Power grid ultra-high harmonic detection method and device based on adaptive compressed sensing

Abstract

The present application provides a method and device for detecting ultra-high order harmonics in a power grid based on adaptive compressive sensing. The detection method includes: Step 1, collecting power signals in the area to be detected in the power grid to obtain discrete power signals; Step 2, sampling the original signal vector; Step 3, performing DFT operation on the sampled signal matrix and performing modulus operation to obtain a modulus value matrix and the maximum modulus value; Step 4, constructing a modulus value distribution matrix according to the distribution of modulus values in the modulus value matrix; Step 5, using the compressive sensing method to compress the original signal vector and performing signal reconstruction to solve the sparse coefficient vector; Step 6, performing Fourier transform on the sparse coefficient vector and selecting multiple largest peaks in the modulus value sequence for characteristic frequency calculation; Step 7, constructing a filter bank according to the characteristic frequencies, filtering the modulus value sequence and then performing transformation to obtain the ultra-high order harmonic detection result. The present application can improve the efficiency and accuracy of detecting ultra-high order harmonics in the power grid.

Description

Technical Field

[0001] This application relates to the field of power system detection technology, and particularly to a method and device for detecting ultra-high order harmonics in a power grid based on adaptive compressive sensing. Background Art

[0002] With the wide application of various power electronic devices, the non-linearity and harmonic problems in the current power system are becoming more and more serious, and the ultra-high order harmonic problem is particularly prominent. Especially the large use of electric vehicle chargers, LED lights and various power electronic switching devices has led to the increasingly prominent ultra-high order harmonic problem in the frequency spectrum range of 2 - 150 kHz. Due to the high frequency, small amplitude and energy concentration in a specific frequency band of ultra-high order harmonics, the ultra-high order harmonics cause great damage to the power equipment in the distribution system. Therefore, how to quickly and accurately evaluate the ultra-high order harmonics in the power grid has become an important issue for the stable operation of the power system.

[0003] In the prior art, frequency domain analysis methods such as DFT (Discrete Fourier Transform) are usually used to detect and evaluate ultra-high order harmonic signals in power grid signals. However, due to the high frequency of ultra-high order harmonics, a high sampling rate and a large number of sampling points are required. The method of using the traditional discrete Fourier transform DFT has low execution efficiency, and requires a large amount of data transmission and computing resources, which is not suitable for real-time detection and evaluation tasks. Moreover, since the 2 - 150 kHz frequency band where ultra-high order harmonics are located is too wide, using the traditional DFT method can only achieve a rough analysis of ultra-high order harmonics, and cannot perform a fine ultra-high order harmonic analysis on specific signals, resulting in difficulty in obtaining the detailed parameters of ultra-high order harmonics, and thus unable to achieve a quick and accurate detection and evaluation of ultra-high order harmonics in the power grid. Summary of the Invention

[0004] The purpose of this application is to provide a method and device for detecting ultra-high order harmonics in a power grid based on adaptive compressive sensing, which can improve the efficiency and accuracy of ultra-high order harmonic detection in the power grid, and can also improve the data transmission efficiency, reduce the data storage requirements and the required computing resources.

[0005] To achieve the above object, the embodiments of this application provide:

[0006] A method for detecting ultra-high order harmonics in a power grid based on adaptive compressive sensing, the steps include:

[0007] Step 1, collect the power signals in the area to be detected in the power grid x ( t ) obtain N n-point discrete power signals x ( n ) n i = 1, 2, …, N n, and form the original signal vectorX = x (1), x (2), x (3), …, x ( N )];

[0008] Step 2. Sample the original signal vector X to construct a sampled signal matrix of ( N / V)* L dimensions , where V represents the sampling multiple, L represents the number of rows of the matrix, corresponding to the number of intervals required to divide the ultra-high harmonic frequency range;

[0009] Step 3. Perform DFT operations on each row vector in the sampled signal matrix to obtain a DFT coefficient vector matrix Z, and perform modulus operation on the DFT coefficient vector matrix Z to obtain a modulus matrix and the maximum modulus V m ;

[0010] Step 4. Divide each modulus in each modulus matrix into multiple intervals according to the maximum modulus V m , and construct a modulus distribution matrix Q according to the distribution of the moduli in the modulus matrix;

[0011] Step 5. Use the adaptive compressive sensing method to compress the original signal vector Q according to the modulus distribution matrix X to obtain a compressed signal vector Y , and solve for the sparse coefficient vector S by signal reconstruction of the compressed signal vector Y . In the adaptive compressive sensing method, the sparsity Q is calculated according to the modulus distribution matrix K ;

[0012] Step 6. Perform Fourier transform on the sparse coefficient vector S to form a complex vector H from the Fourier transform coefficients, and select the largest multiple of the peak values in the modulus sequence H of the complex vector for characteristic frequency calculation;

[0013] Step 7. Construct a non-linear filter bank G according to the calculated characteristic frequencies, and use the non-linear filter bank G to filter the modulus sequence and then obtain the ultra-high harmonic detection result through logarithm and discrete cosine transform.

[0014] In step 4, according to the maximum modulus value V m Adopt the golden section principle to divide each modulus value in each modulus matrix into multiple intervals according to a non-linear division method. The calculation expression for interval division is:

[0015]

[0016] where is the division coefficient, u is the division interval number, and the value range of u is from 1 to L , when u = L , the corresponding divided interval is ;

[0017] Construct a modulus value distribution matrix Q as:

[0018]

[0019] where represents the element in the modulus value distribution matrix Q at the i th row and j th column, corresponding to the number of times the row vector in the sampling signal matrix has a modulus value distribution in the interval.

[0020] As an optional implementation method, in step 5, the adaptive compressive sensing method is used to compress the original signal vector X , including:

[0021] Step 51, construct a M×N -dimensional measurement matrix Φ, where the expression of the measurement matrix Φ is:

[0022]

[0023] where the matrix element satisfies a Gaussian distribution with a mean of 0 and a variance of , M represents the dimension of the compressed signal, K represents the sparsity;

[0024] Step 52, obtain the compressed signal vector according to the measurement matrix Φ.

[0025] As an optional implementation method, in step 5, according to the modulus value distribution matrix Q compress the original signal vector X to obtain the compressed signal vector YExpressed as Y = y (1), y (2), y (3), …, y ( M )], where M The value satisfies:

[0026]

[0027]

[0028] Among them, Represents the sparse degree conversion coefficient corresponding to the element Q In the modulus value distribution matrix , Represents rounding up.

[0029] As an alternative implementation, in step 5, the compressed sampling matching pursuit algorithm is used to perform signal reconstruction on the compressed signal vector Y To solve for the sparse coefficient vector S , and the calculation expression is:

[0030]

[0031]

[0032]

[0033]

[0034] Among them, Φ represents M×N A measurement matrix of dimension Ψ Is a matrix composed of the Gaussian sparse basis ψ , α And β Respectively represent the frequency and time stretching factors, And Respectively represent the frequency interval width and the time interval width, , Respectively correspond to Each element of the matrix The two variables within the () function, that is = , = , , F Represents the highest frequency value to be analyzed, T Represents N A discrete power signal of points x ( n) corresponding time.

[0035] As an alternative implementation, in step 6, the two-point interpolation method is used to calculate the characteristic frequencies corresponding to each peak :

[0036]

[0037] where, and respectively represent the positions of the maximum and the second maximum values of the spectrum near the k th characteristic frequency , represents the correction coefficient, represents the sampling frequency.

[0038] As an alternative implementation, step 7 includes:

[0039] Step 71, construct a filter bank based on each characteristic frequency G , and use the filter bank G to filter the magnitude sequence to obtain the filtered sequence U；

[0040] Step 72, take the logarithm of the sequence U to obtain the sequence L = log| U | 2 , and perform a discrete cosine transform on the sequence L to obtain the ultra-high harmonic detection result V = v (1), v (2), …, v ( k ), …, v ( K )], v ( k ) represents the aggregation energy normalization parameter within the characteristic frequency interval, K represents the number of characteristic frequencies.

[0041] As an alternative implementation, in step 7, the non-linear filter bank G includes K non-linear filters, where the k th filter has the following calculation expression:

[0042]

[0043]

[0044] Wherein, n = 1, 2, 3, …, N , k = 1, 2, 3, …, K , represents the modulus sequence the largest peak value in K the k th peak value from left to right among the represents the k th peak value and k the difference in sequence numbers between the η and μ are respectively the filter aggregation coefficients, and the expressions are respectively:

[0045]

[0046] Wherein, represents the characteristic frequency corresponding to the k th peak point.

[0047] As an alternative implementation, in step 7, the calculation expression of the obtained ultra-high order harmonic detection result is:

[0048]

[0049] Wherein, represents the L th value in the sequence i , K represents the number of filters in the filter bank G , represents the modulus sequence the largest peak value in K the k th peak value and k the difference in sequence numbers between the

[0050] An ultra-high order harmonic detection device for power grid based on adaptive compressive sensing, comprising:

[0051] A processor;

[0052] A memory for storing instructions executable by the processor;

[0053] Wherein, the processor is configured to implement the above method when executing the instructions.

[0054] Compared with the prior art, the advantages of the present application are as follows: The present application first samples the collected original power signal and performs DFT operation to obtain the modulus matrix of the coefficient vector matrix, obtains the modulus distribution matrix according to the distribution of the modulus, and then uses the compressive sensing method to compress the original power signal based on the modulus distribution matrix and then performs signal reconstruction. The sparse coefficient vector is determined by using signal reconstruction and the modulus sequence of the Fourier transform coefficient vector is calculated. After calculating the characteristic frequencies corresponding to multiple peaks, a filter bank is constructed by using the characteristic frequencies, and the filter bank is used to adaptively aggregate the energy of the reconstructed signal spectrum to achieve ultra-high harmonic detection. It can make full use of the adaptive compressive sensing method and adaptive energy aggregation to quickly and accurately evaluate ultra-high harmonics, realizing the grid ultra-high harmonic detection and evaluation method, which can not only improve the detection accuracy and reliability, but also reduce the sampling and storage costs, so as to meet the application requirements of real-time detection of ultra-high harmonics. Description of the Drawings

[0055] In the following, the present invention will be described in more detail based on embodiments and with reference to the drawings. Among them:

[0056] Figure 1 It is a schematic diagram of the implementation process of the grid ultra-high harmonic detection method based on adaptive compressive sensing according to an embodiment of the present application. Detailed Embodiments

[0057] The present invention will be further described in detail below with reference to the drawings of the specification and specific embodiments, but the protection scope of the present invention is not limited thereby.

[0058] The sampling rate required for ultra-high harmonics is relatively high. Therefore, at a high sampling rate, if the traditional discrete Fourier transform (DFT) is used, a large amount of resources are required for calculation, and a large amount of data transmission and computing resources are needed. The compressive sensing (CS) algorithm is a sampling theory based on signal sparsity. By using a small number of linear measurement data to reconstruct the sparse signal and recover the original signal, the storage and transmission volume of data can be significantly reduced. In the compressive sensing algorithm, the sparsity in the transform domain can describe the number or distribution of non-zero coefficients of the signal in the transform domain. The higher the sparsity of the signal in the sparse domain, the fewer the required measurement numbers. Therefore, the sparsity will directly affect the efficiency and quality of signal compression. In the traditional compressive sensing algorithm, the sparsity is usually fixed, but the sparsity of the actual grid signal may change with time and environment. The fixed sparsity cannot adapt to the change of the signal and cannot effectively distinguish the signal and noise, resulting in large reconstruction errors and poor robustness when the sparsity changes.

[0059] The present invention aims at detecting ultra-high order harmonics in the power grid. By adopting the compressive sensing method to compress the power grid signal to be detected, the data volume in the operation process can be effectively reduced, and the detection efficiency of ultra-high order harmonics can be improved. At the same time, an adaptive mechanism is introduced in the compression process. By sampling and separating the original signal and dynamically calculating the sparsity using the modulus value distribution matrix calculated from the sampled signal matrix, the adaptivity of compressive sensing is realized, and the sparsity can be dynamically adjusted according to the sparse characteristics of the signal in real time, effectively improving the accuracy and robustness of real-time signal reconstruction. Thus, on the premise of improving the detection efficiency and real-time performance of ultra-high order harmonics, the detection accuracy and robustness can be ensured.

[0060] In addition, due to the extremely high bandwidth of ultra-high order harmonics, it is necessary to adopt the method of frequency band aggregation to characterize the distribution of ultra-high order harmonics in different frequency bands of the signal. Frequency band aggregation is achieved by dividing the ultra-high order harmonic signal into different frequency bands and analyzing and aggregating the signals within each frequency band, so as to realize the characterization of the distribution of ultra-high order harmonics in different frequency bands of the signal. Through frequency band aggregation, ultra-high order harmonics in the signal can be detected more accurately. Traditional frequency band aggregation methods include discrete wavelet transform (DWT) and sliding window algorithm, etc. However, the frequency band aggregation method of discrete wavelet transform has high computational complexity and highly depends on the selected wavelet basis function, and it is actually difficult to accurately select a suitable wavelet basis function. While the sliding window algorithm has large storage overhead and poor ability to handle burst traffic, and it is difficult to handle a large number of requests or signals that change violently in a short time.

[0061] Based on the compression processing of the power grid signal by the adaptive compressive sensing method, the present invention further calculates the characteristic frequency, and then constructs a non-linear filter bank centered on the characteristic frequency, which can perform aggregation analysis on ultra-high order harmonics in different frequency bands, so as to more accurately reflect the distribution of ultra-high order harmonics, realize adaptive frequency band aggregation, and use the adaptive frequency band aggregation method to reasonably divide the ultra-high order harmonic frequency band according to the change of the intensity of each frequency component in the measured signal, thus effectively improving the accuracy and reliability of ultra-high order harmonic detection.

[0062] See Figure 1 , the method for detecting ultra-high order harmonics in the power grid based on adaptive compressive sensing in this embodiment includes the following steps:

[0063] Step 1, collect the power signal in the detected area of the power grid x ( t ) to obtain N N-point discrete power signals x ( n ), n n = 1, 2, …, N , and form the original signal vector X =x (1), x (2), x (3), …, x ( N )];

[0064] Step 2. Sample the original signal vector X to construct a signal matrix of ( N / V) * L dimensions , where L represents the number of rows of the matrix, corresponding to the number of intervals required to divide the ultra-high harmonic frequency range; V represents the sampling multiple, which can specifically be a multiple of 2;

[0065] Step 3. Perform DFT operations on each row vector in the signal matrix to obtain the DFT coefficient vector matrix Z, and perform modulus operation on the DFT coefficient vector matrix Z to obtain the modulus matrix and the maximum modulus V m ;

[0066] Step 4. Divide each modulus in each modulus matrix into V m intervals according to the maximum modulus L , and construct a modulus distribution matrix Q according to the distribution of the moduli in the modulus matrix;

[0067] Step 5. Use the adaptive compressive sensing method to compress the original signal vector Q according to the modulus distribution matrix X to obtain the compressed signal vector Y , and solve for the sparse coefficient vector S by signal reconstruction of the compressed signal vector Y , where the modulus distribution matrix Q in the adaptive compressive sensing method calculates the sparsity K ;

[0068] Step 6. Perform Fourier transform on the sparse coefficient vector S to form a complex vector H from the Fourier transform coefficients, and select the largest multiple of the modulus sequence H of the complex vector for characteristic frequency calculation;

[0069] Step 7. Construct a non-linear filter bank G according to the calculated characteristic frequencies, and filter the modulus sequence using the non-linear filter bank G and then obtain the ultra-high harmonic detection result through logarithm and discrete cosine transform.

[0070] In the above method of this embodiment, by first sampling the collected original power signal and performing DFT operation to obtain the modulus matrix of the coefficient vector matrix, the modulus distribution matrix is obtained according to the distribution of the modulus. Q , and then based on the modulus distribution matrix Q , the original power signal is compressed by using the adaptive compressive sensing method and then signal reconstruction is performed. By using the dynamically changing modulus distribution matrix Q for sparsity calculation, the amount of data in the operation process can be reduced, and at the same time, the detection efficiency, accuracy and robustness of ultra-high-order harmonics can be effectively improved; by using signal reconstruction to determine the sparse coefficient vector and calculating the modulus sequence of the Fourier transform coefficient vector, after calculating the characteristic frequencies corresponding to multiple peaks, a filter bank is constructed by using the characteristic frequencies, and the filter bank is used to perform adaptive energy aggregation on the reconstructed signal spectrum, which can reasonably divide the ultra-high-order harmonic frequency band adaptively according to the change of the intensity of each frequency component in the measured signal, and improve the accuracy and reliability of ultra-high-order harmonic detection. Therefore, the adaptive compressive sensing method and adaptive energy aggregation can be fully utilized to realize the fast and accurate evaluation of ultra-high-order harmonics for power grid ultra-high-order harmonic detection and evaluation, which can not only improve the detection accuracy and reliability, but also reduce the sampling and storage costs, so as to meet the application requirements of ultra-high-order harmonic real-time detection.

[0071] In step 2 of this embodiment, by sampling the original signal vector X through the sampling method, the sampling frequency can be reduced, thereby reducing the computational complexity and data storage amount. In a possible implementation manner, a signal matrix of ( N / 64)*7 dimensions can be constructed that is, V takes 64, L takes 7, and the constructed signal matrix is as follows:

[0072] (1)

[0073] Among them, represents the i-th row element in the sampling signal matrix , and i = 1 to 7.

[0074] Furthermore, in step 3, the 7 row vectors in the signal matrix to can be subjected to DFT operation to obtain the DFT coefficient vector matrix Z, and then the modulus of the matrix Z is calculated to obtain its modulus matrix, and the maximum modulus is denoted as V m ; then according to this maximum modulus V mDivide each modulus value in each modulus matrix into multiple intervals, and construct a modulus value distribution matrix according to the distribution of the modulus values in the modulus matrix. Q to dynamically calculate the sparsity using the modulus value distribution matrix. K It can adapt to the real-time changes of grid signals and dynamically calculate the matching sparsity, improving the accuracy and robustness of signal compression and reconstruction. The distribution of modulus values is the probability distribution of different modulus values within different value ranges. The modulus value distribution matrix Q is the sampled signal matrix The distribution of the modulus values of each element in, and the modulus value distribution matrix Q Each element in corresponds to the number of each row vector in the sampled signal matrix distributed in each interval.

[0075] In a possible implementation, when dividing each modulus value in each modulus matrix into multiple intervals according to the maximum modulus value V m the golden section principle can be adopted, that is, the calculation expression for interval division is:

[0076] (2)

[0077] where is the division coefficient, which can specifically take , u is the division interval number, and its value range is from 1 to L ; specifically, when u = L , that is, the last division interval, its interval range is:

[0078] .

[0079] For example, L when taking 7, the intervals can be divided according to the above formula (2) as: P 1 = [0, 0.382 V m ), P 2 = [0.382 V m , 0.618 V m ), P 3 = [0.618 V m , 0.764 V m ), P 4 = [0.764 V m , 0.854 V m ), P 5 = [0.854 Vm , 0.91 V m ), P 6 = [0.91 V m , 0.956 V m ), P 7 = [0.956 V m , V m .

[0080] Furthermore, construct a modulus value distribution matrix Q as follows:

[0081] (3)

[0082] Among them, represents the element in the Q th row and i th column of the modulus value distribution matrix, corresponding to the number of times the row vector j in the sampling signal matrix has a modulus value distribution in the interval. i, = 1~ j . For example, L i.e., it represents the number of times the modulus value corresponding to the row vector is distributed in the first interval P . 1

[0083] Then, after sampling according to the method of formula (1), the modulus value distribution matrix Q can be constructed as follows:

[0084] (4)

[0085] In this embodiment, by using the adaptive compressive sensing method to compress the original signal vector X , the high-dimensional power grid signal can be compressed into a low-dimensional sparse signal, improving the data transmission efficiency and effectively reducing the data acquisition and storage requirements. In a possible implementation manner, step 5 uses the compressive sensing method to compress the original signal vector X including:

[0086] Step 51, construct a M×N -dimensional measurement matrix Φ, where the expression of the measurement matrix Φ is:

[0087] (5)

[0088] Among them, the matrix element satisfies that the mean is 0 and the variance is ​Gaussian distribution M represents the dimension of the compressed signal K represents the sparsity, which is dynamically calculated according to the modulus value distribution matrix Q

[0089] Step 52: Use the measurement matrix Φ to obtain the compressed signal vector Y :

[0090] During the compression process using the adaptive compressive sensing method, the selection of the sparse basis will directly affect the accuracy and reliability in compression and reconstruction. In this embodiment, according to the modulus value distribution matrix Q the original signal vector X is compressed to obtain the compressed signal vector Y which can be expressed as Y = y (1), y (2), y (3), …, y ( M )] M represents the dimension of the compressed signal. First, the sparsity Q is dynamically calculated based on the modulus value distribution matrix K , and then the dimension K of the compressed signal is determined using the sparsity M . The value range of can be adaptively adjusted according to the state of the real-time signal, ensuring accurate representation of the original signal using compressive sensing in the case of ultra-high harmonic changes, and further improving the accuracy and robustness of signal reconstruction.

[0091] In a possible implementation, the sparsity x ( n ) of the Gaussian transform domain of the discrete power signal K can be calculated according to the following formula:

[0092] (6)

[0093] where represents the modulus value distribution matrix Q the element at the i -th row and j -th column of corresponding sparsity conversion coefficient, represents rounding up.

[0094] The value of the dimension M of the compressed signal needs to satisfy:

[0095] (7) ​​

[0096] In a possible implementation, in step 5, the compressed sampling matching pursuit algorithm is used to perform signal reconstruction on the compressed signal vector Y to solve for the sparse coefficient vector S , and the calculation expression is:

[0097] (8)

[0098] where Φ represents the M×N -dimensional measurement matrix, and the superscript "-1" represents the pseudo-inverse matrix operation. Ψ is the sparse basis matrix, which is composed of the Gaussian sparse basis ψ . In this embodiment, by adopting a dynamic sparse basis, the form of the sparse basis can be changed according to the frequency change in the measured signal, so that the original signal can be better compressed and reconstructed, further improving the accuracy of ultra-high harmonic analysis.

[0099] For example, the following improved Gaussian sparse basis Ψ can be adopted:

[0100] (9)

[0101] (10)

[0102] where α and β represent the frequency and time scaling factors respectively, and represent the frequency and time interval widths respectively, , correspond to the two variables in the function in each element of the sparse basis matrix, that is, = , = , . For example, for the last element in the first row of the sparse basis matrix, in represents , represents .

[0103] In a possible implementation, and can be calculated respectively according to the following formulas:

[0104] (11)

[0105] whereF Indicates the highest frequency value to be analyzed, T represents the power signal N point x ( n ) corresponding time.

[0106] In this embodiment, the sparse basis is constructed by the above method Ψ , and the shape of the sparse basis Ψ will change with the frequency and time interval width of the real-time measured signal, forming a dynamic sparse basis, so as to adapt to the changes of the real-time signal for accurate compression and reconstruction.

[0107] In a possible implementation manner, in step 6, the two-point interpolation method can be used to calculate the characteristic frequencies corresponding to each peak , for example, the calculation expression can be expressed as:

[0108] (12)

[0109] Wherein, and respectively represent the positions of the maximum and the second maximum values of the spectrum near the k th characteristic frequency , and represents the correction coefficient. That is, by selecting the maximum value of the spectrum near the characteristic frequency and the second maximum value to calculate the characteristic frequency corresponding to the peak, the detailed parameters of the ultra-high-order harmonic can be obtained. Through the characteristic frequency and the corresponding spectral line, the amplitude of the ultra-high-order harmonic can be accurately calculated.

[0110] Considering that the frequency band where the ultra-high-order harmonic exists is too wide and frequency band aggregation is required, in this embodiment, the above characteristic frequency is used to construct a non-linear filter bank centered on the characteristic frequency to perform aggregation analysis on ultra-high-order harmonics in different frequency bands. In a possible implementation manner, step 7 includes: G Step 71, construct a non-linear filter bank based on each characteristic frequency

[0111] , and use the filter bank G to filter the modulus sequence G to obtain the filtered K-dimensional sequence U；

[0112] Step 72, take the logarithm of the sequence U to obtain the sequence L =log|​U | 2 , the discrete cosine transform is performed on the sequence L to obtain the ultra-high harmonic detection result V = v (1), v (2), …, v ( k ), …, v ( K )], v ( k ) represents the aggregation energy normalization parameter within the characteristic frequency interval, K is the determined sparsity, that is, the number of characteristic frequencies.

[0113] In this embodiment, the non-linear filter bank G is based on the peak in the modulus sequence and adopts a non-linear aggregation method, which can efficiently display the characteristics of ultra-high harmonics in different bandwidth frequency bands within the 2 - 150 kHz frequency band. It can adaptively and reasonably divide the ultra-high harmonic frequency band according to the change of the intensity of each frequency component in the measured signal, so as to facilitate the accurate analysis of the impact of ultra-high harmonics on the power system. In one possible implementation, the non-linear filter bank G is specifically composed of K non-linear filters . The expression of the k th filter can be expressed as:

[0114] (13)

[0115] (14)

[0116] In the formula, n = 1, 2, 3, …, N , k = 1, 2, 3, …, K , K is the determined sparsity, represents the th peak with the largest modulus in the Fourier transform coefficient modulus sequence K . The k th peak point from left to right in the peaks corresponds to the serial number, k represents the difference in serial numbers between the k th peak and the p (0) represents the spectral position corresponding to the 2000 Hz frequency, that is, the expression is:

[0117] (15)

[0118] η and μ are the filter aggregation coefficients respectively, and their expressions are respectively:

[0119] (16)

[0120] Wherein, represents the characteristic frequency corresponding to the k th peak point.

[0121] In a possible implementation manner, in step 7, the calculation expression of the obtained ultra-high harmonic detection result can be expressed as:

[0122] (17)

[0123] In the formula, represents the L th value in the sequence i , K represents the determined sparsity, corresponding to the number of filters in the filter bank G , represents the difference in the sequence number between the th peak with the largest peak value and the K rd peak among the k peaks with the largest peak values in the modulus value sequence k -1th peak.

[0124] Since the influence of ultra-high harmonics on the power system is different in different frequency bands, and affected by the non-linear frequency band division, different frequency bands are obtained by superimposing ultra-high harmonics within different bandwidths, so differences will be formed in different frequency bands, which is not conducive to evaluating ultra-high harmonics at the same level. In this embodiment, the aggregation energy normalization parameter in the characteristic frequency interval is extracted by using the cosine transform as the ultra-high harmonic detection result, which can eliminate the difference problem formed by different frequency bands, enable the evaluation of ultra-high harmonics in different frequency bands at the same level, and can efficiently obtain the detailed parameters of ultra-high harmonics, so as to quickly realize the accurate detection of ultra-high harmonics.

[0125] In summary, by adopting the adaptive compressive sensing method, the present invention compresses and reconstructs the high-dimensional power grid signal into a low-dimensional sparse signal, which can effectively improve the data transmission efficiency, reduce the data acquisition and storage requirements. At the same time, by performing spectrum analysis on the reconstructed signal, constructing a non-linear filter bank using the characteristic frequency, and adaptively aggregating the energy of the spectrum of the reconstructed signal by using the filter bank, an accurate and efficient evaluation of the ultra-high harmonics of the power grid is realized, and the fast and accurate evaluation of the ultra-high harmonics of the power grid can be effectively realized.

[0126] In another embodiment of the present application, a power system frequency measurement device based on iterative optimization includes:

[0127] A processor;

[0128] A memory for storing instructions executable by the processor;

[0129] Wherein, when the processor is configured to execute instructions, it implements the above method. The processor and the memory are directly or indirectly electrically connected to achieve data transmission or interaction. For example, these components can be electrically connected through one or more communication buses or signal buses. The foregoing control methods respectively include at least one software function module that can be stored in the memory in the form of software or firmware.

[0130] The processor can be an integrated circuit chip with signal processing capabilities. The processor can be a general-purpose processor, including a CPU (Central Processing Unit), an NP (Network Processor), etc.; it can also be a digital signal processor, an application-specific integrated circuit, a field-programmable gate array, or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. It can implement or execute the various methods, steps, and logic block diagrams disclosed in the embodiments of the present invention. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc.

[0131] The memory can store various software programs and modules, such as the program instructions / modules corresponding to the image processing method and device provided in the embodiments of the present invention. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, implements the method in the embodiments of the present application. The memory can include, but is not limited to, RAM (Random Access Memory), ROM (Read Only Memory), PROM (Programmable Read-Only Memory), EPROM (Erasable Programmable Read-Only Memory), EEPROM (Electric Erasable Programmable Read-Only Memory), etc.

[0132] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can 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.) that contain computer-usable program code.

[0133] The present application is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of the flows and / or blocks in the flowchart and / or block diagram can also be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 one or more of the flows Figure 1 or blocks.

[0134] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including instruction means that implement the functions specified in Figure 1 one or more of the flows Figure 1 or blocks. These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, so that the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in Figure 1 one or more of the flows Figure 1 or blocks.

[0135] Although the present invention has been described with reference to the preferred embodiments, various improvements can be made to it and components therein can be replaced with equivalents without departing from the scope of the present invention. In particular, as long as there is no structural conflict, the technical features mentioned in each embodiment can be combined in any way. The present invention is not limited to the specific embodiments disclosed in the text, but includes all technical solutions falling within the scope of the claims.

Claims

1. A method for detecting ultra-high harmonics in a power grid based on adaptive compressed sensing, characterized in that the steps include: Step 1: Collect power signals from the inspected area in the power grid x ( t )get N Point discrete power signal x ( n ), n =1, 2, …, N , and form the original signal vector X =[ x (1), x (2), x (3), …, x ( N )]; Step 2: The original signal vector X Sampling is performed to construct ( N / V)* L dimensional sampled signal matrix , where V represents the sampling multiple, L The number of rows in the matrix corresponds to the number of intervals required to divide the ultra-high harmonic frequency range; Step 3: Sample signal matrix Perform DFT operation on each row vector in to obtain a DFT coefficient vector matrix Z, and perform modular operation on the DFT coefficient vector matrix Z to obtain a modular value matrix and a maximum modular value V m ; Step 4: According to the maximum modulus value V m Divide each modulus value in each modulus value matrix into multiple intervals, and construct a modulus value distribution matrix according to the distribution of the modulus values ​​in the modulus value matrix Q ; Step 5: According to the module value distribution matrix Q Adopting adaptive compressed sensing method to transform the original signal vector X Perform compression processing to obtain the compressed signal vector Y , and by the compressed signal vector Y Signal reconstruction is performed to solve the sparse coefficient vector S. In the adaptive compressed sensing method, the modulus distribution matrix Q Calculating sparsity K; Step 6: Calculate the sparse coefficient vector S Perform Fourier transform and construct a complex vector from the Fourier transform coefficients H , and select the complex vector H The modulus sequence Calculate the characteristic frequencies of the largest multiple peaks; Step 7: Construct a nonlinear filter bank based on the calculated characteristic frequencies G , using the nonlinear filter bank G to filter the module value sequence After filtering, the ultra-high harmonic detection results are obtained through logarithmic and discrete cosine transform.

2. The method for detecting ultra-high harmonics in a power grid based on adaptive compressed sensing according to claim 1, characterized in that: In step 4, according to the maximum modulus value V m The golden section principle is used to divide each module value in each module value matrix into multiple intervals in a nonlinear partitioning manner. The calculation expression for interval partition is: in, is the division coefficient, u is the partition interval number, and u The value range is 1 to L ,when u = L The corresponding interval is ; Constructing the module distribution matrix Q for: in, Represents the modulus distribution matrix Q Middle i Line j The elements of the column correspond to the sampled signal matrix Middle row vector The corresponding modulus value is distributed in The number of intervals.

3. The method for detecting ultra-high harmonics in a power grid based on adaptive compressed sensing according to claim 1, characterized in that: Step 5: Adopt adaptive compressed sensing method to the original signal vector X Compression processing includes: Step 51: Build M×N The measurement matrix Φ of dimension Φ is expressed as: Among them, the matrix elements The mean is 0 and the variance is Gaussian distribution, M represents the dimension of the compressed signal, K Indicates sparsity; Step 52: Obtain the compressed signal vector according to the measurement matrix Φ .

4. The method for detecting ultra-high harmonics in a power grid based on adaptive compressed sensing according to claim 3 is characterized in that: In step 5, according to the module value distribution matrix Q For the original signal vector X Perform compression processing to obtain the compressed signal vector Y Expressed as Y =[ y (1), y (2), y (3), …, y ( M )],in M The value satisfies: in, Represents the module value distribution matrix Q Middle i Line j Elements of a column The corresponding sparsity conversion coefficient, Indicates rounding up.

5. The method for detecting ultra-high harmonics in a power grid based on adaptive compressed sensing according to claim 1, characterized in that: In step 5, the compressed signal vector is subjected to the compression sampling matching pursuit algorithm. Y Perform signal reconstruction to solve the sparse coefficient vector S , the calculation expression is: Among them, Φ represents M×N dimensional measurement matrix, Gaussian sparse basis ψ The matrix formed, α and β denote the frequency and time stretching factors respectively, and Respectively represent the frequency interval width and time interval width, , Corresponding to Each element of the matrix ψ () function, namely = , = , , F Indicates the highest frequency value that needs to be analyzed. T express N Point discrete power signal x ( n )’s corresponding time.

6. The method for detecting ultra-high harmonics in a power grid based on adaptive compressed sensing according to any one of claims 1 to 5, characterized in that: In step 6, the two-point interpolation method is used to calculate the characteristic frequency corresponding to each peak : in, and Respectively represent k Characteristic frequency The positions of the spectral maximum and the second maximum near represents the correction factor, Indicates the sampling frequency.

7. The method for detecting ultra-high harmonics in a power grid based on adaptive compressed sensing according to any one of claims 1 to 5, characterized in that: Step 7 includes: Step 71: Construct a nonlinear filter bank based on each characteristic frequency G , and adopts nonlinear filter bank G Modulus sequence Filter and get the filtered sequence U； Step 72: Sequence U Take the logarithm to get the sequence L =log| U | 2 , for the sequence L After performing discrete cosine transform, the ultra-high harmonic detection results are obtained V =[ v (1), v (2), …, v ( k ), …, v ( K )], v ( k ) represents the characteristic frequency The aggregate energy normalization parameter in the interval, K Indicates the determined sparsity.

8. The method for detecting ultra-high harmonics in a power grid based on adaptive compressed sensing according to claim 7, characterized in that: In step 7, the nonlinear filter bank G include K nonlinear filter, where the k Filters The calculation expression is: In the formula, n =1, 2, 3, ..., N , k =1, 2, 3, ..., K , p ( k ) represents the modulus sequence The largest peak K The peaks are arranged from left to right k The sequence number corresponding to the peak point, Indicates k Peak value and k -1 The difference in sequence number between peaks, η and μ are the filter clustering coefficients, and their expressions are: in, Indicates k The characteristic frequency corresponding to the peak point.

9. The method for detecting ultra-high harmonics in a power grid based on adaptive compressed sensing according to claim 7, characterized in that: In step 7, the calculation expression of the ultra-high harmonic detection result is obtained as follows: In the formula, Representation sequence L The i values, K Represents the determined sparsity, corresponding to the filter group G The number of internal filters, Represents a sequence of modulo values The largest peak K The peak k Peak value and k -1 The difference in sequence number between peaks.

10. A power grid ultra-high harmonic detection device based on adaptive compressed sensing, characterized in that: include: processor; a memory for storing processor-executable instructions; Wherein, the processor is configured to implement the method described in any one of claims 1 to 9 when executing the instructions.

Images