A Fast Fault Location Method for High Resistance in Distribution Networks Based on Directional Frequency Scanning

Abstract

This invention relates to the field of power system fault detection technology, and discloses a method for rapid location of high-resistivity faults in distribution networks based on directional frequency scanning, including the following steps: S1, intercepting the zero-sequence current signal I collected from various monitoring points in the distribution network. 0j and bus zero-sequence voltage signal U0; S2, determine the fault occurrence time t f S3. Extract the first intrinsic mode function (IMF) and use it directly to reconstruct the original fault signal; S4. Project the IMF signal onto the reference orthogonal vector group respectively; S5. Use a clustering algorithm to distinguish the upstream and downstream of the fault point, and use the clustering results to locate the fault area. The method of constructing a virtual reference direction using orthogonal vector groups eliminates the need to install voltage transformers at all measurement points. Furthermore, its distributed parallel computing mode enables rapid local processing of fault data without uploading complete data to the main station, thereby reducing hardware investment and maintenance costs, decreasing dependence on physical hardware, and significantly improving fault detection speed.

Description

Technical Field

[0001] This invention relates to the field of power system fault detection technology, specifically a method for rapid location of high-resistivity faults in distribution networks based on directional frequency scanning. Background Technology

[0002] High-resistance faults are among the most common and difficult-to-detect fault types in distribution networks. These faults typically manifest as very weak current signals, often only on the order of a few amperes, and are easily masked by various noise signals in the distribution network. Therefore, accurately extracting the characteristics of high-resistance faults and quickly locating the fault location has become a major challenge in the operation and maintenance of distribution networks.

[0003] Traditional methods typically rely on measurement data from current and voltage transformers to identify and locate fault points through centralized calculations. However, these methods are susceptible to noise, leading to unstable fault feature extraction. Furthermore, the large volume of measurement data required for centralized processing at a main station not only increases data transmission overhead but also introduces significant latency, making it difficult to meet the requirements for rapid fault response. Additionally, installing expensive voltage transformers at all measurement points results in substantial hardware investment and high maintenance costs. Summary of the Invention

[0004] To overcome the above shortcomings, this invention provides an image recognition method that aims to improve upon traditional methods that typically rely on measurement data from current and voltage transformers. Installing expensive voltage transformers at all measurement points leads to significant hardware investment and high maintenance costs.

[0005] In a first aspect, the present invention provides the following technical solution: a method for rapid location of high-resistivity faults in distribution networks based on directional frequency scanning, comprising the following steps:

[0006] S1. Signal Acquisition: Extracting zero-sequence current signals I from various monitoring points in the distribution network. 0j The bus zero-sequence voltage signal U0, where j is the corresponding sequence number of the monitoring point, and the zero-sequence current signal interception time is [t]. f -T,t f +5T],t f The fault occurs at a time when T is the power grid frequency period.

[0007] S2. Determination of the fault occurrence time: By analyzing the bus zero-sequence voltage signal U0, the fault occurrence time t is determined. f ;

[0008] S3. Discrete Fourier Decomposition and Fault Characteristic Signal Reconstruction: Using Discrete Fourier Decomposition to reconstruct the zero-sequence current signal I acquired at the monitoring point. 0j A directional frequency scanning strategy is constructed to extract the first intrinsic mode function and directly use it to reconstruct the original fault signal;

[0009] S4. Feature vector construction and projection: Construct a reference orthogonal vector group in the high-dimensional space, project the first intrinsic mode function signal onto the reference orthogonal vector group respectively, and generate three-dimensional feature vectors;

[0010] S5. Feature fusion and clustering analysis: Based on the feature vectors of multiple time scales, use the clustering algorithm to distinguish the upstream and downstream of the fault point, and use the clustering results to locate the fault area.

[0011] Preferably, in the step S2, the determination of the fault occurrence time includes the following steps:

[0012] S201. Take the time scale t corresponding to the sampling serial number that first satisfies U0 = 0.15U n as the reference; set as the reference;

[0013] S202. In the bus zero-sequence voltage sampling sequence of the first 100T before t set , search backward for the sampling point that satisfies the following conditions as t f :

[0014] |U(z) - U(z - 4)| < 9 && |U(z - 15) - U(z - 30)| < 4

[0015] where U(z) is the discrete sampling value of the bus zero-sequence voltage, z is the index number of the corresponding sampling time; U(z - 4) is the discrete sampling value of the bus zero-sequence voltage corresponding to the index number z - 4; U(z - 15) is the discrete sampling value of the bus zero-sequence voltage corresponding to the index number z - 15; U(z - 30) is the discrete sampling value of the bus zero-sequence voltage corresponding to the index number z - 30.

[0016] Preferably, the calculation of the discrete Fourier transform includes:

[0017] Perform Fourier expansion on the discrete time series I 0j [n] with a length of N:

[0018]

[0019] where I 0j [n] refers to the discrete time series of the zero-sequence current signal at the jth monitoring point; DFT(I 0j [0]) refers to the result of the discrete Fourier transform, indicating the frequency components of the signal in the frequency domain; N refers to the actual sampling points of the intercepted signal I 0j ; k refers to the frequency index, indicating the number of the current frequency component, and the value range is 1 ≤ k < N / 2 - 1; n refers to the discrete point index in the time series, and the value range is 1 ≤ n ≤ N; The phase rotation factor in the Fourier transform represents the phase information of the frequency components.

[0020] The real part of the above result is taken as the signal component and used for subsequent feature extraction.

[0021] Preferably, the extraction of the first intrinsic mode function includes:

[0022] First, ensure the zero-sequence current signal I... 0j The length N is even; when the length is odd, the last sample value is discarded, making I... 0j The length N is always an even number;

[0023] For the frequency component f after Fourier transform i [n] The characteristic function is calculated using the following formula:

[0024]

[0025] Among them, f i [n] refers to the amplitude of the i-th frequency component; φ i [n] refers to the phase angle of the i-th frequency component; DFT(I 0k [n]) refers to the result of the Discrete Fourier Transform, representing the frequency components of the signal in the frequency domain; Refers to the phase rotation factor;

[0026] make sure and N / 2≥N i Add 1, then take the real part of the above equation to obtain N / 2-1 characteristic functions, denoted as IMF. m , m=1,2,...,N / 2-1; the first intrinsic mode function IMF1 is taken and directly used to reconstruct the original signal for high-impedance fault feature extraction, and denoted as

[0027] Preferably, in step S4, the feature vector construction includes the following steps:

[0028] S401. Design a set of mutually orthogonal vectors. and For signal projection, the expression is:

[0029]

[0030] Where ω0 is the angular frequency of the power grid. t k ∈[t f ,t f +0.5T×P], where P and k are both positive integers, t k+1 -t k =1 / f c fc This is the actual sampling frequency of the current sensor;

[0031] Typical values ​​are: A1 = A2 = 1. f c =10kHz, t k+1 -t k =10 -4 ;

[0032] S402. Project the first intrinsic mode function signal onto the above vector group, taking P = 1, f c Take the actual sampling frequency of the current sensor, and extract all monitoring points corresponding to [t]. f ,t f The first intrinsic mode function is calculated over the time interval [+0.5T], and the first intrinsic mode function obtained at the j-th monitoring point is denoted as... Will Projecting the orthogonal vector group proposed in step S401 above, we obtain the projection result:

[0033]

[0034] Preferably, in step S4, the result of feature vector generation is represented in polar coordinates:

[0035] S403, The final set of three-dimensional vectors is:

[0036]

[0037] In polar coordinates, it can be represented as a set of J vector pairs:

[0038]

[0039] S404. Take all j = 1, 2, 3, ..., J. If the following inequality holds:

[0040]

[0041] Then take the vector group Use the feature vector set to determine high-resistance faults; otherwise, take the vector set. As a feature vector set for judging high-resistance faults;

[0042] S405. The feature vector set for judging high-resistance faults obtained in the previous step is denoted as...

[0043] S406. Repeat steps "S402~S405" with P=2 and P=4 respectively to obtain two more sets of feature vectors for judging high-resistance faults. and

[0044] Preferably, in the step S5, the feature fusion and clustering analysis include:

[0045] S501. Cluster the generated direction vector groups and using the fuzzy clustering algorithm. The number of clustering categories is set to 2, representing the upstream and downstream of the fault point respectively;

[0046] S502. Calculate the silhouette coefficients corresponding to the above three sets of clustering results, denoted as SC 0.5T 、SC 1T and SC 2T , and the silhouette coefficient formula is:

[0047] where a(i) represents the cohesion of the sample point, and the calculation method is as follows:

[0048]

[0049] where j represents other sample points in the same class as sample i, and distance represents the distance between i and j. Therefore, the smaller a(i) is, the closer the class is; the calculation method of b(i) is similar to that of a(i), but it needs to traverse other clusters to obtain multiple values {b1(i), b2(i), b3(i), …, b m (i)} and select the smallest value as the final result;

[0050] So the original S(i):

[0051]

[0052] It can be found from the above formula that:

[0053] When a(i) < b(i), that is, the distance within the class is less than the distance between classes, the clustering result is more compact, and the value of S will approach 1. The closer it is to 1, the more obvious the silhouette is;

[0054] On the contrary, when a(i) > b(i), the distance within the class is greater than the distance between classes, indicating that the clustering result is very loose, and the value of S will approach -1. The closer it is to -1, the worse the clustering effect is.

[0055] S503. Compare the magnitudes of the three, and take the clustering result corresponding to the maximum value as the basis for judging the fault location, and determine the section location where the fault is located in combination with the actual network topology.

[0056] In the second aspect, the present invention provides the following technical solution. A high-resistance fault rapid location system for a distribution network based on directional frequency scanning includes:

[0057] The signal acquisition module is used to acquire the zero-sequence current signal I at various monitoring points in the power distribution network. 0j The zero-sequence voltage signal U0 of the busbar provides basic data for fault location;

[0058] The fault timing determination module analyzes the bus zero-sequence voltage signal U0 and determines the precise fault occurrence time t by setting a threshold and using a reverse search algorithm. f ;

[0059] The feature extraction module is used to extract the zero-sequence current signal I. 0j Perform a discrete Fourier transform to extract the first intrinsic mode function, which is used to reconstruct the fault signal features;

[0060] The feature vector construction module is used to generate reference vectors in high-dimensional space, project the first intrinsic mode function onto the orthogonal vector group, construct three-dimensional feature vectors, and intuitively represent fault characteristics;

[0061] The feature fusion and analysis module is used to analyze direction vectors based on fuzzy clustering algorithms, calculate contour coefficients, and locate fault sections by combining network topology.

[0062] Thirdly, the invention provides the following technical solution: a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-mentioned method for rapid location of high-resistance faults in distribution networks based on directional frequency scanning.

[0063] Fourthly, the present invention provides the following technical solution: a readable storage medium storing a computer program, wherein when the computer program is executed by a processor, it implements the above-mentioned method for rapid location of high-resistance faults in distribution networks based on directional frequency scanning.

[0064] The present invention has the following beneficial effects:

[0065] 1. In this invention, the method of constructing a virtual reference direction through orthogonal vector groups eliminates the need to install voltage transformers at all measurement points. Furthermore, its distributed parallel computing mode enables rapid local processing of fault data without uploading complete data to the main station, thereby reducing hardware investment and maintenance costs, decreasing dependence on physical hardware, and significantly improving fault detection speed.

[0066] 2. In this invention, by transmitting processed three-dimensional feature vectors instead of complete sampled data, the amount of data transmission and communication latency are significantly reduced compared to traditional methods, thus improving the system's response speed. This is crucial for achieving real-time and efficient fault location, and is especially suitable for situations requiring rapid feedback in large-scale power distribution networks.

[0067] 3. In this invention, the first intrinsic mode function (IMF1) is extracted through adaptive Fourier decomposition (FDM), which can effectively reconstruct the original fault signal in a noisy background. This ensures that the features of high-impedance faults are accurately extracted, improving the fault identification rate and maintaining high accuracy even in complex noisy environments.

[0068] 4. In this invention, edge computing nodes are deployed at each sampling point. Each node independently completes data feature extraction and calculation tasks, utilizing distributed parallel computing to accelerate processing and effectively reduce the computational burden on the main station. This design improves the real-time performance, reliability, and efficiency of the overall system, making it particularly suitable for application scenarios with strict requirements for fault response time. Attached Figure Description

[0069] Figure 1 This is a flowchart of the method for rapid location of high-resistivity faults in distribution networks based on directional frequency scanning according to the present invention.

[0070] Figure 2 This is a schematic diagram of seven zero-sequence current sampling points distributed in a 10kV medium-voltage power distribution system according to the present invention;

[0071] Figure 3 This is a schematic diagram of the ideal zero-sequence current sampling signal at each measuring point of the present invention, without considering the influence of noise interference.

[0072] Figure 4 This is a schematic diagram of the zero-sequence current sampling signal at each measuring point of the present invention under strong noise interference of -2.8dB;

[0073] Figure 5 This is a schematic diagram of the polar coordinate form of the feature vector group under different fault angles according to the present invention;

[0074] Figure 6 This is a schematic diagram of the polar coordinate form of the feature vector group under -2.8dB strong noise interference according to the present invention. Detailed Implementation

[0075] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0076] Example 1

[0077] Reference Figures 1-6 In the first embodiment of the present invention, the present invention provides a method for rapid location of high-resistivity faults in distribution networks based on directional frequency scanning, comprising the following steps:

[0078] S1. Signal Acquisition: Extracting zero-sequence current signals I from various monitoring points in the distribution network. 0j The bus zero-sequence voltage signal U0, where j is the corresponding sequence number of the monitoring point, and the zero-sequence current signal interception time is [t]. f -T,t f +5T],t f The fault occurs at a time when T is the power grid frequency period.

[0079] S2. Determination of the fault occurrence time: By analyzing the bus zero-sequence voltage signal U0, the fault occurrence time t is determined. f ;

[0080] S3. Discrete Fourier Decomposition and Fault Characteristic Signal Reconstruction: Using Discrete Fourier Decomposition to reconstruct the zero-sequence current signal I acquired at the monitoring point. 0j A directional frequency scanning strategy is constructed to extract the first intrinsic mode function and directly use it to reconstruct the original fault signal;

[0081] S4. Feature Vector Construction and Projection: Construct a set of reference orthogonal vectors in a high-dimensional space, and project the first intrinsic mode function signal onto the set of reference orthogonal vectors to generate three-dimensional feature vectors.

[0082] S5. Feature Fusion and Cluster Analysis: Based on feature vectors at multiple time scales, a clustering algorithm is used to distinguish the upstream and downstream of the fault point, and the clustering results are used to locate the fault area.

[0083] Specifically, the overall method flow is described, including core steps such as signal acquisition, fault time determination, discrete Fourier transform and feature extraction, feature vector construction and projection, and feature fusion and cluster analysis. Through directional frequency scanning and extraction of the first intrinsic mode function (IMF1), this method has a high fault feature extraction capability in noisy environments and can reliably capture weak high-resistance fault signal features. The edge computing approach distributes data processing tasks to various sampling points, greatly reducing data transmission volume and the computational burden on the central station, and significantly improving response speed. It eliminates the need to install expensive voltage transformers at all monitoring points, relying on the calculation of virtual reference directions to achieve low-cost and high-precision fault location. The use of multi-timescale analysis and fuzzy clustering technology can effectively improve the robustness of fault detection and location, avoiding misjudgments caused by the randomness of data at a single time point.

[0084] In step S2, determining the time of the fault occurrence includes the following steps:

[0085] S201, so that U0 = 0.15U is satisfied for the first time. n The time stamp t corresponding to the sampling sequence number set Based on;

[0086] S202, at tset Among the sampling sequences of the zero-sequence voltage of the busbar for the first 100T, reverse search for the sampling points that meet the following conditions as t f :

[0087] |U(z) - U(z - 4)| < 9 && |U(z - 15) - U(z - 30)| < 4

[0088] where U(z) is the discrete sampling value of the zero-sequence voltage of the busbar, z is the index number of the corresponding sampling moment; U(z - 4) is the discrete sampling value of the zero-sequence voltage of the busbar corresponding to the index number z - 4; U(z - 15) is the discrete sampling value of the zero-sequence voltage of the busbar corresponding to the index number z - 15; U(z - 30) is the discrete sampling value of the zero-sequence voltage of the busbar corresponding to the index number z - 30.

[0089] Specifically, it uses a specific zero-sequence voltage threshold condition U0 = 0.15U n as a reference, and by reverse searching for the sampling points that meet specific conditions, the precise moment t when the fault occurs is finally determined f， . Its functions and effects include:

[0090] Simplify calculation: By simple voltage threshold comparison, quickly lock the range of potential fault occurrence moments and reduce the calculation complexity;

[0091] Improve accuracy: Reverse search combined with multi-condition constraints (such as |U(z) - U(z - 4)| < 9) effectively avoids moment misjudgment caused by random noise or mismeasurement;

[0092] Enhance adaptability: This method is not only applicable to high-resistance faults but can also adapt to the judgment of other types of fault moments, with strong generality.

[0093] The calculation of the discrete Fourier transform includes:

[0094] For the discrete-time sequence I 0j [n] with length N, perform Fourier expansion:

[0095]

[0096] where I 0j [n] refers to the discrete-time sequence of the zero-sequence current signal at the jth monitoring point; DFT(I 0j [0]) refers to the result of the discrete Fourier transform, representing the frequency components of the signal in the frequency domain; N refers to the actual number of sampling points of the intercepted signal I 0j ; k refers to the frequency index, representing the number of the current frequency component, with a value range of 1 ≤ k < N / 2 - 1; n refers to the discrete point index in the time sequence, with a value range of 1 ≤ n ≤ N; The phase rotation factor in the Fourier transform represents the phase information of the frequency components.

[0097] The real part of the above result is taken as the signal component and used for subsequent feature extraction.

[0098] Specifically, a method for implementing the Discrete Fourier Transform is defined, which extracts high-frequency and low-frequency features by expanding and decomposing the signal into frequency components. Its function and effects are reflected in:

[0099] Multi-band analysis: The zero-sequence current signal is decomposed into frequency components through Fourier transform, laying the foundation for subsequent feature extraction and analysis;

[0100] Preserving signal features: By using real part feature extraction, the original characteristics of the signal can be preserved while reducing computational complexity;

[0101] Wide applicability: This step is not only applicable to high-resistance faults in distribution networks, but can also be applied to other fault detection scenarios involving frequency domain analysis.

[0102] The extraction of the first intrinsic mode function includes:

[0103] First, ensure the zero-sequence current signal I... 0j The length N is even; when the length is odd, the last sample value is discarded, making I... 0j The length N is always an even number;

[0104] For the frequency component f after Fourier transform i [n] The characteristic function is calculated using the following formula:

[0105]

[0106] Among them, f i [n] refers to the amplitude of the i-th frequency component; φ i [n] refers to the phase angle of the i-th frequency component; DFT(I 0k [n]) refers to the result of the Discrete Fourier Transform, representing the frequency components of the signal in the frequency domain; Refers to the phase rotation factor;

[0107] make sure and N / 2≥N i Add 1, then take the real part of the above equation to obtain N / 2-1 characteristic functions, denoted as IMF. m , m=1,2,...,N / 2-1; the first intrinsic mode function IMF1 is taken and directly used to reconstruct the original signal for high-impedance fault feature extraction, and denoted as

[0108] Specifically, the extraction process of the first intrinsic mode function (IMF1) is described, the core of which lies in obtaining the main characteristic components of the signal through frequency scanning. Its functions and effects include:

[0109] Signal feature enhancement: As the first mode of the signal, IMF1 contains the most critical high-frequency components of the signal, which can effectively enhance fault characteristics;

[0110] Noise reduction capability: By selecting frequency monotonicity conditions, frequency components unrelated to the fault are eliminated, effectively suppressing noise interference;

[0111] Optimized feature extraction: Adaptive Fourier decomposition technology is adopted to accurately extract the core features of high-resistivity faults in complex environments.

[0112] In step S4, feature vector construction includes the following steps:

[0113] S401. Design a set of mutually orthogonal vectors. and For signal projection, the expression is:

[0114]

[0115] Where ω0 is the angular frequency of the power grid. t k ∈[t f ,t f +0.5T×P], where P and k are both positive integers, t k+1 -t k =1 / f c f c This is the actual sampling frequency of the current sensor;

[0116] Typical values ​​are: A1 = A2 = 1. f c =10kHz, t k+1 -t k =10 -4 ;

[0117] S402. Project the first intrinsic mode function signal onto the above vector group, taking P = 1, f c Take the actual sampling frequency of the current sensor, and extract all monitoring points corresponding to [t]. f ,t f The first intrinsic mode function is calculated over the time interval [+0.5T], and the first intrinsic mode function obtained at the j-th monitoring point is denoted as... Will Projecting the orthogonal vector group proposed in step S401 above, we obtain the projection result:

[0118]

[0119] Specifically, a method for constructing eigenvectors is proposed, which involves designing orthogonal vector sets. and Projecting the real part of the first intrinsic mode function (IMF1) onto the signal generates a physically meaningful eigenvector. Its functions and effects include:

[0120] Intuitive representation: By projecting orthogonal vector groups, complex signals are decomposed into easily analyzable geometric features, improving the intuitiveness of fault analysis;

[0121] Enhanced spatial characteristics: The feature vectors reveal the distribution characteristics of the fault signal in a high-dimensional space, providing high-dimensional information support for subsequent fault localization;

[0122] Reduced complexity: Orthogonal projection reduces the amount of original signal data, optimizing computational efficiency while maintaining signal integrity.

[0123] In step S4, the result of feature vector generation is represented in polar coordinates:

[0124] S403, The final set of three-dimensional vectors is:

[0125]

[0126] In polar coordinates, it can be represented as a set of J vector pairs:

[0127]

[0128] S404. Take all j = 1, 2, 3, ..., J. If the following inequality holds:

[0129]

[0130] Then take the vector group Use the feature vector set to determine high-resistance faults; otherwise, take the vector set. As a feature vector set for judging high-resistance faults;

[0131] S405. The feature vector set for judging high-resistance faults obtained in the previous step is denoted as...

[0132] S406. Repeat steps "S402~S405" with P=2 and P=4 respectively to obtain two more sets of feature vectors for judging high-resistance faults. and

[0133] Specifically, the calculation methods of the direction vector and the three-dimensional feature vector are further refined, and the polar coordinate representation of the feature vector is proposed. Its functions and effects include:

[0134] High-precision modeling: By calculating the direction vector, the directional characteristics of the fault signal are accurately captured;

[0135] Geometric representation optimization: The polar coordinate representation of the three-dimensional feature vector enhances the spatial distribution information of the signal features and lays a foundation for cluster analysis;

[0136] Improve robustness: The polar coordinate representation avoids the deviation caused by the selection of the coordinate system and improves the expression ability of the fault signal features.

[0137] In step S5, feature fusion and cluster analysis include:

[0138] S501. Use the fuzzy clustering algorithm to cluster the generated direction vector groups and , and set the number of clustering categories to 2, which represent the upstream and downstream of the fault point respectively;

[0139] S502. Calculate the silhouette coefficients corresponding to the above three sets of clustering results, denoted as SC 0.5T , SC 1T and SC 2T , and the silhouette coefficient formula is:

[0140] where a(i) represents the cohesion of the sample point, and the calculation method is as follows:

[0141]

[0142] where j represents other sample points in the same class as sample i, and distance represents the distance between i and j. Therefore, the smaller a(i) is, the closer the class is; the calculation method of b(i) is similar to a(i), but it needs to traverse other clusters to obtain multiple values {b1(i), b2(i), b3(i), …, b m (i)} and select the smallest value as the final result;

[0143] So the original S(i):

[0144]

[0145] It can be found from the above formula that:

[0146] When a(i) < b(i), that is, the distance within the class is less than the distance between classes, the clustering result is more compact, and the value of S will approach 1. The closer it is to 1, the more obvious the silhouette is;

[0147] Conversely, when a(i) > b(i), the intra-class distance is greater than the inter-class distance, indicating that the clustering result is loose. The value of S will approach -1, and the closer it is to -1, the worse the clustering effect is.

[0148] S503. Compare the three values ​​and take the clustering result corresponding to the maximum value as the basis for judging the fault location, and combine it with the actual network topology to determine the segment location where the fault is located.

[0149] Specifically, a feature fusion and analysis method based on fuzzy clustering (FCM) algorithm and silhouette coefficient calculation is defined in detail, and its functions and effects include:

[0150] Fault section identification: By clustering the direction vectors, the upstream and downstream distribution of the fault point can be quickly identified;

[0151] Classification performance optimization: The introduction of silhouette coefficient can quantify the clustering effect and ensure the accuracy of classification results;

[0152] Combining network topology: By combining clustering results with network topology, the faulty segment can be located more precisely;

[0153] Computational flexibility: Fuzzy clustering algorithms are highly adaptable and can handle feature vector distributions in different scenarios.

[0154] Figure 5 The polar coordinate form of the feature vector group under different fault angles is shown in the figure. (a) a = 90°. (b) a = 60°. (c) a = 0°. (d) a = -30°. The vector group in the light green shaded part is the feature vector group that is finally selected as the feature vector group for judging high resistance faults. The other group must be discarded.

[0155] Example 2:

[0156] Reference Figure 1 In a second embodiment of the present invention, the present invention provides a rapid fault location system for high resistance in distribution networks based on directional frequency scanning, comprising:

[0157] The signal acquisition module is used to acquire the zero-sequence current signal I at various monitoring points in the power distribution network. 0j The zero-sequence voltage signal U0 of the busbar provides basic data for fault location;

[0158] The fault timing determination module analyzes the bus zero-sequence voltage signal U0 and determines the precise fault occurrence time t by setting a threshold and using a reverse search algorithm. f ;

[0159] The feature extraction module is used to extract the zero-sequence current signal I. 0j Perform a discrete Fourier transform to extract the first intrinsic mode function, which is used to reconstruct the fault signal features;

[0160] The feature vector construction module is used to generate reference vectors in high-dimensional space, project the first intrinsic mode function onto the orthogonal vector group, construct three-dimensional feature vectors, and intuitively represent fault characteristics;

[0161] The feature fusion and analysis module is used to analyze direction vectors based on fuzzy clustering algorithms, calculate contour coefficients, and locate fault sections by combining network topology.

[0162] Example 3

[0163] In the third embodiment of the present invention, based on the same inventive concept, the present invention proposes a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the steps of the method for rapid location of high-resistivity faults in distribution networks based on directional frequency scanning in the above embodiments.

[0164] Example 4

[0165] In the fourth embodiment of the present invention, based on the same inventive concept, a computer device is proposed, the terminal comprising: a processor and a memory; the processor and the memory communicate with each other; the memory is used to store instructions; the processor is used to execute the instructions in the memory to execute the fast location method for high-resistivity faults in power distribution networks based on directional frequency scanning as described in the above embodiment.

[0166] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0167] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for rapid location of high-resistivity faults in distribution networks based on directional frequency scanning, characterized in that, It includes the following steps: S1. Signal Acquisition: Extracting zero-sequence current signals I from various monitoring points in the distribution network. 0j The bus zero-sequence voltage signal U0, where j is the corresponding sequence number of the monitoring point, and the zero-sequence current signal interception time is [t]. f -T,t f +5T],t f The fault occurs at a time when T is the power grid frequency period. S2. Determination of the fault occurrence time: By analyzing the bus zero-sequence voltage signal U0, the fault occurrence time t is determined. f ; S3. Discrete Fourier Transform and Fault Characteristic Signal Reconstruction: Using the Discrete Fourier Transform to reconstruct the zero-sequence current signal I acquired at the monitoring point. 0j A directional frequency scanning strategy is constructed to extract the first intrinsic mode function and directly use it to reconstruct the original fault signal; S4. Feature vector construction and projection: Construct a reference orthogonal vector group in the high-dimensional space, project the first intrinsic mode function signal onto the reference orthogonal vector group respectively, and generate three-dimensional feature vectors; S5. Feature fusion and clustering analysis: Based on the feature vectors of multiple time scales, use a clustering algorithm to distinguish the upstream and downstream of the fault point, and use the clustering results to locate the fault area; In the step S4, the feature vector construction includes the following steps: S401. Design a set of mutually orthogonal vectors. and For signal projection, the expression is: Where ω0 is the angular frequency of the power grid. t k ∈[t f ,t f +0.5T·P], where P and k are both positive integers, t k+1 -t k =1 / f c f c This is the actual sampling frequency of the current sensor; Typical values ​​are: A1 = A2 = 1. f c =10kHz, t k+1 -t k =10 -4 ; S402. Project the first intrinsic mode function signal onto the above vector group, taking P = 1, f c Take the actual sampling frequency of the current sensor, and extract all monitoring points corresponding to [t]. f ,t f The first intrinsic mode function is calculated over the time interval [+0.5T], and the first intrinsic mode function obtained at the j-th monitoring point is denoted as... Will Projecting the orthogonal vector group proposed in step S401 above, we obtain the projection result: In the step S4, the result of the feature vector generation is represented in polar coordinate form: S403. Finally, a set of three-dimensional vectors obtained is: It can be correspondingly represented as J pairs of vector groups in polar coordinates: S404. Take all j = 1, 2, 3,..., J. If the following inequality can hold: Then take the vector group Use the feature vector set to determine high-resistance faults; otherwise, take the vector set. As a feature vector set for judging high-resistance faults; S405. The feature vector set for judging high-resistance faults obtained in the previous step is denoted as... S406. Repeat steps "S402~S405" for P=2 and P=4 respectively to obtain two more sets of feature vectors for judging high-resistance faults. and 2. The method for rapid location of high-resistivity faults in distribution networks based on directional frequency scanning according to claim 1, characterized in that, In the step S2, the determination of the fault occurrence time includes the following steps: S201, so that U0 = 0.15U is satisfied for the first time. n The time stamp t corresponding to the sampling sequence number set Based on; S202, at t set In the first 100T of the bus zero-sequence voltage sampling sequence, the sampling points that satisfy the following conditions are used as t through reverse searching. f : |U(z) - U(z - 4)| < 9 && |U(z - 15) - U(z - 30)| < 4 Where, U(z) is the discrete sampling value of the zero-sequence voltage of the bus, z is the index number of the corresponding sampling time; U(z - 4) is the discrete sampling value of the zero-sequence voltage of the bus corresponding to the index number z - 4; U(z - 15) is the discrete sampling value of the zero-sequence voltage of the bus corresponding to the index number z - 15; U(z - 30) is the discrete sampling value of the zero-sequence voltage of the bus corresponding to the index number z - 30.

3. The method for rapid location of high-resistivity faults in distribution networks based on directional frequency scanning according to claim 1, characterized in that, The calculation of the discrete Fourier transform includes: For a discrete time series I of length N 0j Perform a Fourier expansion on [n]: Among them, I 0j [n] refers to the discrete-time sequence of zero-sequence current signals at the j-th monitoring point; DFT(I 0j [0]) refers to the result of the discrete Fourier transform, representing the frequency components of the signal in the frequency domain; N refers to the actual number of sampling points of the intercepted signal I 0j ; k refers to the frequency index, representing the number of the current frequency component, with a value range of 1 ≤ k < N / 2 - 1; n refers to the discrete point index in the time series, with a value range of 1 ≤ n ≤ N; refers to the phase rotation factor in the Fourier transform, representing the phase information of the frequency component; Take the real part of the above result as the signal component for subsequent feature extraction.

4. The method for rapid location of high-resistivity faults in distribution networks based on directional frequency scanning according to claim 1, characterized in that, The extraction of the first intrinsic mode function includes: First, ensure the zero-sequence current signal I... 0j The length N is even; when the length is odd, the last sample value is discarded, making I... 0j The length N is always an even number; For the frequency component f after Fourier transform i [n] The characteristic function is calculated using the following formula: Among them, f i [n] refers to the amplitude of the i-th frequency component; φ i [n] refers to the phase angle of the i-th frequency component; DFT(I 0k [n]) refers to the result of the Discrete Fourier Transform, representing the frequency components of the signal in the frequency domain; Refers to the phase rotation factor; make sure and N / 2≥N i Add 1, then take the real part of the above equation to obtain N / 2-1 characteristic functions, denoted as IMF. m , m=1,2,...,N / 2-1; the first intrinsic mode function IMF1 is taken and directly used to reconstruct the original signal for high-impedance fault feature extraction, and denoted as 5. The method for rapid location of high-resistivity faults in distribution networks based on directional frequency scanning according to claim 1, characterized in that, In the step S5, the feature fusion and clustering analysis include: S501, For the generated direction vector group and Fuzzy clustering algorithm is used for clustering, with the number of clusters set to 2, representing the upstream and downstream of the fault point respectively; S502. Calculate the silhouette coefficients corresponding to the above three groups of clustering results, denoted as SC. 0.5T SC 1T and SC 2T The formula for the contour coefficient is: Where, a(i) represents the cohesion of the sample point, and the calculation method is as follows: Where j represents other sample points within the same class as sample i, and distance represents the distance between i and j. Therefore, the smaller a(i) is, the closer the classes are. The calculation method of b(i) is similar to a(i), but it needs to traverse other clusters to obtain multiple values ​​{b1(i), b2(i), b3(i), ..., bj}. m (i)} Select the smallest value from them as the final result; So the original S(i): It can be found from the above formula that: When a(i) < b(i), that is, the distance within the class is less than the distance between classes, the clustering result is more compact, and the value of S will approach 1. The closer it is to 1, the more obvious the contour is; On the contrary, when a(i) > b(i), the distance within the class is greater than the distance between classes, indicating that the clustering result is very loose, and the value of S will approach -1. The closer it is to -1, the worse the clustering effect is; S503. Compare the magnitudes of the three, and take the clustering result corresponding to the maximum value as the basis for judging the fault location, and determine the section location where the fault is located in combination with the actual network topology.

6. A rapid fault location system for high-resistivity distribution networks based on directional frequency scanning, characterized in that, For the fast location method of high-resistance faults in a distribution network based on directional frequency scanning according to any one of claims 1 - 5, it includes: The signal acquisition module is used to acquire the zero-sequence current signal I at various monitoring points in the power distribution network. 0j The zero-sequence voltage signal U0 of the busbar provides basic data for fault location; The fault timing determination module analyzes the bus zero-sequence voltage signal U0 and determines the precise fault occurrence time t by setting a threshold and using a reverse search algorithm. f ; The feature extraction module is used to extract the zero-sequence current signal I. 0j Perform a discrete Fourier transform to extract the first intrinsic mode function, which is used to reconstruct the fault signal features; A feature vector construction module, used to generate a reference vector in the high-dimensional space, project the first intrinsic mode function onto the orthogonal vector group, construct three-dimensional feature vectors, and visually represent the fault features; A feature fusion and analysis module, used to analyze the direction vectors based on the fuzzy clustering algorithm, calculate the silhouette coefficient, and locate the fault section in combination with the network topology structure.

7. A computer device 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, it implements the fast location method of high-resistance faults in a distribution network based on directional frequency scanning according to any one of claims 1 to 5.

8. A readable storage medium, characterized in that, The readable storage medium stores a computer program, which, when executed by a processor, implements the method for rapid location of high-resistivity faults in distribution networks based on directional frequency scanning as described in any one of claims 1 to 5.

Images