A distribution network fault location method and terminal

Abstract

The present invention discloses a distribution network fault location method and terminal. The method comprises the following steps: constructing a distribution network system topology structure, and collecting current signals between sub-node terminals according to the distribution network system topology structure; compressing and collecting the current signals to obtain compressed signals; reconstructing the compressed signals to obtain reconstructed compressed signals, and decomposing the reconstructed compressed signals using an adaptive noise complete set empirical mode decomposition algorithm to obtain a series of intrinsic mode functions; and performing cluster analysis based on the series of intrinsic mode functions using a fuzzy C-means clustering algorithm to obtain fault point location information. The method can complete signal sampling at a lower sampling rate, give full play to the role of useful signal information, reduce information redundancy, use CEEMDAN for signal decomposition to obtain better decomposition effect, and utilize the fuzzy C-means clustering algorithm to quickly and accurately locate the fault line, thereby accurately and efficiently locating the fault point.

Description

Technical Field

[0001] The present invention relates to the technical field of distribution network fault detection, and in particular to a distribution network fault locating method and terminal. Background Art

[0002] The distribution network is a vital component of the power grid, serving as a hub connecting energy production and consumption and a key element in building a new power system. The integration of large-scale distributed renewable energy and various energy storage devices into the grid has not only improved power supply performance but has also had a profound impact on the distribution network's operations, prompting further improvements and enhancements to traditional distribution network protection methods. The distribution network, directly connected to consumers, is a vital public infrastructure serving the public's livelihoods. Therefore, it has more stringent requirements for power quality and power supply security. According to statistics, over 80% of consumer power outages are caused by distribution network faults, of which single-phase ground faults account for over 80%. In medium-voltage distribution networks, even if a single-phase ground fault occurs, the system can maintain normal operation for a short period of time. However, in new distribution networks, the introduction of distributed renewable energy can increase fault currents. If the fault point is not quickly isolated, the fault can easily expand. Therefore, it is crucial to quickly, accurately, and reliably locate the fault point, isolate the fault, repair the faulty section, and restore normal power supply as soon as possible.

[0003] Currently, there are two main methods for locating fault points in distribution networks: direct and indirect. Direct algorithms primarily include matrix algorithms. Matrix algorithms are used to locate faults in distribution networks. They utilize the distribution network topology to form a description matrix representing the distribution network model. After a series of operations, a fault location discriminant matrix is ​​obtained, resulting in the location of the fault point. However, the success of these algorithms is directly dependent on the information collected by sensors, and their fault tolerance is poor. Indirect algorithms for locating faults in distribution networks employ an improved bionic electromagnetic method. This method establishes a solution model based on switching functions and uses the fault section as the solution space to obtain the optimal solution. This method offers good fault tolerance, diverse applications, and fast update speed, but suffers from slow computational speed and a tendency to fall into local optima. Alternatively, convolutional neural networks are used to train distribution network fault data. Neural networks have strong computational capabilities and can handle large amounts of complex data. However, when applied to fault location in distribution networks with distributed generation (DGs), the algorithm struggles to converge due to the volatile output of DGs, which can be switched on and off at any time. Furthermore, neural network-based fault location in distribution networks requires extensive initial training data, resulting in a high computational load. In the fault point detection method based on current signal processing, the existing signal acquisition process mainly consists of four steps: sampling, compression, transmission, and decompression. The sampling process must follow the Shannon-Nyquist sampling theorem. If this process is applied to the fault current data of the distribution network, a large amount of data stacking will be generated, which not only increases the redundancy of information, but also prolongs the processing time, occupies the transmission bandwidth and storage space of the line, and will seriously restrict the efficient use of data and high-speed transmission. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a distribution network fault locating method and terminal, which can accurately and efficiently locate the fault point.

[0005] In order to solve the above technical problems, a technical solution adopted by the present invention is:

[0006] A method for locating a distribution network fault, comprising the steps of:

[0007] Constructing a distribution network system topology structure, and collecting current signals between the terminals of each sub-node according to the distribution network system topology structure;

[0008] compressing and collecting the current signal to obtain a compressed signal;

[0009] Reconstructing the compressed signal to obtain a reconstructed compressed signal, and decomposing the reconstructed compressed signal using an adaptive noise complete set empirical mode decomposition algorithm to obtain a series of intrinsic mode functions;

[0010] Cluster analysis is performed based on the series of intrinsic mode functions using a fuzzy C-means clustering algorithm to obtain fault point location information.

[0011] In order to solve the above technical problems, another technical solution adopted by the present invention is:

[0012] A distribution network fault location terminal includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the following steps are implemented:

[0013] Constructing a distribution network system topology structure, and collecting current signals between the terminals of each sub-node according to the distribution network system topology structure;

[0014] compressing and collecting the current signal to obtain a compressed signal;

[0015] Reconstructing the compressed signal to obtain a reconstructed compressed signal, and decomposing the reconstructed compressed signal using an adaptive noise complete set empirical mode decomposition algorithm to obtain a series of intrinsic mode functions;

[0016] A fuzzy C-means clustering algorithm is used to perform cluster analysis based on the series of intrinsic mode functions to obtain fault point location information.

[0017] The beneficial effects of the present invention are as follows: compressing and collecting the current signal to obtain a compressed signal, reconstructing the compressed signal, and decomposing the reconstructed compressed signal using an adaptive noise complete set empirical mode decomposition algorithm to obtain a series of intrinsic mode functions, clustering analysis is performed using a fuzzy C-means clustering algorithm based on the series of intrinsic mode functions to obtain fault point location information, using a compressed sampling method to sample the current signal, which can complete signal sampling at a lower sampling rate, give full play to the role of useful signal information, and reduce information redundancy, using an adaptive noise complete set empirical mode decomposition algorithm (CEEMDAN) to decompose the signal can suppress the modal aliasing problem existing in EMD and obtain a better decomposition effect, and finally using the fuzzy C-means clustering algorithm to quickly and accurately locate the fault line, thereby accurately and efficiently locating the fault point. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a flow chart of the steps of a distribution network fault location method according to an embodiment of the present invention;

[0019] Figure 2 This is a structural diagram of a distribution network fault location terminal according to an embodiment of the present invention;

[0020] Figure 3 A schematic diagram of a distribution network network architecture in a distribution network fault location method according to an embodiment of the present invention;

[0021] Figure 4 This is a schematic diagram of fault identification and fault isolation corresponding to a certain regional network architecture in a distribution network fault location method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0022] To illustrate the technical content, achieved objectives and effects of the present invention in detail, the following description is given in conjunction with the embodiments and accompanying drawings.

[0023] Please refer to Figure 1 , an embodiment of the present invention provides a distribution network fault location method, comprising the steps of:

[0024] Constructing a distribution network system topology structure, and collecting current signals between the terminals of each sub-node according to the distribution network system topology structure;

[0025] compressing and collecting the current signal to obtain a compressed signal;

[0026] Reconstructing the compressed signal to obtain a reconstructed compressed signal, and decomposing the reconstructed compressed signal using an adaptive noise complete set empirical mode decomposition algorithm to obtain a series of intrinsic mode functions;

[0027] A fuzzy C-means clustering algorithm is used to perform cluster analysis based on the series of intrinsic mode functions to obtain fault point location information.

[0028] From the above description, it can be seen that the beneficial effects of the present invention are: compressing and collecting the current signal to obtain a compressed signal, reconstructing the compressed signal, and decomposing the reconstructed compressed signal using the adaptive noise complete set empirical mode decomposition algorithm to obtain a series of intrinsic mode functions, clustering analysis is performed using the fuzzy C-means clustering algorithm based on the series of intrinsic mode functions to obtain fault point location information, using the compressed sampling method to sample the current signal, it is possible to complete signal sampling at a lower sampling rate, give full play to the role of useful signal information, and reduce information redundancy, using the adaptive noise complete set empirical mode decomposition algorithm (CEEMDAN) for signal decomposition can suppress the modal aliasing problem existing in EMD and obtain a better decomposition effect, and finally using the fuzzy C-means clustering algorithm to quickly and accurately locate the fault line, thereby accurately and efficiently locating the fault point.

[0029] Furthermore, compressing and collecting the current signal to obtain a compressed signal includes:

[0030] Determine the discrete cosine transform basis as a sparse transform basis matrix and initialize the sparse transform dictionary;

[0031] Solve the sparse representation using an orthogonal matching pursuit algorithm according to the current signal to obtain a sparse representation coefficient;

[0032] Iteratively updating the sparse transform dictionary to obtain an updated sparse transform dictionary;

[0033] Performing sparse transformation on the current signal according to the updated sparse transformation dictionary to obtain a sparse signal;

[0034] Determine a matrix in which all column vectors in the sparse transformation basis matrix are uncorrelated and conform to Gaussian distribution as a measurement matrix;

[0035] A compressed signal is obtained according to the measurement matrix and the sparse signal.

[0036] As can be seen from the above description, compressed acquisition is a concept for low-sampling-rate data acquisition. Sampling is completed at a frequency far lower than the Nyquist sampling theorem, reducing information redundancy and network load, thereby improving the efficiency of fault location. The basis matrix used is the discrete cosine transform basis, which can reduce the computational complexity by more than half compared to the discrete Fourier transform, better meeting the timeliness requirements of fault location signal processing. At the same time, signal compression can be completed at a high compression rate and reconstruction can be completed with a low mean square error at the distribution network substation, reducing the amount of data transmitted in the network and reducing the network load.

[0037] Furthermore, the sparse representation is solved using the orthogonal matching pursuit algorithm according to the current signal to obtain the sparse representation coefficient include:

[0038] ;

[0039] ;

[0040] Where, I j,t represents the current signal collected by terminal j at time t, D represents the sparse transformation dictionary, L represents the number of terminals, and N represents the collection time;

[0041] The sparse transformation is performed on the current signal according to the updated sparse transformation dictionary to obtain a sparse signal I s,t for:

[0042] ;

[0043] The compressed signal R obtained according to the measurement matrix and the sparse signal is:

[0044] ;

[0045] Where, Denotes the measurement matrix.

[0046] From the above description, it can be seen that since the current signal is not sparse or the sparsity is not obvious in the time domain, the original current signal needs to be sparsely transformed. The signal after sparse transformation can be processed by the measurement matrix to facilitate subsequent observation, thereby realizing the location of the fault point.

[0047] Furthermore, the use of the adaptive noise complete set empirical mode decomposition algorithm to decompose the reconstructed compressed signal to obtain the intrinsic mode function includes:

[0048] Adding Gaussian white noise to the reconstructed compressed signal and performing adaptive noise complete set empirical mode decomposition to obtain a series of initial intrinsic mode functions;

[0049] Calculating the arithmetic mean of the series of initial eigenmode functions to obtain a first eigenmode function;

[0050] The first intrinsic mode function is removed from the current signal to obtain an updated signal, and the step of adding Gaussian white noise to the reconstructed compressed signal and performing adaptive noise complete set empirical mode decomposition is returned to execute according to the updated signal until a unique residual that does not meet the decomposition conditions and a series of intrinsic mode functions are obtained.

[0051] From the above description, it can be seen that based on the characteristic that the fault point currents do not have similarity, the use of the adaptive noise complete set empirical mode decomposition algorithm to decompose the reconstructed compressed signal can enhance the signal characteristics and improve the accuracy of subsequent fault point location.

[0052] Furthermore, Gaussian white noise is added to the reconstructed compressed signal and adaptive noise complete set empirical mode decomposition is performed to obtain a series of initial intrinsic mode functions:

[0053] ;

[0054] Where, represents the reconstructed compressed signal, represents the standard deviation of Gaussian white noise, N i (t) represents Gaussian white noise with unit variance and mean 0, represents a series of initial eigenmode functions, represents the signal residual, m represents the number of Gaussian white noise groups, and q represents the number of initial intrinsic mode functions;

[0055] The arithmetic mean of the series of initial eigenmode functions is calculated to obtain the first eigenmode function:

[0056] ;

[0057] Where IMF1 represents the first intrinsic mode function.

[0058] As can be seen from the above description, using the above method to decompose the signal can effectively extract useful information from the signal, which facilitates the subsequent judgment of the fault point.

[0059] Furthermore, the cluster analysis is performed based on the series of intrinsic mode functions using the fuzzy C-means clustering algorithm to obtain the fault point location information including:

[0060] Using normalized energy entropy to eliminate false modes in the series of intrinsic mode functions to obtain reorganized intrinsic mode functions;

[0061] Cluster analysis is performed using a fuzzy C-means clustering algorithm based on the reorganized intrinsic mode function to obtain fault point location information.

[0062] From the above description, we can see that after the CEEMDAN algorithm decomposes, there will be a lot of noise and false modes in the early decomposition. Therefore, normalized energy entropy is further used to eliminate false modes, making the information contained in the signal mode more accurate, thereby improving the accuracy of fault location.

[0063] Furthermore, the use of normalized energy entropy to eliminate false modes in the series of intrinsic mode functions to obtain reorganized intrinsic mode functions includes:

[0064] calculating the energy of the series of eigenmode functions;

[0065] Normalizing the energy to obtain normalized energy;

[0066] Calculating normalized energy entropy values ​​corresponding to the series of intrinsic mode functions according to the normalized energy;

[0067] The eigenmode functions of the series of eigenmode functions whose normalized energy entropy values ​​are less than a first preset threshold are determined as false modes, and the false modes are eliminated to obtain reorganized eigenmode functions.

[0068] From the above description, it can be seen that the use of normalized energy entropy can quickly and effectively eliminate false modes and improve the reliability of data.

[0069] Furthermore, the cluster analysis is performed using the fuzzy C-means clustering algorithm based on the reorganized intrinsic mode function to obtain the fault point location information including:

[0070] Determining the reorganized intrinsic mode function as a sample and initializing a membership matrix;

[0071] Determining the cluster center of each sample subset in the sample;

[0072] Determining an objective function of a fuzzy C-means clustering algorithm and constraints corresponding to the objective function based on the cluster centers of the sample subsets;

[0073] Obtaining an iterative change amount of the objective function, and comparing the iterative change amount with a second preset threshold;

[0074] If the iterative change amount is greater than or equal to the second preset threshold, updating the membership matrix to obtain an updated membership matrix, and returning to the step of determining the cluster center of each sample subset according to the updated membership matrix;

[0075] If the iterative change amount is less than the second preset threshold, obtaining the category with the largest membership, and classifying the reorganized intrinsic mode function into the category with the largest membership;

[0076] Determine an upstream detection point and a downstream detection point adjacent to the fault point according to the membership matrix;

[0077] Calculating a difference in adjacent membership degrees according to the upstream detection point and the downstream detection point;

[0078] It is determined whether the difference is greater than a third preset threshold; if so, the section between the detection points corresponding to the adjacent membership degrees is determined to be the fault point location.

[0079] As can be seen from the above description, this can efficiently and accurately locate the fault point, making it easier for the distribution network substation to promptly and accurately cut off the faulty line and protect the safety of the distribution network line.

[0080] Furthermore, the cluster center of each sample subset in the sample is determined as:

[0081] ;

[0082] Where c i represents the cluster center of the i-th sample, n represents the total number of samples, X j represents the reorganized intrinsic mode function, represents the sample x calculated using the membership index h j The degree of membership to class i;

[0083] The objective function of the fuzzy C-means clustering algorithm determined based on the cluster centers of the sample subsets and the constraints corresponding to the objective function are:

[0084] ;

[0085] ;

[0086] In the formula, q represents the number of cluster centers, x j Represents the reorganized intrinsic mode function X j The jth sample in u ij Represents sample x j The degree of membership to class i.

[0087] From the above description, it can be seen that based on the above objective function and its constraints, the location of the fault point can be accurately determined, which can facilitate the staff to perform closing and opening operations on the sectionalizer of the upstream node of the fault according to the location information, thereby protecting the safety of the entire distribution network and improving the working stability of the distribution network.

[0088] Please refer to Figure 2 Another embodiment of the present invention provides a distribution network fault location terminal, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, each step in the above-mentioned distribution network fault location method is implemented.

[0089] The above-mentioned distribution network fault location method and terminal of the present invention can be applied to scenarios where fault detection of the distribution network is required, and are described below through specific implementation methods:

[0090] Example 1

[0091] Please refer to Figure 1 and Figure 3-Figure 4 , a distribution network fault location method of this embodiment includes the steps of:

[0092] S1. Constructing a distribution network system topology structure, and collecting current signals between each sub-node terminal according to the distribution network system topology structure;

[0093] The distribution network system topology defines the hierarchical relationship between distribution network substations and terminals.

[0094] S2. Compressing and collecting the current signal to obtain a compressed signal, specifically comprising:

[0095] S21, determining a discrete cosine transform basis as a sparse transform basis matrix, and initializing a sparse transform dictionary;

[0096] S22. Solve the sparse representation using the Orthogonal Matching Pursuit (OMP) algorithm according to the current signal to obtain a sparse representation coefficient. , specifically:

[0097] ;

[0098] ;

[0099] Where, I j,t represents the current signal collected by terminal j at time t, D represents the sparse transformation dictionary, L represents the number of terminals, and N represents the collection time;

[0100] S23, iteratively updating the sparse transform dictionary to obtain an updated sparse transform dictionary;

[0101] Specifically, each atom in the sparse transformation dictionary is iteratively updated, and an updated sparse transformation dictionary can be obtained after k iterative solutions.

[0102] S24, performing sparse transformation on the current signal according to the updated sparse transformation dictionary to obtain a sparse signal I s,t , specifically:

[0103] ;

[0104] S25. Determine a matrix in which all column vectors in the sparse transformation basis matrix are uncorrelated and conform to Gaussian distribution as a measurement matrix;

[0105] The sparse transformation basis matrix for:

[0106] ;

[0107] ;

[0108] ;

[0109] Where R M×M represents the basis matrix of order MxM, Represents each index , i represents the row of the sparse transformation basis matrix, r represents the column of the sparse transformation basis matrix, M represents the dimension of the M×M order sparse transformation basis matrix, and C represents the coefficient;

[0110] Among them, when i=0, ,when hour, .

[0111] In an optional embodiment, the validity of the measurement matrix is ​​also tested using an equidistance condition to determine whether the original signal can be effectively reconstructed after projecting the generated random Gaussian matrix as the measurement matrix. If the condition is met, a new random matrix satisfying the Gaussian distribution is generated. The reason for using a random Gaussian matrix instead of a direct Gaussian matrix is ​​that it has a certain degree of randomness and adaptability.

[0112] S26. Obtain a compressed signal R according to the measurement matrix and the sparse signal. Specifically:

[0113] ;

[0114] Where, Denotes the measurement matrix.

[0115] S3. Reconstruct the compressed signal to obtain a reconstructed compressed signal, and decompose the reconstructed compressed signal using an adaptive noise complete set empirical mode decomposition algorithm to obtain a series of intrinsic mode functions, specifically including:

[0116] S31, reconstructing the compressed signal to obtain a reconstructed compressed signal;

[0117] The reconstructed compressed signal for:

[0118] ;

[0119] S32, adding Gaussian white noise to the reconstructed compressed signal and performing adaptive noise complete set empirical mode decomposition to obtain a series of initial intrinsic mode functions, specifically:

[0120] ;

[0121] Where, represents the reconstructed compressed signal, represents the standard deviation of Gaussian white noise, N i (t) represents Gaussian white noise with unit variance and mean 0, represents a series of initial eigenmode functions, represents the signal residual, m represents the number of Gaussian white noise groups, and q represents the number of initial intrinsic mode functions;

[0122] S33, calculating the arithmetic mean of the series of initial eigenmode functions to obtain a first eigenmode function, specifically:

[0123] ;

[0124] Where IMF1 represents the first intrinsic mode function.

[0125] S34. Remove the first intrinsic mode function from the current signal to obtain a latest signal, and return to execute S32 to S33 according to the latest signal until a unique residual that does not meet the decomposition condition and a series of intrinsic mode functions are obtained.

[0126] Specifically, from the current signal Ij,t Remove the IMF1 part and get the latest signal v, that is, =v, and then return to execute S32~S33 until the unique residual R(t) that does not meet the decomposition conditions and a series of intrinsic mode functions are obtained.

[0127] For example, after obtaining the first intrinsic mode function, the signal residual r1(t) is calculated as:

[0128] ;

[0129] Then add Gaussian white noise to r1(t) and execute S32~S33, and then calculate the second residual, and so on. k and r k (t), until the decomposition condition cannot be met, the decomposition is terminated, and the residual R(t) is obtained. The corresponding signal can be written as: ;

[0130] After decomposition, n m-dimensional modal energy sequences are obtained, which are expressed as: ;

[0131] Wherein, the decomposition conditions are:

[0132] ;

[0133] Where N z Indicates the number of extreme points, N e represents the number of zero points, fmax(t) represents the upper envelope of the signal, and fmin(t) represents the lower envelope of the signal;

[0134] Specifically, the cubic spline difference algorithm is used to perform curve fitting on the maximum and minimum points respectively to obtain the upper envelope corresponding to the maximum point and the lower envelope corresponding to the minimum point, and the average value of the two is calculated and recorded as m(t). Let y(t) be the difference between the signal and m(t), and judge whether y(t) meets the decomposition condition.

[0135] S4. Perform cluster analysis based on the series of intrinsic mode functions using the fuzzy C-means clustering algorithm to obtain fault point location information, specifically including:

[0136] S41, using normalized energy entropy to eliminate false modes in the series of intrinsic mode functions to obtain reorganized intrinsic mode functions, specifically including:

[0137] S411. Calculate the energy E(X) of the series of intrinsic mode functions. Specifically:

[0138] ;

[0139] Where, Xn (t) represents the series of eigenmode functions.

[0140] S412: Normalize the energy to obtain normalized energy p(n). Specifically:

[0141]

[0142] S413. Calculate the normalized energy entropy values ​​EN corresponding to the series of intrinsic mode functions according to the normalized energy. n , specifically:

[0143] ;

[0144] S414: Determine the eigenmode functions in the series of eigenmode functions whose normalized energy entropy values ​​are less than a first preset threshold as false modes, and eliminate the false modes to obtain reorganized eigenmode functions.

[0145] The energy spectrum of a signal can characterize the relative relationship of the energy occupied by each state variable in the entire system, and is less affected by noise interference within the signal. The intrinsic mode function that is sensitive to the characteristic information of the original signal should occupy the main energy, while the energy of the false mode should occupy a smaller proportion. The first preset threshold is 0.1;

[0146] The reorganized eigenmode function X j for:

[0147] ;

[0148] S42: Perform cluster analysis based on the reorganized intrinsic mode function using a fuzzy C-means clustering algorithm to obtain fault point location information, specifically including:

[0149] S421, determining the reorganized intrinsic mode function as a sample, and initializing a membership matrix;

[0150] Among them, the sum of the membership degrees of the sample set is 1.

[0151] S422: Determine the cluster center of each sample subset in the sample, specifically:

[0152] ;

[0153] Where c i represents the cluster center of the i-th sample, n represents the total number of samples, X j represents the reorganized intrinsic mode function, represents the sample x calculated using the membership index h j The degree of membership to class i;

[0154] S423: Determine the objective function of the fuzzy C-means clustering algorithm and the constraints corresponding to the objective function based on the cluster centers of the sample subsets. Specifically:

[0155] ;

[0156] ;

[0157] In the formula, q represents the number of cluster centers, x j Represents the reorganized intrinsic mode function X j The jth sample in u ij Represents sample x j The degree of membership to class i.

[0158] Among them, the Lagrange multiplier method is used to obtain u ij and c i are interrelated, and u is obtained through iterative operations ij and c i The value of .

[0159] S424: Obtain an iterative change amount of the objective function, and compare the iterative change amount with a second preset threshold;

[0160] S425: If the iterative change amount is greater than or equal to the second preset threshold, the membership matrix is ​​updated to obtain an updated membership matrix, and the process returns to step S422 according to the updated membership matrix;

[0161] Wherein, the updated membership matrix U is:

[0162] ;

[0163] S426. If the iterative change amount is less than the second preset threshold, obtaining the category with the largest membership, and classifying the reorganized intrinsic mode function into the category with the largest membership;

[0164] S427, determining an upstream detection point and a downstream detection point adjacent to the fault point according to the membership matrix;

[0165] S428, calculating the difference between adjacent membership degrees according to the upstream detection point and the downstream detection point;

[0166] S429: Determine whether the difference is greater than a third preset threshold; if so, determine that the section between the detection points corresponding to the adjacent membership degrees is the fault point location; otherwise, it is not a fault point.

[0167] Wherein, the third preset threshold is 0.1;

[0168] The distribution substation can perform closing and opening operations on the sectionalizer of the node upstream of the fault point according to the location of the fault point.

[0169] like Figure 3 As shown, Figure 3 This diagram shows the architecture of a regional distribution network. This architecture consists of several distribution rooms / substations with direct electrical connections, forming a distribution area. Within this area, a distribution substation is deployed to implement the control functions of the master station, while the remaining distribution rooms / substations serve as intelligent terminals.

[0170] like Figure 4 As shown, Figure 4 This diagram illustrates fault identification and isolation for the region's network architecture. Current sensors collect current signals between terminal nodes within the regional centralized distribution network, performing compressed sampling, transmission reconstruction, and cluster analysis. When a line fault occurs, the distribution substation receives the cluster analysis results, controls the disconnection of the local recloser, and simultaneously sends a sectionalizer blocking signal to the upstream node where the fault occurred. Nodes that receive the blocking signal block their sectionalizers, while nodes that do not receive the blocking signal disconnect their sectionalizers. After a delay, the recloser closes, rapidly restoring power to non-faulty areas.

[0171] Figure 4 The relationship between the nodes shown in the figure is as follows: the distribution substation has eight terminal nodes, T1-T8, with point B between T3-T4 representing the faulty section. Upon receiving the fault signal, the distribution substation first controls the K1-1 recloser to open. The substation then issues a sectionalizer blocking signal. Upon receiving the sectionalizer blocking signal, nodes T1 and T2 control sectionalizers K1-2 and K2-2 to complete the blocking operation. T3 then controls the K3-1 recloser to open. After a time delay, the recloser closes, quickly restoring power to the non-faulty section.

[0172] Example 2

[0173] Please refer to Figure 2 A distribution network fault location terminal of this embodiment includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, each step of the distribution network fault location method in the first embodiment is implemented.

[0174] In summary, the present invention provides a distribution network fault location method and terminal, which constructs a distribution network system topology structure, and collects current signals between each sub-node terminal according to the distribution network system topology structure; compresses and collects the current signal to obtain a compressed signal; reconstructs the compressed signal to obtain a reconstructed compressed signal, and uses the adaptive noise complete set empirical mode decomposition algorithm to decompose the reconstructed compressed signal to obtain a series of intrinsic mode functions; based on the series of intrinsic mode functions, a fuzzy C-means clustering algorithm is used to perform cluster analysis to obtain fault point location information. Specifically, due to CEEMDAN After the algorithm decomposition, the problems of a large amount of noise and false modes in the early decomposition are solved. The normalized energy entropy is further used to eliminate the false modes, making the information contained in the signal mode more accurate. The compressed sampling method is used to sample the current signal, which can complete the signal sampling at a lower sampling rate, giving full play to the role of the useful information of the signal and reducing the redundancy of information. The adaptive noise complete set empirical mode decomposition algorithm is used for signal decomposition to suppress the modal aliasing problem of EMD and obtain better decomposition effect. Finally, the fuzzy C-means clustering algorithm can be used to quickly and accurately locate the fault line, thereby accurately and efficiently locating the fault point.

[0175] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent transformations made using the contents of the present invention's description and drawings, or directly or indirectly applied in related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A distribution network fault location method, characterized in that: Including steps: Constructing a distribution network system topology structure, and collecting current signals between the terminals of each sub-node according to the distribution network system topology structure; compressing and collecting the current signal to obtain a compressed signal; Reconstructing the compressed signal to obtain a reconstructed compressed signal, and decomposing the reconstructed compressed signal using an adaptive noise complete set empirical mode decomposition algorithm to obtain a series of intrinsic mode functions; Based on the series of intrinsic mode functions, a fuzzy C-means clustering algorithm is used to perform cluster analysis to obtain the fault point location information; The compressing and collecting the current signal to obtain a compressed signal includes: Determine the discrete cosine transform basis as a sparse transform basis matrix and initialize the sparse transform dictionary; Solve the sparse representation using an orthogonal matching pursuit algorithm according to the current signal to obtain a sparse representation coefficient; Iteratively updating the sparse transform dictionary to obtain an updated sparse transform dictionary; Performing sparse transformation on the current signal according to the updated sparse transformation dictionary to obtain a sparse signal; Determine a matrix in which all column vectors in the sparse transformation basis matrix are uncorrelated and conform to Gaussian distribution as a measurement matrix; Obtaining a compressed signal according to the measurement matrix and the sparse signal; Decomposing the reconstructed compressed signal using the adaptive noise complete set empirical mode decomposition algorithm to obtain the intrinsic mode function includes: Adding Gaussian white noise to the reconstructed compressed signal and performing adaptive noise complete set empirical mode decomposition to obtain a series of initial intrinsic mode functions; Performing arithmetic averaging on the series of initial eigenmode functions to obtain a first eigenmode function; removing the first intrinsic mode function from the current signal to obtain an updated signal, and returning to the step of adding Gaussian white noise to the reconstructed compressed signal and performing adaptive noise complete set empirical mode decomposition according to the updated signal until a unique residual that does not meet the decomposition condition and a series of intrinsic mode functions are obtained; The cluster analysis based on the series of intrinsic mode functions using the fuzzy C-means clustering algorithm to obtain the fault point location information includes: Using normalized energy entropy to eliminate false modes in the series of intrinsic mode functions to obtain reorganized intrinsic mode functions; Cluster analysis is performed using a fuzzy C-means clustering algorithm based on the reorganized intrinsic mode function to obtain fault point location information.

2. A distribution network fault location method according to claim 1, characterized in that: The sparse representation is solved by using the orthogonal matching pursuit algorithm according to the current signal to obtain the sparse representation coefficient include: ； ； Where, I j,t represents the current signal collected by terminal j at time t, D represents the sparse transformation dictionary, L represents the number of terminals, and N represents the collection time; The sparse transformation is performed on the current signal according to the updated sparse transformation dictionary to obtain a sparse signal I s,t for: ； The compressed signal R obtained according to the measurement matrix and the sparse signal is: ； Where, Denotes the measurement matrix.

3. A distribution network fault location method according to claim 1, characterized in that: Gaussian white noise is added to the reconstructed compressed signal and adaptive noise complete set empirical mode decomposition is performed to obtain a series of initial intrinsic mode functions: ； Where, represents the reconstructed compressed signal, represents the standard deviation of Gaussian white noise, N i (t) represents Gaussian white noise with unit variance and mean 0, represents a series of initial eigenmode functions, represents the signal residual, m represents the number of Gaussian white noise groups, and q represents the number of initial intrinsic mode functions; The arithmetic mean of the series of initial eigenmode functions is performed to obtain the first eigenmode function: ； Where IMF1 represents the first intrinsic mode function.

4. A distribution network fault location method according to claim 1, characterized in that: The process of using normalized energy entropy to eliminate false modes in the series of intrinsic mode functions to obtain reorganized intrinsic mode functions includes: calculating the energy of the series of eigenmode functions; Normalizing the energy to obtain normalized energy; Calculating normalized energy entropy values ​​corresponding to the series of intrinsic mode functions according to the normalized energy; The eigenmode functions of the series of eigenmode functions whose normalized energy entropy values ​​are less than a first preset threshold are determined as false modes, and the false modes are eliminated to obtain reorganized eigenmode functions.

5. A distribution network fault location method according to claim 1, characterized in that: The cluster analysis is performed using the fuzzy C-means clustering algorithm based on the reorganized intrinsic mode function to obtain the fault point location information, including: Determining the reorganized intrinsic mode function as a sample and initializing a membership matrix; Determining the cluster center of each sample subset in the sample; Determining an objective function of a fuzzy C-means clustering algorithm and constraints corresponding to the objective function based on the cluster centers of the sample subsets; Obtaining an iterative change amount of the objective function, and comparing the iterative change amount with a second preset threshold; If the iterative change amount is greater than or equal to the second preset threshold, updating the membership matrix to obtain an updated membership matrix, and returning to the step of determining the cluster center of each sample subset according to the updated membership matrix; If the iterative change amount is less than the second preset threshold, obtaining the category with the largest membership, and classifying the reorganized intrinsic mode function into the category with the largest membership; Determine an upstream detection point and a downstream detection point adjacent to the fault point according to the membership matrix; Calculating a difference in adjacent membership degrees according to the upstream detection point and the downstream detection point; It is determined whether the difference is greater than a third preset threshold; if so, the section between the detection points corresponding to the adjacent membership degrees is determined to be the fault point location.

6. A distribution network fault location method according to claim 5, characterized in that: The cluster center of each sample subset in the sample is determined as follows: ； Where c i represents the cluster center of the i-th sample, n represents the total number of samples, X j represents the reorganized intrinsic mode function, represents the sample x calculated using the membership index h j The degree of membership to class i; The objective function of the fuzzy C-means clustering algorithm determined based on the cluster centers of the sample subsets and the constraints corresponding to the objective function are: ； ； In the formula, q represents the number of cluster centers, x j Represents the reorganized intrinsic mode function X j The jth sample in u ij Represents sample x j The degree of membership to class i.

7. A distribution network fault location terminal, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, each step of the distribution network fault location method according to any one of claims 1 to 6 is implemented.

Images