An intelligent fault diagnosis method and system for distribution network

Abstract

This application discloses an intelligent fault diagnosis method and system for a distribution network. The method includes obtaining a real-time potential signal from a node; inputting the real-time potential signal into a pre-trained fault type classifier to obtain the fault type of the node; and performing three-dimensional reconstruction on the real-time potential signal to obtain a three-dimensional visualized potential image; determining the location of the fault point based on the three-dimensional visualized potential image and the spatial topology of the distribution network; and displaying the location of the fault point and the fault type on the three-dimensional visualized potential image. The above method can accurately identify the fault type and locate the fault point, and can also display it intuitively, thereby improving the efficiency and accuracy of fault diagnosis.

Description

Technical Field

[0001] The present application relates to the technical field of distribution network fault location, and in particular to an intelligent fault diagnosis method and system for a distribution network. Background Art

[0002] The stable operation of power systems is crucial to modern society. As the core of energy transmission, grid fault diagnosis and location technologies are directly related to power supply reliability and economic benefits. In recent years, with the expansion and increasing complexity of power grids, the rapid and accurate identification and location of faults has become a research hotspot. Traditional fault diagnosis methods often suffer from slow response speeds and insufficient accuracy, especially in multi-point fault or atypical fault scenarios, resulting in high rates of misjudgment. Summary of the Invention

[0003] The present application provides an intelligent fault diagnosis method and system for a distribution network to accurately identify the fault type and locate the fault point.

[0004] In order to solve the above technical problems, the present application provides an intelligent fault diagnosis method for a distribution network, comprising:

[0005] Get the real-time potential signal of the node;

[0006] Inputting the real-time potential signal into a pre-trained fault type classifier to obtain the fault type of the node, wherein the fault type classifier is obtained by performing feature classification training on a potential difference matrix and a pre-established spatial topology feature library, and the fault type classifier includes a correspondence between the potential difference value of the node and the fault type, as well as a potential distribution of the fault type, and the potential difference matrix is ​​obtained from the potential signal of the node;

[0007] Performing three-dimensional reconstruction on the real-time potential signal to obtain a three-dimensional visualized potential image;

[0008] determining the location of a fault point based on the three-dimensional visualized potential image and the spatial topological structure of the distribution network;

[0009] The location of the fault point and the fault type are displayed on the three-dimensional visualized potential image.

[0010] In some embodiments, inputting the real-time potential signal into a pre-trained fault type classifier to obtain the fault type of the node includes:

[0011] Obtaining a propagation path characteristic of the real-time potential signal according to the real-time potential signal and the spatial topological structure;

[0012] The propagation path characteristics of the real-time potential signal are matched with the fault path characteristics in the fault type classifier to obtain the fault type.

[0013] In some embodiments, determining the location of the fault point based on the three-dimensional visualized potential image and the spatial topology of the power grid includes:

[0014] Obtaining a coverage matrix of the node according to the three-dimensional visualized potential image and the spatial topological structure;

[0015] dividing the three-dimensional visualized potential image into regions according to the coverage matrix;

[0016] If the potential difference value of the divided area matches the fault type, the fault area is marked using a boundary extraction algorithm to obtain a fault area image;

[0017] Digitally processing the fault area image to obtain a digital potential matrix;

[0018] The spatial correspondence between the pixel values ​​of the digitized potential matrix and the fault type is iteratively optimized by a gradient descent algorithm to determine the position of the fault point.

[0019] In some embodiments, the step of pre-establishing a spatial topology feature library is further included, and the step of pre-establishing a spatial topology feature library includes:

[0020] collecting a potential signal of the node, and obtaining a potential change sequence according to the potential signal;

[0021] Obtaining a target sequence according to the potential change sequence and the energy distribution characteristics of the fault signal;

[0022] Obtaining the time-frequency localization characteristics of the target sequence, and obtaining the potential characteristic vector of the node according to the geographical location coordinates and electrical connection type of the node;

[0023] According to the potential characteristic vector of the node, the propagation path characteristics of the time-frequency localization feature in the spatial topological structure are obtained to obtain the spatial time-frequency distribution feature;

[0024] A spatial topological feature library is obtained according to the spatial time-frequency distribution features.

[0025] In some embodiments, the step of obtaining a spatial topological feature library based on the spatial time-frequency distribution features includes:

[0026] Using a density clustering algorithm to calculate the clustering density and separation of the feature points in the spatial time-frequency distribution features to obtain the initial cluster center sequence;

[0027] Using the initial cluster center sequence, a kernel density estimation algorithm is used to calculate the spatial distribution characteristics of the cluster centers to obtain the spatial distribution feature sequence;

[0028] If the degree of separation of the spatial distribution feature sequence is greater than a preset threshold, a minimum spanning tree algorithm is used to construct a topological structure of the distribution network corresponding to the cluster center to obtain the topological connection matrix;

[0029] According to the topological connection matrix, a graph traversal algorithm is used to extract the fault path characteristics of the fault type to obtain the spatial topological feature library.

[0030] In some embodiments, obtaining the potential difference matrix of the nodes from the potential signal includes:

[0031] Using a wavelet transform algorithm to perform time-frequency decomposition on the potential signal to obtain a time-frequency distribution feature sequence;

[0032] According to the time-frequency distribution feature sequence, the potential difference between nodes is calculated using the Euclidean distance algorithm to obtain the potential difference matrix of the nodes.

[0033] In some embodiments, performing three-dimensional reconstruction on the real-time potential signal to obtain a three-dimensional visualized potential image includes:

[0034] Performing time-frequency decomposition on the real-time potential signal to obtain a time-frequency feature sequence;

[0035] The three-dimensional stereoscopic microscopy technique is used to extract spatial distribution data of the time-frequency feature sequence to obtain a three-dimensional visualized potential image.

[0036] In some embodiments, the present application further proposes an intelligent fault diagnosis system for a distribution network, comprising:

[0037] Data acquisition module, used to obtain the real-time potential signal of the node;

[0038] a fault type acquisition module, configured to input the real-time potential signal into a pre-trained fault type classifier to obtain the fault type of the node, wherein the fault type classifier is obtained by performing feature classification training on the potential difference matrix and a pre-established spatial topology feature library, and the fault type classifier includes a correspondence between the potential difference of the node and the fault type, as well as a potential distribution of the fault type, and the potential difference matrix is ​​obtained from the potential signal of the node;

[0039] A three-dimensional visualization module is used to perform three-dimensional reconstruction on the real-time potential signal to obtain a three-dimensional visualized potential image;

[0040] A fault point determination module is used to determine the location of the fault point based on the three-dimensional visualized potential image and the spatial topological structure of the distribution network;

[0041] The three-dimensional visualization module is further configured to display the position of the fault point and the fault type on the three-dimensional visualization potential image.

[0042] The intelligent fault diagnosis method and system for the distribution network of the above embodiment obtains the real-time potential signal of the node; inputs the real-time potential signal into a pre-trained fault type classifier to obtain the fault type of the node; and performs three-dimensional reconstruction on the real-time potential signal to obtain a three-dimensional visualized potential image; determines the location of the fault point based on the three-dimensional visualized potential image and the spatial topology of the distribution network; and displays the location of the fault point and the fault type on the three-dimensional visualized potential image. The above method and system can accurately identify the fault type and locate the fault point, and can also display it intuitively, thereby improving the efficiency and accuracy of fault diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 This is a flow chart of an intelligent fault diagnosis method for a distribution network provided by this application;

[0044] Figure 2 This is a schematic diagram of the structure of an intelligent fault diagnosis system for a distribution network provided in this application. DETAILED DESCRIPTION

[0045] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0046] To solve the above problems, refer to Figure 1 The intelligent fault diagnosis method for distribution network proposed in this application includes the following steps:

[0047] Step 101: Acquire the real-time potential signal of the node.

[0048] The power grid consists of distribution equipment and ancillary facilities, including overhead lines, towers, cables, distribution transformers, switchgear, and reactive power compensation capacitors. Its primary function within the power grid is to distribute electrical energy. From the perspective of the nature of the distribution network, distribution network equipment also includes distribution equipment at substations. A node in the power grid is a collection of physical points at the same potential. In this embodiment, high-precision sensors are used to collect real-time potential signals from the nodes, and synchronized timestamping technology is used to time-stamp the collected data. For example, in node potential signal collection scenarios, high-precision sensors typically employ capacitive or resistive principles, capable of capturing minute potential changes. For example, a high-precision sensor deployed at a node can detect potential fluctuations as small as 0.01 volts, demonstrating high sensitivity and strong interference resistance. Synchronized timestamping technology uses a high-precision clock module to accurately time-stamp each set of potential signal data, for example, in units of 1 microsecond, ensuring consistent timing for subsequent analysis. This approach effectively tracks transient changes in the power grid, such as voltage dips or surges, and provides a basis for fault location.

[0049] Step 102: Input the real-time potential signal into a pre-trained fault type classifier to obtain the fault type of the node, wherein the fault type classifier is obtained by feature classification training using a potential difference matrix and a pre-established spatial topology feature library, and the fault type classifier includes a correspondence between the potential difference of the node and the fault type, as well as a potential distribution template of the fault type, and the potential difference matrix is ​​obtained from the potential signal of the node.

[0050] In this embodiment, a one-dimensional convolutional neural network is used to perform feature classification training on the potential difference matrix of the node and the pre-established spatial topology feature library to obtain a fault type classifier. Among them, the one-dimensional convolutional neural network extracts the local pattern in the matrix through the sliding convolution kernel, and learns the mapping relationship between the potential difference and the fault type. The fault type includes but is not limited to single-point short circuit and multi-point grounding. The fault type classifier also contains potential distribution templates for various fault types. For example, a single-point short circuit fault corresponds to a specific frequency distribution. For example, the potential difference matrix of a power grid is input into a one-dimensional convolutional neural network. The convolutional neural network includes 3 layers of convolution layers and 2 layers of fully connected layers, and a fault type classifier is obtained by training. During the training process, the convolutional neural network extracts features of known fault modes from the spatial topology feature library. For example, the potential difference pattern of a single-point short circuit is concentrated on a certain node pair. The fault type classifier can efficiently identify complex fault modes.

[0051] Step 103 : Perform three-dimensional reconstruction on the real-time potential signal to obtain a three-dimensional visualized potential image.

[0052] In some embodiments, wavelet transform technology or a fast Fourier transform algorithm is used to extract a time-frequency feature sequence from the real-time potential signal. A database of the spatial topology of the distribution network is then queried for this time-frequency feature sequence to obtain a preliminary feature set of the potential signal. If the deviation between the preliminary feature set and the potential distribution of a fault type in a pre-established fault type classifier is less than a preset threshold, for example, less than 5%, the preliminary feature set is spatially reconstructed using three-dimensional stereoscopic microscopy to obtain a three-dimensional visual potential image and construct a three-dimensional model of the potential signal. This three-dimensional image more intuitively displays the spatial variations of the potential signal, providing a visual basis for fault area identification.

[0053] In this embodiment, a spatial topology database stores the geographic locations and connectivity of nodes. For example, a database might contain the coordinates and adjacency information for ten nodes. If the time-frequency feature sequence matches the node attributes in the database, a preliminary feature set containing the node's location and potential characteristics is generated. For example, if a node's time-frequency feature sequence shows an abnormality in a specific frequency component, the database can be queried to confirm that the node is located near a substation. This approach facilitates the preliminary location of the abnormal area.

[0054] Step 104 : determining the location of the fault point based on the three-dimensional visualized potential image and the spatial topological structure of the distribution network.

[0055] In some embodiments, the spatial topology structure of the distribution network includes a spatial topology feature library, which contains node coordinates and the connection relationships between nodes and their neighboring nodes. If the potential difference value of the divided area matches the fault type, a boundary extraction algorithm is used to mark the fault area and generate a fault area image. If the deviation between the potential difference value of the divided area and the potential distribution of a certain fault type in the spatial topology feature library is less than a preset threshold, the location of the fault point is determined using a gradient descent optimization algorithm based on analog-to-digital conversion accuracy and time-frequency localization characteristics.

[0056] Step 105: Display the location of the fault point and the fault type on the three-dimensional visualized potential image.

[0057] In some embodiments, the location and type of the fault point are displayed on the three-dimensional visualized potential image, which not only intuitively displays the fault point but also helps maintenance personnel to accurately locate and monitor the fault point.

[0058] The intelligent fault diagnosis method for a distribution network in the above embodiment obtains a real-time potential signal from a node; inputs the real-time potential signal into a pre-trained fault type classifier to determine the node's fault type; and performs three-dimensional reconstruction of the real-time potential signal to obtain a three-dimensional visualized potential image; determines the location of the fault point based on the three-dimensional visualized potential image and the spatial topology of the distribution network; and displays the location of the fault point and the fault type on the three-dimensional visualized potential image. The above method can accurately identify the fault type and locate the fault point, and can also display it intuitively, thereby improving the efficiency and accuracy of fault diagnosis.

[0059] Furthermore, in some embodiments, the real-time potential signal is input into a pre-trained fault type classifier to obtain the fault type of the node, including:

[0060] Step 201 : Obtain propagation path characteristics of the real-time potential signal according to the real-time potential signal and the spatial topological structure.

[0061] In this embodiment, the propagation path characteristics of the real-time potential signal are generated by analyzing the propagation trajectory of the real-time potential signal in the spatial topological structure.

[0062] Step 202 : Match the propagation path characteristics of the real-time potential signal with the fault path characteristics in the fault type classifier to obtain the fault type.

[0063] In this embodiment, if the propagation path characteristics of the real-time potential signal match those in the spatial topology feature library, the fault type classifier directly outputs the fault type. For example, a fault signal propagating from node A to node C, with a path length of 10 kilometers, matches the single-point short circuit path in the fault type classifier. Based on the matching result, the fault type classifier outputs the fault type, such as "single-point short circuit." This method leverages path characteristics to quickly locate the fault type, improving detection efficiency.

[0064] In some embodiments, determining the location of the fault point based on the three-dimensional visualized potential image and the spatial topology of the power grid includes:

[0065] Step 301: Obtain a coverage matrix of the node according to the three-dimensional visualized potential image and the spatial topological structure.

[0066] The 3D visualized potential image is combined with a pre-established database of the spatial topology of the distribution network to obtain node distribution information. The spatial coverage between nodes is calculated using the Euclidean distance algorithm to generate a node coverage matrix. For example, a power grid has eight nodes, each with three-dimensional coordinates. The Euclidean distance algorithm is used to calculate the geometric distance between nodes, generating an 8×8 node coverage matrix. The matrix elements represent the spatial coverage between nodes. For example, the coverage of a node pair is 2.5 kilometers. This matrix quantifies the spatial relationships between nodes and provides a basis for regional division.

[0067] Step 302: Divide the three-dimensional visualized potential image into regions according to the coverage matrix.

[0068] In this embodiment, an image segmentation algorithm is used to segment the 3D visualized potential image based on the node coverage matrix. The image segmentation algorithm clusters pixels with similar potential values, dividing the image into several regions. For example, a 3D visualized potential image may generate five regions after segmentation, with the average potential value of one region being abnormally high. By comparing the potential value range with a preset fault type, such as the potential value range of 0.8 to 1.2 for a multi-point ground fault, the region is confirmed to match the fault type. This approach allows for precise isolation of fault-related regions.

[0069] Step 303: If the potential difference value of the divided area matches the fault type, the fault area is marked using a boundary extraction algorithm to obtain a fault area image.

[0070] In this embodiment, if the potential value of the divided area matches the preset fault type, the fault area is annotated using a boundary extraction algorithm to generate a fault area image. Otherwise, step 301 is re-executed. For example, if the potential gradient of a fault area is greater than 0.5, an annotated image containing that area is generated. The fault area in the image is highlighted in red, providing a visual representation of the fault area and facilitating subsequent analysis.

[0071] Step 304: digitally process the fault area image to obtain a digital potential matrix.

[0072] In this embodiment, an analog-to-digital conversion algorithm is used to digitize the pixel values ​​of the three-dimensional potential distribution image to obtain a digitized potential matrix.

[0073] Step 305 : Iteratively optimize the spatial correspondence between the pixel values ​​of the digitized potential matrix and the fault type by using a gradient descent algorithm to determine the location of the fault point.

[0074] In this embodiment, if the potential distribution deviation between the three-dimensional visualization image and a certain fault type in the spatial topology feature library is less than a preset threshold, the fault point location is determined through a gradient descent optimization algorithm based on analog-to-digital conversion accuracy and time-frequency localization characteristics.

[0075] In some embodiments, the intelligent fault diagnosis method for the distribution network proposed in this application further includes the step of pre-establishing a spatial topology feature library. This step includes:

[0076] Step 401 : Acquire potential signal acquisition data of a node, and obtain a potential change sequence according to the potential signal acquisition data.

[0077] In some embodiments, a high-precision sensor is used to collect potential signals from the node, and a synchronous timestamp technology is used to time-domain mark the collected data to obtain a raw potential signal sequence with a timestamp. If the amplitude of the raw potential signal sequence is lower than a preset threshold, a signal amplifier is used to perform gain processing on the raw potential signal, and the raw potential signal sequence is adjusted to an adaptation range to obtain an amplified potential signal sequence. Based on the amplified potential signal sequence, a high-precision analog-to-digital converter is used for digitization processing, and the signal is discretized in combination with high sampling frequency technology to generate a digitized potential signal sequence. The digitized potential signal sequence is subjected to frequency domain analysis by Fourier transform. If the noise frequency is detected to be higher than the preset threshold, a low-pass filter is applied for noise reduction processing to obtain an initial potential change sequence.

[0078] In one possible implementation, the amplitude of the original potential signal sequence may be low due to grid load fluctuations. For example, the signal amplitude at a certain node may be only 0.05 volts, below the preset threshold of 0.1 volts. In this case, the signal amplifier uses an adjustable gain module to amplify the signal to 0.2 volts to match the input range of the analog-to-digital converter. It should be noted that the amplification process must avoid introducing additional noise, and a low-noise amplifier is preferred to ensure signal fidelity. This signal amplification step significantly improves the accuracy of subsequent digital processing.

[0079] In one possible implementation, a high-precision analog-to-digital converter (ADC) combined with high sampling frequency technology discretizes the amplified potential signal sequence. For example, a 24-bit ADC with a sampling frequency of 10 kHz can convert the continuous signal into a high-resolution digital sequence. This high sampling frequency can capture rapidly changing potential signals, such as transient fluctuations caused by switching operations in the power grid, providing data support for precise fault location.

[0080] In one possible implementation, a Fourier transform is used to perform frequency domain analysis on the digitized potential signal sequence. For example, the analysis reveals 500 Hz high-frequency noise in the potential signal, exceeding a preset threshold of 300 Hz. A low-pass filter with a cutoff frequency of 400 Hz is applied to effectively filter out the high-frequency noise while retaining the useful signal components. The filtered initial potential change sequence more clearly reflects the actual operating status of the power grid, such as potential fluctuations caused by load changes or short circuits.

[0081] For example, a substation successfully used this solution to detect a minor potential anomaly caused by aging lines, providing an early warning and avoiding a power outage. The overall solution integrates high-precision data acquisition, synchronous marking, signal amplification, digital processing, and noise reduction analysis, forming a complete chain from data acquisition to feature extraction. Each link supports the other, ensuring high signal fidelity and reliable analysis.

[0082] Furthermore, in step 401, if the noise component in the initial potential change sequence exceeds a preset threshold, wavelet transform technology is used to separate the noise spectrum through adaptive threshold and threshold shrinkage rule based on wavelet basis function and multi-scale decomposition to obtain a denoised potential change sequence.

[0083] High-frequency components are obtained from the initial potential change sequence. Wavelet transform technology is used to perform multi-scale decomposition of the sequence based on preselected wavelet basis functions to obtain a wavelet coefficient sequence. The amplitude of each scale coefficient in the wavelet coefficient sequence is calculated. If the amplitude exceeds a preset threshold, it is adjusted using an adaptive threshold algorithm to obtain an adjusted coefficient sequence. Based on the adjusted coefficient sequence, a threshold contraction rule is applied to suppress high-frequency noise components, and wavelet reconstruction technology is used to restore the signal to obtain a preliminary denoised sequence. This preliminary denoised sequence is subjected to frequency domain analysis using Fourier transform. If a residual noise spectrum is detected, a low-pass filter is applied to obtain the final denoised potential change sequence.

[0084] For example, in the processing of node potential signals, wavelet transform technology is used to extract high-frequency components. Wavelet transform can effectively separate high-frequency noise from useful signals by decomposing the signal into frequency components of different scales. The wavelet basis functions used by wavelet transform, such as Daubechies wavelet, are suitable for capturing transient changes in power grid signals due to their orthogonality and compact support characteristics. In one possible implementation, if the potential change sequence of a certain node collected contains high-frequency components of voltage mutations, the db4 wavelet basis function is selected for 4-layer decomposition to obtain a coefficient sequence from low frequency to high frequency. This decomposition method can clearly distinguish between the stable part and transient disturbances in the signal data.

[0085] Furthermore, Fourier transform is used to analyze the frequency domain characteristics of the preliminary denoised sequence. For example, the analysis found that there is still 400 Hz residual noise in the sequence, which exceeds the expected normal range of 300 Hz. Preferably, a low-pass filter is applied with a cutoff frequency of 350 Hz to filter out the high-frequency residual noise and obtain the final denoised potential change sequence. This multi-step processing method of decomposition, adjustment, suppression and filtering forms a complete signal optimization chain, in which each link supports each other and ensures the high fidelity of the potential signal data.

[0086] Step 402: Obtain a target sequence based on the potential change sequence and the energy distribution characteristics of the fault signal.

[0087] In some embodiments, a fast Fourier transform is used to perform frequency domain decomposition on the (denoised) potential change sequence to obtain a frequency domain component sequence of the potential change sequence. A bandpass filter is used to filter the frequency components associated with the fault within this frequency domain component sequence. The portion of the sequence whose component amplitude exceeds a preset threshold is considered the fault-related frequency band sequence. A Hilbert transform is used to extract the energy distribution characteristics of each frequency band in the fault-related frequency band sequence, and the energy value of each frequency band is calculated to obtain an energy distribution sequence. A principal component analysis algorithm is used to perform dimensionality reduction on the energy values ​​of the energy distribution sequence to obtain a target sequence.

[0088] Exemplarily, Fast Fourier Transform (FFT) technology is used to convert the (de-noised) potential change sequence from the time domain to the frequency domain, decomposing it into a frequency domain component sequence. Preferably, the potential change sequence collected at a node contains periodic voltage fluctuations. After applying the FFT, the resulting frequency domain component sequence exhibits a 50 Hz fundamental frequency and several high-frequency harmonics. This decomposition facilitates subsequent analysis of fault-related frequency characteristics. For the frequency domain component sequence, a bandpass filter with upper and lower cutoff frequencies is used to filter out only specific frequency bands, filtering out components related to the fault and eliminating irrelevant components. For example, power grid faults are often accompanied by high-frequency components between 200 and 500 Hz. A bandpass filter with a frequency range of 200 to 500 Hz is set to filter out relevant components. If the component amplitude exceeds a preset threshold, such as 0.05, the component is retained, resulting in the fault-related frequency sequence. For another example, if a 300 Hz component with an amplitude of 0.07 is detected in a node signal, exceeding the threshold, this component may indicate a fault signature and be retained for subsequent analysis.

[0089] For example, the Hilbert transform is used to extract the energy distribution characteristics of the fault-related frequency band sequence, calculate the energy value of the fault-related frequency band sequence, and generate an energy distribution sequence. For example, the energy value of a node signal at the moment of the fault increases from 0.02 to 0.08, which clearly reflects the dynamic characteristics of the fault. The principal component analysis algorithm is used to reduce the dimension of the energy value of the energy distribution sequence to obtain the target sequence of the target. Principal component analysis reduces the data dimension and retains key information by extracting the main change direction of the energy value. For example, an energy distribution sequence contains energy values ​​of multiple frequency bands. The principal component analysis identifies the first two principal components, which explain 90% of the variance and generate a simplified target sequence.

[0090] The target sequence clearly reflects the potential change trend caused by the fault, facilitating subsequent fault location and diagnosis. This multi-step process efficiently extracts fault features through frequency domain decomposition, screening, energy extraction, and dimensionality reduction.

[0091] Step 403: Acquire the time-frequency localization characteristics of the target sequence, and obtain the potential characteristic vector of the node according to the geographical location coordinates and electrical connection type of the node.

[0092] In this embodiment, a time-frequency analysis method is used to extract the time-frequency localization features of the target sequence, and the node potential feature vector is generated by combining the geographical location coordinates and electrical connection types in the node distribution coverage information.

[0093] Exemplarily, the target sequence is decomposed into time and frequency using wavelet transform technology to obtain the distribution characteristics of the target sequence in time and frequency, thereby obtaining a time-frequency localized feature sequence. Based on the geographic location coordinates in the node distribution coverage information, a spatial interpolation algorithm is used to calculate the spatial weight of each node to obtain a weighted time-frequency feature sequence of the time-frequency localized feature sequence. If the characteristic value of any node in the weighted time-frequency feature sequence exceeds a preset threshold, the electrical topology information of the node is fused using a graph convolutional neural network algorithm based on the electrical connection type to obtain a preliminary node potential feature vector. The preliminary node potential feature vector is subjected to dimensionality reduction processing using a principal component analysis algorithm to extract spatial correlation characteristics related to the fault signal to obtain a potential feature vector of the target node. In this embodiment, the electrical connection type and electrical topology information adopt the common types and structures of power grids or distribution stations, such as external electrical connections and internal electrical connections, and the electrical topology information adopts modern power grid control systems, which are not specifically limited here.

[0094] In one possible implementation, a potential change sequence collected at a node contains a transient fault signal. Applying a discrete wavelet transform yields a time-frequency localization feature sequence, revealing the sudden change in the 300 Hz high-frequency component within 0.1 seconds when the fault occurs. This decomposition method facilitates capturing the transient characteristics of the fault.

[0095] A spatial interpolation algorithm is used to calculate the spatial weight of each node, combining the geographic coordinates from the node distribution coverage information. For example, given the geographic coordinates of 10 nodes in a power grid, the inverse distance weighted interpolation algorithm is used to calculate a weight of 0.8 for nodes closer to the fault node and 0.3 for those farther away, generating a weighted time-frequency feature sequence. This method effectively integrates spatial information and enhances the regional representativeness of features. When the eigenvalue of a node in the weighted time-frequency feature sequence exceeds a preset threshold, such as 0.06, it indicates a possible fault signal and requires further analysis. A graph convolutional neural network aggregates the features of neighboring nodes based on the electrical connections between nodes to generate a preliminary node potential feature vector. For example, if a node is connected to three neighboring nodes via a high-voltage line, the graph convolutional neural network analyzes its 300 Hz eigenvalue and, combined with the spatial topology, generates a feature vector containing both spatial and electrical information. This approach fully utilizes the topological characteristics of the power grid and enhances the expressiveness of features. Principal component analysis (PCA) is used to reduce the dimensionality of the preliminary node potential feature vector and extract spatial correlation features related to the fault signal.

[0096] Step 404 : acquiring the propagation path characteristics of the time-frequency localization feature in the spatial topological structure according to the potential feature vector of the node, and obtaining the spatial time-frequency distribution feature.

[0097] Based on the geographic location information in the node's potential feature vector, a Kriging interpolation algorithm is used to calculate the geographic spatial weight of each node to obtain the weighted geographic location sequence. Using this weighted geographic location sequence, a minimum spanning tree algorithm is used to construct the spatial topology of the distribution network to obtain the topological connection matrix. From this topological connection matrix, a graph traversal algorithm is used to extract the propagation path of the time-frequency features to obtain the path characteristic sequence. A kernel density estimation algorithm is used to calculate the spatial distribution pattern of the features in this path characteristic sequence to obtain the spatial time-frequency distribution features.

[0098] For example, in power grid fault analysis, the Kriging interpolation algorithm is used to calculate geographic spatial weights based on the geographic location information in the node potential eigenvector. Kriging interpolation is an interpolation method based on spatial autocorrelation. It takes into account the distance between nodes and the statistical characteristics of data distribution to generate a smooth weight distribution. For example, a power grid contains 8 nodes with known geographic coordinates. Based on the distance between nodes and the variance of potential changes, the Kriging interpolation algorithm is used to calculate the weight of nodes close to the fault area as 0.75 and those far away from the area as 0.25, generating a weighted geographic location sequence. This method makes full use of spatial statistical characteristics to ensure that the weight reflects the geographic correlation of the fault signal.

[0099] In one possible implementation, a minimum spanning tree algorithm is used to construct the spatial topology of a distribution network using a weighted sequence of geographic locations. This algorithm uses the geographic distance between nodes as edge weights, connecting all nodes to form an acyclic graph and generating a topological connectivity matrix. For example, in a given power grid, 10 nodes are mapped using the minimum spanning tree algorithm to generate a topological connectivity matrix based on weighted distances. The matrix elements represent the strength of the connections between nodes, such as a connection weight of 0.9 for a pair of nodes. This topological map clearly reflects the spatial structure of the distribution network and facilitates subsequent analysis.

[0100] A graph traversal algorithm extracts the propagation paths of time-frequency features from the topological connectivity matrix, generating a path feature sequence. The graph traversal algorithm uses a depth-first or breadth-first approach to track the propagation trajectory of fault signals within the topological graph. For example, if a node experiences a fault, its 300 Hz time-frequency features are traversed using a breadth-first approach to generate a feature sequence containing three propagation paths, with path lengths of 5, 7, and 10 kilometers, respectively. This path feature sequence intuitively demonstrates the spatial propagation patterns of the signal.

[0101] In some embodiments, for the path characteristic sequence, the kernel density estimation algorithm is used to obtain the distribution law of the features in space and generate spatial time-frequency distribution features. For example, the path characteristic sequence of a certain power grid shows that the 300 Hz feature is concentrated in a certain area, and the kernel density estimation generates a density map, which shows that the peak value of the feature density in this area is 0.85. This distribution feature clearly reveals the spatial aggregation characteristics of the fault signal and provides a basis for fault location. For example, when a node fails, its weighted geographic location sequence shows a weight of 0.7, the topological connection matrix shows that it has a strong connection with its neighboring nodes, the path characteristic sequence reveals that the signal propagates along the main path, and the kernel density estimation further confirms that the features are concentrated in a certain area. The above method makes full use of geographic and topological information to enhance the comprehensiveness of feature extraction.

[0102] Step 405: Obtain a spatial topology feature library based on the spatial time-frequency distribution features.

[0103] In this embodiment, a density clustering algorithm is used to calculate the cluster density and separation of feature points in the spatial time-frequency distribution characteristics to obtain the initial cluster center sequence. Using this initial cluster center sequence, a kernel density estimation algorithm is used to calculate the spatial distribution characteristics of the cluster centers to obtain the spatial distribution feature sequence. If the separation of the spatial distribution feature sequence is greater than a preset threshold, a minimum spanning tree algorithm is used to construct the power grid topology corresponding to the cluster centers to obtain the topological connection matrix. Based on this topological connection matrix, a graph traversal algorithm is used to extract the fault path characteristics of the fault type to obtain the spatial topological feature library.

[0104] For example, in power grid fault analysis, a density clustering algorithm is used to calculate the cluster density and separation of feature points in the spatial time-frequency distribution to generate an initial cluster center sequence. The density clustering algorithm analyzes the density of feature points in space and identifies high-density areas as cluster centers. For example, the spatial time-frequency distribution characteristics of a power grid contain 1,000 feature points. The density clustering algorithm identifies five high-density areas and generates five initial cluster centers, with a separation of 0.8 between the centers. This method effectively captures the spatial clustering characteristics of feature points and provides a reliable starting point for subsequent analysis.

[0105] In one possible implementation, based on the initial cluster center sequence, a kernel density estimation algorithm further calculates the spatial distribution characteristics of the cluster centers to generate a spatial distribution feature sequence. For example, the five cluster centers of a power grid were estimated using kernel density estimation to generate a spatial distribution feature sequence, showing that the density peak of the center points in a certain area was 0.9. This sequence clearly reflects the spatial distribution pattern of the cluster centers in the power grid, facilitating subsequent topological analysis. If the separation of the spatial distribution feature sequence is greater than a preset threshold, such as 0.7, the minimum spanning tree algorithm is used to construct the power grid topology corresponding to the cluster centers and generate a topological connectivity matrix. The minimum spanning tree algorithm uses the geographic distance between cluster centers as the edge weight and connects all centers to form an acyclic graph. For example, the five cluster centers of a power grid were analyzed using the minimum spanning tree algorithm to generate a topological connectivity matrix based on weighted distance. The matrix elements represent the connection strength between centers, such as the connection weight of a pair of centers being 0.85. This topological structure intuitively demonstrates the spatial relationship between cluster centers.

[0106] Based on the topological connectivity matrix, a graph traversal algorithm extracts the fault path characteristics of each fault type and generates a spatial topological feature library. The graph traversal algorithm uses a breadth-first approach to track the corresponding propagation path characteristics of each fault type in the spatial topology of the distribution network.

[0107] Furthermore, in some embodiments, the potential difference matrix is ​​obtained from the potential signals of the nodes, including:

[0108] Step 501: obtaining time-frequency distribution features from a potential signal, performing time-frequency decomposition on the real-time potential signal using a wavelet transform algorithm, and obtaining a time-frequency distribution feature sequence;

[0109] Step 502 : According to the time-frequency distribution feature sequence, the potential difference between nodes is calculated using the Euclidean distance algorithm to obtain a node potential difference matrix.

[0110] For example, a power grid consists of 10 nodes, each with a time-frequency signature sequence consisting of 100 characteristic points. The Euclidean distance algorithm calculates the distance between each pair of nodes, generating a 10×10 potential difference matrix. Each matrix element represents the potential difference between nodes. For example, the difference between a pair of nodes is 0.75. This matrix intuitively reflects the potential relationship between nodes and provides a key basis for fault location.

[0111] Reference Figure 2 Other embodiments of the present application further provide an intelligent fault diagnosis system for a distribution network, including:

[0112] The data acquisition module 601 is used to obtain the real-time potential signal of the node;

[0113] A fault type acquisition module 602 is configured to input the real-time potential signal into a pre-trained fault type classifier to obtain the fault type of the node, wherein the fault type classifier is obtained by performing feature classification training on the potential difference matrix and a pre-established spatial topology feature library, and the fault type classifier includes the correspondence between the potential difference of the node and the fault type, as well as the potential distribution of the fault type, and the potential difference matrix is ​​obtained from the potential signal of the node;

[0114] A three-dimensional visualization module 603 is used to perform three-dimensional reconstruction on the real-time potential signal to obtain a three-dimensional visualized potential image;

[0115] A fault point determination module 604 is configured to determine the location of the fault point based on the three-dimensional visualized potential image and the spatial topology of the distribution network;

[0116] The three-dimensional visualization module 603 is further configured to display the location of the fault point and the fault type on the three-dimensional visualized potential image.

[0117] It should be noted that the intelligent fault diagnosis system for a distribution network provided in an embodiment of the present application is used to execute all the process steps of the intelligent fault diagnosis method for a distribution network in the above embodiment. The working principles and beneficial effects of the two correspond one to one, so they will not be repeated here.

[0118] The present application also provides an electronic device. The electronic device includes: a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a data acquisition program. When the processor executes the computer program, the steps in the above-mentioned intelligent fault diagnosis method for power distribution networks are implemented, such as Figure 1 Alternatively, when the processor executes the computer program, the functions of the modules / units in the above-mentioned device embodiments are realized, such as the data acquisition module.

[0119] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present application. The one or more modules / units may be a series of computer program instruction segments capable of completing specific functions, which are used to describe the execution process of the computer program in the electronic device.

[0120] The electronic device may be a computing device such as a desktop computer, notebook, PDA, or smart tablet. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will appreciate that the aforementioned components are merely examples of electronic devices and do not constitute a limitation of the electronic device. The electronic device may include more or fewer components than those described above, or a combination of certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, and the like.

[0121] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor. The processor is the control center of the electronic device and connects various parts of the entire electronic device using various interfaces and lines.

[0122] The memory can be used to store the computer programs and / or modules. The processor implements the various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory and accessing the data stored in the memory. The memory may primarily include a program storage area and a data storage area. The program storage area may store an operating system and at least one application required for a function (such as a sound playback function or an image playback function); the data storage area may store data generated based on the use of the mobile phone (such as audio data, a phone book, etc.). Furthermore, the memory may include high-speed random access memory and non-volatile memory, such as a hard disk, internal memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state storage device.

[0123] If the module / unit integrated into the electronic device is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application can implement all or part of the process steps in the above-mentioned method embodiments by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When executed by a processor, the computer program can implement the steps of each of the above-mentioned method embodiments. The computer program includes computer program code, which can be in source code form, object code form, executable file, or some intermediate form. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal, and software distribution medium. It should be noted that the content of the computer-readable medium can be appropriately increased or decreased based on the requirements of legislation and patent practice in a jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media does not include electric carrier signals and telecommunication signals.

[0124] It should be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed across multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment. In addition, in the drawings of the device embodiments provided in this application, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art can understand and implement the present invention without inventive work.

[0125] The specific embodiments described above further illustrate the objectives, technical solutions, and beneficial effects of this application. It should be understood that the above descriptions are merely specific embodiments of this application and are not intended to limit the scope of protection of this application. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of this application by those skilled in the art should be included within the scope of protection of this application.

Claims

1. An intelligent fault diagnosis method for a distribution network, characterized in that: include: Get the real-time potential signal of the node; Inputting the real-time potential signal into a pre-trained fault type classifier to obtain the fault type of the node, wherein the fault type classifier is obtained by performing feature classification training on a potential difference matrix and a pre-established spatial topology feature library, and the fault type classifier includes a correspondence between the potential difference value of the node and the fault type, as well as a potential distribution of the fault type, and the potential difference matrix is ​​obtained from the potential signal of the node; Performing three-dimensional reconstruction on the real-time potential signal to obtain a three-dimensional visualized potential image; determining the location of a fault point based on the three-dimensional visualized potential image and the spatial topological structure of the distribution network; Displaying the location of the fault point and the fault type on the three-dimensional visualized potential image; The real-time potential signal is input into a pre-trained fault type classifier to obtain the fault type of the node, including: Obtaining a propagation path characteristic of the real-time potential signal according to the real-time potential signal and the spatial topological structure; Matching the propagation path characteristics of the real-time potential signal with the fault path characteristics in the fault type classifier to obtain the fault type; The method of determining the location of the fault point according to the three-dimensional visualized potential image and the spatial topological structure of the power grid includes: Obtaining a coverage matrix of the node according to the three-dimensional visualized potential image and the spatial topological structure; dividing the three-dimensional visualized potential image into regions according to the coverage matrix; If the potential difference value of the divided area matches the fault type, the fault area is marked using a boundary extraction algorithm to obtain a fault area image; Digitally processing the fault area image to obtain a digital potential matrix; Iteratively optimizing the spatial correspondence between the pixel values ​​of the digitized potential matrix and the fault type through a gradient descent algorithm to determine the location of the fault point; The step of pre-establishing a spatial topology feature library comprises: collecting a potential signal of the node, and obtaining a potential change sequence according to the potential signal; Obtaining a target sequence according to the potential change sequence and the energy distribution characteristics of the fault signal; Obtaining the time-frequency localization characteristics of the target sequence, and obtaining the potential characteristic vector of the node according to the geographical location coordinates and electrical connection type of the node; According to the potential characteristic vector of the node, the propagation path characteristics of the time-frequency localization feature in the spatial topological structure are obtained to obtain the spatial time-frequency distribution feature; A spatial topological feature library is obtained according to the spatial time-frequency distribution features.

2. The method according to claim 1, characterized in that The step of obtaining a spatial topological feature library according to the spatial time-frequency distribution features comprises: Using a density clustering algorithm to calculate the clustering density and separation of the feature points in the spatial time-frequency distribution features to obtain an initial cluster center sequence; Using the initial cluster center sequence, a kernel density estimation algorithm is used to calculate the spatial distribution characteristics of the cluster centers to obtain a spatial distribution feature sequence; If the separation degree of the spatial distribution feature sequence is greater than a preset threshold, a minimum spanning tree algorithm is used to construct a topological structure of the distribution network corresponding to the cluster center to obtain a topological connection matrix; According to the topological connection matrix, a graph traversal algorithm is used to extract the fault path characteristics of the fault type to obtain the spatial topological feature library.

3. The method according to claim 1, characterized in that The process of obtaining the potential difference matrix includes: Using a wavelet transform algorithm to perform time-frequency decomposition on the potential signal to obtain a time-frequency distribution feature sequence; According to the time-frequency distribution feature sequence, the potential difference between nodes is calculated using the Euclidean distance algorithm to obtain the potential difference matrix of the nodes.

4. The method according to claim 1, wherein Performing three-dimensional reconstruction on the real-time potential signal to obtain a three-dimensional visualized potential image includes: Performing time-frequency decomposition on the real-time potential signal to obtain a time-frequency feature sequence; The three-dimensional stereoscopic microscopy technique is used to extract spatial distribution data of the time-frequency feature sequence to obtain a three-dimensional visualized potential image.

5. An intelligent fault diagnosis system for a distribution network, characterized in that: A method for implementing an intelligent fault diagnosis method for a distribution network as claimed in any one of claims 1 to 4, comprising: Data acquisition module, used to obtain the real-time potential signal of the node; a fault type acquisition module, configured to input the real-time potential signal into a pre-trained fault type classifier to obtain the fault type of the node, wherein the fault type classifier is obtained by performing feature classification training on a potential difference matrix and a pre-established spatial topology feature library, and the fault type classifier includes a correspondence between the potential difference of the node and the fault type, as well as a potential distribution of the fault type, and the potential difference matrix is ​​obtained from the potential signal of the node; A three-dimensional visualization module is used to perform three-dimensional reconstruction on the real-time potential signal to obtain a three-dimensional visualized potential image; A fault point determination module is used to determine the location of the fault point based on the three-dimensional visualized potential image and the spatial topological structure of the distribution network; The three-dimensional visualization module is further configured to display the position of the fault point and the fault type on the three-dimensional visualization potential image.

6. An electronic device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the method for intelligent fault diagnosis of a distribution network according to any one of claims 1 to 4 is implemented.

7. A computer-readable storage medium, characterized in that The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the intelligent fault diagnosis method for the distribution network according to any one of claims 1 to 4.

Images