Power distribution network grounding fault line selection method and system based on current traveling wave spectrum

Abstract

The invention discloses a power distribution network grounding fault line selection method and system based on a current traveling wave spectrum, and the method comprises the steps: S1, obtaining a three-phase current traveling wave signal at the head end of each feeder line, and carrying out the phase-mode conversion, and obtaining a current traveling wave line-mode signal corresponding to each feeder line; s2, frequency spectrum transformation analysis is carried out on the current traveling wave line mode signals of all the feeder lines, and high-frequency spectrum vectors corresponding to all the feeder lines are extracted and constructed; s3, based on a dynamic time warping algorithm, calculating a similar distance between the high-frequency spectrum vectors of every two feeder lines, and constructing a feeder line current traveling wave spectrum similar distance matrix; s4, carrying out spectral clustering-based analysis on the feeder line current traveling wave similar distance matrix; judging a fault position according to a clustering result; if the element values of all the feeder lines are clustered into one class, judging that a bus fault occurs; and if the element values are obviously divided into two types, judging that the feeder line of the smaller cluster has the high-resistance grounding fault. The method has the advantages of quick response, high sensitivity, strong robustness and the like.

Description

TECHNICAL FIELD

[0001] The present application mainly relates to the technical field of power distribution network, and particularly relates to a power distribution network grounding fault line selection method and system based on current traveling wave spectrum. BACKGROUND

[0002] At present, constructing a new power system mainly based on new energy has become a new form of future energy and power development. The new power distribution network is a power distribution network with distributed energy access, such as solar energy, wind energy and other new energy. The new power distribution system is widely distributed in space and has a complex operating environment. The initial fault resistance can be as high as tens of thousands of ohms, and the fault current can be as low as 0.1A, which is only one ten-thousandth of the normal current. Influenced by the coupling of distributed power and power electronic equipment, it is very difficult to correctly select the fault line. High-resistance grounding faults have been "invisible" in the power distribution network for a long time, and if not handled in time, it is likely to expand the fault range, damage power equipment, and even induce forest fires, accidents of electric shock injury, and other accidents.

[0003] The current mainstream line selection technology at home and abroad mainly includes steady-state component method, transient component method and traveling wave method. The main difficulty of the steady-state component method is that the fault steady-state information is weak, which is easily affected by the size of the transition resistance and the system operation mode, and the reliability of the detected signal is not high, and the line selection accuracy is low. Compared with the steady-state component method, the signal of the transient component method is more obvious, although this method improves the line selection accuracy to some extent, but the fault line determination process needs to fully consider the influence of CT saturation. The traveling wave method is not affected by CT saturation and system operation mode, and with the development of wireless communication and signal processing technology, the traveling wave method has great application potential. However, the line selection threshold of the above methods will be affected by the complex actual working conditions generated after the new energy is connected to the grid. In addition, especially in the new power distribution network, the short-time data window weak fault signal provided by the new energy is limited, which affects the reliability of most fault line selection methods based on time domain information in the new power distribution network. SUMMARY

[0004] In view of the technical problems existing in the prior art, the present application provides a current traveling wave spectrum-based power distribution network grounding fault line selection method and system with fast response, high sensitivity and strong robustness.

[0005] To solve the above technical problems, the technical scheme provided by the present application is as follows: A current traveling wave spectrum-based power distribution network grounding fault line selection method, comprising the steps of: S1. Obtain the three-phase current traveling wave signals of each feeder head in the power distribution network, and perform phase-to-magnitude transformation on each of the three-phase current traveling wave signals to obtain the current traveling wave line module signals corresponding to each feeder; S2. Perform spectral transformation analysis on the current traveling wave line mode signal of each feeder, and extract and construct the high-frequency spectrum vector corresponding to each feeder; S3. Based on the dynamic time warping algorithm, calculate the similarity distance between the high-frequency spectrum vectors of every two feeders, and construct... n × n Similarity distance matrix of the spectrum of traveling wave of feeder current. n The total number of feeders is represented by the element values ​​in the matrix, which characterize the degree of difference in the morphology of the spectral waveforms of the corresponding two feeders. S4. Perform spectral clustering analysis on the similarity distance matrix of the feeder current traveling wave to obtain the clustering results; then determine the fault location based on the clustering results: if the element values ​​of all feeders are clustered into one category, it is determined to be a bus fault; if the element values ​​are clearly divided into two categories, it is determined that the outlier feeder or the smaller cluster of feeders has a high-resistance grounding fault.

[0006] Preferably, the current traveling wave line mode signal in step S1 is obtained after performing a Kelvin-Bell phase mode transformation on the three-phase current traveling wave signal. α component signal.

[0007] Preferably, the specific process of step S2 is as follows: S201. Convert the time-domain current traveling wave mode signal i k ( t Discrete signals are obtained by sampling. i k (p) , p Given the number of sampling points for the discrete signal, calculate the discrete signal. i k (p) autocorrelation function R k [ q ]:

[0008] In the formula, N It is a current traveling wave line mode signal i k ( t ) length, q It's a delay; S202. Based on autocorrelation function R k [ q Construct the autocorrelation matrix R k :

[0009] S203. On the autocorrelation matrix R k Perform eigenvalue decomposition:

[0010] In the formula, V k It is the eigenvector matrix. Λ k It is an eigenvalue matrix; Select the largest eigenvalue. m eigenvalues λ 1, λ 2,···, λ m and its corresponding eigenvectors v 1, v 2,···, v m ; S204. Construct a sampling point system using the selected eigenvalues ​​and eigenvectors. m spectral vector S k (f) :

[0011] In the formula, v x (f) These are the eigenvectors.

[0012] Preferably, in step S3, the specific process of calculating the similarity distance between the high-frequency spectral vectors of every two feeders is as follows: S301. Construct a m × m Local cost matrix D Its elements d ( x , y )=| S kx - S jy |, among which x , y =1,2,···, m ; S302. The cumulative cost matrix is ​​calculated using the dynamic time warping algorithm. C Its elements c ( x , y The recursive formula for ) is:

[0013] In the formula, c (1,1)= d (1,1), and set boundary conditions. c (0,:)=c (:,0)=∞; S303. Bottom right element of the cumulative cost matrix c ( m , m The value of ) is the DTW similarity distance. D dtw (S k (f),S j (f)) ; S k (f),S j (f) feeders k and feeder j The spectrum vector.

[0014] Preferably, n × n Similarity distance matrix of traveling wave spectrum of feeder current M Specifically:

[0015] In the formula, the elements in the matrix M kj For feeder k Current traveling wave spectrum vector S k (f) With feeder j Current traveling wave spectrum vector S j (f) The larger the similarity distance, the greater the difference in the current traveling wave spectrum between the feeders.

[0016] Preferably, in step S4, the specific process of performing spectral clustering analysis on the similarity distance matrix of the feeder current traveling wave to obtain the clustering results is as follows: S401. Using the Gaussian kernel function to analyze the similarity distance matrix M Perform the transformation to generate the affinity matrix. A Its elements are defined as:

[0017] In the formula, A kk =0, the scale parameter σ controls the width of the domain, and is taken as a matrix. M Standard deviation of all elements; elements within the matrix M kj For feeder k Current traveling wave spectrum vector S k (f) With feederj Current traveling wave spectrum vector S j (f) The similarity distance; S402. Calculate the degree matrix Λ diagonal elements Λ kk for:

[0018] Then calculate the normalized Laplacian matrix. In the formula I It is the identity matrix. Represents the degree matrix Λ For each diagonal element, first find its inverse and then take its square root; S403. Calculation L norm The eigenvectors corresponding to the first two smallest eigenvalues u 1, u 2∈ R n , R n Let represent an n-dimensional real vector space and form the eigenvector matrix. U =[ u 1, u 2]∈ R n×2 ; S404. Matrix U Each row is considered as a point in two-dimensional space. y i ∈ R 2 , i =1,···, n The K-means++ algorithm was used to cluster these points into two clusters. C 1, C 2.

[0019] Preferably, in step S4, the matrix is... U Each line of input is fed into the K-means clustering algorithm, and the number of clusters is set to 2. K-means assigns a cluster label to each point. This label directly corresponds to the fault identification result. If K-means assigns all points to the same cluster, it means that the spectral characteristics of all feeders are highly consistent, and it is determined to be a bus fault. If the points are divided into two clusters, the cluster with significantly fewer points is the faulty feeder cluster.

[0020] The present invention also discloses a computer program product, comprising a computer program that, when executed by a processor, performs the steps of the method described above.

[0021] The present invention further discloses a computer-readable storage medium having a computer program stored thereon, the computer program executing the steps of the method described above when run by a processor.

[0022] The present invention also discloses a distribution network grounding fault selection system based on current traveling wave spectrum, including a memory and a processor connected to each other. The memory stores a computer program, which executes the steps of the method described above when run by the processor.

[0023] Compared with the prior art, the advantages of the present invention are as follows: This invention innovatively uses the current traveling wave spectrum waveform for distribution network line selection. It utilizes the characteristic that the traveling wave has a transmission speed close to the speed of light and can extract data in a short time period to achieve ultra-high-speed line selection within 1ms in the distribution network. Furthermore, it is not affected by the weak fault signals in the short-time data window provided by new energy sources. It is applicable to common types of distribution network neutral point grounding methods and new energy operating conditions, and is easy to implement directly in engineering applications where new energy is connected.

[0024] This invention, based on DTW and spectral clustering algorithms, establishes a fast, simple, and intuitive fault selection principle. The DTW algorithm can capture the morphological differences of the spectral waveform, is insensitive to waveform distortion and frequency offset, and is more robust than rigid distance metrics. The spectral clustering algorithm can identify clusters of arbitrary shapes and is insensitive to parameter settings, with stable and reliable clustering results, avoiding the dependence of traditional clustering algorithms on thresholds or initial values. By replacing quantitative tuning calculations with qualitative differences in the spectral waveform, it can accurately select feeders in new distribution networks under different fault locations, fault types, transition resistances, and noise interference. This fundamentally improves the problem of difficulty in setting fault selection thresholds under conditions of new energy access and high resistance, achieving highly sensitive and reliable fault selection for new distribution networks while also considering high-resistance grounding faults. Attached Figure Description

[0025] Figure 1 This is a flowchart of an embodiment of the distribution network grounding fault location method based on current traveling wave spectrum of the present invention.

[0026] Figure 2 This is a simulation model diagram of a new 10kV distribution network built in the PSCAD / EMTDC software platform according to the present invention.

[0027] Figure 3 Faults of the present invention F The traveling wave spectrum detection diagram of the current of each feeder in line 1.

[0028] Figure 4 Faults of the present invention F The result of spectral clustering for line selection is shown in Figure 1. Detailed Implementation

[0029] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0030] like Figure 1 As shown in the figure, the distribution network grounding fault location method based on current traveling wave spectrum provided by this invention includes the following steps: S1. Acquire the three-phase current traveling wave signals at the beginning of each feeder in the new distribution network, and perform phase-mode transformation on each three-phase current traveling wave signal to obtain the current traveling wave line-mode signal corresponding to each feeder; wherein the current traveling wave line-mode signal is obtained after performing Kelvin-Bell phase-mode transformation on the three-phase current traveling wave signal. α component signal; S2. Perform high-resolution spectral transformation analysis on the current traveling wave line mode signal of each feeder, and extract and construct the high-frequency spectral vector corresponding to each feeder; S3. Based on the Dynamic Time Warping (DTW) algorithm, calculate the similarity distance between the high-frequency spectral vectors of every two feeders, and construct... n × n Similarity distance matrix of the spectrum of traveling wave of feeder current. n The total number of feeders is represented by the element values ​​in the matrix, which characterize the degree of difference in the morphology of the spectral waveforms of the corresponding two feeders. S4. Perform spectral clustering analysis on the similarity distance matrix of the feeder current traveling wave to obtain the clustering results; then determine the fault location based on the clustering results: if the element values ​​of all feeders are clustered into one class (in one cluster), it is determined to be a bus fault; if the element values ​​are clearly divided into two classes (there are outlier feeders), it is determined that the outlier feeder or the smaller cluster feeder has a high-resistance grounding fault.

[0031] In one specific embodiment, step S2 uses the Pisarenko spectrum estimation algorithm to extract high-frequency spectrum vectors from the current traveling wave line mode signal, which can effectively suppress noise and accurately extract spectrum features under short data windows.

[0032] feeder k High-frequency spectrum vector S k ( f ) is represented as: ,element S kx Representing the x The amplitude of the spectral signal corresponding to each sampling point f For frequency, t For time, x The values ​​are 1, 2, ... m , m The number of elements; the high-frequency spectrum vector is obtained based on Pisalenko spectral transform analysis.S k (f) The process is as follows: S201. For current traveling wave mode signals represented in the time domain. i k ( t Discrete signals are obtained by sampling. i k (p) , p The discrete signal is sampled at points, and then the autocorrelation function of the discrete signal is calculated. R k [ q ]:

[0033] In the formula, N It is a current traveling wave line mode signal i k ( t ) length, q It's a delay; S202. Autocorrelation function based on discrete signals R k [ q Construct the autocorrelation matrix R k :

[0034] S203. Perform eigenvalue decomposition on the autocorrelation matrix and select the m largest eigenvalues ​​and their corresponding eigenvectors;

[0035] In the formula, V k It is the eigenvector matrix. Λ k It is an eigenvalue matrix, and then the largest eigenvalue is selected. m eigenvalues λ 1, λ 2,···, λ m and its corresponding eigenvectors v 1, v 2,···, v m ; H This represents the conjugate transpose of the matrix. k Representing the k The signal detected by the feeder.

[0036] S204. Using the selected eigenvalues ​​and eigenvectors, construct a sampling point system with the following number of points. m High-frequency spectrum vectorS k (f) :

[0037] In the formula, v x (f) These are the eigenvectors.

[0038] In one specific embodiment, in step S3 n × n Similarity distance matrix of traveling wave spectrum of feeder current M Represented as:

[0039] In the formula, the similarity distance matrix M elements M kj For feeder k Current traveling wave high frequency spectrum vector S k (f) With feeder j Current traveling wave high frequency spectrum vector S j (f) The larger the DTW distance, the greater the difference in the current traveling wave spectrum between the feeders. k , j The value can be 1, 2, ... n .

[0040] Specifically, the high-frequency spectrum vector of the current traveling wave S k (f) and S j (f) The similarity distance can be expressed as: feeder k High-frequency spectrum vector S k (f) Represented as: feeder j High-frequency spectrum vector S j (f) Represented as: ,element S kx , S jy They represent feeders respectively. k No. x Each sampling point and feeder j No. y The amplitude of the spectral signal corresponding to each sampling point f For frequency,t For time, x and y The values ​​are all 1, 2, ... m , m The number of elements; If the similarity distance is DTW distance, first construct a m × m The local distance matrix, matrix elements d ( x , y )=| S kx - S jy |;Then find a path from (1,1) to ( in the local distance matrix. m , m The regular path with the minimum cumulative distance. W The path must satisfy boundary conditions, continuity, and monotonicity constraints; this minimum cumulative distance is the DTW distance. D dtw (S k (f), S j (f)) And as a similarity distance matrix M elements M kj .

[0041] Using DTW distance can effectively overcome the nonlinear scaling and phase shift that may occur in the frequency spectrum waveform. It evaluates the similarity of the overall shape of the waveform by finding the optimal regular path, and is more robust than rigid distance metrics (such as Manhattan distance).

[0042] In one specific embodiment, step S4 specifically includes: converting the matrix M Converted to an affinity matrix using a Gaussian kernel function. A A normalized Laplacian matrix is ​​constructed and eigenvalue decomposition is performed. The eigenvectors corresponding to the first two smallest eigenvalues ​​are taken to form a new feature space. Finally, the K-means++ algorithm is used to perform clustering (k=2) in this space. After clustering, if all feeder points are clustered into one class, it indicates that the spectral characteristics are highly consistent and the fault is determined to be a bus fault. If they are clearly divided into two classes, in the two-dimensional feature space, the faulty feeder points will obviously gather in the region far away from the non-faulty feeder points. At this time, the cluster with fewer points is the faulty line.

[0043] The spectral clustering algorithm used in step S4 is based on graph theory. By converting data points into points in the feature vector space before clustering, it can effectively identify clusters with non-spherical distribution and complex shapes. It is also insensitive to the local structure of the similarity matrix, and the clustering results are usually more stable than density-based clustering algorithms (such as DBSCAN).

[0044] This invention addresses the limitations of existing distribution network fault location methods in new distribution networks, specifically their failure to consider the weak fault signals within short-time data windows provided by renewable energy sources. It provides a novel ground fault location method based on the current traveling wave spectrum. This innovative method utilizes the current traveling wave spectrum for fault location, fully considering that traveling waves are broadband signals. By extracting higher-frequency current traveling waves, it avoids low-frequency interference generated by renewable energy integration, reducing the impact of renewable energy in new distribution networks. While acknowledging that the fault signals provided by renewable energy are short and weak, they still contain rich frequency domain fault information. Therefore, extracting the current frequency traveling wave signal from the fault signal is unaffected by the weak fault signals within short-time data windows provided by renewable energy. Furthermore, it employs a Dynamic Time Warping (DTW) algorithm that accurately measures waveform shape similarity, combined with Spectral Clustering, which effectively identifies complex data distributions. The Clustering algorithm replaces quantitative tuning calculations with qualitative differences in the spectrum waveform, which improves the problem of fault selection threshold tuning caused by the access of new energy sources from the source, and realizes high sensitivity and high reliability of the new distribution network while taking into account high ground resistance grounding fault selection.

[0045] This invention introduces the current traveling wave spectrum waveform, using the qualitative differences of the spectrum waveform to replace the quantitative tuning calculation, which is not affected by the weak fault signals of the short-time data window provided by new energy sources. Furthermore, through cutting-edge similarity and clustering algorithms, it improves the problem of difficulty in tuning the fault selection threshold and clustering stability caused by the access of new energy sources from the source.

[0046] This invention further discloses a high-resistivity grounding fault location system for distribution networks based on the current traveling wave spectrum, comprising: The current traveling wave line mode signal acquisition module is used to acquire the three-phase current traveling wave signals at the head end of each feeder in the new distribution network, and to perform phase mode transformation on each three-phase current traveling wave signal to obtain the current traveling wave line mode signal corresponding to each feeder. The high-frequency spectrum vector construction module is used to perform spectrum transformation analysis on the current traveling wave line mode signal of each feeder, and extract and construct the high-frequency spectrum vector corresponding to each feeder. The similarity distance calculation module is used to calculate the similarity distance between the high-frequency spectral vectors of every two feeders based on the dynamic time warping algorithm, and to construct... n ×n Similarity distance matrix of the spectrum of traveling wave of feeder current. n The total number of feeders is represented by the element values ​​in the matrix, which characterize the degree of difference in the morphology of the spectral waveforms of the corresponding two feeders. The intelligent clustering and discrimination module is used to perform spectral clustering analysis on the similarity distance matrix of the feeder current traveling wave to obtain the clustering results; then, the fault location is determined based on the clustering results: if the element values ​​of all feeders are clustered into one category, it is determined to be a bus fault; if the element values ​​are clearly divided into two categories, it is determined that the outlier feeder or the smaller cluster of feeders has a high-resistance grounding fault. The result output and alarm module is used to display the fault selection results and issue an alarm signal when the fault is determined to be a feeder fault.

[0047] It should be noted that the specific implementation process of each module is detailed in the above method and will not be repeated here. It should be understood that the above division of functional modules is merely a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the integrated units mentioned above can be implemented in hardware or as software functional units.

[0048] This invention innovatively employs the current traveling wave spectrum waveform for distribution network fault location, fully utilizing the characteristic of traveling waves that their transmission speed is close to the speed of light and that they can extract data over a short period. By extracting and analyzing the high-frequency components of fault signals, it can effectively avoid interference from weak fault signals within short-term data windows provided by new energy sources. It is unaffected by weak fault signals within short-term data windows provided by new energy sources, exhibiting characteristics of fast response, high sensitivity, and strong robustness. It is particularly suitable for fault detection in power systems with new energy integration, applicable to common types of distribution network neutral grounding methods and new energy operating conditions, and easy to implement directly in engineering applications involving new energy integration.

[0049] This invention, based on DTW and spectral clustering algorithms, establishes a fast, simple, and intuitive fault selection principle. The DTW algorithm can capture the morphological differences of the spectral waveform, is insensitive to waveform distortion and frequency offset, and is more robust than rigid distance metrics. The spectral clustering algorithm can identify clusters of arbitrary shapes and is insensitive to parameter settings, with stable and reliable clustering results, avoiding the dependence of traditional clustering algorithms on thresholds or initial values. By replacing quantitative tuning calculations with qualitative differences in the spectral waveform, it can accurately select feeders in new distribution networks under different fault locations, fault types, transition resistances, and noise interference. This fundamentally improves the problem of difficulty in setting fault selection thresholds under conditions of new energy access and high resistance, achieving highly sensitive and reliable fault selection for new distribution networks while also considering high-resistance grounding faults.

[0050] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0051] When any feeder in a new distribution network experiences a ground fault, especially a high-resistance ground fault, the spectrum waveforms of the faulty feeder and the non-faulty feeder are significantly different due to the transmission and frequency-dependent variation mechanism of the current traveling wave. The spectrum waveforms of the non-faulty feeders are not significantly different. Based on this inherent difference mechanism, reliable line selection in the distribution network can be achieved.

[0052] like Figure 1 As shown, the method for selecting the fault location of a distribution network based on the current traveling wave spectrum provided by this invention includes the following steps: S1. Possesses n After a high-resistance ground fault occurs in a new type of distribution network with multiple feeders, the current traveling wave detection equipment installed at the beginning of each feeder is activated. The equipment samples at a frequency of 100kHz and captures the time-domain fault current traveling wave line-mode signal within 1ms. i k ( t ),in k =1,2,···, n This represents the feeder number.

[0053] It should be understood that the present invention does not limit the method of extracting the current traveling wave mode signal. Furthermore, this embodiment preferably utilizes Karenbell phase-mode transformation for the feeder. k The detected three-phase current traveling wave signal is processed to obtain... α Components as current traveling wave line mode signals i k (t) This method allows for the acquisition of more stable, short-time data windows of weak fault current signals provided by new energy sources, increasing the success rate of high-frequency signal extraction.

[0054] S2. Utilize spectrum transformation to analyze the current traveling wave mode signal of each feeder. i k (t) Perform spectral transformation analysis to obtain the corresponding spectral vectors of each feeder. S k (f) The current traveling wave spectrum waveform was plotted. The frequency window of the spectrum waveform was set to a high-frequency band window of 20-100kHz to avoid the influence of a large number of low-frequency harmonics generated by the nonlinear characteristics of new energy power electronic devices, thereby enhancing the adaptability of the proposed method in new power distribution networks.

[0055] The embodiments of the present invention do not limit the method of extracting the current traveling wave spectrum waveform.

[0056] According to the Fast Fourier Transform, the feederk Current traveling wave line mode signal i k (t) Decomposed into sampling points of m spectral vector S k (f) : (1) In the formula, the elements within the vector S kx The magnitude represents the first spectral signal. x The amplitude corresponding to each sampling point.

[0057] However, this embodiment of the invention preferably employs the Pisareenko spectrum transform technology. Through its high-resolution characteristics, it effectively avoids the spectrum distortion problem caused by short-time data windows in new energy scenarios. Simultaneously, it can accurately suppress interference from weak fault signals and significantly improve the extraction capability of high-frequency signals, ensuring high-precision spectrum analysis under complex operating conditions. The specific process is as follows: S201. Convert the time-domain signal i k ( t Discrete signals are obtained by sampling. i k ( p ), p Given the number of sampling points in the discrete signal, calculate the autocorrelation function of the discrete signal. R k [ q This effectively overcomes the spectral aliasing problem of short-time data windows in new energy scenarios and avoids energy leakage caused by data truncation in traditional methods. (2) In the formula, N It is a current traveling wave line mode signal i k ( t ) length, q It's a delay.

[0058] S202. Constructing the Autocorrelation Matrix R k : (3) S203. Perform eigenvalue decomposition on the autocorrelation matrix and select the m largest eigenvalues ​​and their corresponding eigenvectors: (4) In the formula, V k It is the eigenvector matrix. Λ kIt is an eigenvalue matrix; the largest eigenvalue is selected. m eigenvalues λ 1, λ 2,···, λ m and its corresponding eigenvectors v 1, v 2,···, v m .

[0059] S204. Construct a sampling point system using the selected eigenvalues ​​and eigenvectors. m spectral vector S k (f) Effectively addresses interference from short-term data windows and weak fault signals in new energy sources, and extracts high-frequency signals: (5) In the formula, v x (f) These are the eigenvectors.

[0060] Through the reciprocal of the eigenvalue 1 / λ x The weighting mechanism optimizes the signal energy ratio within a short data window, suppressing spectral leakage; it utilizes feature vectors. v x (f) The orthogonal projection characteristics enhance the anti-interference ability of weak fault signals; combined with the sparse distribution characteristics of the feature spectrum, high-frequency transient signals are accurately extracted, breaking through the frequency domain limitations of traditional methods.

[0061] S3. Upon detecting a new power distribution network n Current traveling wave spectrum of a single feeder S 1 (f),S 2 (f),···,S k (f),···,S n (f) Based on this, the similarity distance between these vectors is calculated.

[0062] With feeder k and feeder j For example, this invention uses the Dynamic Time Warping (DTW) algorithm to calculate the spectral vectors of two feeders. S k (f) and S j (f) similar distance between D dtw (S k (f),Sj (f)) The specific steps are as follows: S301. Construct a m × m Local cost matrix D Its elements d ( x , y )=| S kx - S jy |, among which x , y =1,2,···, m ; S302. Calculate the cumulative cost matrix using the DTW algorithm. C Its elements c(x,y) The recursive formula is: (6) In the formula, c (1,1)= d (1,1), and set boundary conditions. c (0,:)= c (:,0)=∞; S303. Bottom right element of the cumulative cost matrix c ( m , m The value of ) is the DTW distance. D dtw (S k (f),S j (f)) ; After calculating the DTW distance between all feeder pairs, the network can be constructed. n × n 3D similarity distance matrix M ,in M kj = D dtw (S k (f),S j (f)) .

[0063] Based on this, we can obtain n × n Similarity distance matrix of traveling wave of feeder current M : (7) In the formula, the elements in the matrix Mkj For feeder k Current traveling wave spectrum vector S k (f) With feeder j Current traveling wave spectrum vector S j (f) The larger the similarity distance, the greater the difference in the current traveling wave spectrum between the feeders.

[0064] In this embodiment of the invention, DTW distance is preferred as a similarity distance metric. Traditional distance metrics (such as Manhattan distance) are sensitive to small offsets of the spectral waveform on the frequency axis. By finding the optimal regularized path that minimizes the cumulative difference between two sequences, DTW can effectively align local stretching or compression of the waveform on the frequency axis, thereby more accurately reflecting the similarity of the overall waveform shape.

[0065] S4. Obtain the similarity distance matrix. M Subsequently, spectral clustering algorithm is used for fault identification. Spectral clustering transforms the clustering problem into a graph partitioning problem, effectively handling non-convex data distributions. Its specific steps include: matrix M Converted to an affinity matrix using a Gaussian kernel function. A A normalized Laplacian matrix is ​​constructed and eigenvalue decomposition is performed. The eigenvectors corresponding to the two smallest eigenvalues ​​are taken to form a new feature space. Finally, the K-means++ algorithm is used to perform clustering in this space. k =2); After clustering, if all feeder points cluster into one category, it indicates that the spectral characteristics are highly consistent, and the fault is determined to be a bus fault; if they are clearly divided into two categories, then in the two-dimensional feature space, the faulty feeder points will obviously cluster in the region far away from the non-faulty feeder points, and the cluster with fewer points is the faulty line. The specific process is as follows: S401. Constructing the affinity matrix A Using Gaussian kernel function to analyze similarity distance matrices M Perform the transformation to generate the affinity matrix. A Its elements are defined as: (8) In the formula, A kk =0, the scale parameter σ controls the width of the domain, and can be taken as a matrix. M The standard deviation of all elements.

[0066] S402. Constructing the normalized Laplacian matrix: First, calculate the degree matrix. Λ (Diagonal matrix), diagonal elements Λ kk for: (9) Then calculate the normalized Laplacian matrix. In the formula I It is the identity matrix. Represents the degree matrix Λ For each diagonal element, first find its inverse and then take its square root.

[0067] S403. Characteristic Decomposition: Since this invention clearly distinguishes between "fault" and "non-fault" categories, the calculation... L norm The eigenvectors corresponding to the first two smallest eigenvalues u 1, u 2∈ R n , R n Let represent an n-dimensional real vector space and form the eigenvector matrix. U =[ u 1, u 2]∈ R n×2 ; S404. Clustering: merging matrices U Each row is considered as a point in two-dimensional space. y i ∈ R 2 , i =1,···, n The K-means++ algorithm was used to cluster these points into two clusters. C 1, C 2.

[0068] Specifically, the matrix U Each row (i.e., each two-dimensional point) is input into the K-means clustering algorithm, with the number of clusters set to 2. K-means assigns a cluster label (0 or 1) to each point. This label directly corresponds to the fault identification result; if K-means assigns all points to the same cluster, it indicates that the spectral characteristics of all feeders are highly consistent, and the fault is determined to be a bus fault; if the points are divided into two clusters, the cluster with significantly fewer points is the faulty feeder cluster.

[0069] The high-resistivity grounding fault location method for distribution networks based on the current traveling wave spectrum of this invention has the following advantages compared with existing distribution network location methods based on steady-state and transient characteristics: 1. This invention innovatively uses the current traveling wave spectrum waveform for distribution network line selection. It utilizes the characteristic that the traveling wave has a transmission speed close to the speed of light and can extract data in a short time period to achieve ultra-high-speed line selection within 1ms in the distribution network. Moreover, it is not affected by the weak fault signals in the short-time data window provided by new energy sources. It is applicable to common types of distribution network neutral point grounding methods and new energy operating conditions, and is easy to implement directly in engineering applications where new energy is connected. 2. Based on DTW and spectral clustering algorithms, this invention establishes a fast, simple, and intuitive fault selection principle. The DTW algorithm can capture the morphological differences of the spectral waveform, is insensitive to waveform distortion and frequency offset, and is more robust than rigid distance metrics. The spectral clustering algorithm can identify clusters of arbitrary shapes and is insensitive to parameter settings, with stable and reliable clustering results, avoiding the dependence of traditional clustering algorithms on thresholds or initial values. By replacing quantitative tuning calculations with qualitative differences in the spectral waveform, it can accurately select feeders in new distribution networks under different fault locations, fault types, transition resistances, and noise interference. This fundamentally improves the problem of difficulty in setting fault selection thresholds under conditions of new energy access and high resistance, achieving highly sensitive and reliable fault selection for new distribution networks while also considering high-resistance grounding faults.

[0070] Figure 2 This is a simulation model of a new 10kV distribution network built in the PSCAD / EMTDC simulation environment according to the present invention. The model uses 5 lines... L 1- L Five lines with different power supply methods are included: two overhead-cable hybrid lines with branches, one overhead line with a branch, one 14km overhead line, and one 15km underground cable line. The parameters for both cables and overhead lines are based on frequency-varying parameter models. The lengths and end numbers of each line are also specified. E 1- E 17 All in Figure 2 The project was successfully tendered. The new energy DG1-DG3 units have capacities of 4, 8, and 8MW respectively, with a maximum output current of 1.5 times the rated current. The control strategy is conventional PQ control, and it has low voltage ride-through capability.

[0071] In respectively Figure 2 Four sets of high-resistance grounding faults are set in the feeder: L Section 1 E 1 E 2 distance E A fault is set at 1.2km on one end. F 1; in the feeder L Section 2 E 8 E 9 distance E A fault was set at 2.7km from point 8. F 2; Set a fault at the busbar F3; in the feeder L 4 E 12 Terminal setting fault F 4. Twelve typical case simulation tests were conducted on four groups of faults under different operating conditions (including different transition resistances and fault types). Current traveling wave detection equipment installed at the beginning of each feeder acquired the current traveling wave line mode spectrum waveform of each feeder. Difference analysis was performed on the spectrum waveforms between the feeders to obtain the DTW similarity distance matrix. The elements in the matrix were then subjected to spectral clustering to determine the location of the faulty line. Throughout the test, the traveling wave detection equipment initiated the line selection process at the moment it detected the current traveling wave signal of each feeder. The line selection test results are shown in Table 1.

[0072] Table 1. Results of high-resistivity grounding fault location under different operating conditions

[0073] With fault F Taking the first case as an example, the current traveling wave mode spectrum waveform detected at the beginning of each feeder is as follows: Figure 3 As shown, the DTW similarity distance matrix calculated at this time M F1 As shown in equation (10): (10) Performing spectral clustering on the matrix elements in equation (10), the clustering results are as follows: Figure 4 As shown (the horizontal and vertical axes have no actual meaning). The clustering results indicate two clusters, one of which is the feeder. L 2. L 3. L 4 and L The similar distance data of 5 belong to the same cluster, feeder L The similarity distance data of 1 belongs to an outlier cluster, indicating that the feeder... L 1. A high-resistance grounding fault occurred.

[0074] Further considering the impact of new energy power source type, capacity, and noise interference on the method described in this invention, the capacity and type of DG3 are changed to... F Taking the high-resistance grounding fault of phase A through a 9kΩ transition resistor at 4 locations as an example, the line selection results are shown in Table 2.

[0075] Table 2. Results of high-resistivity grounding fault location under different new energy sources and noise interference.

[0076] As can be seen from Tables 1 and 2, the method can reliably achieve fault location for high-resistance grounding faults in new distribution networks under all circumstances. The reasons are as follows: 1. Fault type, transition resistance, and noise changes mainly affect the amplitude of the current traveling wave. Under certain fault conditions, they have little impact on the waveform shape and only slightly affect the similarity distance calculation value of the spectrum waveform, without affecting the fault location result; 2. High-frequency spectrum traveling waves can effectively avoid the influence of low-frequency harmonics generated by new energy access. Changes in the capacity and type of new energy do not affect the similarity distance calculation value and the fault location result.

[0077] Extensive simulation tests have shown that the principle of this invention is simple. It uses the qualitative difference of the spectrum waveform to replace the quantitative setting calculation, which can accurately select the feeder of the new distribution network under different fault locations, fault types, transition resistance and noise interference. It improves the problem of difficulty in setting the fault selection threshold under the conditions of new energy access and high resistance from the source, and realizes the high sensitivity and high reliability of the new distribution network while taking into account the fault selection of high resistance grounding.

[0078] The present invention also discloses a computer program product, comprising a computer program that, when executed by a processor, performs the steps of the method described above.

[0079] The present invention further discloses a computer-readable storage medium having a computer program stored thereon, the computer program executing the steps of the method described above when run by a processor.

[0080] The present invention also discloses a distribution network grounding fault selection system based on current traveling wave spectrum, including a memory and a processor connected to each other. The memory stores a computer program, which executes the steps of the method described above when run by the processor.

[0081] The products, media, and systems of the present invention, corresponding to the methods described above, also possess the advantages described above.

[0082] The present invention can implement all or part of the processes in the methods of the above embodiments, or it can be implemented by hardware related to computer program instructions. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, it can implement the steps of the above method embodiments. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form. The computer-readable storage medium includes: any entity or device capable of carrying computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. The memory is used to store computer programs and / or modules. The processor implements various functions by running or executing the computer programs and / or modules stored in the memory, and by calling data stored in the memory. The memory may include high-speed random access memory, as well as non-volatile memory, such as hard disks, RAM, plug-in hard disks, smart media cards (SMC), secure digital (SD) cards, flash cards, at least one disk storage device, flash memory device, or other volatile solid-state storage devices.

[0083] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should be considered within the scope of protection of the present invention.

Claims

1. A method for selecting the fault location of a distribution network based on the traveling wave spectrum of current, characterized in that, Including the following steps: S1. Acquire the three-phase current traveling wave signal at the head end of each feeder in the distribution network, and perform phase mode transformation on each of the three-phase current traveling wave signals to obtain the current traveling wave line mode signal corresponding to each feeder; S2. Perform spectral transformation analysis on the current traveling wave line mode signal of each feeder, and extract and construct the high-frequency spectrum vector corresponding to each feeder; S3. Based on the dynamic time warping algorithm, calculate the similarity distance between the high-frequency spectrum vectors of every two feeders, and construct... n × n Similarity distance matrix of the spectrum of traveling wave of feeder current. n The total number of feeders is represented by the element values ​​in the matrix, which characterize the degree of difference in the morphology of the spectral waveforms of the corresponding two feeders. S4. Perform spectral clustering analysis on the similarity distance matrix of the feeder current traveling wave to obtain the clustering results; then determine the fault location based on the clustering results: if the element values ​​of all feeders are clustered into one category, it is determined to be a bus fault; if the element values ​​are clearly divided into two categories, it is determined that the outlier feeder or the smaller cluster of feeders has a high-resistance grounding fault.

2. The method for selecting the fault location of a distribution network based on the current traveling wave spectrum according to claim 1, characterized in that, The current traveling wave line mode signal in step S1 is obtained after performing a Kelvin-Bell phase mode transformation on the three-phase current traveling wave signal. α component signal.

3. The method for selecting the fault location of a distribution network based on the current traveling wave spectrum according to claim 1, characterized in that, The specific process of step S2 is as follows: S201. Convert the time-domain current traveling wave mode signal i k ( t Discrete signals are obtained by sampling. i k (p) , p Given the number of sampling points for the discrete signal, calculate the discrete signal. i k (p) autocorrelation function R k [ q ]: In the formula, N It is a current traveling wave line mode signal i k (t) Length, q It's a delay; S202. Based on autocorrelation function R k [ q Construct the autocorrelation matrix R k : S203. On the autocorrelation matrix R k Perform eigenvalue decomposition: In the formula, V k It is the eigenvector matrix. Λ k It is an eigenvalue matrix; Select the largest eigenvalue. m eigenvalues λ 1, λ 2,···, λ m and its corresponding eigenvectors v 1, v 2,···, v m ; S204. Construct a sampling point system using the selected eigenvalues ​​and eigenvectors. m spectral vector S k (f) : In the formula, v x (f) These are the eigenvectors.

4. The method for selecting the fault location of a distribution network based on the current traveling wave spectrum according to claim 1, 2, or 3, characterized in that, In step S3, the specific process of calculating the similarity distance between the high-frequency spectral vectors of every two feeders is as follows: S301. Construct a m × m Local cost matrix D Its elements d(x,y) =| S kx - S jy |, among which x , y =1,2,···, m ; S302. The cumulative cost matrix is ​​calculated using the dynamic time warping algorithm. C Its elements c ( x , y The recursive formula for ) is: In the formula, c (1,1)= d (1,1), and set boundary conditions. c (0,:)= c (:,0)=∞; S303. Bottom right element of the cumulative cost matrix c ( m , m The value of ) is the DTW similarity distance. D dtw (S k (f),S j (f)) ; S k (f),S j (f) feeders k and feeder j The spectrum vector.

5. The method for selecting the fault location of a distribution network based on the current traveling wave spectrum according to claim 4, characterized in that, n × n Similarity distance matrix of traveling wave spectrum of feeder current M Specifically: In the formula, the elements in the matrix M kj For feeder k Current traveling wave spectrum vector S k (f) With feeder j Current traveling wave spectrum vector S j (f) The larger the similarity distance, the greater the difference in the current traveling wave spectrum between the feeders.

6. The method for selecting the fault location of a distribution network based on the current traveling wave spectrum according to claim 1, 2, or 3, characterized in that, In step S4, the specific process of performing spectral clustering analysis on the similarity distance matrix of the feeder current traveling wave to obtain the clustering results is as follows: S401. Using the Gaussian kernel function to analyze the similarity distance matrix M Perform the transformation to generate the affinity matrix. A Its elements are defined as: In the formula, A kk =0, the scale parameter σ controls the width of the domain, and is taken as a matrix. M Standard deviation of all elements; elements within the matrix M kj For feeder k Current traveling wave spectrum vector S k (f) With feeder j Current traveling wave spectrum vector S j (f) The similarity distance; S402. Calculate the degree matrix Λ diagonal elements Λ kk for: Then calculate the normalized Laplacian matrix. In the formula I It is the identity matrix. Represents the degree matrix Λ For each diagonal element, first find its inverse and then take its square root; S403. Calculation L norm The eigenvectors corresponding to the first two smallest eigenvalues u 1, u 2∈ R n , R n Let represent an n-dimensional real vector space and form the eigenvector matrix. U =[ u 1, u 2]∈ R n×2 ; S404. Matrix U Each row is considered as a point in two-dimensional space. y i ∈ R 2 , i =1,···, n The K-means++ algorithm was used to cluster these points into two clusters. C 1, C 2.

7. The method for selecting the fault location of a distribution network based on the current traveling wave spectrum according to claim 6, characterized in that, In step S4, the matrix U Each line of input is fed into the K-means clustering algorithm, and the number of clusters is set to 2. K-means assigns a cluster label to each point. This label directly corresponds to the fault identification result. If K-means assigns all points to the same cluster, it means that the spectral characteristics of all feeders are highly consistent, and it is determined to be a bus fault. If the points are divided into two clusters, the cluster with significantly fewer points is the faulty feeder cluster.

8. A computer program product, comprising a computer program, characterized in that, The computer program is executed by the processor to perform the steps of the method as described in any one of claims 1-7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, The computer program, when run by a processor, performs the steps of the method as described in any one of claims 1-7.

10. A distribution network grounding fault location system based on current traveling wave spectrum, comprising a memory and a processor interconnected, wherein the memory stores a computer program, characterized in that, The computer program, when run by a processor, performs the steps of the method as described in any one of claims 1-7.

Images