Non-stationary power harmonic detection method and system, electronic device, storage medium

Abstract

The application provides a non-stationary power harmonic detection method and system, electronic equipment and a storage medium, and relates to the technical field of signal detection. The method comprises the following steps: acquiring a continuous electrical signal of a power system; performing full-phase fast Fourier transform on uniform sampling data of the continuous electrical signal to obtain corresponding amplitude spectrum and phase spectrum; determining the minimum fitting order of the uniform sampling data based on the local peak value of the amplitude spectrum and the phase change of the phase spectrum in the region corresponding to the local peak value; and estimating the signal parameters of each fitting component corresponding to the uniform sampling data by using a prony algorithm based on the minimum fitting order, and taking the signal parameters of each fitting component as the signal parameters of each harmonic of the continuous electrical signal to complete the harmonic detection of the continuous electrical signal. The application solves the problems of low detection precision, excessive calculation amount and poor real-time performance of the existing non-stationary power harmonic detection.

Description

Technical Field

[0001] This invention relates to the field of signal detection technology, specifically to a non-stationary power harmonic detection method and system, electronic device, and storage medium. Background Technology

[0002] As power systems become increasingly complex, the proliferation of power electronic equipment and various impulsive loads has led to a wealth of harmonics and interharmonics in the power grid. Simultaneously, the integration of various new energy sources and the use of energy storage devices have resulted in a large number of non-stationary harmonic / interharmonic signals in the grid. These complex signals not only interfere with the operation of electrical equipment but may also damage it, posing a serious threat to the safe and stable operation of the power grid. Therefore, it is urgent to manage these interference signals in the power system. However, achieving effective management requires rapid and accurate detection of these signals.

[0003] In harmonic / interharmonic analysis algorithms in power systems, Esprit and Prony algorithms are generally used for detecting non-steady-state signals. Esprit performs well in noisy conditions, but it involves multiple matrix calculations, resulting in a large computational load that is difficult to meet the needs of real-time analysis scenarios. Prony, on the other hand, is sensitive to noise, and its computational load is too large for high-order algorithms, making it difficult to meet the needs of real-time analysis scenarios. While its computational load is small for low-order algorithms, it is prone to information loss, leading to low accuracy. Summary of the Invention

[0004] To overcome the above-mentioned technical problems, the present invention provides a non-stationary power harmonic detection method and system, electronic device, and storage medium.

[0005] On one hand, the present invention provides a method for detecting non-stationary power harmonics, comprising:

[0006] Acquire continuous electrical signals from the power system, and perform full-phase fast Fourier transform on the uniformly sampled data of the continuous electrical signals to obtain the corresponding amplitude spectrum and phase spectrum;

[0007] Based on the local peaks of the amplitude spectrum and the phase changes of the phase spectrum in the region corresponding to the local peaks, the minimum fitting order of the uniformly sampled data is determined.

[0008] Based on the minimum fitting order, the signal parameters of each fitted component corresponding to the uniform sampling data are estimated using the Prony algorithm, and the signal parameters of each fitted component are used as the signal parameters of each harmonic of the continuous electrical signal to complete the harmonic detection of the continuous electrical signal.

[0009] Optionally, the minimum fitting order of the uniformly sampled data is determined based on a number of local peaks of the amplitude spectrum and a number of phase mutations of the phase spectrum in a region corresponding to the local peaks, and the minimum fitting order of the uniformly sampled data is determined based on a number of local peaks of the amplitude spectrum and a number of phase mutations of the phase spectrum in a region corresponding to the local peaks.

[0010] The minimum fitting order of the uniformly sampled data is determined based on a number of local peaks of the amplitude spectrum and a number of phase mutations of the phase spectrum in a region corresponding to the local peaks.

[0011] The phase mutation is a phase change greater than a preset threshold in a specified frequency region.

[0012] Optionally, the minimum fitting order of the uniformly sampled data is:

[0013] ;

[0014] wherein, is the minimum fitting order of the uniformly sampled data, is the number of local peaks of the amplitude spectrum, is the sum of the number of phase mutations of the phase spectrum in the region corresponding to the local peaks.

[0015] Optionally, the uniformly sampled data of the continuous electrical signal is subjected to a full-phase fast Fourier transform to obtain the corresponding amplitude spectrum and phase spectrum, and the method comprises:

[0016] A convolution window is constructed based on a convolution of two symmetrical windows, and the uniformly sampled data is subjected to windowing processing by using the convolution window to obtain a windowed signal.

[0017] Data symmetric to a center point in the windowed signal is added to obtain a new signal.

[0018] The new signal is subjected to a fast Fourier transform to obtain the amplitude spectrum and phase spectrum of the uniformly sampled data.

[0019] Optionally, signal parameters of each fitting component corresponding to the uniformly sampled data are estimated by using a prony algorithm based on the minimum fitting order, and the method comprises:

[0020] A linear fitting equation set is constructed based on the minimum fitting order.

[0021] A coefficient matrix of the linear fitting equation set is solved by using a least square method to estimate the signal parameters of each fitting component.

[0022] Optionally, the linear fitting equation set is as follows:

[0023] ;

[0024] , ;

[0025] wherein, is any sampling point of the fitting signal, is the sampling point index, is the sampling time interval, represents the imaginary part, is the i-th element of the coefficient matrix of the linear fitting equation set, and p is the minimum fitting order, is the i-th observation vector of the linear fitting equation set; , , , are the frequency, the attenuation factor, the amplitude and the initial phase of the i-th fitting component, respectively.

[0026] Optionally, the coefficient matrix of the linear fitting equation set is solved by using the least square method to estimate the signal parameters of each fitting component, including:

[0027] Based on the uniform sampling data and the minimum fitting order, a self-correlation matrix is constructed;

[0028] The eigenvalues and eigenvectors of the self-correlation matrix are solved, a Toeplitz matrix is constructed based on the eigenvalues and eigenvectors, and the linear fitting equation set is solved by using the least square method to obtain the prony pole;

[0029] The prony pole is used to construct a Vandermonde matrix, and based on the uniform sampling data and the Vandermonde matrix, the coefficient matrix of the linear fitting equation set is estimated, and then the signal parameters of each fitting component are obtained.

[0030] Optionally, the self-correlation matrix is as follows:

[0031] ;

[0032] ;

[0033] wherein, the matrix dimension is , p is the minimum fitting order of the uniform sampling data, is half of the data length of the uniform sampling data, and the superscript T is the matrix transpose; is a Hankel matrix, the matrix dimension is , , ,… are the values of each sampling point of the current time of the uniform sampling data, respectively, The N-1th sampling point of the current moment of the uniform sampling data.

[0034] Optionally, the solving of the eigenvalue and eigenvector of the autocorrelation matrix, the construction of the Toeplitz matrix based on the eigenvalue and eigenvector, and the solving of the linear fitting equation set by using the least square method to obtain the prony pole include:

[0035] Eigenvalue decomposition is performed on the autocorrelation matrix to obtain an eigenvalue vector and an eigenvector matrix of the autocorrelation matrix. The matrix dimensions are and respectively.

[0036] The column vectors in the eigenvector matrix are reordered in a descending order of eigenvalues in the eigenvalue vector , and the first column vectors of the reordered matrix are extracted to form a new eigenvector matrix . The matrix dimension is .

[0037] The first rows and the last rows of the new eigenvector matrix are respectively taken as preprocessed matrices and . The matrix dimensions are .

[0038] Toeplitz matrices are constructed based on the preprocessed matrices and , and the linear fitting equation set is solved by using the least square method to obtain discrete poles . The i-th discrete pole is , the eigenvalue decomposition is , and the superscript -1 is the inverse of the matrix.

[0039] The discrete poles are converted into continuous-time poles, i.e., prony poles .The i-th prony pole is , and the sampling frequency is .

[0040] Optionally, the coefficient matrix of the linear fitting equation set is as follows:

[0041] ;

[0042] ;

[0043] , ;

[0044] wherein, is a coefficient matrix of the linear fitting equation set; is a Vandermonde matrix, is an N-1 order of the pth prony pole; is a fitting signal obtained by taking N sampling points of the uniformly sampled data;

[0045] The signal parameters of each fitting component are as follows:

[0046] ;

[0047] wherein, , , , are the frequency, the attenuation factor, the amplitude and the initial phase of the ith fitting component respectively, represents taking the real part of a complex number, represents taking the imaginary part of a complex number.

[0048] Optionally, after obtaining the signal parameters of each fitting component, the method further comprises:

[0049] eliminating the fitting component whose amplitude is less than a specified amplitude and the fitting component whose attenuation factor absolute value is greater than a specified value;

[0050] setting the attenuation factor of the integer harmonic greater than a specified order to zero.

[0051] In another aspect, the present application also provides a non-stationary power harmonic detection system, comprising:

[0052] a time-frequency transformation module, configured to acquire a continuous electrical signal of a power system, and perform full-phase fast Fourier transformation on uniformly sampled data of the continuous electrical signal to obtain corresponding amplitude spectrum and phase spectrum;

[0053] an order determination module, configured to determine a minimum fitting order of the uniformly sampled data based on local peak values of the amplitude spectrum and phase variation of the phase spectrum in a region corresponding to the local peak values;

[0054] an analysis and detection module, configured to estimate signal parameters of each fitting component corresponding to the uniformly sampled data by using a prony algorithm based on the minimum fitting order, and take the signal parameters of each fitting component as signal parameters of each harmonic of the continuous electrical signal to complete harmonic detection of the continuous electrical signal.

[0055] Optionally, the order determination module is specifically configured to:

[0056] determine the minimum fitting order of the uniformly sampled data based on the number of local peaks of the amplitude spectrum and the number of phase mutations of the phase spectrum in the region corresponding to the local peaks.

[0057] wherein the phase mutation is a phase change greater than a preset threshold in a specified frequency region.

[0058] Optionally, the minimum fitting order of the uniformly sampled data is:

[0059] ;

[0060] wherein, is the minimum fitting order of the uniformly sampled data, is the number of local peaks of the amplitude spectrum, is the sum of the number of phase mutations of the phase spectrum in the region corresponding to the local peaks.

[0061] Optionally, the time-frequency transformation module comprises:

[0062] a windowing submodule configured to construct a convolution window based on convolution of two symmetrical windows, perform windowing processing on the uniformly sampled data using the convolution window to obtain a windowed signal, and add data symmetrically relative to a center point in the windowed signal to obtain a new signal;

[0063] a transformation submodule configured to perform fast Fourier transform on the new signal to obtain the amplitude spectrum and the phase spectrum of the uniformly sampled data.

[0064] Optionally, the analysis and detection module comprises:

[0065] a fitting equation construction submodule configured to construct a linear fitting equation set based on the minimum fitting order;

[0066] a parameter estimation submodule configured to solve a coefficient matrix of the linear fitting equation set using a least square method to estimate signal parameters of each fitting component.

[0067] Optionally, the linear fitting equation set is as follows:

[0068] ;

[0069] , ;

[0070] wherein, is any sampling point of the fitting signal, is a sampling point index, is a sampling time interval, represents an imaginary part, is an i-th element of a coefficient matrix of a linear fitting equation set, p is a minimum fitting order of uniformly sampled data, is an i-th observation vector of the linear fitting equation set; , , , are a frequency, a damping factor, an amplitude and an initial phase of an i-th fitting component, respectively.

[0071] Optionally, the parameter estimation submodule comprises:

[0072] a self-correlation matrix construction subunit configured to construct a self-correlation matrix based on the uniformly sampled data and the minimum fitting order;

[0073] a pole solving subunit configured to solve eigenvalues and eigenvectors of the self-correlation matrix, construct a Toeplitz matrix based on the eigenvalues and the eigenvectors, and solve the linear fitting equation set using a least square method to obtain prony poles;

[0074] a parameter estimation subunit configured to construct a Vandermonde matrix using the prony poles, estimate a coefficient matrix of the linear fitting equation set based on the uniformly sampled data and the Vandermonde matrix, and further obtain signal parameters of each fitting component.

[0075] Optionally, the self-correlation matrix is as follows:

[0076] ;

[0077] ;

[0078] wherein, a matrix dimension is , is a data length of the uniformly sampled data, and a superscript T represents a matrix transpose; is a Hankel matrix, a matrix dimension is , , ,… are respective sampling point values of the uniformly sampled data at a current time, is an N-1-th sampling point of the uniformly sampled data at the current time.

[0079] Optionally, the pole solving subunit is specifically configured to:

[0080] solve the eigenvalues and the eigenvectors of the self-correlation matrix perform eigenvalue decomposition to obtain an eigenvalue vector of the autocorrelation matrix and an eigenvector matrix ; and The matrix dimensions are and ;

[0081] According to the descending order of the eigenvalues in the eigenvalue vector , the column vectors in the eigenvector matrix are reordered, and the first column vectors of the reordered matrix are extracted to form a new eigenvector matrix , The matrix dimensions are ;

[0082] The first rows and the last rows of the new eigenvector matrix are respectively taken as the preprocessed matrix and , and The matrix dimensions are ;

[0083] Based on the preprocessed matrix and , a Toeplitz matrix is constructed, and the linear fitting equation set is solved using the least square method to obtain the discrete poles ; is the i-th discrete pole, is the eigenvalue decomposition, and the superscript -1 is the inverse of the matrix;

[0084] The discrete poles are converted into continuous-time poles, i.e. prony poles ; is the i-th prony pole, is the sampling frequency.

[0085] Optionally, the coefficient matrix of the linear fitting equation set is as follows:

[0086] ;

[0087] ;

[0088] , ;

[0089] wherein, is the coefficient matrix of the linear fitting equation set; is the Vandermonde matrix, The p-th prony pole is raised to the power of N-1. This is the fitted signal obtained by taking N sampling points of the uniformly sampled data;

[0090] The signal parameters of each fitted component are as follows:

[0091] ;

[0092] in, , , , Let be the frequency, attenuation factor, amplitude, and initial phase of the i-th fitted component, respectively. To take the real part of a complex number, This indicates taking the imaginary part of a complex number.

[0093] Optionally, the analysis and detection module further includes a false component removal submodule, which is used for:

[0094] Remove fitted components whose amplitude is less than a specified amplitude and fitted components whose absolute value of the attenuation factor is greater than a specified value;

[0095] Set the attenuation factor of harmonics of integer orders greater than a specified order to zero.

[0096] On the other hand, the present invention also provides an electronic device, comprising: at least one processor and a memory; the memory and the processor are connected via a bus;

[0097] The memory is used to store one or more programs;

[0098] When the one or more programs are executed by the at least one processor, the method described in any of the foregoing is implemented.

[0099] On the other hand, the present invention also provides a readable storage medium having an executable program stored thereon, wherein when the executable program is executed, it implements the method described in any one of the above.

[0100] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0101] The application provides a non-stationary power harmonic detection method and system, which obtains corresponding amplitude spectrum and phase spectrum by performing full-phase fast Fourier transform on uniformly sampled data of continuous electrical signals of a power system; determines the minimum fitting order of the uniformly sampled data based on local peak values of the amplitude spectrum and phase change conditions near the local peak values; estimates signal parameters of each fitting component corresponding to the uniformly sampled data by using a prony algorithm based on the minimum fitting order, and takes the signal parameters of each fitting component as signal parameters of each harmonic of the continuous electrical signals. The data is transformed by full-phase fast Fourier transform, improving data processing efficiency; the number of harmonic components in the signal is accurately determined by combining the amplitude-frequency local peak value characteristics of the transformed signal and the phase change characteristics near the local peak values, so that the minimum fitting order of the prony algorithm is accurately determined, the calculation amount is greatly reduced while ensuring that no information is lost and the analysis accuracy is ensured, the contradiction between precision and efficiency in traditional harmonic / inter-harmonic detection is solved, and real-time analysis scenarios can be met. BRIEF DESCRIPTION OF DRAWINGS

[0102] Figure 1 A flowchart of a non-stationary power harmonic detection method of the application;

[0103] Figure 2 A schematic diagram of each part of an example analog power grid signal of the application;

[0104] Figure 3 An amplitude spectrum and phase spectrum diagram obtained after full-phase fast Fourier transform is performed on an analog power grid signal of an example of the application;

[0105] Figure 4 An amplitude spectrum comparison diagram of an example analysis result and an original analog signal of the application;

[0106] Figure 5 A phase spectrum comparison diagram of an example analysis result and an original analog signal of the application;

[0107] Figure 6 A comparison result of a fitting signal obtained by using the method of the application and an original analog signal;

[0108] Figure 7 An error analysis result of a detection analysis result of an example of the application;

[0109] Figure 8 An electronic device structure block diagram of the application. DETAILED DESCRIPTION

[0110] The specific embodiments of the application will be further described in detail below with reference to the accompanying drawings.

[0111] Example 1

[0112] The present invention provides a method for detecting non-stationary power harmonics, the schematic diagram of which is shown below. Figure 1 As shown, the method includes:

[0113] Step S110: Obtain the continuous electrical signal of the power system, and perform a full-phase fast Fourier transform on the uniformly sampled data of the continuous electrical signal to obtain the corresponding amplitude spectrum and phase spectrum;

[0114] Step S120: Based on the local peaks of the amplitude spectrum and the phase changes of the phase spectrum in the region corresponding to the local peaks, determine the minimum fitting order of the uniformly sampled data;

[0115] Step S130: Based on the minimum fitting order, the signal parameters of each fitted component corresponding to the uniform sampling data are estimated using the Prony algorithm, and the signal parameters of each fitted component are used as the signal parameters of each harmonic of the continuous electrical signal to complete the harmonic detection of the continuous electrical signal.

[0116] In this example implementation, the continuous electrical signal can be a current signal or a voltage signal in the power system. In power systems with various new energy sources and energy storage devices connected and impact loads present, this current / voltage signal contains non-stationary harmonic / interharmonic signals. For example, a current transformer can be used to collect the continuous voltage / current signal of the power system, and the collected voltage / current signal can be uniformly sampled at a sampling frequency to obtain uniformly sampled data. For example, uniform sampling yields... Each sampled data ,in, For sampling point index, For any sampling point, To ensure uniform sampling data length, Generally, the order is taken as an integer power of 2. All-phase Fast Fourier Transform (apFFT) smooths the signal through full-phase preprocessing, reducing spectral leakage and improving frequency resolution. On one hand, by finding local peaks in the amplitude and phase spectra, the local peaks and their nearby phase changes are determined, and based on this, the minimum fitting order (generally less than 30) is determined. The number of harmonic components can then be extracted through amplitude-frequency characteristic analysis. Using this minimum fitting order as the order of the Prony algorithm to analyze uniformly sampled signals ensures the matching degree between the fitting order and the harmonic components, guaranteeing the accuracy of the analysis results. On the other hand, compared to the traditional Prony algorithm (which generally selects half the number of sampling points as the order, such as...),... The minimum fitting order is much smaller than that of traditional analysis methods, which greatly reduces the order of the Prony algorithm and thus greatly reduces the amount of computation in the harmonic analysis process. This makes it suitable for real-time analysis scenarios in power systems.

[0117] In some example embodiments, full-phase fast Fourier transform is performed on the uniformly sampled data of the continuous electrical signal to obtain corresponding amplitude spectrum and phase spectrum, including:

[0118] A convolution window is constructed based on convolution of two symmetrical windows; and the uniformly sampled data is windowed using the convolution window to obtain a windowed signal;

[0119] Data symmetrically located relative to a center point in the windowed signal is added to obtain a new signal;

[0120] Fast Fourier transform is performed on the new signal to obtain amplitude spectrum and phase spectrum of the uniformly sampled data.

[0121] In the example embodiments, the construction process of the convolution window is as follows:

[0122] (1)

[0123] wherein, , are two symmetrical rectangular windows, and the length of each window is , m is the data point of the rectangular window, is the convolution window, and the length of the convolution window is .

[0124] The windowing process of the uniformly sampled data using the convolution window is as follows:

[0125] (2)

[0126] wherein, is the windowed signal of any sampling point in the uniformly sampled data.

[0127] Then, the windowed signal is symmetrically added with corresponding data before and after the symmetrical point to obtain a new signal of points as follows:

[0128] (3)

[0129] wherein, is the new signal, , is the th data point in the new signal.

[0130] Fast Fourier transform is performed on the signal to obtain:

[0131] (4)

[0132] wherein, is the harmonic order, is the signal is the full-phase Fourier transform result of the signal, and are the amplitude and initial phase of the mth harmonic, respectively; denotes the frequency-domain convolution window function, and M is the total number of harmonics. Therefore, the amplitude spectrum and the phase spectrum in the full-phase Fourier transform result can be expressed as:

[0133]

[0134]

[0135] In some example embodiments, determining the minimum fitting order of the uniformly sampled data based on local peaks of the amplitude spectrum and phase variation of the phase spectrum in a region corresponding to the local peaks comprises:

[0136] determining the minimum fitting order of the uniformly sampled data based on the number of local peaks of the amplitude spectrum and the number of phase mutations of the phase spectrum in a region corresponding to the local peaks;

[0137] wherein a phase mutation is a phase variation greater than a preset threshold in a specified frequency region.

[0138] In the example embodiment, local peaks are searched in the amplitude spectrum and the frequency points of the local peaks are recorded as , is the frequency point of the mth local peak, and the number of local peaks is determined as , which is generally less than 10. The phase variation of a region (e.g., the entire frequency range of the local peak) corresponding to each local peak frequency point in the phase spectrum is determined, such as the number of times of phase variation greater than a preset threshold (e.g., 10°), and the number of times of variation of all local peak frequency points is obtained by adding the number of times of variation of each local peak frequency point, denoted as , which is generally less than 5; thereby obtaining the total phase variation. The harmonic condition is evaluated by the number of local peaks and the total phase variation, so as to determine the minimum fitting order, thereby ensuring the harmonic analysis accuracy based on the minimum fitting order and avoiding information loss.

[0139] Exemplarily, the minimum fitting order of the uniformly sampled data is:

[0140] ​​ ;

[0141] wherein, is the minimum fitting order of the uniform sampling data, and is also the number of signals; is generally less than 30; is the number of local peaks of the amplitude spectrum, is the sum of phase mutations of the phase spectrum in the region corresponding to the local peaks.

[0142] In some example embodiments, S130 includes estimating the signal parameters of each fitting component corresponding to the uniform sampling data by using the Prony algorithm based on the minimum fitting order, including:

[0143] constructing a linear fitting equation set based on the minimum fitting order;

[0144] solving the coefficient matrix of the linear fitting equation set by using the least square method to estimate the signal parameters of each fitting component.

[0145] In the present example embodiment, after determining the minimum fitting order of the Prony algorithm, a linear fitting equation set of the Prony algorithm can be constructed as follows:

[0146] (5)

[0147] wherein, is any sampling point of the fitting signal, is the i-th element of the coefficient matrix of the linear fitting equation set, p is the minimum fitting order, is the i-th observation vector of the linear fitting equation set.

[0148] The uniform sampling signal can be expressed as:

[0149] (6)

[0150] Therefore, if , , , is the sampling time interval, represents the imaginary part; by solving the coefficient matrix of the linear fitting equation set by using the least square method, the signal parameters of each fitting component can be estimated , , , .

[0151] Exemplarily, solving the coefficient matrix of the linear fitting equation set by using the least square method to estimate the signal parameters of each fitting component includes:

[0152] constructing an autocorrelation matrix based on the uniformly sampled data and the minimum fitting order;

[0153] solving eigenvalues and eigenvectors of the autocorrelation matrix, constructing a Toeplitz matrix based on the eigenvalues and eigenvectors, and solving the linear fitting equation set by using a least square method to obtain prony poles;

[0154] constructing a Vandermonde matrix based on the prony poles, estimating a coefficient matrix of the linear fitting equation set based on the uniformly sampled data and the Vandermonde matrix, and further obtaining signal parameters of each fitting component.

[0155] In the example embodiment, a Hankel matrix of dimensions may be constructed according to the last N-1 sampling points of the uniformly sampled data in combination with the minimum fitting order p, where each element of the Hankel matrix represents a sampling point value at the current time.

[0156]

[0157] An autocorrelation matrix is constructed based on the Hankel matrix ,which is a matrix of dimensions .

[0158] After the autocorrelation matrix is constructed, eigenvalue decomposition is performed on the autocorrelation matrix to obtain an eigenvalue vector and an eigenvector matrix of the autocorrelation matrix. The matrix dimensions of and are and respectively. The column vectors (eigenvectors) in the eigenvector matrix are reordered in descending order of eigenvalues in the eigenvalue vector , and the first column vectors of the reordered matrix are extracted to form a new eigenvector matrix ,which has a matrix dimension of . To eliminate the influence of noise, the first rows and the last rows of the new eigenvector matrix are respectively taken as preprocessed matrices and ,which have a matrix dimension of . And there are two staggered sub-windows; based on the preprocessed matrix and Construct the Toeplitz matrix and solve the linear fitting equations using the least squares method to obtain the discrete poles. ; Let i be the i-th discrete pole. For eigenvalue decomposition, the superscript -1 represents the inverse of the matrix; the discrete poles are transformed into continuous-time poles, i.e., Prony poles. ; For the i-th prony pole, The sampling frequency is used. This example uses two interleaved sub-windows to construct a preprocessed matrix, reducing the impact of noise on the detection results, improving the noise sensitivity of the Prony algorithm, and enhancing the accuracy of the detection results.

[0159] After obtaining the Prony poles, the frequency and attenuation factor of the corresponding fitted component can be calculated based on the expression of the observed vector as follows:

[0160] .

[0161] Utilizing Prony poles Construct the Vandermonde matrix , for one Matrix;

[0162]

[0163] in, It is the N-1 power of the p-th prony pole.

[0164] Pick ,but The amplitude can be obtained. and initial phase ;

[0165] ;

[0166] Finally, the signal parameters of each fitted component are used as the signal parameters of each harmonic of the continuous electrical signal.

[0167] In some implementations, after obtaining the signal parameters of each fitted component, the method further includes:

[0168] Remove fitted components whose amplitude is less than a specified amplitude and fitted components whose absolute value of the attenuation factor is greater than a specified value;

[0169] Set the attenuation factor of harmonics of integer orders greater than a specified order to zero.

[0170] In the example embodiment, the specified amplitude can be determined according to the fundamental amplitude, such as 0.05%-0.1% of the fundamental amplitude, and the specified value can be set according to experience, such as 10. The specified order is generally 3, and can also be set to other integers according to detection requirements. Exemplarily, components with an amplitude less than 0.05% of the fundamental amplitude are removed, components with an attenuation factor with an absolute value greater than 10 are removed, and the attenuation factor of integer harmonics greater than 3rd order harmonics is set to zero. The false components in the detection result can be removed by the above component removal and attenuation factor setting to zero, so as to ensure the correctness of the detection result. After the above false removal operation is completed, each order harmonic (such as harmonics and interharmonics) of the non-stationary electrical signal can be obtained through signal reconstruction to complete the harmonic detection of the continuous electrical signal.

[0171] The most representative harmonic / interharmonic analysis algorithm in a power system is the method based on FFT, such as a windowed FFT interpolation correction algorithm, a multi-layer DFT (Discrete Fourier Transform) interpolation correction algorithm, an estimation algorithm based on multi-layer elimination of spectrum leakage interference, an FFT algorithm based on Newton iteration frequency compensation, and the like. These algorithms have excellent performance in the detection of steady-state harmonics / interharmonics, but are powerless when non-steady-state harmonics / interharmonics are involved. For the detection of non-steady-state signals, wavelet transform, Esprit algorithm, and prony algorithm are generally used. The advantage of wavelet transform is filtering and denoising, but it is not convenient to solve characteristic parameters by using the wavelet transform. The Esprit algorithm performs well in the presence of noise, but the algorithm involves multiple matrix calculations, which is computationally intensive and is not suitable for situations with high time requirements. The conventional prony algorithm is sensitive to noise, and the order has a greater impact on the correctness of the result. A high order has a large calculation amount and high precision, but may also bring some false components. A low order has a small calculation amount, but is easy to lose information and thus has low precision.

[0172] To solve the technical problems of low precision and large calculation amount of the existing non-stationary harmonic / interharmonic detection method in the related art, the present application first performs apFFT transformation on the data, accurately determines the number of harmonic components in the signal in combination with the amplitude-frequency characteristics and phase-frequency characteristics after the transformation, accurately determines the order of the prony algorithm, achieves the dual purposes of accurate analysis and reduced calculation amount, and reduces the influence of noise on the detection result in combination with the improved prony algorithm. The method is suitable for high-precision and fast analysis of non-stationary harmonics / interharmonics in a power system.

[0173] Simulation experiment

[0174] A non-stationary voltage signal of a power grid is simulated, which mainly consists of three parts. The first part is a steady harmonic / interharmonic (such as Figure 2The stable part), the second part is the non-stationary harmonics / interharmonics (such as... Figure 2 The first part is the time-varying part, and the second part is white noise with a signal-to-noise ratio of 60. The three parts are superimposed to obtain the noisy signal, as shown below. Figure 2 The bottom part is shown. The signal parameters of the first part of the simulated non-stationary voltage signal, which contains stationary harmonics / interharmonics, are shown in Table 1, and the signal parameters of the second part of the simulated non-stationary voltage signal, which contains non-stationary harmonics / interharmonics, are shown in Table 2.

[0175] Table 1

[0176]

[0177] Table 2

[0178]

[0179] The noisy signal is used as a simulated non-stationary voltage signal. This noisy signal is uniformly sampled at a sampling rate of 1200 samples per 0.6 seconds. The method of this invention is used to perform harmonic detection on this non-stationary voltage signal. The obtained amplitude spectrum and phase spectrum are as follows: Figure 3 As shown, from Figure 3 The intermediate amplitude spectrum shows the number of local peaks. The phase spectrum shows the changes in phase magnitude near local peak frequencies, as follows: Figure 3 The rectangular box indicates that the first local peak experienced two phase abrupt changes (from 30° to 0°, then from 0° to approximately 40°), the second local peak experienced one phase abrupt change (from 0° to 100°), and the region corresponding to the fifth local peak corresponds to a plateau region, indicating that no phase abrupt change occurred in the frequency region of that peak. Therefore, the minimum fitting order of the Prony algorithm for this set of signals is determined as follows: .Compare Figure 3 As can be seen from the data in Tables 1 and 2, this invention can analyze stationary harmonics / interharmonics by the number of local peaks and analyze non-stationary harmonics / interharmonics by the number of phase abrupt changes, thus ensuring the accuracy of the analysis.

[0180] After calculating the signal parameters of each fitted component, components with amplitudes less than 0.05% of the fundamental amplitude and components with attenuation factors greater than 10 were removed; the attenuation factors of integer harmonics greater than the third harmonic were set to zero. The final signal parameters are shown in Tables 3 and 4, where Table 3 shows the results of stationary harmonic analysis and Table 4 shows the results of non-stationary harmonic analysis.

[0181] Table 3

[0182]

[0183] Table 4

[0184]

[0185] Comparing Tables 3 and 1, and Tables 2 and 4, it can be seen that the error between the harmonic signal parameter analysis results of this invention and the original signal is very small, indicating that the detection accuracy of the method of this invention is high. The comparison graph of the analysis results of this invention and the signal parameters of the original analog signal is shown below. Figure 4 and Figure 5 As shown, Figure 4 This is an amplitude spectrum comparison chart. Figure 5 For phase spectrum comparison diagram, from Figure 4 and Figure 5 As can be seen, the analysis results of this invention almost completely coincide with the amplitude and phase values ​​of the original analog signal, indicating that the detection accuracy of this invention is high.

[0186] Based on the analysis results of this invention, signal reconstruction is performed to obtain a fitted signal. The comparison result between the fitted signal and the original analog signal is as follows: Figure 6 As shown, from Figure 6 As can be seen, the two signals are basically overlapping, indicating that the detection accuracy of the present invention is high.

[0187] Error analysis was performed on the analytical results of this invention, and the results are as follows: Figure 7 As shown, Figure 7 From top to bottom, the diagrams show the relative error of amplitude, the relative error of frequency, and the relative error of phase. Figure 7 It can be seen that the frequency error is less than 4*10. -5 Amplitude error less than 4*10 -3 The phase error is less than 0.03, and the fitting accuracy is very high.

[0188] Example 2

[0189] Based on the same inventive concept, the present invention also provides a non-stationary power harmonic detection method system, comprising:

[0190] The time-frequency conversion module is used to acquire the continuous electrical signal of the power system, and to perform a full-phase fast Fourier transform on the uniformly sampled data of the continuous electrical signal to obtain the corresponding amplitude spectrum and phase spectrum.

[0191] The order determination module is used to determine the minimum fitting order of the uniformly sampled data based on the local peaks of the amplitude spectrum and the phase changes of the phase spectrum in the region corresponding to the local peaks.

[0192] The analysis and detection module is configured to estimate signal parameters of each fitting component corresponding to the uniformly sampled data by using a Prony algorithm based on the minimum fitting order, and take the signal parameters of each fitting component as signal parameters of each harmonic of the continuous electrical signal, so as to complete harmonic detection of the continuous electrical signal.

[0193] In a possible implementation, the order determination module is specifically configured to:

[0194] determine the minimum fitting order of the uniformly sampled data based on a number of local peaks of the amplitude spectrum and a number of phase mutations of the phase spectrum in a region corresponding to the local peaks.

[0195] The phase mutation is a phase change greater than a preset threshold in a specified frequency region.

[0196] In a possible implementation, the minimum fitting order of the uniformly sampled data is:

[0197] ;

[0198] wherein, is the minimum fitting order of the uniformly sampled data, is the number of local peaks of the amplitude spectrum, is a sum of the number of phase mutations of the phase spectrum in the region corresponding to the local peaks.

[0199] In a possible implementation, the time-frequency conversion module comprises:

[0200] a windowing submodule configured to construct a convolution window based on convolution of two symmetrical windows, perform windowing processing on the uniformly sampled data by using the convolution window to obtain a windowed signal, and add data symmetrically located relative to a center point in the windowed signal to obtain a new signal;

[0201] a transformation submodule configured to perform fast Fourier transform on the new signal to obtain an amplitude spectrum and a phase spectrum of the uniformly sampled data.

[0202] In a possible implementation, the analysis and detection module comprises:

[0203] a fitting equation construction submodule configured to construct a linear fitting equation set based on the minimum fitting order;

[0204] a parameter estimation submodule configured to solve a coefficient matrix of the linear fitting equation set by using a least square method to estimate signal parameters of each fitting component.

[0205] In a possible implementation, the linear fitting equation set is as follows:

[0206] ;

[0207] , ;

[0208] wherein, is any sampling point of the fitting signal, is a sampling time interval, represents an imaginary part, is an i-th element of a coefficient matrix of the linear fitting equation set, and p is a minimum fitting order, is an i-th observation vector of the linear fitting equation set; , , , are a frequency, a decay factor, an amplitude and an initial phase of an i-th fitting component respectively.

[0209] In a possible implementation, the parameter estimation submodule comprises:

[0210] a self-correlation matrix construction subunit configured to construct a self-correlation matrix based on the uniform sampling data and the minimum fitting order;

[0211] a pole solution subunit configured to solve eigenvalues and eigenvectors of the self-correlation matrix, construct a Toeplitz matrix based on the eigenvalues and the eigenvectors, and solve the linear fitting equation set by using a least square method to obtain prony poles;

[0212] a parameter estimation subunit configured to construct a Vandermonde matrix by using the prony poles, estimate a coefficient matrix of the linear fitting equation set based on the uniform sampling data and the Vandermonde matrix, and further obtain signal parameters of each fitting component.

[0213] In a possible implementation, the self-correlation matrix is as follows:

[0214] ;

[0215] ;

[0216] wherein, a matrix dimension is , is a data length of the uniform sampling data, and a superscript T represents a matrix transpose; is a Hankel matrix, a matrix dimension is , , ,… respectively are the values of the sampling points of the uniform sampling data at the current time, is the (N-1)th sampling point of the uniform sampling data at the current time.

[0217] In a possible implementation, the pole solving subunit is specifically configured to:

[0218] perform eigenvalue decomposition on the autocorrelation matrix to obtain an eigenvalue vector and an eigenvector matrix of the autocorrelation matrix; and the matrix dimensions are and respectively;

[0219] reorder the column vectors in the eigenvector matrix in a descending order of eigenvalues in the eigenvalue vector , and extract the first column vectors of the reordered matrix to form a new eigenvector matrix , the matrix dimension is ;

[0220] use the first rows and the last rows of the new eigenvector matrix as preprocessed matrices and , and the matrix dimensions are ;

[0221] construct a Toeplitz matrix based on the preprocessed matrices and , and solve the linear fitting equation set by using a least square method to obtain discrete poles ; is the ith discrete pole, is the eigenvalue decomposition, and the superscript -1 is the inverse of the matrix;

[0222] convert the discrete poles into continuous-time poles, i.e., prony poles ; is the ith prony pole, is the sampling frequency.

[0223] In a possible implementation, the coefficient matrix of the linear fitting equation set is as follows:

[0224] ;

[0225] ;

[0226] , ;

[0227] in, This is the coefficient matrix of the linearly fitted system of equations; For the Vandermonde matrix, The p-th prony pole is raised to the power of N-1. This is the fitted signal obtained by taking N sampling points of the uniformly sampled data;

[0228] The signal parameters of each fitted component are as follows:

[0229] ;

[0230] in, , , , Let be the frequency, attenuation factor, amplitude, and initial phase of the i-th fitted component, respectively. To take the real part of a complex number, This indicates taking the imaginary part of a complex number.

[0231] In one possible implementation, the analysis and detection module further includes a spurious component removal submodule, which is used for:

[0232] Remove fitted components whose amplitude is less than a specified amplitude and fitted components whose absolute value of the attenuation factor is greater than a specified value;

[0233] Set the attenuation factor of harmonics of integer orders greater than a specified order to zero.

[0234] Example 3

[0235] like Figure 8 As shown, the present invention also provides an electronic device, which may be a computer device, a microcontroller device, a smart mobile device, etc. The electronic device in this embodiment may include a processor, a memory, a transceiver component, etc. The memory, processor, and transceiver component are connected via a bus; the memory can be used to store executable programs, and an exemplary executable program may include instructions; the processor is used to execute the instructions stored in the memory. The memory can also be used to store data, which can be accessed and / or modified when instructions are executed.

[0236] The processor can be a central processing unit (CPU), and can also be other general-purpose processors, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, etc., which are the computing core and control core of the terminal, and are suitable for implementing one or more instructions, and are specifically suitable for loading and executing one or more instructions in the storage medium to implement a corresponding method flow or a corresponding function, so as to implement the steps of the non-stationary power harmonic detection method in the above embodiment.

[0237] Embodiment 4

[0238] Based on the same inventive concept, the application further provides a readable storage medium, specifically an electronic device readable storage medium (Memory). The electronic device readable storage medium is a memory device in the electronic device, and is used to store programs and data. It can be understood that the storage medium herein can include a built-in storage medium in the electronic device, and of course can also include an expansion storage medium supported by the electronic device. The storage medium provides a storage space, and the storage space stores an operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space, and the instructions can be one or more execution programs (including program codes). It should be noted that the storage medium herein can be a high-speed RAM memory, or a non-volatile memory such as at least one disk memory. The processor loads and executes one or more instructions stored in the storage medium, and the steps of the non-stationary power harmonic detection method in the above embodiment can be implemented.

[0239] Those skilled in the art should understand that the embodiments of the application can be provided as a method, a system, or a computer program product. Therefore, the application can adopt a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the application can adopt the form of a computer program product implemented on one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer usable program codes.

[0240] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart or flows and / or block diagram block or blocks. Figure 1 one or more flow or flows and / or block diagram block or blocks. Figure 1 one or more flow or flows and / or block diagram block or blocks.

[0241] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart or flows and / or block diagram block or blocks. Figure 1 one or more flow or flows and / or block diagram block or blocks. Figure 1 one or more flow or flows and / or block diagram block or blocks.

[0242] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart or flows and / or block diagram block or blocks. Figure 1 one or more flow or flows and / or block diagram block or blocks. Figure 1 one or more flow or flows and / or block diagram block or blocks.

[0243] Finally, it should be noted that the above embodiments are merely used to illustrate the technical solutions of the present application, rather than to limit the scope of protection of the present application. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that, after reading the present application, they can make various changes, modifications or equivalent replacements to the specific embodiments of the application. However, these changes, modifications or equivalent replacements are all within the scope of protection of the claims of the present application.

Claims

1. A non-stationary power harmonic detection method, characterized in that, The method comprises the following steps: Obtaining continuous electrical signals of a power system, performing full-phase fast Fourier transform on uniformly sampled data of the continuous electrical signals to obtain corresponding amplitude spectrum and phase spectrum; Based on the local peak value of the amplitude spectrum and the phase change of the phase spectrum in the region corresponding to the local peak value, the minimum fitting order of the uniformly sampled data is determined; Based on the minimum fitting order, the signal parameters of each fitting component corresponding to the uniformly sampled data are estimated by using the prony algorithm, and the signal parameters of each fitting component are taken as the signal parameters of each harmonic of the continuous electrical signals to complete the harmonic detection of the continuous electrical signals; Wherein, based on the number of local peak values of the amplitude spectrum and the number of phase mutations of the phase spectrum in the region corresponding to the local peak value, the minimum fitting order of the uniformly sampled data is determined; the phase mutation is that the phase change in the specified frequency region is greater than the preset threshold; The minimum fitting order of the uniformly sampled data is: ； wherein, is the minimum fitting order of the uniform sampling data, is the number of local peaks of the amplitude spectrum, is the sum of phase mutations of the phase spectrum in the region corresponding to the local peaks.

2. The method of claim 1, wherein, Performing full-phase fast Fourier transform on the uniformly sampled data of the continuous electrical signals to obtain corresponding amplitude spectrum and phase spectrum, comprising: Based on the convolution of two symmetric windows, a convolution window is constructed; the uniformly sampled data is windowed by using the convolution window to obtain a windowed signal; Adding the data symmetric to the center point in the windowed signal to obtain a new signal; Performing fast Fourier transform on the new signal to obtain the amplitude spectrum and phase spectrum of the uniformly sampled data.

3. The method of claim 1, wherein, Based on the minimum fitting order, the signal parameters of each fitting component corresponding to the uniformly sampled data are estimated by using the prony algorithm, comprising: Based on the minimum fitting order, a linear fitting equation set is constructed; The coefficient matrix of the linear fitting equation set is solved by using the least square method to estimate the signal parameters of each fitting component.

4. The method of claim 3, wherein, The linear fitting equation set is as follows: ； ， ； wherein, is any sample point of the fitted signal, is the sample point index, is the sample time interval, represents the imaginary part, is the i-th element of the coefficient matrix of the linear fitting equation system, p is the minimum fitting order of the uniformly sampled data, is the i-th observation vector of the linear fitting equation system; , , , are the frequency, the attenuation factor, the amplitude and the initial phase of the i-th fitted component, respectively.

5. The method of claim 3, wherein, The coefficient matrix of the linear fitting equation set is solved by using the least square method to estimate the signal parameters of each fitting component, comprising: Based on the uniformly sampled data and the minimum fitting order, a self-correlation matrix is constructed; The eigenvalues and eigenvectors of the self-correlation matrix are solved, a Toeplitz matrix is constructed based on the eigenvalues and eigenvectors, and the linear fitting equation set is solved by using the least square method to obtain prony poles; A Vandermonde matrix is constructed by using the prony poles, and the signal parameters of each fitting component are obtained based on the uniformly sampled data and the Vandermonde matrix.

6. The method of claim 5, wherein, The autocorrelation matrix As follows: ； ； wherein, matrix dimension is , p is the minimum fitting order of the uniformly sampled data, is half of the data length of the uniformly sampled data, and the superscript T represents matrix transposition; is a Hankel matrix, matrix dimension is , , ,… are respectively each sampling point value of the uniformly sampled data at the current time, is the N-1th sampling point of the uniformly sampled data at the current time.

7. The method of claim 6, wherein, The eigenvalues and eigenvectors of the self-correlation matrix are solved, a Toeplitz matrix is constructed based on the eigenvalues and eigenvectors, and the linear fitting equation set is solved by using the least square method to obtain prony poles, comprising: performing eigenvalue decomposition on the autocorrelation matrix to obtain an eigenvalue vector of the autocorrelation matrix and an eigenvector matrix ; and the matrix dimensions are and ; According to the eigenvalue vector , the column vectors in the eigenvector matrix are reordered in a descending order of eigenvalues, and the first column vectors of the reordered matrix are extracted to form a new eigenvector matrix , The matrix dimension is ; The new feature vector matrix is divided into two parts, the front row and the back row, and the front row and the back row are respectively taken as the preprocessed matrix and , and the matrix dimensions are ; based on the pre-processed matrix and constructing a Toeplitz matrix, solving the linear fitting equations using a least square method to obtain the discrete poles ; for the i-th discrete pole, is an eigenvalue decomposition, and the superscript -1 is an inverse of the matrix; converting the discrete poles into continuous-time poles, i.e. prony poles ; for the i-th prony pole, is the sampling frequency.

8. The method of claim 7, wherein, The coefficient matrix of the linear fitting equation set is as follows: ； ； ， ； wherein, is a coefficient matrix of the linear fitting equation set, is the i-th element of is a Vandermonde matrix, is the N-1 power of the p-th prony pole; is a fitting signal obtained by taking N sampling points of the uniformly sampled data; The signal parameters of each fitting component are as follows: ； wherein , , , are the frequency, the damping factor, the amplitude and the initial phase of the i-th fitted component, respectively, denotes taking the real part of a complex number, denotes taking the imaginary part of a complex number.

9. The method of claim 1, wherein, After obtaining the signal parameters of each fitting component, it further comprises the following steps: Eliminate the fitting components with amplitude less than a specified amplitude and the fitting components with absolute value of attenuation factor greater than a specified value; Set the attenuation factor of the integer harmonic greater than a specified order to zero.

10. A non-stationary power harmonic detection system characterized by, The method comprises the following steps: The time-frequency conversion module is configured to obtain a continuous electrical signal of a power system, perform full-phase fast Fourier transform on uniformly sampled data of the continuous electrical signal, and obtain a corresponding amplitude spectrum and a phase spectrum; The order determination module is configured to determine a minimum fitting order of the uniformly sampled data based on local peak values of the amplitude spectrum and phase variation of the phase spectrum in regions corresponding to the local peak values. The analysis and detection module is configured to estimate signal parameters of each fitting component corresponding to the uniformly sampled data by using a prony algorithm based on the minimum fitting order, and take the signal parameters of each fitting component as signal parameters of each harmonic of the continuous electrical signal, so as to complete harmonic detection of the continuous electrical signal. The minimum fitting order of the uniformly sampled data is determined based on a number of the local peak values of the amplitude spectrum and a number of phase mutations of the phase spectrum in regions corresponding to the local peak values; a phase mutation refers to a phase variation greater than a preset threshold in a specified frequency region. The minimum fitting order of the uniformly sampled data is: ； wherein, is the minimum fitting order of the uniform sampling data, is the number of local peaks of the amplitude spectrum, is the sum of phase mutations of the phase spectrum in the region corresponding to the local peaks.

11. The system of claim 10, wherein, The time-frequency conversion module comprises: A windowing submodule is configured to construct a convolution window based on convolution of two symmetrical windows, perform windowing processing on the uniformly sampled data by using the convolution window to obtain a windowed signal, and add data symmetrically relative to a center point in the windowed signal to obtain a new signal. A transformation submodule is configured to perform fast Fourier transform on the new signal to obtain an amplitude spectrum and a phase spectrum of the uniformly sampled data.

12. The system of claim 10, wherein, The analysis and detection module comprises: A fitting equation construction submodule is configured to construct a linear fitting equation set based on the minimum fitting order. A parameter estimation submodule is configured to solve a coefficient matrix of the linear fitting equation set by using a least square method to estimate signal parameters of each fitting component.

13. The system of claim 12, wherein, The linear fitting equation set is as follows: ； ， ； wherein, is any sample point of the fitted signal, is the sample point index, is the sample time interval, represents the imaginary part, is the i-th element of the coefficient matrix of the linear fitting equation system, p is the minimum fitting order, is the i-th observation vector of the linear fitting equation system; , , , are the frequency, the attenuation factor, the amplitude and the initial phase of the i-th fitted component, respectively.

14. The system of claim 12, wherein, The parameter estimation submodule comprises: An autocorrelation matrix construction subunit is configured to construct an autocorrelation matrix based on the uniformly sampled data and the minimum fitting order. A pole solving subunit is configured to solve eigenvalues and eigenvectors of the autocorrelation matrix, construct a Toeplitz matrix based on the eigenvalues and eigenvectors, and solve the linear fitting equation set by using a least square method to obtain prony poles. A parameter estimation subunit is configured to construct a Vandermonde matrix by using the prony poles, estimate a coefficient matrix of the linear fitting equation set based on the uniformly sampled data and the Vandermonde matrix, and further obtain signal parameters of each fitting component.

15. The system of claim 14, wherein, The autocorrelation matrix As follows: ； ； wherein, matrix dimension is , p is the minimum fitting order of the uniformly sampled data, is half of the data length of the uniformly sampled data, and the superscript T is the matrix transpose; is a Hankel matrix, matrix dimension is , , ,… are respectively each sampling point value of the uniformly sampled data at the current time, is the N-1th sampling point of the uniformly sampled data at the current time.

16. The system of claim 15, wherein, The pole solving subunit is specifically configured to: performing eigenvalue decomposition on the autocorrelation matrix to obtain an eigenvalue vector of the autocorrelation matrix and an eigenvector matrix ; and the matrix dimensions are and ; According to eigenvalue vector The column vectors in the characteristic vector matrix are reordered in descending order of eigenvalues, and the first column vectors of the reordered matrix are extracted to form a new characteristic vector matrix , The matrix dimension is ; The new feature vector matrix is divided into a front row and a back row, and the front row and the back row are respectively taken as a preprocessed matrix and , and The matrix dimensions are ; based on the pre-processed matrix and constructing a Toeplitz matrix, solving the linear fitting equations using a least square method to obtain the discrete poles ; for the i-th discrete pole, is an eigenvalue decomposition, and the superscript -1 is an inverse of the matrix; converting the discrete poles into continuous-time poles, i.e. prony poles ; for the i-th prony pole, is the sampling frequency.

17. The system of claim 16, wherein, The coefficient matrix of the linear fitting equation set is as follows: ； ； ， ； wherein, is a coefficient matrix of the linear fitting equation set; is a Vandermonde matrix, is the N-1 power of the pth prony pole; is a fitting signal obtained by taking N sampling points of the uniformly sampled data; The signal parameters of each fitting component are as follows: ； wherein , , , are the frequency, the damping factor, the amplitude and the initial phase of the i-th fitting component, respectively, denotes taking the real part of a complex number, denotes taking the imaginary part of a complex number.

18. The system of claim 10, wherein, The analysis and detection module further comprises a false component elimination submodule, which is configured to: Eliminate fitting components with an amplitude less than a specified amplitude and fitting components with an absolute value of an attenuation factor greater than a specified value. Set attenuation factors of integer order harmonics greater than a specified order to zero.

19. An electronic device, comprising: Comprise: At least one processor and a memory; The memory and the processor are connected through a bus; The memory is configured to store one or more programs; The one or more programs, when executed by the at least one processor, implement the method of any one of claims 1 to 9.

20. A readable storage medium, characterized by, Having a program stored thereon, which program, when executed, implements the method of any one of claims 1 to 9.

Images