Traveling wave waveform detection method and device based on complementary set empirical mode decomposition and least square QR decomposition algorithm

Through the method based on complementary set empirical modal decomposition and least squares QR decomposition algorithm, the primary traveling wave signal of the power grid is accurately extracted, solving the problem that the secondary traveling wave signal in the existing technology does not truly reflect the primary traveling wave signal, and improving the Accuracy of fault location and system stability.

CN119089690APending Publication Date: 2024-12-06CHINA SOUTHERN POWER GRID COMPANY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411211187.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-30
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

The existing traveling wave detection technology cannot accurately detect the primary traveling wave signal, resulting in the waveform characteristics of the secondary traveling wave signal not truly reflecting the characteristics of the primary traveling wave signal of the power grid, affecting the stable operation of the system.

Method used

Using the method based on complementary set empirical modal decomposition and least squares QR decomposition algorithm, the transfer function model of the traveling wave sensor is established to perform complementary set empirical modal decomposition on the quadratic traveling wave signal, and using Lanczos diagonalization and QR decomposition The algorithm performs inversion, obtains the inherent modal components of the primary traveling wave signal, and superimposes them to obtain the inverted primary traveling wave signal.

Benefits of technology

It realizes effective extraction of voltage primary traveling wave signal, solves the problem of inconsistency between the voltage primary traveling wave signal and the secondary traveling wave signal obtained through voltage traveling wave sensor, improves the accuracy of fault positioning and reliable protection of transmission lines. sex.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119089690A_ABST
    Figure CN119089690A_ABST
Patent Text Reader

Abstract

The invention discloses a traveling wave waveform detection method and device based on complementary set empirical mode decomposition and a least square QR decomposition algorithm. The method comprises the following steps: establishing a transfer function model of a traveling wave sensor; performing complementary set empirical mode decomposition on a secondary traveling wave signal actually detected and output by the traveling wave sensor to obtain a plurality of intrinsic mode components; in the decomposition process, whether the intrinsic mode component obtained through current decomposition is effective or not is judged based on the multi-sequence permutation entropy; performing inversion on each intrinsic mode component by using Lanczos diagonalization and a QR decomposition algorithm to obtain a corresponding intrinsic mode component of the primary traveling wave signal; and superposing all the intrinsic mode components obtained by inversion to obtain an inverted primary traveling wave signal. According to the method, the IMF component containing the feature information is selected by adopting the multi-sequence permutation entropy and the boundary local continuation method, the random characteristic of the signal is truly reflected, the endpoint effect problem existing in CEEMD decomposition is solved, and the decomposition accuracy is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of traveling wave detection, and in particular to a traveling wave waveform detection method and device based on complementary set empirical mode decomposition and least squares QR decomposition algorithm. Background Art

[0002] In recent years, traveling wave technology has become a research hotspot in the field of relay protection at home and abroad due to its unique advantages such as high fault location accuracy, fast protection action speed, and no influence from transition resistance, current transformer saturation, system oscillation and long-line distributed capacitance. This technology completes the detection and location of faults by analyzing and processing high-frequency transient traveling wave signals containing fault information, thereby improving the stability and safety of transmission network operation.

[0003] The key to the practical application of traveling wave technology lies in the accurate detection of traveling wave signals. Traveling wave signals are wide-band transient signals that appear in the form of current and voltage in faulty power grids. Currently available traveling wave detection technologies show that conventional current transformers have good transmission capabilities for transient current traveling waves, and voltage transformers can transmit the highest frequency band signal and polarity of the traveling wave head without delay. However, the primary traveling wave signal will be distorted when passing through the traveling wave sensor, and the use of the distorted secondary traveling wave signal for fault location will produce large errors, which will seriously affect the stable operation of the system. In short, the waveform characteristics of the secondary traveling wave signal detected by the existing traveling wave detection technology cannot truly reflect the waveform characteristics of the primary traveling wave signal of the power grid. Therefore, accurate detection of high-frequency transient traveling wave signals is the key to the practical application of traveling wave technology. Summary of the invention

[0004] The present invention provides a traveling wave waveform detection method and device based on complementary set empirical mode decomposition and least squares QR decomposition algorithm, which solves the problem of inconsistency between a real primary traveling wave signal and a secondary traveling wave signal obtained by voltage traveling wave sensor transmission.

[0005] In order to achieve the above technical objectives, the present invention adopts the following technical solutions:

[0006] A traveling wave waveform detection method based on complementary set empirical mode decomposition and least squares QR decomposition algorithm, comprising:

[0007] S1, establish the transfer function model of the traveling wave sensor;

[0008] S2, performing complementary set empirical mode decomposition on the secondary traveling wave signal actually detected and output by the traveling wave sensor to obtain multiple intrinsic mode components; during the decomposition process, judging whether the intrinsic mode components currently decomposed are valid based on the multi-sequence permutation entropy, and re-decomposing if invalid;

[0009] S3, for each obtained intrinsic mode component, inversion is performed using Lanczos diagonalization and QR decomposition algorithm to obtain the corresponding intrinsic mode component of the primary traveling wave signal;

[0010] S4, superimposing all the natural mode components obtained in step S3 to obtain an inverted primary traveling wave signal.

[0011] Furthermore, step S1 constructs the transfer function model of the traveling wave sensor as follows:

[0012]

[0013] Wherein, H(s) represents the transfer function model, s=jω, U1(s) is the primary voltage traveling wave signal, and U2(s) is the secondary voltage traveling wave signal; A1, A2, A3, and A4 are coefficients of the transfer function model respectively; the power grid fault is simulated multiple times by the power grid fault traveling wave generator, and the primary and secondary traveling wave signals of the traveling wave sensor are measured, and the coefficients A1, A2, A3, and A4 are calculated, and then the transfer function model of the traveling wave sensor is obtained.

[0014] Furthermore, the complementary set empirical mode decomposition is performed on the secondary traveling wave signal, specifically including:

[0015] (1) Add Gaussian white noise n(t) with opposite signs to the secondary traveling wave signal y0(t) in pairs to obtain a pair of signals y 1+ (t), y 1- (t), that is:

[0016] y 1+ (t) = y0(t) + n(t)

[0017] y 1- (t) = y0(t) - n(t)

[0018] (2) Using the boundary local feature extension method to calculate the signal y 1+ (t) and y 1- (t) is processed to obtain a pair of initial decomposed signals y' 1+ (t) and y' 1- (t); let m = 1;

[0019] (3) Using the ensemble empirical mode decomposition algorithm, we can respectively calculate y' m+ (t) and y' m- (t) is decomposed to obtain the corresponding intrinsic modal component K m+ (t), K m- (t); then take K m+ (t), K m- The mean value of (t) is taken as the mth natural mode component K of the secondary traveling wave signalm (t);

[0020] (4) Calculate the mth natural modal component K m (t); if the calculated multiple sequence permutation entropy is greater than a given threshold, it is considered that the mth intrinsic modal component K m (t) If valid, continue to step (5); otherwise, return to step (1);

[0021] (5) First update the remaining components Then, it is determined whether the orthogonality index between the remaining component and the reconstructed signal obtained by the first m intrinsic modal components meets a preset standard;

[0022] If the preset standard is reached, the decomposition is terminated; otherwise: update m=m+1, update the signal to be decomposed Return to step (3).

[0023] Furthermore, the calculation method of the multi-sequence permutation entropy of the intrinsic modal component is:

[0024] (1) With a given scale factor s, the intrinsic modal component time series y = {y(1), y(2), …, y(N)} is coarsened to obtain the coarse-grained series y s ={x s (1),x s (2),…,x s (J)}, where:

[0025]

[0026] Where J is the amount of data in the coarse-grained time series,

[0027] (2) Reconstruct the phase space of the coarse-grained sequence and obtain the following reconstruction matrix:

[0028]

[0029] Where d is the embedding dimension; τ is the delay factor; the jth row in the matrix is ​​recorded as the jth reconstruction component There are K reconstruction components, K = J-(d-1)τ;

[0030] (3) Reconstruction component Index all elements in and arrange them in ascending order to obtain k non-repeating index number sequences:

[0031] S(g)=(j1,j2,…,j d );g=1,2,…,k;k≤d!

[0032] Where S(g) represents the g-th index number sequence, j1,j2,…,j d Represents the reconstruction component The index number of each element after the d elements are arranged in ascending order;

[0033] (4) Count the probability of each index number sequence appearing Then the multi-sequence permutation entropy of the intrinsic modal component is calculated

[0034]

[0035] Furthermore, Lanczos diagonalization and QR decomposition algorithms are used for inversion, specifically:

[0036] (1) Based on the transfer function model H(s) of the traveling wave sensor, the Toeplitz matrix A generated by the unit impulse response of the traveling wave sensor is obtained;

[0037] (2) The lth effective natural mode component of the secondary traveling wave signal is expressed as y l , the corresponding natural modal component of the primary traveling wave signal is expressed as x l , then the inversion relationship is expressed as y l =Ax l , at this time, the least squares method is used to solve the above inversion problem, which is expressed as: min||Ax-y||2, where ||·||2 represents the L2 norm of the vector;

[0038] (3) Perform Lanczos bidiagonalization on matrix A to convert it into a bidiagonal matrix B k , and then for the matrix B k Perform QR decomposition and finally update iteratively; take the lth effective intrinsic modal component y of the secondary traveling wave signal l And iteratively obtain the intrinsic mode component x l The minimum relative error between them is used as the convergence criterion to obtain the intrinsic modal component x of the first traveling wave signal obtained by the final inversion. l .

[0039] A traveling wave waveform detection device based on complementary set empirical mode decomposition and least squares QR decomposition algorithm, comprising:

[0040] The transfer function model building module is used to: establish the transfer function model of the traveling wave sensor;

[0041] The modal decomposition module is used to: perform complementary set empirical mode decomposition on the secondary traveling wave signal actually detected and output by the traveling wave sensor to obtain multiple intrinsic modal components; during the decomposition process, determine whether the intrinsic modal components currently decomposed are valid based on the multi-sequence permutation entropy, and decompose again if invalid;

[0042] The component inversion module is used to: perform inversion on each obtained intrinsic mode component by using Lanczos diagonalization and QR decomposition algorithm to obtain the corresponding intrinsic mode component of the primary traveling wave signal;

[0043] The component synthesis module is used to superimpose all the intrinsic mode components obtained by the component inversion module to obtain an inverted primary traveling wave signal.

[0044] Furthermore, the complementary set empirical mode decomposition is performed on the secondary traveling wave signal, specifically including:

[0045] (1) Add Gaussian white noise n(t) with opposite signs to the secondary traveling wave signal y0(t) in pairs to obtain a pair of signals y 1+ (t), y 1- (t), that is:

[0046] y 1+ (t) = y0(t) + n(t)

[0047] y 1- (t) = y0(t) - n(t)

[0048] (2) Using the boundary local feature extension method to calculate the signal y 1+ (t) and y 1- (t) is processed to obtain a pair of initial decomposed signals y' 1+ (t) and y' 1- (t); let m = 1;

[0049] (3) Using the ensemble empirical mode decomposition algorithm, we can respectively calculate y' m+ (t) and y' m- (t) is decomposed to obtain the corresponding intrinsic modal component K m+ (t), K m- (t); then take K m+ (t), K m- The mean value of (t) is taken as the mth natural mode component K of the secondary traveling wave signal m (t);

[0050] (4) Calculate the mth natural modal component K m (t); if the calculated multiple sequence permutation entropy is greater than a given threshold, it is considered that the mth intrinsic modal component K m (t) If valid, continue to step (5); otherwise, return to step (1);

[0051] (5) First update the remaining components Then, it is determined whether the orthogonality index between the remaining component and the reconstructed signal obtained by the first m intrinsic modal components meets a preset standard;

[0052] If the preset standard is reached, the decomposition is terminated; otherwise: update m=m+1, update the signal to be decomposed Return to step (3).

[0053] Furthermore, the calculation method of the multi-sequence permutation entropy of the intrinsic modal component is:

[0054] (1) With a given scale factor s, the intrinsic modal component time series y = {y(1), y(2), …, y(N)} is coarsened to obtain the coarse-grained series y s ={x s (1),x s (2),…,x s (J)}, where:

[0055]

[0056] Where J is the amount of data in the coarse-grained time series,

[0057] (2) Reconstruct the phase space of the coarse-grained sequence and obtain the following reconstruction matrix:

[0058]

[0059] Where d is the embedding dimension; τ is the delay factor; the jth row in the matrix is ​​recorded as the jth reconstruction component There are K reconstruction components, K = J-(d-1)τ;

[0060] (3) Reconstruction component Index all elements in and arrange them in ascending order to obtain k non-repeating index number sequences:

[0061] S(g)=(j1,j2,…,j d );g=1,2,…,k;k≤d!

[0062] Where S(g) represents the g-th index number sequence, j1,j2,…,j d Represents the reconstruction component The index number of each element after the d elements are arranged in ascending order;

[0063] (4) Count the probability of each index number sequence appearing Then the multi-sequence permutation entropy of the intrinsic modal component is calculated

[0064]

[0065] Furthermore, Lanczos diagonalization and QR decomposition algorithms are used for inversion, specifically:

[0066] (1) Based on the transfer function model H(s) of the traveling wave sensor, the Toeplitz matrix A generated by the unit impulse response of the traveling wave sensor is obtained;

[0067] (2) The lth effective natural mode component of the secondary traveling wave signal is expressed as y l , the corresponding natural modal component of the primary traveling wave signal is expressed as x l , then the inversion relationship is expressed as y l =Ax l , at this time, the least squares method is used to solve the above inversion problem, which is expressed as: min||Ax-y||2, where ||·||2 represents the L2 norm of the vector;

[0068] (3) Perform Lanczos bidiagonalization on matrix A to convert it into a bidiagonal matrix B k , and then for the matrix B k Perform QR decomposition and finally update iteratively; take the lth effective intrinsic modal component y of the secondary traveling wave signal l And iteratively obtain the intrinsic mode component x l The minimum relative error between them is used as the convergence criterion to obtain the intrinsic modal component x of the first traveling wave signal obtained by the final inversion. l .

[0069] Compared with the prior art, the advantages of the present invention are:

[0070] The present invention uses CEEMD decomposition to decompose the secondary traveling wave signal into IMF components of different frequency bands, and uses multiple sequence permutation entropy and boundary local extension method to select IMF components containing characteristic information, which truly reflects the random characteristics of the signal and solves the endpoint effect problem existing in CEEMD decomposition, thereby improving the accuracy of decomposition. Simulation and experimental results show that the present invention can effectively extract the primary traveling wave signal of voltage, and effectively solve the problem of inconsistency between the primary traveling wave signal of voltage and the secondary traveling wave signal obtained by the voltage traveling wave sensor. It is helpful to improve the reliability of the traveling wave protection technology for transmission line faults. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 It is a schematic flow diagram of the method of the present invention;

[0072] Figure 2 is a flow chart of the LSQR algorithm of the present invention;

[0073] Figure 3 It is a circuit model diagram of the present invention;

[0074] Figure 4 It is the simulated primary and secondary traveling wave waveform of the present invention;

[0075] Figure 5 This is a schematic diagram of the decomposition result of the secondary traveling wave by CEEMD of the present invention;

[0076] Figure 6 It is a schematic diagram of each inherent modal component of a traveling wave signal obtained by inversion of the present invention;

[0077] Figure 7 It is a waveform diagram of a traveling wave signal obtained by inversion of the present invention. DETAILED DESCRIPTION

[0078] The following is a detailed description of an embodiment of the present invention. This embodiment is based on the technical solution of the present invention, and provides a detailed implementation method and a specific operation process to further explain the technical solution of the present invention.

[0079] like Figure 1 As shown, the traveling wave waveform accurate detection method based on complementary set empirical mode decomposition (CEEMD) and least squares QR decomposition algorithm (LSQR) of the present invention comprises the following steps:

[0080] S1, establish the transfer function model of the traveling wave sensor and obtain:

[0081]

[0082] In the formula, H(s) represents the transfer function model, s=jω, U1(s) is the primary voltage traveling wave signal, U2(s) is the secondary voltage traveling wave signal; A1, A2, A3, and A4 are the coefficients of the transfer function model. The power grid fault traveling wave generator is used to simulate the power grid fault multiple times, and the primary and secondary traveling wave signals of the traveling wave sensor are measured. The coefficients A1, A2, A3, and A4 are calculated, and then the transfer function model of the traveling wave sensor is obtained.

[0083] S2, the complementary set empirical mode decomposition is performed on the secondary traveling wave signal actually detected and output by the traveling wave sensor to obtain multiple intrinsic mode components; during the decomposition process, the validity of the intrinsic mode components currently decomposed is judged based on the multi-sequence permutation entropy, and if invalid, it is decomposed again.

[0084] Among them, the complementary set empirical mode decomposition is performed on the secondary traveling wave signal, specifically including:

[0085] (1) Add Gaussian white noise n(t) with opposite signs to the secondary traveling wave signal y0(t) in pairs to obtain a pair of signals y 1+ (t), y 1- (t), that is:

[0086] y 1+ (t) = y0(t) + n(t)

[0087] y 1- (t) = y0(t) - n(t)

[0088] (2) Using the boundary local feature extension method to transform the signal y 1+ (t) and y 1- (t) is processed to obtain a pair of initial decomposed signals y' 1+ (t) and y' 1- (t); let m = 1;

[0089] The local feature extension method of the boundary: when it is impossible to find similar blocks to match the boundary features, it is necessary to extract the extreme point features at the left and right endpoints of the signal, use this feature to add extreme points outside the left and right endpoints of the signal, and involve them in the envelope fitting process, and finally estimate the mean curve. Since in the high frequency band of the traveling wave signal, the oscillation amplitude is relatively large within a cycle of the fault initiation, and the internal regularity is relatively weak, the local information at the edge of the fault is considered, and the signal y is selected 1+ (t) and y 1- (t) At the extreme point within a specific segment length, the amplitude characteristics and time characteristics of the extreme point within the segment are extracted as the basis for extension, so as to 1+ (t) and y 1- (t) adds a pair of maximum and minimum points at both ends. 1+ (t) and y 1- (t) The processing at the endpoint can solve the endpoint effect problem existing in EEMD decomposition and improve the accuracy of decomposition.

[0090] (3) Using the ensemble empirical mode decomposition algorithm, we can respectively calculate y' m+ (t) and y' m- (t) is decomposed to obtain the corresponding intrinsic modal component K m+ (t), K m- (t); then take K m+ (t), K m- The mean value of (t) is taken as the mth natural mode component K of the secondary traveling wave signal m (t):

[0091] K m (t) = 0.5 (K m+ (t)+K m- (t))

[0092] (4) Calculate the mth natural modal component K m(t); if the calculated multiple sequence permutation entropy is greater than a given threshold, it is considered that the mth intrinsic modal component K m (t) If valid, continue to step (5); otherwise, return to step (1);

[0093] (5) First update the remaining components Then, it is determined whether the orthogonality index between the remaining component and the reconstructed signal obtained by the first m intrinsic modal components meets a preset standard;

[0094] If the preset standard is reached, the decomposition is terminated; otherwise: update m=m+1, update the signal to be decomposed Return to step (3).

[0095] Among them, the multi-sequence permutation entropy of each intrinsic modal component is calculated as follows:

[0096] (1) With a given scale factor s, the intrinsic modal component time series y = {y(1), y(2), …, y(N)} is coarsened to obtain the coarse-grained series y s ={x s (1),x s (2),…,x s (J)}, where:

[0097]

[0098] Where J is the amount of data in the coarse-grained time series,

[0099] (2) Reconstruct the phase space of the coarse-grained sequence and obtain the following reconstruction matrix:

[0100]

[0101] Where d is the embedding dimension; τ is the delay factor; the jth row in the matrix is ​​recorded as the jth reconstruction component There are K reconstruction components, K = J-(d-1)τ;

[0102] (3) Reconstruction component Index all elements in and arrange them in ascending order to obtain k non-repeating index number sequences:

[0103] S(g)=(j1,j2,…,j d );g=1,2,…,k;k≤d!

[0104] Where S(g) represents the g-th index number sequence, j1,j2,…,j d Represents the reconstruction component The index number of each element after the d elements are arranged in ascending order;

[0105] (4) Count the probability of each index number sequence appearing That is, the ratio of the number of occurrences of each index number sequence to the sum of the number of occurrences of all index number sequences, and Then the multi-sequence permutation entropy of the intrinsic modal component is calculated

[0106]

[0107] S3, for each obtained intrinsic mode component, use Lanczos diagonalization and QR decomposition algorithm to perform inversion to obtain the corresponding intrinsic mode component of the primary traveling wave signal. Including:

[0108] (1) Based on the transfer function model H(s) of the traveling wave sensor, the Toeplitz matrix A generated by the unit impulse response of the traveling wave sensor is obtained;

[0109] (2) The lth effective natural mode component of the secondary traveling wave signal is expressed as y l , the corresponding natural modal component of the primary traveling wave signal is expressed as x l , then the inverse problem is expressed as y l =Ax l , at this time, the least squares method is used to solve the inverse problem and we get:

[0110] min||Ax l -y l ||2

[0111] Where ||·||2 represents the L2 norm of the vector.

[0112] (3) Figure 2 It is the flow chart of LSQR algorithm in the example of the present invention, as Figure 2 As shown, given the initial conditions such as u1, v1, α1, β1, w1, x0, the matrix A is Lanczos bidiagonalized according to the iterative relationship of the following three equations:

[0113] U k+1 (β1e1)=b

[0114] AV k =U k+1 B k

[0115]

[0116] Get the matrix B in bidiagonal form k :

[0117]

[0118] In the formula, α1, α2, …, α k and β1,β2,…,β k is the matrix B k k parameters of the upper diagonal and lower diagonal, and each parameter satisfies;

[0119]

[0120] ||u i ||=||v i ||=1 i=1,2…

[0121] (4) For matrix B k Perform QR decomposition to further reduce the amount of calculation:

[0122]

[0123] Where e1 is a unit vector; e k+1 is the last column of the k+1-dimensional identity matrix; θ k+1 is the matrix B k+1 The k+1th diagonal element of Q k =Q k,k+1 ···Q 2,3 Q 1,2 is a plane rotation matrix, which is used to eliminate B k The lower diagonal elements of β2, β3, ···.

[0124] Finally, it is determined whether the iteration stop condition or the maximum number of iterations is met. If not, the iteration continues.

[0125] S4, superimposing all the natural mode components obtained in step S3 to obtain an inverted primary traveling wave signal.

[0126] The specific application examples of the traveling wave waveform inversion method proposed in this embodiment are as follows:

[0127] (1) Constructing the transfer function model of traveling wave sensor:

[0128]

[0129] (2) Figure 3 As shown in the figure, a simulation model of a 500kV transmission line is established by using the simulation software Simulink. In this model, the length of the transmission line MN is 300km, and the sampling frequency is set to 1MHz. First, a single-phase grounding fault is set at 100km of the 500kV transmission line MN, and a primary traveling wave signal is obtained by setting different transition resistances Rg and appropriate noise interference. Then, the primary traveling wave signal is passed through the voltage traveling wave sensor simulation model to obtain a simulated secondary traveling wave signal. The simulated primary and secondary traveling wave waveforms are shown in Figure 1. Figure 4 As shown in the figure, the primary and secondary traveling wave waveforms obtained by simulation are obviously different.

[0130] (3) Perform CEEMD decomposition on the measured secondary traveling wave signal to obtain 10 IMF components sorted from high to low frequency, and then extract the waveform of each IMF component, such as Figure 5 shown.

[0131] (4) The 10 IMF components are inverted using the LSQR algorithm in sequence. After inversion, the IMF′ components are as follows: Figure 6 shown.

[0132] (5) Finally, all IMF′ components are superimposed to obtain a primary traveling wave waveform, such as Figure 7 shown.

[0133] The simulation analysis results show that the present invention can accurately extract the primary voltage traveling wave through the inversion of the secondary voltage traveling wave, effectively solving the problem of waveform distortion of the secondary traveling wave of the dedicated voltage traveling wave sensor. The LSQR algorithm proposed in this paper has good convergence performance and anti-noise ability, effectively solves the pathological problems existing in the waveform inversion process, and improves the accuracy and stability of the inversion results. The simulation and experiments under different fault resistances, fault initial phase angles, and signal-to-noise ratios show that this method is not affected by the modal aliasing effect and has strong anti-noise ability. The fault information of the primary traveling wave inverted by this method has a small error with the fault information of the measured primary traveling wave, and the use of the obtained inverted primary signal can effectively improve the fault location accuracy.

[0134] The above embodiments are preferred embodiments of the present application. Ordinary technicians in this field can also make various changes or improvements on this basis. Without departing from the overall concept of the present application, these changes or improvements should fall within the scope of protection required by the present application.

Claims

1. A traveling wave waveform detection method based on complementary set empirical mode decomposition and least squares QR decomposition algorithm, characterized in that: include: S1, establish the transfer function model of the traveling wave sensor; S2, performing complementary set empirical mode decomposition on the secondary traveling wave signal actually detected and output by the traveling wave sensor to obtain multiple intrinsic mode components; During the decomposition process, the validity of the intrinsic modal components obtained by the current decomposition is determined based on the multi-sequence permutation entropy. If invalid, the decomposition is repeated. S3, for each obtained intrinsic mode component, inversion is performed using Lanczos diagonalization and QR decomposition algorithm to obtain the corresponding intrinsic mode component of the primary traveling wave signal; S4, superimposing all the natural mode components obtained in step S3 to obtain an inverted primary traveling wave signal.

2. The traveling wave waveform detection method according to claim 1, characterized in that: Step S1 constructs the transfer function model of the traveling wave sensor as: Wherein, H(s) represents the transfer function model, s=jω, U1(s) is the primary voltage traveling wave signal, and U2(s) is the secondary voltage traveling wave signal; A1, A2, A3, and A4 are coefficients of the transfer function model respectively; the power grid fault is simulated multiple times by the power grid fault traveling wave generator, and the primary and secondary traveling wave signals of the traveling wave sensor are measured, and the coefficients A1, A2, A3, and A4 are calculated, and then the transfer function model of the traveling wave sensor is obtained.

3. The traveling wave waveform detection method according to claim 1, characterized in that: The complementary set empirical mode decomposition is performed on the secondary traveling wave signal, including: (1) Add Gaussian white noise n(t) with opposite signs to the secondary traveling wave signal y0(t) in pairs to obtain a pair of signals y 1+ (t), y 1- (t), that is: y 1+ (t)=y0(t)+n(t) y 1- (t)=y0(t)-n(t) (2) Using the boundary local feature extension method to calculate the signal y 1+ (t) and y 1- (t) is processed to obtain a pair of initial decomposed signals y' 1+ (t) and y' 1- (t); let m = 1; (3) Using the ensemble empirical mode decomposition algorithm, we can respectively calculate y' m+ (t) and y' m- (t) is decomposed to obtain the corresponding intrinsic modal component K m+ (t), K m- (t); then take K m+ (t), K m- The mean value of (t) is taken as the mth natural mode component K of the secondary traveling wave signal m (t); (4) Calculate the mth natural modal component K m (t); if the calculated multiple sequence permutation entropy is greater than a given threshold, it is considered that the mth intrinsic modal component K m (t) If valid, continue to step (5); otherwise, return to step (1); (5) First update the remaining components Then, it is determined whether the orthogonality index between the remaining component and the reconstructed signal obtained by the first m intrinsic modal components meets a preset standard; If the preset standard is reached, the decomposition is terminated; otherwise: update m=m+1, update the signal to be decomposed Return to step (3).

4. The traveling wave waveform detection method according to claim 3, characterized in that: The calculation method of the multi-sequence permutation entropy of the intrinsic modal component is: (1) With a given scale factor s, the intrinsic modal component time series y = {y(1), y(2), …, y(N)} is coarsened to obtain the coarse-grained series y s ={x s (1),x s (2),…,x s (J)}, where: Where J is the amount of data in the coarse-grained time series, (2) Reconstruct the phase space of the coarse-grained sequence and obtain the following reconstruction matrix: Where d is the embedding dimension; τ is the delay factor; the jth row in the matrix is ​​recorded as the jth reconstruction component There are K reconstruction components, K = J-(d-1)τ; (3) Reconstruction component Index all elements in and arrange them in ascending order to obtain k non-repeating index number sequences: S(g)=(j1,j2,…,j d );g=1,2,…,k;k≤d! Where S(g) represents the g-th index number sequence, j1,j2,…,j d Represents the reconstruction component The index number of each element after the d elements are arranged in ascending order; (4) Count the probability of each index number sequence appearing Then the multi-sequence permutation entropy of the intrinsic modal component is calculated 5. The traveling wave waveform detection method according to claim 1, characterized in that: The inversion is performed using Lanczos diagonalization and QR decomposition algorithms, specifically: (1) Based on the transfer function model H(s) of the traveling wave sensor, the Toeplitz matrix A generated by the unit impulse response of the traveling wave sensor is obtained; (2) The lth effective natural mode component of the secondary traveling wave signal is expressed as y l , the corresponding natural modal component of the primary traveling wave signal is expressed as x l , then the inversion relationship is expressed as y l =Ax l , at this time, the least squares method is used to solve the above inversion problem, which is expressed as: min||Ax-y||2, where ||·||2 represents the L2 norm of the vector; (3) Perform Lanczos bidiagonalization on matrix A to convert it into a bidiagonal matrix B k , and then for the matrix B k Perform QR decomposition and finally update the iteration; Take the lth effective natural mode component y of the secondary traveling wave signal l And iteratively obtain the intrinsic mode component x l The minimum relative error between them is used as the convergence criterion to obtain the intrinsic modal component x of the first traveling wave signal obtained by the final inversion. l .

6. A traveling wave waveform detection device based on complementary set empirical mode decomposition and least squares QR decomposition algorithm, characterized in that: include: The transfer function model building module is used to: establish the transfer function model of the traveling wave sensor; The modal decomposition module is used to: perform complementary set empirical modal decomposition on the secondary traveling wave signal actually detected and output by the traveling wave sensor to obtain multiple inherent modal components; During the decomposition process, the validity of the intrinsic modal components obtained by the current decomposition is determined based on the multi-sequence permutation entropy. If invalid, the decomposition is repeated. The component inversion module is used to: perform inversion on each obtained intrinsic mode component by using Lanczos diagonalization and QR decomposition algorithm to obtain the corresponding intrinsic mode component of the primary traveling wave signal; The component synthesis module is used to superimpose all the intrinsic mode components obtained by the component inversion module to obtain an inverted primary traveling wave signal.

7. The traveling wave waveform detection device according to claim 6, characterized in that: The complementary set empirical mode decomposition is performed on the secondary traveling wave signal, including: (1) Add Gaussian white noise n(t) with opposite signs to the secondary traveling wave signal y0(t) in pairs to obtain a pair of signals y 1+ (t), y 1- (t), that is: y 1+ (t)=y0(t)+n(t) y 1- (t)=y0(t)-n(t) (2) Using the boundary local feature extension method to calculate the signal y 1+ (t) and y 1- (t) is processed to obtain a pair of initial decomposed signals y' 1+ (t) and y' 1- (t); let m = 1; (3) Using the ensemble empirical mode decomposition algorithm, we can respectively calculate y' m+ (t) and y' m- (t) is decomposed to obtain the corresponding intrinsic modal component K m+ (t), K m- (t); then take K m+ (t), K m- The mean value of (t) is taken as the mth natural mode component K of the secondary traveling wave signal m (t); (4) Calculate the mth natural modal component K m (t); if the calculated multiple sequence permutation entropy is greater than a given threshold, it is considered that the mth intrinsic modal component K m (t) If valid, continue to step (5); otherwise, return to step (1); (5) First update the remaining components Then, it is determined whether the orthogonality index between the remaining component and the reconstructed signal obtained by the first m intrinsic modal components meets a preset standard; If the preset standard is reached, the decomposition is terminated; otherwise: update m=m+1, update the signal to be decomposed Return to step (3).

8. The traveling wave waveform detection device according to claim 6, characterized in that: The calculation method of the multi-sequence permutation entropy of the intrinsic modal component is: (1) With a given scale factor s, the intrinsic modal component time series y = {y(1), y(2), …, y(N)} is coarsened to obtain the coarse-grained series y s ={x s (1),x s (2),…,x s (J)}, where: Where J is the amount of data in the coarse-grained time series, (2) Reconstruct the phase space of the coarse-grained sequence and obtain the following reconstruction matrix: Where d is the embedding dimension; τ is the delay factor; the jth row in the matrix is ​​recorded as the jth reconstruction component There are K reconstruction components, K = J-(d-1)τ; (3) Reconstruction component Index all elements in and arrange them in ascending order to obtain k non-repeating index number sequences: S(g)=(j1,j2,…,j d );g=1,2,…,k;k≤d! Where S(g) represents the g-th index number sequence, j1,j2,…,j d Represents the reconstruction component The index number of each element after the d elements are arranged in ascending order; (4) Count the probability of each index number sequence appearing Then the multi-sequence permutation entropy of the intrinsic modal component is calculated 9. The traveling wave waveform detection device according to claim 6, characterized in that: The inversion is performed using Lanczos diagonalization and QR decomposition algorithms, specifically: (1) Based on the transfer function model H(s) of the traveling wave sensor, the Toeplitz matrix A generated by the unit impulse response of the traveling wave sensor is obtained; (2) The lth effective natural mode component of the secondary traveling wave signal is expressed as y l , the corresponding natural modal component of the primary traveling wave signal is expressed as x l , then the inversion relationship is expressed as y l =Ax l , at this time, the least squares method is used to solve the above inversion problem, which is expressed as: min||Ax-y||2, where ||·||2 represents the L2 norm of the vector; (3) Perform Lanczos bidiagonalization on matrix A to convert it into a bidiagonal matrix B k , and then for the matrix B k Perform QR decomposition and finally update the iteration; Take the lth effective natural mode component y of the secondary traveling wave signal l And iteratively obtain the intrinsic modal component x l The minimum relative error between them is used as the convergence criterion to obtain the intrinsic modal component x of the first traveling wave signal obtained by the final inversion. l .