A power grid type fly point detection method and system based on intelligent clustering

Abstract

A method and system for detecting flying points in power grids based on intelligent clustering is disclosed. The method extracts the planned path coordinates, actual geographic coordinates of equipment, and topological connections of the target line from the power grid GIS system, and converts the coordinate data to a unified planar coordinate system. The planned path coordinates are then subjected to sliding window clustering and merged based on overlap to obtain multiple planned clusters. The actual geographic coordinate data are grouped and clustered by cluster radius and minimum number of equipment, outputting multiple equipment clusters. The spatial offset distance between each equipment cluster and its nearest planned cluster is obtained, and the initial spatial offset distance is corrected using shape similarity. If the corrected offset distance is greater than a preset offset distance threshold for the voltage level, and all equipment in the equipment cluster has a topological connection, then the equipment cluster is determined to be a set of flying point-like equipment. This invention improves the technical accuracy and reliability of flying point detection on the power grid GIS graphical side.

Description

Technical Field

[0001] This invention belongs to the field of power grid data anomaly detection technology, specifically relating to a power grid-type flying point detection method and system based on intelligent clustering. Background Technology

[0002] Equipment coordinate flying point detection is a core component in ensuring the data quality of power grid GIS (Geographic Information System). Flying points typically refer to abnormal locations where the geographical coordinates of equipment deviate significantly from the planned path or normal distribution range. Currently, mainstream flying point detection methods on the power grid GIS graphic side mainly employ two approaches: detection methods based on fixed distance thresholds and detection methods based on single topology verification.

[0003] For example, patent CN109885624B employs a fixed-radius buffer analysis method, generating a buffer zone centered on the planned path and identifying device coordinates outside the buffer zone as flying points. Patent CN113111892A proposes a fixed threshold setting method based on line segmentation, dividing the line into different segments according to voltage levels and presetting corresponding distance thresholds to improve the targeting of flying point detection for line equipment at different voltage levels. And patent CN108764321A constructs a topology association matrix for power grid equipment, verifies the topological connection relationship between the equipment and its associated line, and identifies isolated equipment without associated records in the matrix as flying points.

[0004] However, the existing technologies mentioned above are insufficient for detecting fly-point-like points. The core characteristic of fly-point-like points is that while the topological connections between devices are normal, their overall positions are offset. Single topology verification can only identify isolated fly-points with topological anomalies, completely failing to distinguish between fly-point-like points with normal topology and normal points. Furthermore, the actual laying environment of power grid lines varies significantly. In densely populated urban areas, devices may form reasonably dispersed clusters on both sides of the planned path due to terrain constraints; while in open suburban areas, devices may be considered fly-point-like points even if they are offset from the planned path but do not exceed the threshold. Fixed thresholds cannot adapt to the device distribution characteristics in different scenarios, easily misjudging reasonably dispersed normal points as fly-points, or failing to detect overall offset fly-point-like points that do not exceed the threshold, resulting in insufficient detection accuracy. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for detecting power grid-like flying points based on intelligent clustering. This addresses the problem in existing technologies where single topology verification can only identify isolated flying points with topological anomalies, leading to misjudgments of scattered normal points or flying points that have not exceeded thresholds. This invention achieves accurate differentiation between flying points and normal points, solves the problem of accurately separating abnormal locations from normal equipment functions, and improves the accuracy of power grid-like flying point detection.

[0006] The first aspect of this invention discloses a method for detecting power grid-related flying points based on intelligent clustering, employing the following technical solution: The associated data of the target line to be detected is extracted from the power grid GIS system, including the planned path coordinate data, the actual geographic coordinate data of all equipment, and the topological connection data between the equipment and the line; the planned path coordinate data and the actual geographic coordinate data are converted to a plane coordinate system. The system presets the window length and step size, performs sliding window clustering on the planned path data, and generates multiple initial planned clusters for the target route to be detected. The initial planned clusters in adjacent windows are merged according to the overlap of data points within the clusters, and multiple planned clusters are output. For the actual geographic coordinate data of the same type of equipment, after grouping by equipment type, cluster them according to the cluster radius and minimum number of equipment in the current scenario, and output only multiple equipment clusters that are of the same type and spatially adjacent; Obtain the spatial offset distance between each device cluster and its nearest planned cluster, including obtaining the distance between the center point of each device cluster and the center point of each planned cluster, and correct the initial spatial offset distance based on shape similarity; If the spatial offset distance of the device cluster is greater than the preset offset distance threshold corresponding to the voltage level of the target line to be detected, and there is a topological connection between any two devices in the device cluster, then the device cluster is determined to be a set of flying point devices.

[0007] Furthermore, the preset window length and step size include: For the target route to be detected, based on the total length of the planned path Set window length to Step size is Calculate the path curvature based on the sequence of path points along the planned path; If the path curvature is greater than the preset curvature threshold, it is determined to be a curved segment, and the window length of the curved segment is adjusted to... Step size adjusted to ; and These are the window length adjustment factor and the step size adjustment factor, respectively. and All are greater than 1; If the path curvature is less than or equal to the preset curvature threshold, it is determined to be a straight line segment, and the window length is maintained. , length is .

[0008] Furthermore, the step of obtaining the planned clusters includes: For any two adjacent windows, take one of the initial planned clusters from each window to form a cluster pair, and calculate the overlap of the cluster pair. When the overlap is greater than the preset overlap threshold, the cluster pair is merged into a single planned cluster. The merged planned cluster includes all coordinate points from the two initial planned clusters. Merge the initial planned clusters in any two adjacent windows based on their overlap; traverse all adjacent sliding windows to output multiple consecutive planned clusters.

[0009] Furthermore, the actual geographic coordinate data of similar devices are grouped; including: Obtain the actual geographic coordinates and device types of all devices within the normal clustering area; The actual geographic coordinate data is grouped according to the device type to form several device groups; each device type corresponds to one device group. For each device group, clustering is performed based on the corresponding cluster radius and the minimum number of devices, resulting in multiple device clusters.

[0010] Furthermore, the process of determining the cluster radius and minimum number of devices for each device group includes: exist For any area where similar equipment is normally clustered, Calculate the distance between any two devices. The maximum device distance in the range is denoted as Then calculate all The mean of the maximum distances , calculate With safety factor Multiplying them gives the cluster radius; exist For each normal cluster of similar equipment, obtain the number of equipment in each cluster. Count the number A of normal clusters where the number of equipment falls within the interval [c,d]. If the ratio of A to M exceeds 95%, then the lower limit c is taken as the minimum number of equipment.

[0011] Furthermore, the process of determining the spatial offset distance between each device cluster and its nearest planned cluster includes: Obtain the coordinates of the center points of the device clusters and the planned clusters. The coordinates of the center points are the average x-coordinates and average y-coordinates of all device coordinate points or planned coordinate points within the cluster. For each device cluster, iterate through the distances between it and all planned clusters, and take the minimum distance as the initial spatial offset distance between the device cluster and its nearest planned cluster; Determine the minimum bounding rectangle of the equipment cluster and the minimum bounding rectangle of the nearest planned cluster of the equipment cluster, and calculate the shape similarity based on the minimum bounding rectangles of the two clusters; For each device cluster, if its initial spatial offset distance is not greater than a preset offset distance threshold and its shape similarity is less than a preset shape similarity threshold, then the initial spatial offset distance is corrected. The shape similarity is a weighted sum of the size ratio similarity and the orientation ratio similarity.

[0012] Furthermore, the process for determining whether the device cluster is a set of flying point devices includes: If the spatial offset distance of the device cluster is greater than the preset offset distance threshold corresponding to the voltage level of the target line to be detected, and there is a topological connection between any two devices in the device cluster, then the device cluster is determined to be a set of flying point devices.

[0013] Furthermore, the process for determining the topological connection relationship includes topological logical connectivity analysis and geographic logical connectivity analysis, and the steps include: Construct a topology graph with devices as nodes and connection relationships as edges; The process of topological logical connectivity analysis is as follows: for device clusters Extract all devices and their identifiers to form a complete system. The collection of devices; from Choose any device from the set of devices as the starting point, find all device nodes that can be reached from that starting point in the topology graph, and store them in the set. If set Device nodes in If all device nodes in a device set are identical, then the device is determined to be a device cluster. All devices within the system are logically interconnected in terms of topology; Determine the device cluster Once all devices within the network are logically connected to each other in terms of topology, a geographic logical connectivity analysis is then performed.

[0014] Furthermore, the geographic logical connectivity analysis further identifies the device clusters as a set of flying point-like devices, including: Each edge of the topology graph is assigned a weight to generate a spatially weighted topology graph, from which device clusters are extracted. Weighted subgraph ;exist Run the maximum spanning tree algorithm on it; If the generated maximum tree contains a weighted subgraph If all device nodes are identified, then the device cluster is determined. Through geographical logical connectivity analysis, then All devices in the cluster belong to the same connected component, thus confirming that there is a topological connection between any two devices in the cluster. The edge weights are determined as follows: Equipment at both ends of the side and equipment Actual geographic distance in a plane projection coordinate system ,when hour, ,otherwise ;in, For equipment The maximum distance threshold for the device type.

[0015] The second aspect of this invention discloses a power grid-type flying point detection system based on intelligent clustering, which implements the power grid-type flying point detection method described in the first aspect of this invention. The system includes: The data extraction unit extracts the associated data of the target line to be detected from the power grid GIS system, including the planned path coordinate data, the actual geographic coordinate data of all equipment, and the topological connection data between the equipment and the line; and converts the planned path coordinate data and the actual geographic coordinate data to a plane coordinate system. The transformation unit, with a preset window length and step size, performs sliding window clustering on the planned path data to generate multiple initial planned clusters of the target route to be detected; for the initial planned clusters in adjacent windows, it merges them according to the overlap of data points within the clusters to output multiple planned clusters. The clustering unit groups the actual geographic coordinate data of devices of the same type by device type, and then clusters them according to the clustering radius and minimum number of devices corresponding to that device type in the current scenario. The output contains only multiple spatially adjacent device clusters of the same type. The distance calculation unit obtains the spatial offset distance between each device cluster and its nearest planned cluster, including obtaining the distance between the center point of each device cluster and the center point of each planned cluster, and corrects the initial spatial offset distance based on shape similarity. If a device cluster is determined to be a set of flying point devices, and the spatial offset distance is greater than a preset offset distance threshold corresponding to the voltage level of the target line to be detected, and any two devices in the device cluster have a topological connection relationship, then the device cluster is determined to be a set of flying point devices.

[0016] The beneficial effects of this invention are that, compared with the prior art, 1. By using preset window length and step size to slide cluster along the planned path, the continuous linear planning path is decomposed into several planning clusters corresponding to specific spatial regions, avoiding the problem of poor adaptability of traditional fixed thresholds to different road segments; then, by using density clustering, the actual equipment is aggregated into equipment clusters according to spatial distance, accurately capturing the spatial characteristics of relatively concentrated local equipment, providing morphological basis for distinguishing between fly-point clusters with overall offset and normal clusters with reasonable local dispersion.

[0017] 2. By calculating the distance between the center point of the device cluster and the nearest planned cluster, the spatial difference between the actual and planned locations is transformed into a quantifiable indicator. Compared with traditional global distance judgment, the local cluster-to-cluster distance calculation can accurately capture the characteristics of the overall device offset. If the devices are only locally dispersed, their clusters will maintain a relatively close distance to the corresponding planned clusters, while the clusters of fly-point-like locations will exhibit large distance characteristics due to the overall offset.

[0018] 3. The final accurate definition of flying points is completed by verifying the topology connection relationship. Based on the characteristic that the flying point has a normal topology but an abnormal position, the topological connectivity within the equipment cluster is verified to eliminate isolated flying points with topological breaks and scattered points caused by equipment failure. Only the clusters with excessive offset distance and normal topology are identified as flying point sets, thus achieving accurate separation between abnormal position and normal equipment function.

[0019] This invention overcomes the limitations of fixed threshold scene adaptation and makes up for the deficiency of single topology verification in being unable to identify location anomalies. It achieves accurate differentiation between flying points and normal points, fundamentally improving the technical accuracy and reliability of flying point detection on the power grid GIS graphic side. Attached Figure Description

[0020] Figure 1 The flowchart of the power grid-type flying point detection method based on intelligent clustering provided in the embodiments of the present invention is shown below. Figure 2 The diagram shows the structure of a power grid-type flying point detection system based on intelligent clustering, as provided in an embodiment of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of this invention. The embodiments described in this invention are merely some embodiments of this invention, and not all embodiments. Based on the spirit of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of this invention.

[0022] As an embodiment of the present invention, a specific implementation method for a power grid-like flying point detection method based on intelligent clustering is disclosed. The execution flow of the method embodiment is as follows: Figure 1.

[0023] S100: As one implementation method, the associated data of the target line to be detected is extracted from the power grid GIS system, including at least the planned path coordinate data, the actual geographic coordinate data of all equipment, and the topological connection data between the equipment and the line. Specifically: S110: Power Grid GIS System refers to the Power Grid Geographic Information System, which is a digital platform that integrates the spatial location, attribute information, and topological relationships of power grid equipment. It can store data such as line planning drawings, measured coordinates of equipment, and topological connection records.

[0024] (1) The planned route coordinate data is a set of latitude and longitude coordinates of the ideal laying route determined by the design institute during the design phase based on the needs of terrain, safety distance, user distribution, etc., such as the latitude and longitude sequence of the planned route of the "Chengdong Line" along Chengdong Avenue (a total of 120 coordinate points, with a longitude range of 118.72°-118.78° and a latitude range of 31.25°-31.29°).

[0025] (2) The actual geographic coordinate data of all devices refers to the actual geographic coordinate data of all devices included in the target line to be detected. The actual geographic coordinate data is the latitude and longitude coordinates of the actual installation location of the devices obtained by GPS positioning, UAV mapping and other methods.

[0026] (3) The topology connection data of equipment and line refers to the structured data that records the electrical connection relationship between equipment and line, and between equipment and adjacent equipment, including at least the equipment identifier, the connected line identifier and the parent equipment identifier.

[0027] In a further implementation, the extracted data is formatted uniformly, converting the planned path coordinate data and actual geographic coordinate data into a decimal format of longitude (°) - latitude (°), and the topology table is indexed by device identifier. Before the conversion, the device coordinate set is traversed, and records with null coordinates or those exceeding the administrative boundaries of the target route are removed.

[0028] S120: In a further embodiment, the planned path coordinate data and the actual geographic coordinate data of all devices are converted to the same plane projection coordinate system to obtain the key planned path coordinate data and key actual geographic coordinate data in the plane coordinate system.

[0029] In a further specific implementation, a suitable planar projection coordinate system, such as the Gauss-Kruger projection coordinate system, is selected to convert latitude and longitude coordinates on the Earth's ellipsoid into a plane rectangular coordinate system. The coordinate unit in the plane coordinate system is the meter.

[0030] In a further specific implementation, the planned path coordinate data (latitude and longitude) and the actual geographic coordinate data (latitude and longitude) extracted by S100 are converted into a plane coordinate system under a plane projection coordinate system.

[0031] Because distance calculations in a planar projected coordinate system are unaffected by the Earth's ellipsoid parameters, they offer high accuracy, especially within the same projection zone. This meets the high-precision requirements for power grid equipment offset detection. The specific conversion method can be any method known to those skilled in the art, and will not be elaborated upon here.

[0032] S200, as one implementation method of the embodiment, involves performing sliding window clustering on the key planning path coordinate data according to a preset window length and a preset step size to obtain several planning clusters; wherein each planning cluster corresponds to a continuous spatial region on the target route planning path. Specifically: S210: In a further implementation, based on the total length of the planned path Set window length to , This is used to break down the planned path into multiple short segments. The window length should be adapted to the distribution range of the equipment; for example, a window length of 500 meters can be set for the spacing between equipment on a power distribution line of 100-300 meters. Preset step size. Set as This ensures a 50% overlap between adjacent windows, preventing path segmentation and breakage. Those skilled in the art can scale or fine-tune the W and S ratios proportionally according to actual accuracy requirements.

[0033] In a further implementation, the planned path is divided into multiple sliding windows according to a preset window length and a preset step size, and the coordinate data of the planned path within each window are clustered. Taking the starting point of the planned path as the starting end of the window, a density-based clustering algorithm (such as the DBSCAN algorithm) is used, with a neighborhood radius set as... Local density adapted to window scale, and minimum number of core points. Filter out random noise.

[0034] Move the window by a step size S and repeat the above clustering process until the window covers the end point of the path. The clustering results within each window generate several initial planned clusters, and each cluster corresponds to a continuous path region.

[0035] S220: In a further embodiment, the curved sections in the path are specially treated. Specifically: As described above, the planned path is a continuous linear trajectory. The planned path for straight segments is typically stable, with relatively uniform spacing between coordinate points. However, in curved segments, the distribution of coordinate points changes significantly, potentially causing clustering algorithms to mistakenly group points from straight and curved segments into the same cluster, thus affecting the comparison of power grid equipment and subsequent analysis. Therefore, this invention employs special processing for curved segments to avoid chaotic clustering in these areas.

[0036] In a further specific implementation, the curvature of the path point sequence is calculated for the curved sections of the path. In one specific implementation, the curvature of the curved section is calculated by taking the reciprocal of the radius of the circumcircle of three points. The larger the curvature value, the more severe the path curvature at the location of the intermediate point; the smaller the value, the smoother the path, approaching a straight line.

[0037] If the curvature of the path point sequence is greater than a preset curvature threshold (e.g., 0.001 / meter), then the window length of the curved segment of the path will be adjusted to [value missing]. Step size adjusted to The remaining path segments retain their original window length and step size. It should be noted that the preset curvature threshold is the optimal value obtained through multiple trials.

[0038] The window slides along the planned path, repeatedly splitting and clustering the planned path until the end of the path. Finally, several initial planned clusters are output.

[0039] S230: Further, taking the starting point of the planned path as the starting end of the window, the overlap of the initial planned clusters in adjacent sliding windows is calculated to determine whether two adjacent clusters need to be merged: In two adjacent sliding windows A and B, you can choose one initial planning cluster A1 and B1 respectively. (A1, B1) constitutes a cluster pair. The overlap between A1 and B1 is the ratio of the number of intersection points of the coordinate points in the two clusters to the number of points in the key cluster. The key cluster is the cluster with fewer points in A1 and B1.

[0040] As an example, if A1 contains There are coordinate points, and B1 contains... coordinate points, and Then the overlap of (A1, B1) for ; Let A1 be the number of intersection points of A1 and B1.

[0041] like If the initial clustering is successful, then A1 and B1 will be merged into one cluster; otherwise, the two independent initial clustering clusters will be retained. The merged cluster contains all points from the two initial clustering clusters, and the cluster center is the mean of the coordinates of the union. The preset overlap threshold is set to 80% in this embodiment. Window A and B are traversed to merge the initially planned clusters within the two windows.

[0042] In further explanation, if the initial planned clusters of three adjacent windows A, B, and C are A1, B1, and C1, and All greater than If so, A1, B1, and C1 will be directly merged into a single cluster.

[0043] It should be noted that the sliding window moves along the planned route, and the planned coordinates of adjacent windows may overlap spatially. In complex road sections (such as branches, intersections, and turns), multiple initial planned clusters may appear in both the current window and the next window. If each window has multiple initial planned clusters, then multiple cluster pairs are formed.

[0044] The clusters in all windows are merged to output a set of continuous planned clusters that precisely correspond to the segmented spatial regions where the equipment should be installed.

[0045] In step S100, since the planned path is a continuous linear trajectory, directly performing global clustering on the coordinate points of the entire path can easily lead to the overlap of coordinate points from different road segments into a single cluster due to path curvature and branching (e.g., coordinate points of a straight segment and an adjacent curved segment are grouped into the same cluster), making it impossible to provide segmented benchmarks for equipment comparison. Sliding window clustering, however, divides the linear path into several segmented clusters: window splitting ensures the range of each cluster is controllable, local clustering ensures spatial continuity of coordinate points within a cluster, and overlapping and merging ensures no gaps between adjacent clusters. The resulting planned clusters can accurately correspond to the segmented spatial areas where the equipment should be installed.

[0046] S300: As one implementation method of this embodiment, based on the cluster radius and the minimum number of devices, the actual geographic coordinate data is clustered on a device basis to obtain several device clusters.

[0047] S310: In a further implementation, the actual geographic coordinate data of all devices are grouped according to device type to obtain several device groups, with each device group corresponding to one device type. This ensures that the spatial distribution characteristics of different devices are taken into account during clustering, avoiding clustering errors caused by differences in device type. The cluster radius and minimum number of devices are determined as follows: S311: Determine the initial range of clustering parameters based on domain knowledge, and establish the theoretical range of parameters as the basis and boundary for subsequent clustering parameter solving. Specifically: The initial range of the cluster radius is set based on the functional attributes, installation specifications, and power supply requirements of the equipment. For example, the initial range of the cluster radius for distribution transformers is set to 30-60 meters, which is suitable for equipment spaced 300-500 meters apart, ensuring that transformers with close spacing within the same power supply area can be grouped together; the initial range of the cluster radius for meter boxes is set to 15-30 meters, which is suitable for equipment spaced 50-100 meters apart, providing centralized service to building users; and the initial range of the cluster radius for transmission towers is set to 80-120 meters, which is suitable for equipment spaced 500-1000 meters apart.

[0048] The initial range for the minimum number of devices is set based on the normal distance scale of the devices. For example, the initial range for the minimum number of devices for distribution transformers is 2-4 units; the initial range for the minimum number of devices for meter boxes is 3-6 units; and the initial range for the minimum number of devices for transmission towers is 4-8 units.

[0049] S312: Calculate clustering parameters based on historical data and compare them with the initial intervals for verification. Specifically: (1) For cluster radius: Extract historical installation data of equipment in normal clustered areas (such as equipment coordinates, associated lines, and commissioning time) from the power grid GIS system, and optimize the cluster radius by grouping key indicators according to equipment type. Specifically: The maximum spacing between normally clustered devices of the same type is statistically analyzed, and the average value is calculated. This average is then multiplied by a safety factor to determine the final cluster radius. For example, in... For any area where similar equipment is normally clustered, Calculate the distance between any two of these devices. The maximum device distance in the range is denoted as Then calculate all The mean of the maximum distances .

[0050] Then the cluster radius is , To ensure the security factor for the tightness of devices within the cluster, this embodiment sets it to 0.8.

[0051] (2) For the minimum number of devices: In In areas where similar equipment is normally clustered, if more than 95% of the equipment numbers fall within the cd range, then the lower limit c will be used as the minimum equipment number. .

[0052] S313: Verify the cluster radius and minimum number of devices obtained in S312 using the initial range.

[0053] like If it falls within the initial range corresponding to the device type, then retain it; if If the value is less than the lower limit of the initial interval for the corresponding device type, then the lower limit of the initial interval will be assigned to... ;like If the value is greater than the initial range upper limit for the corresponding device type, then the initial range upper limit will be assigned to... .

[0054] Similarly, the minimum number of devices required for execution Verification.

[0055] Based on the obtained cluster radius and minimum number of devices For each device group, the DBSCAN algorithm is called to perform density clustering.

[0056] Further explanation is provided, where the initial range of clustering parameters can be adjusted for different power grid scenarios (dense urban areas, suburbs, rural areas) to ensure adaptability to varying scenarios. For example, in densely populated urban areas, where equipment installation space is limited and the spacing between similar devices may be smaller (e.g., the spacing between distribution transformers is reduced to 200-300 meters), the initial range of the clustering radius needs to be appropriately reduced (e.g., set to 20-40 meters) to avoid misclassifying devices from different power supply areas into one cluster. In contrast, in rural areas, where equipment is more dispersed and the spacing between similar devices may be larger (e.g., the spacing between distribution transformers is expanded to 400-600 meters), the initial range of the clustering radius needs to be appropriately increased (e.g., set to 40-70 meters) to ensure that devices within the same power supply area are correctly clustered. The minimum number of devices also needs to have its initial range set according to the scenario-specific clustering scale.

[0057] When extracting device data from historically normal clustered areas to calculate clustering parameters, select similar scenarios for extraction.

[0058] S314: Organize the final determined equipment type, scenario, cluster radius, and minimum equipment quantity correspondence into a dynamic mapping table. For example: Distribution transformers in densely populated urban areas: cluster radius 30 meters, minimum equipment quantity 3 units; Distribution transformers in rural areas: cluster radius 45 meters, minimum equipment quantity 2 units; Meter boxes in densely populated urban areas: cluster radius 20 meters, minimum equipment quantity 4 units; Then, every 1-2 years, update the parameters in the mapping table based on new equipment installation data, power grid planning adjustments, and scenario changes (such as new urban areas and rural power grid renovations) to ensure that the parameters always match the actual situation.

[0059] It should be noted that the normal spatial distribution density of different types of power grid equipment varies significantly. For example, distribution transformers, needing to cover a certain power supply radius, are typically installed at intervals of 300-500m; meter boxes, serving a specific building's users, are usually densely installed within a 20-50m range. If clustering is done directly without grouping by equipment type, errors can easily occur, such as grouping distribution transformers and meter boxes into the same cluster due to their proximity (e.g., if there are 10 meter boxes within 50m of a distribution transformer, direct clustering would group one transformer and 10 meter boxes into one cluster, failing to reflect the true clustering characteristics). Grouping by equipment type and setting appropriate clustering radii and minimum equipment numbers for different types of equipment ensures that each equipment cluster contains only the same type of spatially close equipment, accurately capturing the local relative concentration characteristics of the equipment. The core characteristic of "flying points" is overall offset but local concentration, while normal points are locally reasonably dispersed. The shape of the equipment clusters can initially distinguish between the two scenarios.

[0060] S400: As one implementation of the embodiment, obtain the spatial offset distance between each device cluster and its nearest planned cluster.

[0061] S410: In a further embodiment, the coordinates of the center point of each device cluster and each planned cluster are obtained; the center point coordinates are the coordinates composed of the average x-coordinate and the average y-coordinate of all device coordinate points or planned coordinate points within the cluster, which are used to represent the center position of the entire cluster.

[0062] S420: Further, the distance between the center point of each device cluster and the center point of each planned cluster is calculated using the Euclidean distance formula to obtain the initial spatial offset distance between clusters; specifically: for each device cluster, the distances between it and all planned clusters are traversed, and the minimum value is taken as the initial spatial offset distance between the device cluster and its nearest planned cluster. .

[0063] S430: Further, obtain the minimum bounding rectangle of each device cluster and the minimum bounding rectangle of the nearest planned cluster for that device cluster. The minimum bounding rectangle (MBR) is a rectangle that can completely enclose all points within the cluster, and the sides of the rectangle are parallel or perpendicular to the x-axis and y-axis of the plane coordinate system.

[0064] Calculate the shape similarity between the minimum bounding rectangle of each device cluster and the minimum bounding rectangle of the planned cluster, based on size ratio similarity. Similarity in direction ratio Composition. Calculation method: ; ; in, and These are the major axes of the equipment cluster MBR and the planning cluster MBR, respectively; and These are the rotation angles for the equipment cluster MBR and the planning cluster MBR, respectively.

[0065] Final shape similarity , and These are the weighting coefficients for size proportion similarity and orientation proportion similarity, respectively, satisfying... .

[0066] In further explanation, since the planned path of the power grid line has clear morphological characteristics, such as the major axis of the straight section being much larger than the minor axis, and the ratio of the major axis to the minor axis of the curved section being adapted to the turning radius, it is not necessary to consider the ratio of the minor axis.

[0067] S440: Correct the initial spatial offset distance based on shape similarity; If the initial spatial offset distance Less than or equal to the preset offset distance threshold, and If the shape similarity between the two device cluster MBRs and the planned cluster MBR is less than a preset shape similarity threshold, then the spatial offset distance is corrected. The correction method is as follows: ; in, This is the corrected spatial offset distance.

[0068] To avoid missing latent flying points, shape similarity is used to adjust the spatial offset distance. The adjusted spatial offset distance must be greater than the initial spatial offset distance. The result is less than 1, meaning the adjusted spatial offset distance is less than the initial spatial offset distance. Therefore, the formula is written as follows: The first 1 is the foundation, ensuring the overall result is increased. It is the adjustment amount. Because It is inversely proportional to the spatial offset distance. The smaller the value, the greater the morphological difference, and the greater the final spatial offset distance.

[0069] In this invention, shape similarity is introduced to verify the morphological matching degree between device clusters and planned clusters by using rotation angles and major axes, thereby avoiding the missed detection of hidden flying points that are close in center distance but have large morphological differences. This method simplifies calculations, reduces unnecessary complexity, and ensures detection accuracy.

[0070] S500: As one implementation method of the embodiment, detection of a flying point device is performed.

[0071] The detection method is as follows: if the spatial offset distance of the device cluster relative to the planned cluster... If the offset distance exceeds a preset threshold, and the topological connection data between the devices and the lines indicates that any two devices within the corresponding device cluster have a topological connection, then the device cluster is classified as a set of flying point-like devices. Specifically: S510: Spatial distance exceeding limit detection; First, not all lines use the same preset offset distance threshold. Instead, the preset offset distance threshold is obtained by looking up a preset offset distance mapping table based on the voltage level of the target line to be tested. For example, the threshold is 100m for 220kV, 80m for 110kV, and 50m for below 10kV.

[0072] Further explanation is provided: the voltage level of the target line under test is directly proportional to the preset offset distance threshold; that is, the higher the voltage level, the greater the allowable deviation distance. This is because a higher voltage level requires a larger safety and functional space for the line, allowing for greater allowable equipment construction errors and a wider service radius, thus increasing the upper limit of the reasonable equipment offset.

[0073] If a certain device cluster Spatial offset distance It is greater than the preset offset distance threshold corresponding to its line, that is If the spatial offset distance exceeds the limit, it indicates that the equipment has deviated from the planned path, meeting the conditions for abnormal location. If the spatial offset distance of the equipment cluster does not exceed the limit, then even if there are differences in shape, it is within the range of normal construction errors. Normal topology connection of equipment within the cluster indicates that the electrical connection between the equipment meets the design requirements and meets the conditions for normal function. If the topology is broken (such as a certain equipment not being connected to the line), it is a faulty equipment, not a flypoint (flypoints do not require equipment repair, only coordinate correction).

[0074] S520: Topology connection detection, including topology logical connectivity analysis and geographic logical connectivity analysis.

[0075] In a further implementation, based on the topology connection data of devices and lines obtained from the GIS system in S100, all devices and their topology connection relationships attached to the target line to be detected are extracted, including at least the device identifier, the attached line identifier, and the parent device identifier. Based on this data, a device topology connection graph is constructed. This topology connection graph is an undirected graph, with devices as nodes and direct connections between devices (based on the one-to-one correspondence between parent and child device identifiers) as edges.

[0076] First, a rapid screening of topological logical connectivity is performed. Based on the topological connection records of the power grid GIS system, it quickly determines whether all devices within a device cluster form a logically connected whole. This stage ignores spatial distance and only verifies the "existence" of connections, aiming to efficiently eliminate obviously disconnected device clusters.

[0077] Specifically: extraction equipment clusters All equipment and equipment identification in the system constitute The set of devices. In the device topology connection diagram, for Connectivity analysis is performed on the set of devices. Do all devices within belong to the same component? from We can choose any device from the set of devices as the starting point and use either Breadth-First Search (BFS) or Depth-First Search (DFS) to find all device nodes reachable from that starting point in the topology graph. Here, we set this as a set. If set Device nodes in If all device nodes in a device set are identical, then the device is determined to be a device cluster. All devices within the system are logically interconnected.

[0078] If the devices are clustered If the topological logical connectivity analysis passes, proceed to the next step, geographic logical connectivity analysis. Otherwise, determine the device cluster. Topologically disconnected, a set of non-flying points.

[0079] In a further implementation, geographic logical connectivity analysis is performed. Based on the topological logical connectivity analysis, the actual geographic distance between devices is introduced as a weight to verify... Whether the connectivity within is based on reasonable spatial proximity.

[0080] Based on the device topology connection graph constructed above, each edge is assigned a weight. This weight is determined by the actual geographical distance D between the two devices connected by the edge. Specifically: When the device and equipment Actual geographical location in a plane projection coordinate system Less than or equal to At that time, equipment and equipment The edge weights between them are ;in, For equipment The maximum distance threshold for the equipment type, which is the setting value based on the maximum reasonable spacing for normal installation of equipment of that type, is set to 1.2 times the maximum reasonable spacing in this invention; it needs to be emphasized again that the equipment and equipment They belong to the same type of equipment. When hour, .

[0081] When the two devices are very close together A value approximately equal to 1 indicates that the two devices are not only electrically connected but also spatially close, exhibiting the highest topological geographical correlation. As the distance between the two devices gradually increases... Follow The linear decrease in the value of indicates that although the two devices are still within a reasonable connection range, their spatial proximity decreases, and their topological geographical correlation declines accordingly. When the distance between the two devices exceeds 100 km / h... hour, The minimum weight indicates that although the two devices may have an electrical connection record, the spatial distance is far beyond the normal installation range of this type of device, which is likely an abnormal association.

[0082] Furthermore, after determining the weight of each edge in the device topology connection graph, a spatially weighted topology graph is generated, and device clusters are extracted from it. Weighted subgraph ;exist Run the maximum spanning tree algorithm (Kruskal or Prim) on it.

[0083] If the generated maximum tree contains a weighted subgraph If all device nodes are identified, then the device cluster is determined. Through geographical logical connectivity analysis, then All devices in the cluster belong to the same connected component, thus confirming that there is a topological connection between any two devices in the cluster. The generated maximum spanning tree failed to contain a weighted subgraph. If all device nodes in the graph are divided into multiple connected components, the cluster is deemed to fail the verification, thus determining that at least two devices in the device cluster do not have a topological connection relationship.

[0084] It should be noted that the topology data of the power grid GIS system may contain redundant connection records (such as incorrectly entered cross-regional equipment associations). This can lead to edges in the equipment subgraph that are electrically connected but geographically far apart (for example, equipment A is in the east of the city, and equipment B is in the west of the city, but are mistakenly recorded as directly connected). If we do not consider weights and determine connectivity solely based on whether a tree can be generated, we will misclassify geographically dispersed equipment clusters connected only by incorrect topology records as locally concentrated fly-like points, violating the core definition of spatial clustering of fly-like points. The highest total weight means that the edges selected for the spanning tree are those that are close in distance and have high weights, ensuring that connected equipment is not only electrically connected but also highly concentrated spatially, which conforms to the characteristic of fly-like points that are abnormally located but locally concentrated. If there are false connection edges (far away and low weight), these edges will be automatically discarded when generating the tree with the highest weight; if discarding them does not include all nodes, the tree will be determined to be disconnected, thus eliminating the erroneous judgment of false connectivity.

[0085] This embodiment uses inverse spatial distance weighting to allow topology analysis to consider both electrical connectivity and geographical proximity, making it more suitable for flying point detection scenarios. Furthermore, combining electrical topology and geographical distribution enhances robustness in complex power grid environments.

[0086] In a further approach, when devices are clustered... Spatial offset distance relative to the nearest planned cluster If the offset distance is greater than a preset threshold, and there is a topological connection between any two devices within the cluster, then the devices are clustered into a group. It was identified as a set of flying point devices.

[0087] refer to Figure 2 , Figure 2 This is a block diagram of a power grid-type flying point detection system based on intelligent clustering.

[0088] As an embodiment of the present invention, a power grid-type flying point detection system based on intelligent clustering is disclosed. Employing the specific implementation method described above for power grid-type flying point detection, the system includes: The data extraction unit extracts the associated data of the target line to be detected from the power grid GIS system, including the planned path coordinate data, the actual geographic coordinate data of all equipment, and the topological connection data between the equipment and the line; and converts the planned path coordinate data and the actual geographic coordinate data to a plane coordinate system. The transformation unit, with a preset window length and step size, performs sliding window clustering on the planned path data to generate multiple initial planned clusters of the target route to be detected; for the initial planned clusters in adjacent windows, it merges them according to the overlap of data points within the clusters to output multiple planned clusters. The clustering unit groups the actual geographic coordinate data of devices of the same type by device type, and then clusters them according to the clustering radius and minimum number of devices corresponding to that device type in the current scenario. The output contains only multiple spatially adjacent device clusters of the same type. The distance calculation unit obtains the spatial offset distance between each device cluster and its nearest planned cluster, including obtaining the distance between the center point of each device cluster and the center point of each planned cluster, and corrects the initial spatial offset distance based on shape similarity. If a device cluster is determined to be a set of flying point devices, and the spatial offset distance is greater than a preset offset distance threshold corresponding to the voltage level of the target line to be detected, and any two devices in the device cluster have a topological connection relationship, then the device cluster is determined to be a set of flying point devices.

[0089] As an embodiment of the present invention, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is loaded onto the processor, it employs the specific implementation method described above for the power grid-type flying point detection method.

[0090] As an embodiment of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing a computer program, which, when executed by a processor, adopts the specific implementation method described above for the power grid flying point detection method.

[0091] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A method for detecting power grid-type flying points based on intelligent clustering, characterized in that, include: Extract the associated data of the target line to be detected from the power grid GIS system, including the planned path coordinate data, the actual geographic coordinate data of all equipment, and the topological connection data between the equipment and the line; Convert the planned path coordinate data and the actual geographic coordinate data to a plane coordinate system; The window length and step size are preset, and sliding window clustering is performed on the planned path data to generate multiple initial planned clusters of the target route to be detected; For the initial planned clusters in adjacent windows, merge them according to the overlap of data points within the clusters to output multiple planned clusters; For the actual geographic coordinate data of the same type of equipment, after grouping by equipment type, cluster them according to the cluster radius and minimum number of equipment in the current scenario, and output only multiple equipment clusters that are of the same type and spatially adjacent; Obtain the spatial offset distance between each device cluster and its nearest planned cluster, including obtaining the distance between the center point of each device cluster and the center point of each planned cluster, and correct the initial spatial offset distance based on shape similarity; If the spatial offset distance of a device cluster is greater than the preset offset distance threshold corresponding to the voltage level of the target line to be detected, and there is a topological connection between any two devices in the device cluster, then the device cluster is determined to be a set of flying point devices. The process for determining whether a device cluster is a set of flying point devices includes: if the spatial offset distance of the device cluster is greater than a preset offset distance threshold corresponding to the voltage level of the target line to be detected, and there is a topological connection between any two devices in the device cluster, then the device cluster is determined to be a set of flying point devices. The process for determining the topological connection relationship includes topological logical connectivity analysis and geographic logical connectivity analysis, and the steps include: Construct a topology graph with devices as nodes and connection relationships as edges; The process of topological logical connectivity analysis is as follows: for device clusters Extract all devices and their identifiers to form a complete system. The collection of devices; from Choose any device from the set of devices as the starting point, find all device nodes that can be reached from that starting point in the topology graph, and store them in the set. If set Device nodes in If all device nodes in a device set are identical, then the device is determined to be a device cluster. All devices within the system are logically interconnected in terms of topology; Determine the device cluster Once all devices within the network are logically connected to each other in terms of topology, a geographic logical connectivity analysis is then performed.

2. The method for detecting power grid-type flying points based on intelligent clustering according to claim 1, characterized in that, Preset window length and step size, including: For the target route to be detected, based on the total length of the planned path Set window length to Step size is Calculate the path curvature based on the sequence of path points along the planned path; If the path curvature is greater than the preset curvature threshold, it is determined to be a curved segment, and the window length of the curved segment is adjusted to... Step size adjusted to ; and These are the window length adjustment factor and the step size adjustment factor, respectively. and All are greater than 1; If the path curvature is less than or equal to the preset curvature threshold, it is determined to be a straight line segment, and the window length is maintained. , length is .

3. The method for detecting power grid flying points based on intelligent clustering according to claim 1, characterized in that, The steps for obtaining the planned clusters include: For any two adjacent windows, take one of the initial planned clusters from each window to form a cluster pair, and calculate the overlap of the cluster pair. When the overlap is greater than the preset overlap threshold, the cluster pair is merged into a single planned cluster. The merged planned cluster includes all coordinate points from the two initial planned clusters. Merge the initial planned clusters in any two adjacent windows based on their overlap; traverse all adjacent sliding windows to output multiple consecutive planned clusters.

4. The method for detecting power grid-type flying points based on intelligent clustering according to claim 1, characterized in that, Grouping the actual geographic coordinate data of similar equipment; including: Obtain the actual geographic coordinates and device types of all devices within the normal clustering area; The actual geographic coordinate data is grouped according to the device type to form several device groups; each device type corresponds to one device group. For each device group, clustering is performed based on the corresponding cluster radius and minimum number of devices, resulting in multiple device clusters.

5. The method for detecting power grid flying points based on intelligent clustering according to claim 4, characterized in that, The process of determining the cluster radius and minimum number of devices for each device group includes: exist For any area where similar equipment is normally clustered, Calculate the distance between any two devices. The maximum device distance in the range is denoted as Then calculate all The mean of the maximum distances , calculate Multiply by safety factor Obtain the cluster radius; exist For each normal cluster of similar equipment, obtain the number of equipment in each cluster. Count the number A of normal clusters where the number of equipment falls within the interval [c,d]. If the ratio of A to M exceeds 95%, then the lower limit c is taken as the minimum number of equipment.

6. The method for detecting power grid flying points based on intelligent clustering according to claim 1, characterized in that, The process of determining the spatial offset distance between each device cluster and its nearest planned cluster includes: Obtain the coordinates of the center points of the device clusters and the planned clusters. The coordinates of the center points are the average x-coordinates and average y-coordinates of all device coordinate points or planned coordinate points within the cluster. For each device cluster, iterate through the distances between it and all planned clusters, and take the minimum distance as the initial spatial offset distance between the device cluster and its nearest planned cluster; Determine the minimum bounding rectangle of the equipment cluster and the minimum bounding rectangle of the nearest planned cluster of the equipment cluster, and calculate the shape similarity based on the minimum bounding rectangles of the two clusters; For each device cluster, if its initial spatial offset distance is not greater than a preset offset distance threshold and its shape similarity is less than a preset shape similarity threshold, then the initial spatial offset distance is corrected. The shape similarity is a weighted sum of the size ratio similarity and the orientation ratio similarity.

7. The method for detecting power grid flying points based on intelligent clustering according to claim 1, characterized in that, The geographic logical connectivity analysis further identifies the device clusters as a set of flying point-like devices, including: Each edge of the topology graph is assigned a weight to generate a spatially weighted topology graph, from which device clusters are extracted. Weighted subgraph ;exist Run the maximum spanning tree algorithm on it; If the generated maximum tree contains a weighted subgraph If all device nodes are identified, then the device cluster is determined. Through geographical logical connectivity analysis, then All devices in the cluster belong to the same connected component, thus confirming that there is a topological connection between any two devices in the cluster. The edge weights are determined as follows: Equipment at both ends of the side and equipment Actual geographic distance in a plane projection coordinate system ,when hour, ,otherwise ;in, For equipment The maximum distance threshold for the device type.

8. A power grid-type flying point detection system based on intelligent clustering, executing the power grid-type flying point detection method as described in any one of claims 1-7, characterized in that, The system includes: The data extraction unit extracts the associated data of the target line to be detected from the power grid GIS system, including the planned path coordinate data, the actual geographic coordinate data of all equipment, and the topological connection data between the equipment and the line; and converts the planned path coordinate data and the actual geographic coordinate data to a plane coordinate system. The transformation unit, with a preset window length and step size, performs sliding window clustering on the planned path data to generate multiple initial planned clusters of the target route to be detected; for the initial planned clusters in adjacent windows, it merges them according to the overlap of data points within the clusters to output multiple planned clusters. The clustering unit groups the actual geographic coordinate data of devices of the same type by device type, and then clusters them according to the clustering radius and minimum number of devices corresponding to that device type in the current scenario. The output contains only multiple spatially adjacent device clusters of the same type. The distance calculation unit obtains the spatial offset distance between each device cluster and its nearest planned cluster, including obtaining the distance between the center point of each device cluster and the center point of each planned cluster, and corrects the initial spatial offset distance based on shape similarity. If a device cluster is determined to be a set of flying point devices, and the spatial offset distance is greater than a preset offset distance threshold corresponding to the voltage level of the target line to be detected, and any two devices in the device cluster have a topological connection relationship, then the device cluster is determined to be a set of flying point devices.

Images