Power cable head end anti-radio frequency domain suppression method and device based on dynamic mode decomposition, computer storage medium and electronic equipment

Abstract

The invention belongs to the technical field of power cable monitoring, and discloses a power cable head end anti-radio frequency domain suppression method and device based on dynamic mode decomposition, a computer storage medium and electronic equipment, and the method comprises the steps: obtaining a cable frequency domain head end reflection coefficient spectrum reflection signal, and carrying out the preprocessing to obtain a signal; constructing a spatio-temporal data matrix and a corresponding time migration matrix according to the signals, and performing dynamic mode decomposition and reconstruction; and calculating the frequency and attenuation rate of each reconstruction mode, screening the reconstruction modes, and reconstructing the interference-removed signal according to the screened modes. According to the method, the dynamic mode can be adaptively extracted, the head end reflection interference is effectively separated and eliminated, and the defect is accurately positioned.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of electric power engineering, and relates to power cable detection technology, in particular to a power cable head-end reflection frequency domain suppression method and device based on dynamic modal decomposition, a computer storage medium and an electronic device. BACKGROUND

[0002] Traditional frequency domain reflectometry (FDR) realizes power cable defect positioning and diagnosis by injecting a sweep signal into the cable. However, because the characteristic impedance of the cable body and the characteristic impedance of the test signal line are different in size, the sweep signal injected by the frequency domain reflectometry will be multiple times folded and reflected at the connection point of the test signal line and the head end of the cable body, resulting in the collected cable frequency domain reflection coefficient spectrum containing a large amount of head-end harmonic information. If the cable frequency domain reflection coefficient spectrum containing the head-end harmonic information is directly subjected to fast Fourier transform (FFT) analysis, false peaks are likely to occur in the FDR test results, seriously affecting the defect diagnosis results and reducing the diagnosis accuracy and positioning accuracy. Therefore, it is necessary to study a method to suppress the harmonic noise caused by the cable head end and eliminate false peak interference.

[0003] Existing head-end reflection frequency domain suppression methods are all based on static modal decomposition processing methods, mainly including wavelet transform, empirical mode decomposition (EMD), etc. These methods extract the head-end frequency domain harmonic noise from the components by the inherent harmonic properties of the head end in the frequency domain, and suppress the interference of the head-end reflection in the frequency domain from a mathematical level. Wavelet transform is a non-adaptive decomposition method, and its decomposition result is affected by the mother wavelet and the signal decomposition level, and there are problems of over-decomposition or under-decomposition in different cable applications, resulting in the inability to filter out the head-end reflection and serious interference to the test results. The empirical mode decomposition (EMD) method automatically decomposes the frequency domain reflection coefficient spectrum into a limited number of modal components according to the signal characteristics, and performs well in processing linear and stationary signals. However, the cable insulation state, length, number of joints and environmental factors are different, which makes the cable reflection coefficient spectrum present time-varying nonlinear characteristics, and further causes the EMD to have problems of multi-modal aliasing and end effect in processing the cable frequency domain reflection coefficient spectrum, resulting in the inability to effectively separate the head-end reflection peak and affecting the cable defect diagnosis and positioning. Therefore, the current head-end reflection suppression method based on static modal decomposition is limited by the different operating characteristics of different cables, resulting in multi-modal signal aliasing in the algorithm processing, the inability to effectively separate the head-end reflection, and serious interference to the judgment.

[0004] Traditional dynamic modal decomposition is mainly applied in radar field to realize target main modal recognition. However, due to different cable terminal types, running states and laying lengths, the reflection signals collected at the cable head end are significantly different. If the dynamic modal decomposition is directly used for the cable reflection signals, it will cause insufficient or excessive decomposition, and seriously interfere with defect judgment. In order to adapt the dynamic modal decomposition to the characteristics of different cables, realize adaptive matching of parameters and effectively suppress the head end reflection, the application provides a power cable head end reflection frequency domain suppression method based on dynamic modal decomposition. SUMMARY

[0005] The application aims to solve the above problems and provides a power cable head end reflection frequency domain suppression method based on dynamic modal decomposition. The method can simultaneously process the aliasing signals formed by superposition of multiple dynamic processes without presetting the number of modal components, and can perform signal noise reduction before separating the modes, greatly reducing the interference caused by complex noise, so as to effectively separate and output the head end reflection peak. The core idea of dynamic modal decomposition is to find an optimal linear approximation model for the observed nonlinear system, and to discover the dominant dynamic mode of the system through data-driven method. The dynamic modal decomposition method firstly constructs the collected signals into two time-delayed data matrices, then performs dimension reduction on the original cable frequency domain spectrum data through singular value decomposition, and then constructs a low-dimensional linear operator to describe the dynamic characteristics of the system. By performing characteristic decomposition on the linear operator, the modes representing the dynamic characteristics of the system and their corresponding growth / decay rates and oscillation frequencies can be obtained. These modes represent different dynamic components in the signal, and can effectively separate the characteristic frequencies caused by the cable head end and the interference frequencies caused by multiple refractions.

[0006] Compared with the traditional method, the dynamic modal decomposition has several significant advantages: first, the dynamic modal decomposition is a completely data-driven method, which does not need to assume the number of modalities of the signal or select the base function, and has stronger adaptability for processing cables with different characteristics; second, the dynamic modal decomposition can provide the spatial structure and time evolution law of the mode, which can not only identify the defect characteristic frequency, but also analyze its dynamic characteristics; third, the dynamic modal decomposition is established through strict mathematical derivation, and the results of modal decomposition correspond to the characteristic frequencies of the impedance mismatch points in the cable system; finally, the dynamic modal decomposition algorithm has high calculation efficiency and is suitable for processing cable frequency domain data under complex conditions.

[0007] In cable defect detection, the dynamic modal decomposition can adaptively separate the aliasing signals according to the running state characteristics of different cables, accurately extract the characteristic frequency of the defect position, especially the head end reflection with strong reflection energy, and suppress the interference frequency caused by multiple refractions, thereby improving the accuracy of defect positioning.

[0008] Based on the above analysis, the application provides a power cable head end reflection frequency domain suppression method based on dynamic modal decomposition, which comprises the following steps: Obtain a cable frequency domain head end reflection coefficient spectrum reflection signal S(i) and pre-process the signal S(i) to obtain a signal S'(i); Construct a time-space data matrix X and a corresponding time offset matrix X' according to the signal S'(i); Perform dynamic modal decomposition on the time-space data matrix X and the corresponding time offset matrix X' and reconstruct; Calculate the frequency and attenuation rate of each reconstructed modal;

[0009] According to the frequency and attenuation rate of each reconstructed modal, the reconstructed modal is screened, and a disturbance signal is reconstructed according to the screened modal.

[0010] In an implementable manner, the obtained cable frequency domain head end reflection coefficient spectrum reflection signal S(i) is pre-processed, including denoising and normalization processing. Specifically, the collected head end reflection coefficient spectrum reflection signal S(i) is denoised by using wavelet threshold denoising or Kalman filtering to suppress measurement noise. The amplitude of the denoised reflection signal is normalized to eliminate the influence of device gain. The signal after denoising and normalization processing is S'(i).

[0011] In an implementable manner, the Hankel matrix is constructed according to the signal S'(i). Specifically, the signal S'(i) is converted into a time domain reflection signal x(t), and the time-space data matrix X and the corresponding time offset matrix X' are constructed according to the time domain reflection signal x(t). Specifically, the conversion formula of the frequency domain signal sequence S'(i) into the time domain reflection signal x(t) is as follows: ; Wherein, f min is the starting frequency of the dispersion characteristic, which is determined by the dispersion characteristic of the cable material, and the XLPE cable is usually 1-50MHz, that is, f min =1MHz, is an equal interval sampling time, N is the number of frequency domain sampling points, k is the time index, and i is the frequency index.

[0012] The time domain signal x(t) is converted into a time-space data matrix X, which is represented as: ; The time-space data matrix X is represented as: ; Wherein, m and n respectively represent the embedding dimension and the total length of the time series.

[0013] In an implementation, the step of performing DMD (Dynamic Mode Decomposition) modal decomposition and reconstruction on the spatiotemporal data matrix X and the corresponding temporal shift matrix X' is as follows:

[0014] Performing truncated SVD (Singular Value Decomposition) decomposition on the spatiotemporal data matrix X to obtain left singular vectors, a singular value matrix, and right singular vectors;

[0015] Constructing an approximate linear dynamic matrix according to the temporal shift matrix X', the left singular vectors, the right singular vectors, and the singular value matrix;

[0016] Performing eigenvalue decomposition on the approximate linear dynamic matrix to obtain an eigenvalue matrix and an eigenvector matrix;

[0017] Reconstructing a dynamic modal decomposition mode according to the temporal shift matrix X', the right singular vectors, the singular value matrix, and the eigenvector matrix.

[0018] Performing SVD decomposition on the spatiotemporal data matrix X, which is expressed as: ; wherein U r ∈R m×r represents left singular vectors (spatial modes); Σ r ∈R r×r represents a singular value matrix (energy weight); and V r ∈R n×r represents right singular vectors (temporal evolution).

[0019] The truncated rank r can be determined according to the singular value decay curve, and the main dynamic components are retained.

[0020] In an implementation, the approximate linear dynamic matrix is calculated according to the following formula: ; wherein represents the transpose of U r ; and represents the inverse matrix of Σ r . This matrix describes the evolution law of the system in a low-dimensional space.

[0021] Then, the approximate linear dynamic matrix is subjected to eigenvalue decomposition, which is expressed as: ; wherein Λ represents an eigenvalue matrix, the eigenvalue λ j represents the dynamic characteristics of the mode; and W represents an eigenvector matrix.

[0022] In an implementation, the dynamic modal decomposition modes are reconstructed according to the following formula based on the time offset matrix X', the right singular vector, the singular value matrix and the eigenvector: ;

[0023] Each mode corresponds to a dynamic behavior in the cable system (such as defect oscillation, reflection interference, etc.).

[0024] In an implementation, the frequency and damping rate of each reconstructed mode are calculated according to the following formula based on the eigenvalue λ j (discrete-time eigenvalue): Frequency: ; Damping rate: .

[0025] In an implementation, the reconstructed modes are screened according to the frequency and damping rate of each reconstructed mode, and the interference signal is reconstructed based on the screened modes, with the purpose of effective mode screening and interference signal reconstruction according to the physical characteristics of the cable. The specific steps are as follows:

[0026] Modes with a damping rate less than a preset stability threshold are set to zero;

[0027] Modes outside a given frequency range are set to zero;

[0028] The interference signal is reconstructed according to the following formula based on the remaining screened modes: ; Where S is the set of screened modes, j is the mode index, is the jth screened mode, b j is the initial amplitude of the mode, t is the time variable, λ j is the eigenvalue of the jth mode, is the sampling time interval.

[0029] In an implementation, modes with a damping rate less than a preset stability threshold are set to zero, which is for stability screening, keeping approximately neutral stable modes and setting to zero fast-decaying or growing interference modes, represented as: ; Where ε stable is the preset stability threshold, generally 0.01-0.05 s -1 When |γ j |≈0, the mode amplitude is constant, representing the true physical oscillation of the cable defect. When |γ j |>0, the mode decays or grows rapidly, which is a head-end reflection interference or non-physical noise.

[0030] In one feasible approach, zeroing out modes outside a given frequency range is done in order to select a mode frequency f that matches the dispersion characteristics of the cable material. j The modal frequencies must satisfy: f min ≤f j ≤f max ; [f min ,f max The frequency dispersion characteristics of XLPE cables are determined by the frequency distribution properties of the cable material; the range is 1-50MHz, i.e., f min =1MHz, f max =50MHz. Since the oscillation frequency of a real defect must conform to the physical laws of cable propagation constant, and the frequency of interference modes often exceeds the effective bandwidth of the material, modes outside the given frequency range are set to zero.

[0031] In one feasible approach, in the reconstructed de-interference signal, b j Calculate using the following formula: ; in, For all modes The matrix formed by these elements, where x1 is the first column of the spatiotemporal data matrix X, is [x(t1), x(t2), ..., x(t...]. m )] T † is the inverse pseudo-operator.

[0032] The present invention also provides a frequency domain suppression system for the head-end reflection of power cables based on dynamic mode decomposition, comprising: The signal acquisition module is used to acquire the reflection signal S(i) of the reflection coefficient spectrum at the beginning of the cable in the frequency domain and to preprocess it to obtain the signal S′(i). The Hankel matrix construction module is used to construct the spatiotemporal data matrix X and the corresponding time offset matrix X′ based on the signal S′(i); The DMD mode decomposition and reconstruction module is used to perform DMD mode decomposition and reconstruction on the spatiotemporal data matrix X and the corresponding time offset matrix X′. The calculation unit is used to calculate the frequency and attenuation rate of each reconstructed mode; The interference removal signal reconstruction module is used to filter the reconstructed modes based on the frequency and attenuation rate of each reconstructed mode, and reconstruct the interference removal signal based on the selected modes.

[0033] The present invention also provides a computer storage medium storing a plurality of instructions adapted for loading by a processor and executing the steps of the power cable head-end reflection frequency domain suppression method as described in any of the preceding implementable embodiments.

[0034] The present invention also provides an electronic device, including a processor and a computer storage medium as described above; the processor executes steps of the power cable head-end reflection frequency domain suppression method as described in any of the preceding implementable embodiments, stored in the computer storage medium.

[0035] Compared with the prior art, the present invention has the following technical effects:

[0036] 1) This invention is based on dynamic mode decomposition and reconstruction, which can achieve adaptive extraction of dynamic modes;

[0037] 2) Based on the adaptively extracted dynamic modes, this invention can obtain the real dynamic behavior of the cable system corresponding to each mode (such as defect oscillation and interference reflection), which has physical interpretability;

[0038] 3) This invention effectively separates and eliminates head-end reflection interference through modal stability analysis, and accurately locates defects. Attached Figure Description

[0039] Figure 1 This is a schematic diagram of the frequency domain suppression method for the first-end reflection of power cables based on dynamic mode decomposition provided in Embodiment 1 of the present invention;

[0040] Figure 2 This is a flowchart illustrating the process of dynamic modal decomposition and reconstruction of a spatiotemporal data matrix.

[0041] Figure 3 A schematic diagram of the interference removal signal reconstruction process;

[0042] Figure 4 This is a model of a cable with defects.

[0043] Figure 5 The image shows the cable location spectrum; where (a) corresponds to the identification result of the traditional FDR method, and (b) corresponds to the identification result of the FDR method based on the reconstructed signal of the present invention. Detailed Implementation

[0044] The technical solutions of various embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0045] Example 1

[0046] This embodiment provides a frequency domain suppression method for power cable head-end reflections based on dynamic mode decomposition. By using dynamic mode decomposition technology to separate cable defect signals from head-end multiple reflection interference signals, the defect location accuracy of FDR is improved.

[0047] The frequency domain suppression method for power cable head-end reflection based on dynamic mode decomposition provided in this embodiment is as follows: Figures 1-3 As shown, it includes the following steps:

[0048] S1, obtain the reflection signal S(i) of the first end reflection coefficient spectrum of the cable in the frequency domain, and perform preprocessing to obtain signal S′(i).

[0049] A wideband sweep frequency signal (frequency range: 1MHz-100MHz, adjustable step size) is injected into the cable under test, and the reflection coefficient spectrum reflection signal S(i) of the cable's head end is collected. The reflection signal includes the characteristic reflection signal of the cable defect (target signal) and multiple reflection signals from the cable head end and connector (interference signal).

[0050] Then, the collected cable frequency domain first-end reflection coefficient spectrum reflection signal S(i) is denoised and normalized.

[0051] Denoising: Wavelet thresholding or Kalman filtering is used to process the acquired first-end reflection coefficient spectrum reflection signal S(i) to suppress measurement noise.

[0052] Normalization: The amplitude of the denoised reflected signal is normalized to eliminate the influence of device gain. The signal after denoising and normalization is S′(i).

[0053] S2, construct the spatiotemporal data matrix X and the corresponding time offset matrix X′ based on the signal S′(i).

[0054] This step constructs the Hankel matrix, converting the time-domain signal into a spatiotemporal data matrix.

[0055] The signal S′(i) is converted into a time-domain reflected signal x(t), and a spatiotemporal data matrix X and a corresponding time offset matrix X′ are constructed based on the time-domain reflected signal x(t).

[0056] The frequency domain signal sequence S′(i) is further transformed into a time domain reflected signal x(t): ; Among them, f min The starting frequency of the dispersion characteristics is determined by the dispersion characteristics of the cable material. For XLPE cables, it is typically 1-50MHz, or f. min =1MHz, The sampling time is at equal intervals, N is the number of frequency domain sampling points, k is the time index, and i is the frequency index.

[0057] The time-domain signal x(t) is converted into a spatiotemporal data matrix X, represented as: ; The time offset matrix X′ of the spatiotemporal data matrix X is represented as: ; Where m and n represent the embedding dimension and the total length of the time series, respectively.

[0058] The time offset matrix X′ is used to describe the dynamic evolution of the system.

[0059] S3 performs dynamic mode decomposition and reconstruction on the spatiotemporal data matrix X and the corresponding time offset matrix X′.

[0060] In this step, singular value decomposition (SVD) is performed on the data matrix to reduce its dimensionality and construct a low-dimensional linear dynamic model. Then, the dynamic mode decomposition mode is reconstructed.

[0061] Step S3 includes the following sub-steps:

[0062] S31. Perform truncated SVD decomposition on the spatiotemporal data matrix X to obtain the left singular vector, singular value matrix and right singular vector.

[0063] The spatiotemporal data matrix X is decomposed using SVD, and expressed as: ; Among them, U r ∈R m×r Represents the left singular vector (spatial mode); Σ r ∈R r×r V represents the singular value matrix (energy weights); r ∈R n×r This represents a right singular vector (time evolution).

[0064] The truncated rank r is determined based on the singular value decay curve, preserving the main dynamic components.

[0065] S32, construct an approximate linear dynamic matrix based on the time offset matrix X′, the left singular vector, the right singular vector, and the singular value matrix.

[0066] The approximate linear dynamic matrix is ​​calculated using the following formula: ; in, U r transpose; Represents Σ r The inverse matrix of the system. This matrix describes the evolution of the system in low-dimensional space.

[0067] S33, perform eigenvalue decomposition on the approximate linear dynamic matrix to obtain the eigenvalue matrix and eigenvector matrix.

[0068] For approximate linear dynamic matrix Eigenvalue decomposition is performed, which is expressed as: ; Where Λ represents the eigenvalue matrix, and the eigenvalues ​​λ j W represents the dynamic characteristics of the mode; W represents the eigenvector matrix.

[0069] S34. Reconstruct the dynamic mode decomposition mode based on the time offset matrix X′, the right singular vector, the singular value matrix, and the eigenvector matrix.

[0070] Reconstruct the dynamic mode decomposition mode according to the following formula: ; Each mode This corresponds to a dynamic behavior in a cable system (such as defect oscillation, reflection interference, etc.).

[0071] S4, calculate the frequency and decay rate of each reconstructed mode.

[0072] This step calculates the dynamic characteristics corresponding to each reconstructed mode, such as growth / decay rate and oscillation frequency.

[0073] Specifically, based on the eigenvalue λ j (That is, discrete-time eigenvalues), calculate the frequency and decay rate of each reconstructed mode according to the following formula: frequency: ; Attenuation rate: .

[0074] S5: Based on the frequency and attenuation rate of each reconstructed mode, the reconstructed modes are screened, and the interference signal is reconstructed based on the screened modes.

[0075] This step involves effective mode screening and interference-free signal reconstruction based on the physical characteristics of the cable. True modes related to cable defects are screened based on physical characteristics (such as modal stability and frequency matching), while interfering modes caused by head-end reflections are eliminated.

[0076] Step S5 specifically includes the following sub-steps:

[0077] S51 sets the attenuation rate of modes with a decay rate less than the preset stability threshold to zero.

[0078] Stability screening is performed based on the decay rate, retaining approximately neutral and stable modes, while setting rapidly decaying or growing disturbance modes to zero, as follows: ; Where, ε stable The preset stability threshold is typically set to 0.01~0.05s. -1 When |γ j When |≈0, the modal amplitude is constant, representing the true physical oscillation of the cable defect. When |γ j When |>0, the mode decays or grows rapidly, which is due to head-end reflection interference or non-physical noise.

[0079] S52 sets the modes outside the given frequency range to zero.

[0080] Select a modal frequency f that matches the dispersion characteristics of the cable material. j The modal frequencies must satisfy: f min ≤f j ≤f max ; f min f max These are the starting and ending frequencies of the dispersion characteristic, determined by the dispersion characteristics of the cable material. For XLPE cables, this ranges from 1-50MHz, or f. min =1MHz, f max =50MHz. Since the oscillation frequency of a real defect must conform to the physical laws of cable propagation constant, and the frequency of interference modes often exceeds the effective bandwidth of the material, modes outside the given frequency range are set to zero.

[0081] S53, based on the remaining selected modes, reconstruct the interference-removing signal according to the following formula: ; Where S is the filtered modality set, and j is the modality index. For the j-th mode being selected, b j λ is the initial amplitude of the mode, t is the time variable, and λ is the initial amplitude of the mode. j Let j be the eigenvalue of the j-th mode. This represents the sampling time interval.

[0082] b is calculated as follows j : ; in, For all modes The matrix formed by these elements, where x1 is the first column of the spatiotemporal data matrix X, is [x(t1), x(t2), ..., x(t...]. m )] T † is the inverse pseudo-operator.

[0083] The cable was simulated using the method described above, such as... Figure 4As shown, the cable length is set to l = 300 m, the starting position of the shielding layer damage defect is x1 = 100 m, the ending position is x2 = 101 m, the capacitance change factor (relative to the cable body) is set to 1.5 times, and a 0-100MHz sinusoidal sweep frequency signal is input at the head end.

[0084] The localization spectrum obtained directly using the traditional FDR method is as follows: Figure 5 As shown in (a), the traditional FDR method (see Novel Local Defect Location Method for Power Cables Based on Reflection Coefficient Spectrum; Xie Min, Zhou Kai, Zhao Shilin, He Min, Zhang Fuzhong; Power System Technology, 2017, No. 9, pp. 3083-3089).

[0085] The reflected signal is first reconstructed using the method described in this embodiment, and then the localization spectrum is obtained using the traditional FDR method, such as... Figure 5 As shown in (b).

[0086] from Figure 5 (a) It can be seen that although the cable defect can be detected at 100m from the detection results obtained from the original signal, a false detection result appears at 200m due to the influence of multiple refractions of the signal. The peak value of the false peak is relatively large, making it easy to be misjudged as a defect. However, the detection curve obtained after reconstructing the signal (such as...) Figure 5 As shown in (b), the influence of the spurious peak at 200 m on defect detection was eliminated. Furthermore, the influence of front-end reflection on detection was also eliminated, making the detection results more reliable.

[0087] Example 2

[0088] This embodiment provides a power cable head-end reflection frequency domain suppression system based on dynamic mode decomposition, which is used to implement the power cable head-end reflection frequency domain suppression method based on dynamic mode decomposition in Embodiment 1.

[0089] The power cable head-end reflection frequency domain suppression system based on dynamic mode decomposition includes: The signal acquisition module is used to acquire the reflection signal S(i) of the reflection coefficient spectrum at the beginning of the cable in the frequency domain and to preprocess it to obtain the signal S′(i). The Hankel matrix construction module is used to construct the spatiotemporal data matrix X and the corresponding time offset matrix X′ based on the signal S′(i); The DMD mode decomposition and reconstruction module is used to perform DMD mode decomposition and reconstruction on the spatiotemporal data matrix X and the corresponding time offset matrix X′. The calculation unit is used to calculate the frequency and attenuation rate of each reconstructed mode; The interference removal signal reconstruction module is used to filter the reconstructed modes based on the frequency and attenuation rate of each reconstructed mode, and reconstruct the interference removal signal based on the selected modes.

[0090] The signal acquisition module, Hankel matrix construction module, DMD mode decomposition and reconstruction module, calculation unit and interference removal signal reconstruction module are operated according to the steps given in the power cable head-end reflection frequency domain suppression method based on dynamic mode decomposition in Example 1.

[0091] Example 3

[0092] This embodiment provides a computer storage medium that stores multiple instructions, which are adapted to be loaded by a processor and executed by the processor to perform the steps of the power cable head-end reflection frequency domain suppression method based on dynamic mode decomposition provided in Embodiment 1.

[0093] Example 4

[0094] This embodiment provides an electronic device, including a processor and a computer storage medium. The processor executes the steps of the power cable head-end reflection frequency domain suppression method based on dynamic mode decomposition provided in Embodiment 1, which are stored in the computer storage medium.

[0095] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the principles of the invention, and should be understood that the scope of protection of the invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations based on the technical teachings disclosed in this invention without departing from the spirit of the invention, and these modifications and combinations are still within the scope of protection of this invention.

Claims

1. A method for suppressing the frequency domain reflection at the beginning of a power cable based on dynamic mode decomposition, characterized in that, Includes the following steps: The reflection signal S(i) of the first end of the cable in the frequency domain is obtained and preprocessed to obtain the signal S′(i); Construct the spatiotemporal data matrix X and the corresponding time offset matrix X′ based on the signal S′(i); Perform DMD mode decomposition and reconstruction on the spatiotemporal data matrix X and the corresponding time offset matrix X′; Calculate the frequency and decay rate of each reconstructed mode; Based on the frequency and attenuation rate of each reconstructed mode, the reconstructed modes are screened, and the interference signal is reconstructed based on the screened modes.

2. The method for suppressing the frequency domain reflection at the beginning of a power cable based on dynamic mode decomposition according to claim 1, characterized in that, The obtained cable frequency domain first-end reflection coefficient spectrum reflection signal S(i) is preprocessed, including: using wavelet threshold denoising or Kalman filtering to denoise the acquired first-end reflection coefficient spectrum reflection signal S(i); and performing amplitude normalization processing on the denoised reflection signal.

3. The method for suppressing the frequency domain reflection at the beginning of a power cable based on dynamic mode decomposition according to claim 1, characterized in that, The specific steps for constructing the spatiotemporal data matrix X and the corresponding time offset matrix X′ based on the signal S′(i) are as follows: convert the signal S′(i) into a time-domain reflected signal x(t), and construct the spatiotemporal data matrix X and the corresponding time offset matrix X′ based on the time-domain reflected signal x(t).

4. The method for suppressing the frequency domain reflection at the beginning of a power cable based on dynamic mode decomposition according to claim 3, characterized in that, The conversion formula for further transforming the frequency domain signal sequence S'(i) into the time domain reflected signal x(t) is as follows: ； Among them, f min The starting frequency of the dispersion characteristic. The sampling time is at equal intervals, N is the number of frequency domain sampling points, k is the time index, and i is the frequency index; The time-domain signal x(t) is converted into a spatiotemporal data matrix X, represented as: ； The time offset matrix X′ of the spatiotemporal data matrix X is represented as: ； Where m and n represent the embedding dimension and the total length of the time series, respectively.

5. The method for suppressing the frequency domain reflection at the beginning of a power cable based on dynamic mode decomposition according to claim 1, characterized in that, The steps for performing DMD mode decomposition and reconstruction on the spatiotemporal data matrix X and the corresponding time offset matrix X′ are as follows: The spatiotemporal data matrix X is subjected to truncated SVD decomposition to obtain the left singular vector, the singular value matrix and the right singular vector; An approximate linear dynamic matrix is ​​constructed based on the time offset matrix X′, the left singular vector, the right singular vector, and the singular value matrix. Eigenvalue matrix and eigenvector matrix are obtained by performing eigenvalue decomposition on the approximate linear dynamic matrix; The dynamic mode decomposition mode is reconstructed based on the time offset matrix X′, the right singular vector, the singular value matrix, and the eigenvector matrix.

6. The method for suppressing the frequency domain reflection at the beginning of a power cable based on dynamic mode decomposition according to claim 5, characterized in that, The spatiotemporal data matrix X is decomposed using SVD, and expressed as: ； Among them, U r ∈R m×r Represents the left singular vector; Σ r ∈R r×r V represents the singular value matrix; r ∈R n×r This represents a right singular vector.

7. The method for suppressing the frequency domain reflection at the beginning of a power cable based on dynamic mode decomposition according to claim 5, characterized in that, The approximate linear dynamic matrix is ​​calculated using the following formula: ； in, U r transpose; Represents Σ r The inverse matrix.

8. The method for suppressing the frequency domain reflection at the beginning of a power cable based on dynamic mode decomposition according to claim 7, characterized in that, For approximate linear dynamic matrix Eigenvalue decomposition is performed, which is expressed as: ； Where Λ represents the eigenvalue matrix, and the eigenvalues ​​λ j W represents the dynamic characteristics of the mode; W represents the eigenvector matrix.

9. The method for suppressing frequency domain reflections at the beginning of power cables based on dynamic mode decomposition according to claim 8, characterized in that, Based on the time offset matrix X′, the right singular vector, the singular value matrix, and the eigenvectors, the dynamic mode decomposition mode is reconstructed according to the following formula: 。 10. The method for suppressing the frequency domain reflection at the beginning of a power cable based on dynamic mode decomposition according to claim 9, characterized in that, Based on the eigenvalue λ j The frequency and decay rate of each reconstructed mode are calculated according to the following formula: frequency: ; Attenuation rate: .

11. The method for suppressing the frequency domain reflection at the head end of a power cable based on dynamic mode decomposition according to any one of claims 1 to 10, characterized in that, Based on the frequency and attenuation rate of each reconstructed mode, the reconstructed modes are screened, and the interference signal is reconstructed based on the screened modes. The steps are as follows: Modes with attenuation rates less than a preset stability threshold are set to zero; Set modes outside the given frequency range to zero; Based on the remaining selected modes, the interference-removing signal is reconstructed according to the following formula: ； Where S is the filtered modality set, and j is the modality index. For the j-th mode being selected, b j λ is the initial amplitude of the mode, t is the time variable, and λ is the initial amplitude of the mode. j Let j be the eigenvalue of the j-th mode. This represents the sampling time interval.

12. The method for suppressing the frequency domain reflection at the beginning of a power cable based on dynamic mode decomposition according to claim 11, characterized in that, In the reconstructed interference-free signal, b j Calculate using the following formula: ； in, For all modes The matrix formed by these elements, where x1 is the first column of the spatiotemporal data matrix X, is [x(t1), x(t2), ..., x(t...]. m )] T † is the inverse pseudo-operator.

13. A frequency domain suppression system for head-end reflection of power cables based on dynamic mode decomposition, characterized in that, Includes the following steps: The signal acquisition module is used to acquire the reflection signal S(i) of the reflection coefficient spectrum at the beginning of the cable in the frequency domain and to preprocess it to obtain the signal S'(i). The Hankel matrix construction module is used to construct the spatiotemporal data matrix X and the corresponding time offset matrix X′ based on the signal S'(i); The DMD mode decomposition and reconstruction module is used to perform DMD mode decomposition and reconstruction on the spatiotemporal data matrix X and the corresponding time offset matrix X′. The calculation unit is used to calculate the frequency and attenuation rate of each reconstructed mode; The interference removal signal reconstruction module is used to filter the reconstructed modes based on the frequency and attenuation rate of each reconstructed mode, and reconstruct the interference removal signal based on the selected modes.

14. A computer storage medium, characterized in that, The computer storage medium stores multiple instructions, which are adapted to be loaded by a processor and executed by the processor to perform the steps of the power cable head-end reflection frequency domain suppression method based on dynamic mode decomposition as described in any one of claims 1-12.

15. An electronic device, characterized in that, The device includes a processor and the computer storage medium of claim 14; the processor executes the steps of the power cable head-end reflection frequency domain suppression method based on dynamic mode decomposition as described in any one of claims 1-12, stored in the computer storage medium.

Images