency components:

[0059] (12)

[0060] In Equation (12), is the low-frequency component, is the high-frequency component, and are the scaling function and the wavelet function respectively, j , f are the scale number and the translation number respectively;

[0061] Further transform Equation (12), and the sparse signal representation of the sliding acquisition signal under the discrete wavelet transform is:

[0062] (13)

[0063] In Equation (13), is the frequency-domain transform matrix under the discrete wavelet transform.

[0064] Furthermore,

[0065] The calculation formula of the sensing matrix D is:

[0066] (14)

[0067] In Equation (14), is the frequency-domain transform matrix, is the partial orthogonal Gaussian measurement matrix.

[0068] Furthermore,

[0069] According to the compressed signal and the sensing matrix D, use the compressive sampling matching pursuit reconstruction algorithm CoSaMP-ESP based on optimal sparsity estimation to iteratively solve the sparse signal .

[0070] Furthermore,

[0071] The said according to the compressed signal and the sensing matrix D, use the compressive sampling matching pursuit reconstruction algorithm CoSaMP-ESP based on optimal sparsity estimation to iteratively solve the sparse signal , including:

[0072] Input the compressed signal the sensing matrix D, and the error threshold The compressive sampling matching pursuit reconstruction algorithm CoSaMP-ESP based on optimal sparsity estimation includes:

[0073] ①Initialization: Set the index value set E to be an empty set, F to be an empty set, the residual v = , the sparsity K = 5, and the number of iteration rounds d = 0;

[0074] ②d = d + 1, calculate the inner product of the current residual and the sensing matrix D, and select the column of the sensing matrix D that best meets the conditions according to the fuzzy threshold c and store it in the set E:

[0075] (15)

[0076] In formula (15), D j is the j-th column of matrix D, v d-1 is the residual of the (d - 1)-th iteration;

[0077] ③According to the set E, obtain the support set :

[0078] (16)

[0079] ④According to the support set Solve the least squares solution:

[0080] (17)

[0081] In formula (17), , represents the column in the sensing matrix D corresponding to the set , and I is the identity matrix;

[0082] ⑤Prune the support set, select the K terms with the largest absolute values in and store them in z(k), and update the corresponding , then update the sparsity estimation value K:

[0083] (18)

[0084] In formula (18), K is the sparsity, is the support set of the d-th iteration, h(m) is the compressed data, D is the sensing matrix, z(k) is the sparse signal to be solved, and the adjustable p-norm is used to control the sparsity;

[0085] ⑥Update the residual and relevance :

[0086] (19)

[0087] (20)

[0088] Among them, is the residual, is the relevance, representing the column in the sensing matrix D D j with the highest relevance to the current residual;

[0089] ⑦ If , stop the iteration and output the sparse signal , otherwise return to step ② to recalculate the residual.

[0090] Furthermore,

[0091] the value range of the fuzzy threshold c is 0.6 - 0.8.

[0092] Furthermore,

[0093] In formula (18), the p - value benchmark is 1, and the p - value range is 0.3 - 3.

[0094] Furthermore,

[0095] the reconstructed data is calculated by the formula:

[0096] (21)

[0097] In formula (21), is the frequency - domain transformation matrix, is the sparse signal obtained by iterative solution.

[0098] On the second aspect, a current signal compression acquisition and reconstruction device is also provided. The device includes:

[0099] A sliding acquisition module, which is used to perform sliding acquisition on the current signal using a flexible Blackman window to obtain a sliding acquisition signal ;

[0100] A matrix construction module, which is used to construct a partial orthogonal Gaussian measurement matrix ;

[0101] A compression acquisition module, which is used to perform compression acquisition on the sliding acquisition signal using the partial orthogonal Gaussian measurement matrix to obtain a compressed signal ;

[0102] A sparse transformation module for performing sparse transformation on the sliding acquisition signal to obtain a frequency-domain transformation matrix through solution ;

[0103] A calculation and solution module for calculating a sensing matrix D by using the frequency-domain transformation matrix and the partial orthogonal Gaussian measurement matrix , then iteratively solving the sparse signal according to the compressed signal and the sensing matrix D by using an algorithm ;

[0104] A data reconstruction module for calculating reconstructed data according to the frequency-domain transformation matrix and the sparse signal ; .

[0105] Based on the same inventive concept, the present invention also provides a computer-readable storage medium storing one or more programs, which can implement the foregoing method for compressed acquisition and reconstruction of current signals when the one or more programs are executed.

[0106] Based on the same inventive concept, the present invention also provides an electronic device, including a processor, a communication interface, the computer-readable storage medium as described above, and a communication bus; wherein, the processor, the communication interface, and the computer-readable storage medium communicate with each other through the communication bus; the processor is configured to execute the programs stored in the computer-readable storage medium.

[0107] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0108] 1. It can effectively break through the Nyquist high-frequency sampling limitation, realize low-speed dynamic compressed sampling and accurate reconstruction of fault arc high-frequency current signals, retain the effective characteristics of fault arcs in current signals, provide data acquisition technical support for fault arc identification in power grid operation, and have good engineering application prospects;

[0109] 2. Use a flexible Blackman window to perform sliding acquisition on the current signal, with lower sidelobe peaks and faster sidelobe attenuation rates, and higher data spectrum acquisition; use a flexible Blackman window to adjust the window width according to the degree of electrical fluctuation, which can effectively collect the original information of the current signal and retain the high-frequency characteristics of the fault arc;

[0110] 3. Use the CoSaMP-ESP reconstruction algorithm, with fast reconstruction speed and high reconstruction accuracy, which can provide strong technical support for fault arc identification;

[0111] 4. The compressed sensing technology is used to compress and collect the current signal, reducing the amount of collected current signals and accelerating the transmission efficiency of power data.

[0112] Other features and advantages of the present invention will be described in the following specification, and, in part, will be obvious from the specification, or will be understood by implementing the present invention. The objectives and other advantages of the present invention can be achieved and obtained through the structures pointed out in the specification, claims, and drawings.

[0113] The present invention will be further described below in conjunction with the accompanying drawings. Description of the Drawings

[0114] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the accompanying drawings required for the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0115] Figure 1 It is a schematic diagram of the steps of a method for compressed acquisition and reconstruction of current signals according to an embodiment of the present invention;

[0116] Figure 2 It is a schematic flowchart of a method for compressed acquisition and reconstruction of current signals according to an embodiment of the present invention;

[0117] Figure 3 It is a schematic structural diagram of a device for compressed acquisition and reconstruction of current signals according to an embodiment of the present invention;

[0118] Figure 4 It is a schematic diagram showing the change of the reconstruction error MSE of the reconstruction algorithm in the embodiment of the present invention and other algorithms with the compression ratio CR;

[0119] Figure 5 It is a schematic diagram showing the change of the reconstruction time T of the reconstruction algorithm in the embodiment of the present invention and other algorithms with the signal sampling volume;

[0120] Figure 6 It is a schematic structural diagram of an electronic device according to an embodiment of the present invention. Detailed Embodiments

[0121] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some, rather than all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0122] As Figures 1 to 2 shown, the first embodiment of the present invention provides a method for compressing, collecting, and reconstructing current signals, and the method includes the following steps:

[0123] S1. Use a flexible Blackman window to slide-collect the current signal to obtain a sliding-collected signal ;

[0124] This step is the sliding collection of the current signal. Specifically, use a flexible Blackman window to slide-collect the current signal and introduce a window adjustment parameter to collect data .

[0125] S2. Construct a partial orthogonal Gaussian measurement matrix ;

[0126] This step is to construct a partial orthogonal Gaussian measurement matrix. Specifically, according to the restricted isometry property (RIP) in compressive sensing technology and combined with the characteristic that the correlation between the required measurement matrix and the sparse basis is very low, construct a partial orthogonal Gaussian measurement matrix with independent and identically distributed and orthogonal characteristics .

[0127] S3. Use the partial orthogonal Gaussian measurement matrix to compress-collect the sliding-collected signal to obtain a compressed signal ;

[0128] This step is the compressive sensing of the current signal. Specifically, use the partial orthogonal Gaussian measurement matrix obtained in step S2 to compress-collect the windowed current signal according to compressive sensing technology, and the compressed data is .

[0129] S4. Perform sparse transformation on the sliding-collected signal to solve for the frequency-domain transformation matrix ;

[0130] This step is the signal sparse transformation. Specifically, perform sparse transformation on the windowed current signal in step S1 to solve for the frequency-domain transformation matrix under the DWT (Discrete Wavelet Transform) transform .

[0131] S5. Use the frequency-domain transformation matrix and the partial orthogonal Gaussian measurement matrix The sensing matrix D is calculated, and then the sparse signal is iteratively solved using the compression signal and the sensing matrix D ;

[0132] This step is to iteratively solve the sparse signal. Specifically, the frequency-domain transformation matrix in step S4 and the partial orthogonal Gaussian measurement matrix are used to calculate the sensing matrix D, and then the CoSaMP-ESP improved compressive sampling matching pursuit reconstruction algorithm is used to solve the sparse signal .

[0133] S6. Reconstructed data is calculated according to the frequency-domain transformation matrix and the sparse signal .

[0134] This step is the reconstruction of the current signal. Specifically, according to the sparse signal iteratively solved in step S5 , the reconstructed data is calculated .

[0135] In the above technical solution, through the compressive acquisition and reconstruction of the high-frequency current signal that can characterize the characteristics of the fault arc, low-volume data acquisition is realized, and the high-frequency fault arc characteristics in the data are effectively retained, solving the problem of missing fault arc characteristics after traditional current signal acquisition.

[0136] The embodiment of the present invention actually provides a method for compressive acquisition and reconstruction of current signals considering fault arc identification. The method has the following characteristics: aiming at the problem of difficult feature acquisition and monitoring when various power facilities such as high-voltage equipment, substations, and transmission lines have fault arcs in modern power grids, the embodiment of the present invention proposes a method for compressive acquisition and reconstruction of current signals considering fault arc identification to effectively break through the Nyquist high-frequency sampling limitation, realize low-speed dynamic compressive sampling and accurate reconstruction of high-frequency current signals of fault arcs, retain the effective characteristics of fault arcs in the current signal, provide data acquisition technical support for fault arc identification in power grid operation, and have good engineering application prospects.

[0137] The following will further explain each step of the above method.

[0138] 1. Sliding acquisition of current signals

[0139] The embodiment of the present invention is based on a flexible Blackman window for the current signal characterizing the characteristics of the fault arc ​Perform sliding window acquisition. Compared with traditional windows such as orthogonal Gaussian windows and Hanning windows, the Blackman window has lower sidelobe peaks and a faster sidelobe attenuation rate, and has a higher advantage in data spectrum acquisition, which can effectively retain the original characteristics of the current signal. The flexible Blackman window function is specifically expressed as:

[0140] (1)

[0141] In formula (1), t is the current signal acquisition time, T is the adjustable window width.

[0142] Use the above window function to acquire the current signal, obtain the current signals of different time windows, and get the acquired signal , specifically as follows:

[0143] (2)

[0144] (3)

[0145] Among them, is the number of discrete data points obtained by acquisition, is the sampling interval, is the pulse function, .

[0146] Adjust the window width according to the signal deviation of adjacent windows of the current signal T , and the neighborhood reconstruction deviation is:

[0147] (4)

[0148] In formula (4), 、 are the (h - 1)-th and h-th windows respectively.

[0149] The above flexible Blackman window uses flexible feedback modulation technology. The window width of the traditional Blackman window is fixed, but in the embodiments of the present invention, the window width can be adjusted according to the signal deviation of adjacent windows of the current signal T , so it is called a flexible Blackman window. The advantage of using a flexible Blackman window is that when an arc fault occurs, the current harmonic frequency increases, resulting in increasing, then the window is narrowed, which can effectively capture the instantaneous change of the current over time and improve the arc fault feature acquisition ability; when decreases, the window is expanded to effectively retain the signal spectrum characteristics. Using a flexible acquisition window (i.e., a flexible Blackman window) to adjust the window width according to the degree of electrical fluctuation can effectively acquire the original information of the current signal and retain the high-frequency characteristics of the fault arc.

[0150] 2. Construct a partially orthogonal Gaussian measurement matrix

[0151] The measurement matrix in compressive sensing technology needs to satisfy the restricted isometry property (RIP) condition, which requires a very low correlation between the measurement matrix and the sparse basis. Therefore, in the embodiments of the present invention, a partially orthogonal Gaussian measurement matrix is constructed based on a Gaussian measurement matrix with independent and identically distributed elements, so as to improve the correlation control ability of the matrix through partial orthogonality while retaining the characteristics of the standard Gaussian random matrix, and reduce the computational complexity during reconstruction.

[0152] The random variable in the selected Gaussian measurement matrix satisfies the following distribution:

[0153] (5)

[0154] In formula (5), is a Gaussian distribution with a mean of 0 and a variance of .

[0155] The standard Gaussian measurement matrix is as follows, with a size of :

[0156] (6)

[0157] Extract the first columns of the matrix to form the matrix , The value benchmark is 0.2× , and it is dynamically adjusted according to the neighborhood reconstruction deviation in formula (4). When increases, the value of is decreased; when decreases, the value of is increased. The value range of is (0.1~0.3)× :

[0158] (7)

[0159] Perform orthogonalization on the matrix , select the first column as the orthogonal basis vector , and the j -th column needs to subtract its projection on all previous orthogonal basis vectors and be normalized:

[0160] (8)

[0161] Obtain the orthogonalized matrix:

[0162] (9)

[0163] The orthonormalized column matrix is combined with the remaining non-orthonormalized columns A k'+1:n to form a combination, that is, the first columns of the matrix are replaced with matrix to obtain a partially orthonormal Gaussian matrix , which can improve the correlation control ability of the matrix and reduce the computational amount during reconstruction:

[0164] (10).

[0165] 3. Compressive Sensing of Current Signals

[0166] In the signal processing of the embodiments of the present invention, compressive sensing technology is adopted. Compressive Sensing (CS), as an emerging signal processing technology, provides a brand-new solution. The compressive sensing theory is based on the sparsity of signals, and the original signal can be reconstructed through a small amount of uncorrelated measurement data. This theory breaks through the limitations of the traditional Nyquist sampling theorem, can significantly reduce the amount of data during the sampling stage, and ensure the integrity and accuracy of the signal at the reconstruction end. The compressive sensing technology provides an efficient solution for the compressive acquisition of current signals, and can significantly improve the monitoring and diagnosis capabilities of current signals before and after the occurrence of arc faults.

[0167] Using the partially orthonormal Gaussian measurement matrix obtained in step 2 above, the windowed current signal is compressed according to the compressive sensing technology, and the compressed data is . After compression, the initial n data are compressed into m data, and the specific formula is as follows:

[0168] (11).

[0169] Using the compressive sensing technology for compressive acquisition of current signals reduces the amount of collected current signals and speeds up the transmission efficiency of power data.

[0170] 4. Data Sparse Transformation

[0171] Perform sparse transformation on the windowed current signal in step 1 . Perform DWT transformation (i.e., discrete wavelet transform) on , and use the Daubechies wavelet basis of multi-Bezier wavelets to perform N - 1 decompositions to obtain 1 low-frequency component (approximate coefficient) and N - 1 high-frequency components (detail coefficients):

[0172] (12)

[0173] In Equation (12), is the low-frequency component, is the high-frequency component, and are the scaling function and the wavelet function respectively, j and f are the scale number and the translation number respectively.

[0174] Arrange the right side of Equation (12), and decompose the sparse signal z(k) and the frequency-domain transformation matrix according to the signal sparsity principle , and then further perform inverse processing on both sides of the matrix, then the windowed current signal The sparse signal representation under the DWT transform is:

[0175] (13)

[0176] In Equation (13), is the frequency-domain transformation matrix under the DWT transform, obtained by inverting the matrix on the left side of the expression. Here, the matrix in Equation (13) is obtained by transforming and arranging Equation (12).

[0177] In this embodiment, Step 4 belongs to the compression end, and the frequency-domain transformation matrix is obtained by representing the sparse signal for subsequent reconstruction of the current signal; Step 5 belongs to the reconstruction end, z(k) is unknown, and only the compressed data h(m) and the sensing matrix D are obtained. It is necessary to first iteratively solve z(k), and then obtain the original signal x(n) according to the matrix and Equation (13).

[0178] 5. Iteratively solve the sparse signal

[0179] First, use the DWT frequency-domain transformation matrix and the partial orthogonal Gaussian measurement matrix in the above steps to calculate the sensing matrix D:

[0180] (14)

[0181] Then, use the Compressive Sampling Matching Pursuit based on Optimal Sparsity Estimation (CoSaMP-ESP) reconstruction algorithm to solve the sparse signal , the original fault arc characteristics in the current signal can be effectively reconstructed. Using the CoSaMP-ESP reconstruction algorithm, the reconstruction speed is fast and the reconstruction accuracy is high, which can provide strong technical support for fault arc identification.

[0182] Input the compressed and collected current signal , the sensing matrix D and the error threshold . The CoSaMP-ESP algorithm process is as follows:

[0183] ① Initialization, set the index value set E to be an empty set, F to be an empty set, and the residual v = , the sparsity K = 5, and the number of iteration rounds d = 0; here K = 5 is an empirical value, and the K value is iteratively updated in step ⑤ later, so the initial value of K has little impact on the final algorithm reconstruction effect.

[0184] ② d = d + 1, calculate the inner product of the current residual and the sensing matrix D, and select the columns of the matrix D that best meet the conditions according to the fuzzy threshold c (taking the value of 0.6 - 0.8 in this embodiment), and store them in E, thereby reducing the possibility of incorrect columns being included in the support set, and thus reducing the computational complexity and reconstruction error:

[0185] (15)

[0186] In formula (15), D j is the j-th column of the matrix D, v d-1 is the residual of the (d - 1)-th iteration.

[0187] Among them, the "fuzzy threshold" refers to an adjustment mechanism introduced to reduce the possibility of incorrect atoms (i.e., inaccurate atoms) being included in the support set during the candidate atom selection process. The setting of the fuzzy threshold can reduce the problem of possible incorrect selections in the candidate atom set. The traditional CoSaMP algorithm selects the 2K atoms with the largest absolute value of the inner product of the maximum residual and the sensing matrix each time, and this fixed selection quantity method is prone to bring incorrect atoms in the case of a large sparsity, thereby increasing the computational complexity and reconstruction error.

[0188] ③ According to the set E calculated by the fuzzy threshold in the previous step, obtain the support set :

[0189] (16)

[0190] ④ According to the support set Solve the least squares solution, and find a solution that minimizes the error by minimizing the sum of the squares of the errors, that is, find the estimated value of the reconstructed signal, so that the error between this estimated value and the actual observed data is minimized:

[0191] (17)

[0192] In formula (17), , represents the column in matrix D corresponding to the set , represents the coefficient vector in the current set , and the dimension of this vector corresponds to the number of elements in the set . When performing operations on the sensing matrix, the identity matrix I is introduced to solve the problem that the inverse of the matrix cannot be obtained during reconstruction.

[0193] ⑤ Prune the support set. Select the K terms with the largest absolute values in and store them in z(k) to ensure that the support set only contains those columns of D that are most important for signal reconstruction, and update the corresponding . The difference between the embodiment of the present invention and the traditional reconstruction algorithm is that the sparsity estimate value K is updated, so as to dynamically control the size of the support set , adapt to the sparse characteristics of the signal, improve the accuracy and sparsity effect of signal reconstruction, and at the same time prevent too many irrelevant columns of D from entering the support set, reducing the reconstruction error:

[0194] (18)

[0195] In formula (18), K is the sparsity, is the support set of the d-th iteration, h(m) is the compressed data, D is the sensing matrix, z(k) is the sparse signal to be solved, and the adjustable p-norm is used to control the sparsity. When p < 1, the norm tends to promote sparsity, making more small coefficients tend to zero; when p > 1, the norm is more inclined to reduce sparsity. In practical applications, different current data frequencies have different sparse structures. When the current data frequency is high, the p value is increased, and when the current data frequency decreases, the p value is decreased. By adjusting the appropriate p value, the model can better adapt to the sparse structure of the data, improving the performance and generalization ability of the model. The reference value of the p value is 1, and the variation range generally takes 0.3 - 3.

[0196] ⑥ Update the residual and the relevance :

[0197] (19)

[0198] (20)

[0199] Among them, is the residual, is the relevance, represents all columns of matrix D D jThe column with the highest correlation with the current residual reflects the maximum correlation between the reconstructed signal and D j in the current iteration, and is used to guide the column selection of matrix D in the next iteration. The higher the correlation, D j the stronger the correlation between them and the residual. Therefore, these D j can be preferentially considered to be included in the support set in subsequent iterations, so as to reduce the residual faster and more effectively and improve the reconstruction accuracy.

[0200] ⑦ If , stop the iteration and output the sparse signal , otherwise return to recalculate the residual starting from step ②.

[0201] In some embodiments, the error threshold = 10 -3 , and using this value can better meet the calculation requirements of the CoSaMP-ESP algorithm.

[0202] 6. Current signal reconstruction

[0203] According to the coefficient signal estimated by iterating according to the above steps, the reconstructed data is calculated as:

[0204] (21)

[0205] In formula (21), is the frequency domain transformation matrix, and is the sparse signal solved by iteration.

[0206] In summary, in the embodiments of the present invention, first, the current signal that can characterize the characteristics of the fault arc is slidingly collected using a Blackman window, and the flexible feedback modulation technology is introduced to adjust the window width according to the neighborhood reconstruction deviation to realize the effective monitoring of the high-frequency fluctuations of the current signal; secondly, an improved partial orthogonal Gaussian measurement matrix is introduced, and the compressive sensing technology is used to realize the compressive acquisition of the current signal; then, the DWT discrete wavelet transform technology is used to realize the sparse transform of the collected current signal; finally, a compressive sampling matching pursuit reconstruction algorithm based on optimal sparsity estimation (CoSaMP-ESP) is proposed to realize the accurate iterative reconstruction of the low-dimensional data.

[0207] The second embodiment of the present invention also provides a device for compressive acquisition and reconstruction of current signals, as Figure 3 shown, the device includes:

[0208] A sliding acquisition module for using a flexible Blackman window for the current signal Perform sliding acquisition to obtain a sliding acquisition signal ;

[0209] A matrix construction module for constructing a partial orthogonal Gaussian measurement matrix ;

[0210] A compression acquisition module for using the partial orthogonal Gaussian measurement matrix to perform compression acquisition on the sliding acquisition signal to obtain a compressed signal ;

[0211] A sparse transformation module for performing sparse transformation on the sliding acquisition signal to solve and obtain a frequency-domain transformation matrix ;

[0212] A calculation and solution module for using the frequency-domain transformation matrix and the partial orthogonal Gaussian measurement matrix to calculate and obtain a sensing matrix D, and then iteratively solve a sparse signal according to the compressed signal and the sensing matrix D using an algorithm ;

[0213] A data reconstruction module for calculating and obtaining reconstructed data according to the frequency-domain transformation matrix and the sparse signal .

[0214] Regarding the device in the above embodiments, the specific manners in which each module performs operations have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0215] The following will further elaborate with specific embodiments.

[0216] T According to a specific embodiment, taking the historical data of current signals generated before and after the normal operation and arc fault of a substation in a certain city in China as a verification data set, the initial sampling time window width is 100 ms, the time window width range is set from 1 power frequency cycle to 20 power frequency cycles, that is, the window adjustment range is from 20 ms to 400 ms, the original data sampling frequency is 150 kHz, and the sampling time is 5 s. The initial window width (i.e., the initial sampling time window width) and the width adjustment range (i.e., the time window width range) here are the initial width and adjustment range of the flexible Blackman window in step 1; the number of discrete points n collected in step 1 can be calculated according to the sampling frequency and sampling time.

[0217] ​​To verify the current signal compression acquisition and reconstruction effect of the method in the embodiments of the present invention during arc fault occurrence, the Compressive Sampling Matching Pursuit (CoSaMP) algorithm, the Variable Step Sparsity Self-Estimating Subspace Pursuit (VSSESP) algorithm, and the CoSaMP-ESP algorithm proposed in the embodiments of the present invention are used to reconstruct and compare the acquired signals. Taking the A-phase current signal data as an example, the reconstruction accuracy and efficiency are respectively as follows Figure 4 and Figure 5 shown. The CoSaMP algorithm and the VSSESP algorithm mentioned above both belong to the prior art and will not be elaborated here.

[0218] It can be seen from Figure 4 that when the compression ratio CR (compressed data / original data) is between 0.9 and 0.1, the mean square error (MSE) of the reconstruction by the CoSaMP-ESP algorithm proposed in the embodiments of the present invention is less than that of the CoSaMP and VSSESP algorithms, proving that the method proposed in the embodiments of the present invention has higher reconstruction accuracy; it can be seen from Figure 5 that under the same data volume, the time required for the CoSaMP-ESP reconstruction proposed in the embodiments of the present invention is less than that of the other two methods, indicating that the reconstruction efficiency of the method in the embodiments of the present invention is higher.

[0219] The reconstructed current signals above are respectively divided into a training set and a test set according to the random sampling principle in a ratio of 7:3. The typical fault arc recognition model - CNN model is used to train and verify the data set. The CNN network structure is set as the input layer, fully connected layer, convolutional layer (2×1×1×16, 1 channel, 16 convolutional kernels), batch normalization layer, activation layer (Rule function), convolutional layer (2×1×16×32, 16 channels, 32 convolutional kernels), batch normalization layer, activation layer (Rule function), fully connected layer, Softmax layer, and output layer; the optimizer uses the Adam algorithm, the regularization parameter is set to 1e-1, the learning rate is 1e-3, and the number of iterations is 3000 times. Training and testing are carried out. The identification accuracy results of the test sets of different algorithms are shown in Table 1. The results show that under the reconstruction method in the embodiments of the present invention, the arc fault identification accuracy is higher, that is, the retention of arc fault characteristics in the current signals before and after data compression and reconstruction is better.

[0220] Table 1 Comparison results of fault arc identification accuracy under different reconstruction algorithms

[0221]

[0222] Based on the same inventive concept, the present invention also provides a computer-readable storage medium storing one or more programs, which, when executed, can implement the aforementioned method for compressive acquisition and reconstruction of current signals.

[0223] Based on the same inventive concept, the present invention also provides an electronic device, as Figure 6 shown, comprising a processor, a communication interface, the aforementioned computer-readable storage medium as described above, and a communication bus; wherein, the processor, the communication interface, and the computer-readable storage medium communicate with each other through the communication bus; the processor is configured to execute the program stored in the computer-readable storage medium.

[0224] In several embodiments provided by the present application, it should be understood that the disclosed apparatus and method can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For example, the division of the modules is only a logical function division. In actual implementation, there may be other division methods. For example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection to each other can be through some interfaces, and the indirect coupling or communication connection of the apparatus or module can be in electrical, mechanical or other forms.

[0225] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, that is, they can be located in one place, or distributed to multiple network modules. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. In addition, in each embodiment of the present invention, the functional modules can be integrated in a processing module, or each module can exist physically alone, or two or more modules can be integrated in one module. The above-mentioned integrated modules can be implemented in the form of hardware or in the form of software functional modules.

[0226] When the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs.

[0227] It should be noted that for the foregoing method embodiments, for the sake of simplicity of description, they are all expressed as a series of action combinations. However, those skilled in the art should be aware that the present invention is not limited by the described order of actions, because according to the present invention, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to the present invention.

[0228] In the above embodiments, the descriptions of the various embodiments each have their own emphases. For the parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0229] The parts not covered in the above embodiments are the same as the prior art or can be implemented using the prior art, and will not be further described herein.

[0230] Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method for compressing, collecting and reconstructing current signals, characterized in that, The method includes: Use a flexible Blackman window for the current signal to perform sliding acquisition to obtain a sliding acquisition signal , where the flexible Blackman window adjusts the window width according to the reconstructed signal deviation of adjacent windows of the current signal; Construct a partially orthogonal Gaussian measurement matrix ; Using the partial orthogonal Gaussian measurement matrix to perform compressive acquisition on the sliding acquisition signal to obtain a compressed signal after compressive acquisition ; Perform sparse transformation on the sliding acquisition signal and solve to obtain the frequency-domain transformation matrix ; Using the frequency domain transformation matrix and the partial orthogonal Gaussian measurement matrix to calculate the sensing matrix D, and then iteratively solving for the sparse signal according to the compressed signal and the sensing matrix D using an algorithm ; According to the frequency domain transformation matrix and the sparse signal the reconstructed data is calculated .

2. A method for compressed acquisition and reconstruction of current signals according to claim 1, characterized in that The functional expression of the flexible Blackman window is: （1） In formula (1), t is the current signal acquisition time, T is the adjustable window width.

3. A method for compressed acquisition and reconstruction of current signals according to claim 2, characterized in that Use the flexible Blackman window for the current signal to perform sliding acquisition to obtain a sliding acquisition signal , specifically as follows: （2） （3） Among them, is the number of discrete data points obtained by acquisition, is the sampling interval, is the pulse function, .

4. A method for compressed acquisition and reconstruction of current signals according to claim 2, characterized in that Adjusting the window width according to the reconstructed signal deviation of adjacent windows of the current signal T , the reconstructed signal deviation of adjacent windows of the current signal is as follows: （4） In formula (4), and are respectively the reconstructed signals of the h -1st h window; When an arc fault occurs, the current harmonic frequency increases, resulting in increases, then the window width is reduced T ; When decreases, the window width is increased T .

5. A method for compressed acquisition and reconstruction of current signals according to claim 4, characterized in that The constructed partial orthogonal Gaussian measurement matrix comprises: Based on the standard Gaussian measurement matrix with independent and identically distributed elements, a partially orthogonal Gaussian measurement matrix is constructed ; The random variables in the standard Gaussian measurement matrix satisfy the following distribution: (5) In formula (5), is a Gaussian distribution with a mean of 0 and a variance of .

6. A method for compressed acquisition and reconstruction of current signals according to claim 5, characterized in that The standard Gaussian measurement matrix is represented as the following matrix A , with size : (6) Extraction matrix The first columns form the matrix : (7) Orthogonalize the matrix and select the first column as the orthogonal basis vector . The j th column needs to subtract its projections onto all previous orthogonal basis vectors and be normalized: (8) An orthogonalized matrix is obtained: (9) Replace the first columns of the matrix with the matrix to obtain a partially orthogonal Gaussian matrix : (10)。 7. A method for compressed acquisition and reconstruction of current signals according to claim 6, characterized in that When extracting the first columns of the matrix to form the matrix , the value reference is 0.2× , and the deviation of the reconstructed signal of adjacent windows of the current signal is dynamically adjusted according to Equation (4) ; When increases, decrease value; When decreases, increase value; The value selection range is (0.1~0.3)× .

8. A method for compressed acquisition and reconstruction of current signals according to claim 1, characterized in that According to compressive sensing, the compressed signal is calculated as follows: (11)。 9. A method for compressed acquisition and reconstruction of current signals according to claim 1, characterized in that Perform sparse transformation on the sliding acquisition signal using discrete wavelet transform to solve for the frequency-domain transformation matrix .

10. A method for compressed acquisition and reconstruction of current signals according to claim 9, characterized in that Performing sparse transformation on the sliding acquisition signal by using discrete wavelet transform to solve and obtain a frequency-domain transformation matrix , including: Perform - 1 level decomposition on the sliding acquisition signal using the Daubechies wavelet basis of the Daubechies wavelet, obtaining 1 low - frequency component and and N - 1 high - frequency components: N - 1 (12) In formula (12), is the low-frequency component, is the high-frequency component, and are the scaling function and the wavelet function respectively, j , f are the scale number and the translation number respectively; Further transform Equation (12), the sliding acquisition signal The sparse signal representation under discrete wavelet transform is as follows: (13) In formula (13), is the frequency domain transformation matrix under discrete wavelet transform.

11. A method for compressed acquisition and reconstruction of current signals according to claim 1, characterized in that The calculation formula of the sensing matrix D is: (14) In formula (14), is a frequency-domain transformation matrix, is a partial orthogonal Gaussian measurement matrix.

12. A method for compressed acquisition and reconstruction of current signals according to claim 1, characterized in that According to the compressed signal and the sensing matrix D, the sparse signal is iteratively solved using the compressive sampling matching pursuit reconstruction algorithm CoSaMP-ESP based on optimal sparsity estimation .

13. A method for compressed acquisition and reconstruction of current signals according to claim 12, characterized in that According to the compressed signal and the sensing matrix D, the sparse signal is iteratively solved using the compressive sampling matching pursuit reconstruction algorithm CoSaMP-ESP based on optimal sparsity estimation , including: Input the compressed signal and the sensing matrix D and the error threshold , and the compressive sampling matching pursuit reconstruction algorithm CoSaMP-ESP based on optimal sparsity estimation includes: ① Initialize, set the index value set E as an empty set, F as an empty set, the residual v = , the sparsity K = 5, the number of iteration rounds d = 0; ② d = d +1, calculate the inner product of the current residual and the sensing matrix D, and select the columns of the sensing matrix D that best meet the conditions according to the fuzzy threshold c and store them in the set E : (15) In formula (15), D j is the j th column of matrix D, v d-1 is the residual at the d -1-th iteration; ③According to the said set E , the support set is obtained as follows: (16) ④According to the support set Solve the least squares solution: (17) In formula (17), , represents the column in the sensing matrix D corresponding to the set , and I is the identity matrix; ⑤ Crop the support set, and among select the term with the largest absolute value and store it in K ( z ), and update the corresponding k , then update the sparsity estimate : K : (18) In formula (18), K is the sparsity, is the support set of the d -th iteration, h(m) is the compressed data, D is the sensing matrix, z ( k ) is the sparse signal to be solved, and the adjustable p norm is used to control the sparsity; ⑥Update the residual and relevance : (19) (20) Among them, is the residual, is the relevance, representing all columns of the sensing matrix D D j with the highest relevance to the current residual among them; ⑦ If , stop the iteration and output the sparse signal , otherwise return to step ② to recalculate the residual.

14. A method for compressed acquisition and reconstruction of current signals according to claim 13, characterized in that The value range of the fuzzy threshold c is 0.6 to 0.

8.

15. A method for compressed acquisition and reconstruction of current signals according to claim 13, characterized in that In formula (18), p The value reference is 1, p The value range is taken as 0.3 to 3.

16. A method for compressed acquisition and reconstruction of current signals according to any one of claims 1-15, characterized in that The reconstructed data has the following calculation formula: (21) In formula (21), is a frequency domain transformation matrix, is a sparse signal obtained by iterative solution.

17. A current signal compression acquisition and reconstruction device, characterized in that, The device includes: A sliding acquisition module, which is used to perform sliding acquisition on the current signal using a flexible Blackman window to obtain a sliding acquisition signal , and the width of the flexible Blackman window is adjusted according to the signal deviation of the reconstructed signal of adjacent windows of the current signal; A matrix construction module for constructing a partial orthogonal Gaussian measurement matrix ; A compression acquisition module, which is used to utilize the partial orthogonal Gaussian measurement matrix to perform compression acquisition on the sliding acquisition signal and obtain a compressed signal after compression acquisition ; Sparse transformation module, used for the sliding acquisition signal to perform sparse transformation and solve to obtain a frequency-domain transformation matrix ; A calculation and solution module, which is used to utilize the frequency domain transformation matrix and the partial orthogonal Gaussian measurement matrix to calculate the sensing matrix D, and then iteratively solve the sparse signal according to the compressed signal and the sensing matrix D using an algorithm ; A data reconstruction module for calculating reconstructed data according to the frequency domain transformation matrix and the sparse signal . .

18. A computer-readable storage medium storing one or more programs, characterized in that When the one or more programs are executed, the method for compressed acquisition and reconstruction of current signals according to any one of claims 1-16 can be implemented.

19. An electronic device, comprising a processor, a communication interface, the computer-readable storage medium according to claim 18, and a communication bus; wherein, The processor, communication interface, and computer-readable storage medium communicate with each other through a communication bus; characterized in that the processor is used to execute the programs stored in the computer-readable storage medium.

Images