A risk prevention and control method and system for power transmission site proximity electric operation

Abstract

The application discloses a kind of for the risk prevention and control method and system of power transmission site adjacent electric operation, it is related to power transmission line field technical field.The method steps include: by laser radar scanning the point cloud data of power transmission site and pre-processing, obtain the point cloud data of power transmission site;Through the ground point cloud in the improved progressive encryption triangle net algorithm of point cloud data of power transmission site, obtain the non-ground point cloud data of power transmission site;Through density clustering method, the non-ground point cloud data of power transmission site is clustered, obtains charged body point cloud data and adjacent electric operation body point cloud data;Through the nearest neighbor method based on variable precision weighting, the minimum distance search is carried out to charged body point cloud data and adjacent electric operation body point cloud data, obtains adjacent electric operation risk value, and risk grade is divided based on adjacent electric operation risk value and early warning is carried out, to realize the risk prevention and control of adjacent electric operation.

Description

Technical Field

[0001] This invention relates to the field of power transmission field technology, specifically to a risk prevention and control method and system for near-electricity work at power transmission sites. Background Technology

[0002] In the operation and maintenance of power transmission lines, hoisting and installation work often needs to be carried out near live equipment; this is known as near-electricity work. The core safety risk of this type of work lies in the insufficient spatial distance between the working machinery, personnel, or tools and the live conductors, which may lead to serious accidents such as discharge and electric shock. Therefore, achieving real-time and accurate monitoring and risk warning of the minimum distance between the work object and the live conductor is a key technical requirement for ensuring the safety of power transmission site operations.

[0003] In risk control, traditional methods first acquire 3D point cloud data of the work environment and distinguish between the point cloud of energized equipment and the point cloud of the work body. Then, the Euclidean distance from each point in the work body point cloud to all points in the energized equipment point cloud is calculated, and the minimum value is taken as the current closest distance between the two. This closest distance is directly used as the benchmark value for risk assessment, and by comparing it with a preset safety threshold, the safety of the current work status is determined.

[0004] However, traditional methods directly use the minimum Euclidean distance as the risk value when calculating the nearest neighbor. This fails to effectively distinguish whether the nearest neighbor represents a real charged surface or accidental measurement noise or a tiny floating object. Consequently, single-point noise can cause abnormal abrupt changes in the risk value, triggering false alarms. Ultimately, this reduces the reliability of risk warnings. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a risk prevention and control method and system for near-electricity work at power transmission sites, thereby resolving the problems existing in the background technology.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a risk prevention and control method and system for near-electricity work at power transmission sites, comprising the following steps: Step S1: Scan the point cloud data of the power transmission site using lidar and perform preprocessing to obtain the point cloud data of the power transmission site; Step S2: Remove the ground point cloud from the power transmission site point cloud data using the improved progressive encryption triangular mesh algorithm to obtain the non-ground point cloud data of the power transmission site; Step S3: Cluster the non-ground point cloud data of the power transmission site using density clustering to obtain point cloud data of energized bodies and point cloud data of adjacent power operation bodies; Step S4: By using the nearest neighbor method based on variable precision weighting, perform minimum distance search on the point cloud data of the charged body and the point cloud data of the adjacent power operation body to obtain the risk value of the adjacent power operation. Based on the risk value of the adjacent power operation, classify the risk level and issue an early warning, thereby realizing risk prevention and control of the adjacent power operation.

[0007] Preferably, the step of scanning the point cloud data of the power transmission site with lidar and preprocessing it to obtain the point cloud data of the power transmission site includes the following specific steps: In the power transmission site operation area, the lidar is fixedly installed at the center high point of the rotating platform of the operating machinery, and the external parameters of the lidar are calibrated. The lidar scans the power transmission site operation area to obtain point cloud data of the power transmission site operation area, and the point cloud data is preprocessed to obtain the power transmission site point cloud data.

[0008] Preferably, the step of removing ground point cloud data from the power transmission site point cloud data using the improved progressive encryption triangulation algorithm to obtain non-ground point cloud data of the power transmission site includes the following steps: The power transmission site point cloud data is a set containing N data points. , , where the i-th point cloud Includes three-dimensional spatial coordinates, Represents the x-coordinate of the i-th point cloud. Represents the ordinate of the i-th point cloud, Represents the vertical coordinate of the i-th point cloud; Calculate the elevation tolerance threshold, and obtain a candidate point set based on the elevation tolerance threshold; perform density clustering based on Euclidean distance on the candidate point set to obtain an initial ground point seed set; Based on the initial ground point seed set, the vertical distance and maximum angle are calculated; the power transmission site point cloud data is then filtered and removed based on the vertical distance and maximum angle to obtain the non-ground point cloud data of the power transmission site.

[0009] Preferably, the calculation of the elevation tolerance threshold includes the following steps: Calculate the minimum point cloud elevation of power transmission site point cloud data. , = Then calculate the elevation standard deviation of the point cloud data at the power transmission site. Finally, the elevation tolerance threshold is calculated. , , This is the tolerance coefficient.

[0010] Preferably, the step of calculating the vertical distance and the maximum included angle based on the initial ground point seed set includes the following steps: Based on the initial ground point seed set, a triangulation network is constructed; for unclassified points in the power transmission site point cloud data. ( , , ), with points ( , , Using a point as the center, divide the 3D search area, and filter the triangular faces within the 3D search area in the triangular mesh to form a candidate triangular face set; calculate the points. ( , , ) to each triangle in the candidate triangle set vertical distance Simultaneously calculate points ( , , Projected onto the triangular facet The projection point and the triangular facet Maximum angle between vertices : ; ; in, Vertical distance The maximum included angle, , , Points ( , , The x-coordinate, y-coordinate, and y-coordinate of triangle A, B, C, and D are given. The coefficients of the plane equation, and For the projection point and the triangular facet The vector formed by the vertices, with max() being the function to find the maximum value.

[0011] Preferably, the step of clustering the non-ground point cloud data of the power transmission site using density clustering to obtain the point cloud data of the energized body and the point cloud data of the adjacent power operation body includes the following steps: Non-ground point cloud data of power transmission sites , ={ }, M is, where each point ( , , It includes three-dimensional spatial coordinates; The neighborhood radius of density clustering is The minimum number of points is MinPts; for the j-th point in the non-ground point cloud data of the power transmission site Calculation points neighborhood set Based on neighborhood set Density-reachable clustering is performed, ultimately yielding K clusters and a set of noise points; The K clusters and noise point sets are divided to obtain the point cloud data of the charged body and the point cloud data of the adjacent electrical work body.

[0012] Preferably, the process of dividing the K clusters and the noise point set includes the following specific steps: The feature index of each cluster is extracted and compared with a preset classification threshold to divide the cluster into charged body point cloud and nearby electrical work body point cloud, and the noise point set is classified into the nearby electrical work body point cloud.

[0013] Preferably, the step of obtaining the near-field operation risk value by performing a minimum distance search on the point cloud data of the charged body and the point cloud data of the adjacent power operation body using a variable-precision weighted nearest neighbor method is as follows: Construct a KD-tree, and perform a range search based on the KD-tree to obtain the set of nearest neighbors; Calculate the final weight of each nearest neighbor in the nearest neighbor set; The weighted average of the Euclidean distances of all nearest neighbors is calculated based on the final weights to obtain the weighted distance; the minimum value of the weighted distance is taken as the risk value of nearby electrical work.

[0014] Preferably, the calculation of the final weight of each nearest neighbor in the nearest neighbor set specifically involves: ; in, Indicates the first The final weight of the nearest neighbor points, The number of nearest neighbors. For nearest neighbor index, For variable precision parameters.

[0015] A risk control system for near-electricity work at power transmission sites includes a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the above method.

[0016] This invention provides a risk prevention and control method for near-electricity work at power transmission sites, involving machine learning and deep learning technologies, which has the following beneficial effects: (1) The improved progressive encryption triangulation algorithm has shown significant benefits in the processing of point cloud data at power transmission sites. This algorithm accurately filters the initial ground seed set by dynamically calculating the elevation tolerance threshold and combining it with Euclidean distance clustering, effectively eliminating interference from isolated small-sized non-ground objects such as gravel or low vegetation. In the iterative encryption stage, the algorithm adaptively adjusts the distance and angle thresholds based on the standard deviation of point cloud elevation and introduces local terrain continuity verification to ensure reliable differentiation between ground and non-ground points under both flat and undulating terrain. This series of improvements significantly enhances the accuracy and robustness of ground point cloud removal, laying a clean and high-quality non-ground point cloud data foundation for the subsequent accurate classification of energized bodies and operational objects.

[0017] (2) The nearest neighbor method provides an efficient computational framework for minimum distance search by constructing a KD-tree spatial index structure. The KD-tree recursively partitions the space based on the variance of point cloud coordinates to form a hierarchical binary tree, which significantly optimizes the nearest neighbor search process from nearby electrical work points to the point cloud of the energized body. The best-first search algorithm is adopted to prioritize traversing the subspace with the closest distance and quickly locate the K nearest neighbor points, which greatly improves the search speed, meets the stringent requirements of real-time risk monitoring at the power transmission site for processing efficiency, and ensures that the system can respond to the dynamically changing working environment in a timely manner.

[0018] (3) The nearest neighbor method based on variable precision weighting further enhances the robustness and accuracy of risk value calculation. This method, based on the traditional inverse square distance weighting, introduces a variable precision parameter β to normalize the weights, appropriately diluting the dominant influence of a single nearest neighbor point through a controllable smoothing term. This flexible weighting mechanism effectively suppresses potential misjudgments of risk values ​​caused by measurement noise or abnormal nearest neighbors in point cloud data, improving the algorithm's anti-interference capability in complex real-world operating environments. Finally, a stable and reliable risk value is obtained through weighted averaging, and combined with multi-level threshold triggering for precise early warning, achieving closed-loop risk control from monitoring to intervention, ensuring the safe operation of near-electricity work. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart of the steps of a risk prevention and control method for near-electricity work at a power transmission site proposed in this invention; Figure 2This is a step hierarchy diagram of obtaining point cloud data of energized bodies and point cloud data of adjacent power operation bodies in a risk prevention and control method for near-electricity operation at power transmission sites proposed in this invention; Figure 3 This is a step hierarchy diagram of obtaining the risk value of near-electricity work in a risk prevention and control method for near-electricity work at power transmission sites proposed in this invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Please see Figures 1-3 The present invention provides a technical solution: a risk prevention and control method for near-electricity work at power transmission sites.

[0023] Step S1: Scan the point cloud data of the power transmission site using lidar and perform preprocessing to obtain the point cloud data of the power transmission site.

[0024] In the power transmission site operation area, a pulsed lidar is fixedly installed at the center high point of the rotating platform of the operating machinery. The three-dimensional coordinates of this installation point are determined by measuring with a total station. The installation plane is adjusted to a horizontal state, and its levelness is calibrated by a total station or a high-precision electronic level to ensure that the scanning reference plane is parallel to the horizontal reference plane of the set operation area. After installation, the lidar is calibrated for external parameters. With the machinery stationary, at least four calibration target balls with known relative positions are arranged around the operating machinery (the three-dimensional coordinates of the center of each target ball in the same world coordinate system are accurately measured and recorded in advance using a total station). The machinery is controlled to maintain a stable attitude. The lidar scans the target ball group to obtain the point cloud coordinates of the target ball centers in the lidar coordinate system. Combined with the world coordinate system coordinates obtained by the target balls in advance, the least squares method is used to fit and solve the rotation matrix and translation vector from the lidar coordinate system to the world coordinate system, thus completing coordinate system one. The scanning parameters of the lidar can be set to a scanning frequency of 20 Hz, a horizontal viewing angle of 360 degrees, a vertical viewing angle of 30 degrees, an angular resolution of 0.1 degrees, a maximum ranging distance of 200 meters, and a ranging accuracy of ±2 centimeters. These parameters ensure that the spatial position information of key equipment such as power transmission lines, towers, and insulator strings can be effectively captured during continuous scanning. The scanning parameters of the lidar can be set according to the actual situation and are not limited here.

[0025] During the scanning operation, the lidar emits laser beams at a fixed frequency and receives the echoes, acquiring raw point cloud data containing environmental features of the power transmission site in real time. Each data point includes three-dimensional coordinates, intensity information, and a timestamp. The raw point cloud data obtained from the scan is transmitted via a gigabit Ethernet interface. The transport layer uses the TCP protocol to ensure reliable data delivery, and the application layer data packet format follows the LAS 1.4 standard. Both together ensure the integrity of the point cloud data and its correct parsing by subsequent processes.

[0026] The received raw point cloud data undergoes a preprocessing process. First, coordinate transformation is performed using the calibration-derived rotation matrix and translation vector to transform the point cloud from the local coordinate system of the lidar to a unified operating machinery coordinate system with the machine's rotation center as the origin and the north-east-ground (ENU) direction as the coordinate axes, thus unifying the spatial reference of all point cloud data. Next, a statistical filtering algorithm is used to remove outlier noise points, setting the neighborhood number to 20 and the standard deviation factor to 1.5 to filter out discrete points caused by measurement noise and random errors. Then, a voxel grid filter is used to uniformly downsample the point cloud, setting the voxel side length to 0.1 meters, thereby significantly reducing the total data volume while maintaining the macroscopic geometric structure of the scene. Finally, the point cloud data undergoes temporal alignment and sorting to ensure that subsequent analysis can proceed according to the correct time sequence, with inter-frame temporal accuracy controlled within 10 milliseconds. After the above preprocessing steps, clean, lightweight, and spatiotemporally unified power transmission site point cloud data is output as input for step S2.

[0027] Step S2: Remove the ground point cloud from the power transmission site point cloud data using the improved progressive encryption triangular network algorithm to obtain the non-ground point cloud data of the power transmission site.

[0028] After completing the lidar data acquisition and preprocessing in step S1 to obtain power transmission site point cloud data with a unified spatiotemporal reference, the process proceeds to step S2, the ground point cloud removal process. This step uses the complete point cloud data output from the previous step as direct input. Its goal is to accurately separate the ground points representing the terrain from the ground point cloud containing key targets such as power equipment and operating machinery through an improved progressively encrypted triangulation algorithm, thereby extracting the non-ground point cloud data of the power transmission site necessary for subsequent risk identification.

[0029] The specific implementation process of step S2 is as follows: The power transmission site point cloud data is a set containing N data points. , , where the i-th point cloud Includes three-dimensional spatial coordinates, Represents the x-coordinate of the i-th point cloud. Represents the ordinate of the i-th point cloud, This represents the vertical coordinate of the i-th point cloud.

[0030] The improved progressive encrypted triangulation algorithm separates ground points from non-ground points through iterative encrypted triangulation: first, it calculates the minimum elevation of the point cloud. , = Then calculate the elevation standard deviation of the point cloud data at the power transmission site. Calculate the elevation tolerance threshold. , , This is the tolerance factor, with a default value of 0.1. This is used to initially filter out elevation values ​​that meet the requirements. To eliminate potentially isolated small-sized non-ground object points (such as scattered gravel or insulator fragments) from the candidate point set, density clustering based on Euclidean distance is performed (e.g., neighborhood radius of 0.3m, minimum number of points 15). Clusters with fewer than 15 points after clustering are considered noise and removed, ultimately yielding a purified initial ground point seed set S, where S = { | ≤ + And the number of cluster points to which it belongs is ≥15}.

[0031] It should be noted that the elevation standard deviation is used. Calculate the elevation tolerance threshold By combining Euclidean distance clustering to eliminate small clusters of pseudo-ground points, this method is adapted to scenarios with gentle terrain and occasional isolated obstacles in power transmission sites, thereby improving the accuracy of the initial ground seed set and the elevation standard deviation. Reflecting the dispersion of point cloud elevations at power transmission sites, the elevation distribution of actual ground points exhibits strong continuity. (usually smaller) Using values ​​of 0.1 or 0.1 times the standard deviation can accurately define the "elevation very close" range. "Dense point clusters" are used to avoid including false ground points such as gravel and low vegetation due to excessively wide thresholds, or missing real ground points in gentle slope areas due to excessively narrow thresholds.

[0032] An initial triangulation network is constructed using the Delaunay triangulation algorithm based on the initial ground point seed set S. This triangulation network consists of a series of triangular patches. Each triangular facet is composed of a plane formed by three seed points, and its plane equation coefficients are (A, B, C, D).

[0033] The process then proceeds to the iterative encryption phase, which addresses unclassified points in the point cloud. ( , , To improve computational efficiency, point-based... ( , , Using x as the center, divide a three-dimensional search range. For example, take x∈[ -1, +1],y∈[ -1, +1]、z∈[ -0.5, Within a spatial range of +0.5, select triangular faces within this range in the current triangulation network to form a candidate triangular face set (reducing invalid computation); calculate unclassified points in the point cloud. ( , , ) to each triangle in the candidate triangle set vertical distance At the same time, count the unclassified points. ( , , Projected onto the triangular facet The projection point and the triangular facet Maximum angle between vertices : ; ; in, Vertical distance The maximum included angle, , , Points ( , , The x-coordinate, y-coordinate, and y-coordinate of triangle A, B, C, and D are given. The coefficients of the plane equation, and For the projection point and the triangular facet A vector formed by vertices.

[0034] Distance threshold Dynamically adjusted based on the standard deviation of point cloud elevation: =0.2× , The elevation standard deviation of the point cloud at the power transmission site is given, and the range of values ​​is limited to 0.3m ≤ ≤0.5m (e.g., leveling work area) Small, Take a 0.3m section and precisely remove low-lying vegetation; gentle slope area big, (Take 0.5m to adapt to terrain undulations).

[0035] Angle threshold =10°-0.5× , To achieve more stringent angle determination as the terrain becomes flatter, the standard deviation of the plane coordinates calculated based on the currently classified ground points is used.

[0036] To satisfy ≤ and ≤ point ( , , If the geometric features match the ground points, then the local terrain continuity verification is initiated. ( , , Given a neighborhood set of points (neighborhood radius 0.3m), calculate the standard deviation of elevation for all points within that neighborhood. (Reflecting the continuity of local terrain); if If the elevation difference is greater than 0.2m (a typical characteristic of low vegetation / gravel at power transmission sites: sudden changes in local elevation), it is determined to be a non-ground point and excluded; if Points ≤0.2m (with locally gentle terrain, conforming to ground characteristics) are classified as ground points; points classified as ground points are... ( , , Add the new vertex set to the new triangulation set, and re-execute the Delaunay triangulation algorithm on the updated vertex set to generate a new triangulation to replace the original one. Record the number of new ground points added in each iteration. ,when The iteration terminates when the value is less than 0.001. The final ground point cloud data of the power transmission site is obtained by removing the ground point cloud data from the power transmission site point cloud data.

[0037] After improving the progressive encryption triangulation algorithm, the non-ground point cloud data of the power transmission site was successfully separated. This dataset contains point clouds of all objects above the ground extracted from the original point cloud. However, the obtained non-ground point cloud is still a mixture of various ground targets, including both live equipment requiring key monitoring (such as conductors, insulators, and towers) and moving targets requiring early warning, such as operating machinery and construction personnel. This mixed state of non-ground point cloud data cannot be directly used for accurate distance calculation and risk assessment because different types of objects have different safety distance requirements and risk levels. To achieve precise prevention and control of risks associated with nearby power operations, further refined classification processing of this non-ground point cloud is needed.

[0038] Step S3: Cluster the non-ground point cloud data of the power transmission site using density clustering to obtain point cloud data of energized bodies and point cloud data of adjacent power operation bodies.

[0039] Non-ground point cloud data of power transmission sites Given a set containing M point cloud points, ={ }, where each point =( , , It contains three-dimensional spatial coordinates.

[0040] To classify the mixed non-ground point cloud into two categories—live equipment and nearby electrical work sites—a density-based spatial clustering algorithm is used to analyze the non-ground point cloud data from the power transmission site. For initial segmentation, the core parameters of the algorithm include the neighborhood radius. The minimum number of points, MinPts, can be pre-calibrated using point cloud density (e.g., by setting the neighborhood radius). The value is 0.3m, and the minimum number of points (MinPts) is 10.

[0041] It should be noted that the term "nearby electrical work object" specifically refers to a moving target that requires real-time monitoring and early warning and is operating near energized equipment at a power transmission site. Its corresponding point cloud data (i.e., near-electrical work object point cloud data) mainly includes the three-dimensional spatial location information of the working machinery (such as cranes and insulated bucket trucks), construction personnel, tools, and other temporary obstacles that may intrude into the safety distance. The term "energized equipment" specifically refers to static facilities with high voltage at the power transmission site that require maintaining a safe distance. Its corresponding point cloud data (i.e., energized body point cloud data) mainly includes the three-dimensional spatial location and structural information of fixed power facilities such as conductors, insulator strings, towers, busbars, and instrument transformers. This concept is defined relative to the dynamic "nearby electrical work object" and serves as the hazard source benchmark in risk assessment.

[0042] For non-ground point cloud data of power transmission sites Each point in Calculate its neighborhood radius as neighborhood set When | If |≥MinPts, then mark it. As the core point; when | | <MinPts, but core points exist Make ∈ Then mark If it is a boundary point, otherwise mark it. This is a noise point.

[0043] Clustering relationships are established using density reachability: Starting from any unvisited core point, all points density-reachable from the core point are identified to form clusters. This process is iterated until all points are visited, resulting in K clusters. , ,..., ,..., } and noise point set Noise, Let represent the k-th cluster. For each cluster... Calculate its characteristic indicators, including the number of points in the cluster. Point density = (in For clusters The volume of the axis-aligned bounding box is calculated by multiplying the differences between the maximum and minimum values ​​of points within the cluster along the x, y, and z dimensions. The principal axis direction vector is also considered. (Construct a covariance matrix by taking the coordinates of all points within the cluster, perform eigenvalue decomposition on the covariance matrix, and take the unit eigenvector corresponding to the largest eigenvalue as the principal axis direction vector).

[0044] Based on prior knowledge, classification thresholds are set: Electrified equipment mainly exhibits two typical forms: "vertical towers" and "horizontal conductors," and has a large spatial scale and high point cloud density. Therefore, three classification thresholds are preset, including a point quantity threshold. Point density threshold and geometric feature threshold Specifically, the number of points threshold A point density threshold of 500 to 1000 can be used (the point cloud size of transmission towers and conductors is significantly larger than that of targets such as workers and tools). A density of 80 to 120 points / cubic meter can be used (for equipment with flat surfaces and dense point cloud distribution, the point cloud density of operating machinery / personnel is typically ≤60 points / cubic meter), geometric feature threshold. Including vertical angle threshold Threshold of the angle between the horizontal direction and the vertical direction Vertical angle threshold A value of 3° can be used as the threshold for the horizontal angle. A deviation of 1.5° can be taken (usually, the verticality deviation of the tower is ≤3°, and the horizontality deviation of the conductor is ≤1.5°).

[0045] If clusters satisfy ≥ and ≥ And the principal axis direction vector The angle with the vertical direction (positive z-axis) ≤ or principal axis direction vector The angle with the horizontal direction (positive x-axis) ≤ Then the cluster will be Classified as charged body point cloud If clusters If the above conditions are not met, it is classified as a point cloud of nearby electrical work bodies. Furthermore, the noise point set during the clustering process is also classified into the point cloud of the neighboring electrical work area. (To avoid noise points interfering with subsequent risk calculations), the final result is the point cloud data of the charged body and the point cloud data of the adjacent electrical work body.

[0046] It should be noted that classifying noise points into the nearby power work point cloud is a conservative approach based on the integrity and safety design of risk prevention and control for nearby power work at the transmission site. From a data logic perspective, the non-ground point cloud needs to achieve full classification coverage. Discarding noise points individually would lead to data attribution breaks, failing to meet the core requirements of the technical solution: "reproducibility and unambiguity." From a risk prevention and control perspective, although the vast majority of noise points are useless targets such as lidar measurement noise and isolated debris, it is impossible to completely exclude tiny work-related targets (such as small tools in the hands of workers or insulator fragments) that are few in number in the point cloud. Classifying them into the nearby power work point cloud avoids risk omissions caused by direct discarding. At the same time, the isolated nature of noise points means that in the subsequent minimum distance search, they will either be far from the energized body, causing the distance to exceed the safety threshold, or their weight will be so small that they will not affect the final risk value calculation, thus avoiding false alarms. This ensures both the closed loop of the classification logic and the accuracy of risk prevention and control.

[0047] Step S4: By using the nearest neighbor method based on variable precision weighting, perform minimum distance search on the point cloud data of the charged body and the point cloud data of the adjacent power operation body to obtain the risk value of the adjacent power operation. Based on the risk value of the adjacent power operation, classify the risk level and issue an early warning, thereby realizing risk prevention and control of the adjacent power operation.

[0048] Charged body point cloud data Given a set containing H data points, ={ Point cloud data of nearby electrical work sites. Let H' be a set containing H' data points. ={ }, where each point and It includes three-dimensional spatial coordinates. The goal of the variable-precision weighted nearest neighbor method is to search for... Each point in arrive Midpoint cloud The nearest neighbor ( The value range is [5, 10], which is adapted to the point cloud density and risk calculation accuracy at the power transmission site. It can be adjusted according to the specific scenario (no limit is made here), and is based on... The weighted shortest distance is calculated from the distances of the nearest neighbors and used as the risk value.

[0049] To accelerate the processing of point cloud data of nearby power operation sites The nearest neighbor search for each point in the charged body point cloud data requires... Construct a KD-tree spatial index structure. The KD-tree construction process is as follows: using the entire charged body point cloud data... As the current point set, select the dimension with the largest variance (the variance of coordinates in the x, y, and z dimensions) as the splitting dimension. Calculate the median of all points along this splitting dimension as the splitting value. Divide the current point set into two subsets according to the splitting dimension: points whose splitting dimension coordinates are less than or equal to the splitting value are assigned to the left subset, and points whose splitting dimension coordinates are greater than the splitting value are assigned to the right subset. Use the left and right subsets as new current point sets, and recursively execute the process of "selecting splitting dimension → calculating splitting value → dividing subsets" until the number of points in each subset is less than a preset threshold. ( The range can be [5,10]. When the KD tree's search efficiency and accuracy are balanced, the splitting stops, and the current subset is taken as the leaf node of the KD tree. This construction process divides the three-dimensional space hierarchically as the KD tree is generated, eventually forming a binary tree structure. The internal nodes of the KD tree store the splitting dimension and splitting value of the corresponding level, while the leaf nodes store the point set information of their corresponding spatial regions.

[0050] Point cloud data of nearby electrical work sites Each point in The constructed KD-tree is used for nearest neighbor search. The search process employs a best-first search algorithm: for a single query point... First, initialize an empty current nearest neighbor set and a priority queue, enqueuing the root node of the KD tree. Retrieve nodes from the queue and begin traversal. If a node is internal, calculate the distance from the query point to its corresponding splitting hyperplane, prioritizing the search for subspaces with closer distances. Repeat this splitting hyperplane distance calculation and subspace selection operation for each internal node. Maintain the priority queue to store nodes to be searched, sorted in ascending order by the minimum possible distance between the node's corresponding subspace and the query point (this distance is the absolute difference between the query point's coordinates on the splitting dimension and the splitting value). During the search, when a leaf node is encountered, calculate the Euclidean distance from the query point to all points within that node, and update the current nearest neighbor set according to the rules (initially empty, distances are added to the set sequentially; when the set reaches a certain number of nodes...). At that time, remove the point with the largest distance and keep only the point with the largest distance. (Number of nearest neighbors). During backtracking, calculate the minimum possible distance to the untraversed sibling subtrees of the current node. If this distance is less than the maximum distance in the current nearest neighbor set, add it to the search queue and continue searching until all nodes to be searched have been traversed.

[0051] Finally, it was obtained through KD-tree search. exist In The nearest neighbors form the nearest neighbor set. ,in It is the first Nearest neighbor points.

[0052] calculate To each neighboring point Euclidean distance Based on the principle that "the closer the distance, the greater the impact on risk assessment," a basic weight is assigned to each nearest neighbor using the inverse square of the distance. The weight of each nearest neighbor is then calculated using a variable-precision weighting strategy. ; in, Indicates the first The weights of the nearest neighbors Point No. The Euclidean distance between the nearest neighbors.

[0053] Introducing variable precision parameters (0≤) ≤0.5), this parameter defines the algorithm's tolerance for "approximate" nearest neighbors. The base weights are normalized using variable precision to obtain the final weight for each nearest neighbor. ; in, Indicates the first The final weight of the nearest neighbor points, The number of nearest neighbors. For nearest neighbor index, For variable precision parameters.

[0054] It should be noted that a variable precision parameter is introduced in the final weight calculation. Its core significance lies in introducing a controllable smoothing term. This transforms the rigid weight allocation in traditional nearest neighbor algorithms, which is entirely determined by geometric distance, into a flexible weighting mechanism that balances accuracy and robustness. This parameter allows the algorithm to tolerate a certain degree of distance measurement uncertainty or noise interference. When the value is greater than 0, the dominant influence of extremely close single neighbors on the final weighted result is moderately diluted, thereby effectively suppressing misjudgments of risk values ​​caused by occasional abnormal nearest neighbors (such as measurement noise or small floating objects) in the point cloud data. This improves the overall stability and reliability of risk measurement, ensuring robust performance in complex real-world operating environments. This is achieved by introducing [a certain feature] into the weighted normalization denominator. Through a total number of neighbors A proportional smoothing term is used to establish an adjustable robust control mechanism. Among these, This represents the algorithm's average tolerance for distance measurement errors to each nearest neighbor point, while This ensures that the overall tolerance level is proportional to the number of nearest neighbors (the evidence upon which it is based). The fundamental function of this design is to mathematically weaken the excessive dominance of a few extremely close points (which may be noise or outliers) on the final weighted distance, so that even if a few coincidental, extremely small distances exist, When the value is greater than 0, the distortion effect is effectively diluted, thereby improving the overall stability and reliability of risk value calculation and preventing frequent and severe false alarms caused by single points of anomaly. In short, The term is a specific implementation of the "variable precision" concept in continuous value calculation, which can achieve noise resistance and robustness while pursuing accuracy.

[0055] The query point is obtained by weighting the Euclidean distances of all nearest neighbors based on the final weights. Point cloud data of nearby electrical work sites Weighted distance: ; in, For weighted distance, The number of nearest neighbors. For nearest neighbor indexes.

[0056] Take point cloud data of nearby power operation sites The minimum value of the variable-precision weighted distance of all points is used as the risk value (Risk) for the nearby electric work monitored in this case. This risk value (Risk) is dynamically calculated based on the continuously scanned point cloud data stream through a real-time processing pipeline consisting of steps S1 to S4, and is continuously updated as the work scene and machine posture change, thus providing the latest risk situation assessment at the current moment. To achieve precise prevention and control, considering on-site measurement errors and buffer margins, multiple risk thresholds are set for dynamic early warning. Specifically, when Risk ≥ safe distance threshold (e.g., 1.0 meter), it is judged as safe level (green), and only routine monitoring and data recording are performed; when the warning distance threshold (e.g., 0.7 meters) ≤ Risk < safe distance threshold, it is judged as warning level (yellow), the system automatically activates the audible and visual warning device, and highlights the risk area on the operation terminal interface, prompting "Pay attention to maintaining a safe distance"; when Risk < warning distance threshold, it is judged as dangerous level (red), immediately triggering a continuous high-intensity audible and visual alarm, and simultaneously sending a forced speed limit or stop command to the working machine through the control interface, and displaying the warning message "Stop work immediately and retreat" and real-time distance data on the terminal. All early warning events are recorded simultaneously, including timestamps, risk values, risk levels, and associated point cloud frames, for post-event tracing and safety analysis. Through this tiered response mechanism, quantified risk values ​​are transformed into step-by-step prevention and control actions, forming a closed loop of "monitoring-assessment-early warning-intervention," thereby achieving real-time and proactive prevention and control of risks associated with nearby electrical work.

[0057] Furthermore, based on the above method embodiments, the present invention also provides a control system, including a memory, a processor, and computer programs stored in the memory, which are adapted to be loaded and executed by the processor to implement the above-described risk prevention and control method for near-electricity work at power transmission sites.

[0058] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, the phrase "comprising an element defined as..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0059] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A risk prevention and control method for near-electricity work at power transmission sites, characterized in that: Includes the following steps: Step S1: Scan the point cloud data of the power transmission site using lidar and perform preprocessing to obtain the point cloud data of the power transmission site; Step S2: Remove the ground point cloud from the power transmission site point cloud data using the improved progressive encryption triangular mesh algorithm to obtain the non-ground point cloud data of the power transmission site; Step S3: Cluster the non-ground point cloud data of the power transmission site using density clustering to obtain point cloud data of energized bodies and point cloud data of adjacent power operation bodies; Step S4: By using the nearest neighbor method based on variable precision weighting, perform minimum distance search on the point cloud data of the charged body and the point cloud data of the adjacent power operation body to obtain the risk value of the adjacent power operation. Based on the risk value of the adjacent power operation, classify the risk level and issue an early warning, thereby realizing risk prevention and control of the adjacent power operation.

2. The risk prevention and control method for near-electricity work at power transmission sites according to claim 1, characterized in that: The process of obtaining point cloud data of the power transmission site by scanning the point cloud data with lidar and performing preprocessing includes the following specific steps: In the power transmission site operation area, the lidar is fixedly installed at the center high point of the rotating platform of the operating machinery, and the external parameters of the lidar are calibrated. The lidar scans the power transmission site operation area to obtain point cloud data of the power transmission site operation area, and the point cloud data is preprocessed to obtain the power transmission site point cloud data.

3. A risk prevention and control method for near-electricity work at power transmission sites according to claim 2, characterized in that: The process of removing ground point cloud data from power transmission site point cloud data using an improved progressive encryption triangulation algorithm to obtain non-ground point cloud data for the power transmission site includes the following steps: The power transmission site point cloud data is a set containing N data points. , , where the i-th point cloud Includes three-dimensional spatial coordinates, Represents the x-coordinate of the i-th point cloud. Represents the ordinate of the i-th point cloud, Represents the vertical coordinate of the i-th point cloud; Calculate the elevation tolerance threshold, and obtain a candidate point set based on the elevation tolerance threshold; perform density clustering based on Euclidean distance on the candidate point set to obtain an initial ground point seed set; Based on the initial ground point seed set, the vertical distance and maximum angle are calculated; the power transmission site point cloud data is then filtered and removed based on the vertical distance and maximum angle to obtain the non-ground point cloud data of the power transmission site.

4. A risk prevention and control method for near-electricity work at power transmission sites according to claim 3, characterized in that: The calculation of the elevation tolerance threshold includes the following steps: Calculate the minimum point cloud elevation of power transmission site point cloud data. , = Then calculate the elevation standard deviation of the point cloud data at the power transmission site. Finally, the elevation tolerance threshold is calculated. , , This is the tolerance coefficient.

5. A risk prevention and control method for near-electricity work at power transmission sites according to claim 4, characterized in that: The calculation of vertical distance and maximum angle based on the initial ground point seed set includes the following steps: Based on the initial ground point seed set, a triangulation network is constructed; for unclassified points in the power transmission site point cloud data. ( , , ), with points ( , , Using a point as the center, divide the 3D search area, and filter the triangular faces within the 3D search area in the triangular mesh to form a candidate triangular face set; calculate the points. ( , , ) to each triangle in the candidate triangle set vertical distance Simultaneously calculate points ( , , Projected onto the triangular facet The projection point and the triangular facet Maximum angle between vertices : ； ； in, Vertical distance The maximum included angle, , , Points ( , , The x-coordinate, y-coordinate, and y-coordinate of triangle A, B, C, and D are given. The coefficients of the plane equation, and For the projection point and the triangular facet The vector formed by the vertices, with max() being the function to find the maximum value.

6. A risk prevention and control method for near-electricity work at power transmission sites according to claim 5, characterized in that: The method of clustering non-ground point cloud data of power transmission sites using density clustering to obtain point cloud data of energized bodies and point cloud data of adjacent power operation bodies includes the following steps: Non-ground point cloud data of power transmission sites , ={ }, M is, where each point ( , , It includes three-dimensional spatial coordinates; The neighborhood radius of density clustering is The minimum number of points is MinPts; for the j-th point in the non-ground point cloud data of the power transmission site Calculation points neighborhood set Based on neighborhood set Density-reachable clustering is performed, ultimately yielding K clusters and a set of noise points; The K clusters and noise point sets are divided to obtain the point cloud data of the charged body and the point cloud data of the adjacent electrical work body.

7. A risk prevention and control method for near-electricity work at power transmission sites according to claim 6, characterized in that: The process of dividing the K clusters and the noise point set includes the following specific steps: The feature index of each cluster is extracted and compared with a preset classification threshold to divide the cluster into charged body point cloud and nearby electrical work body point cloud, and the noise point set is classified into the nearby electrical work body point cloud.

8. A risk prevention and control method for near-electricity work at power transmission sites according to claim 7, characterized in that: The method employs a variable-precision weighted nearest neighbor approach to perform a minimum distance search on the point cloud data of the charged body and the point cloud data of the adjacent power operation site, thereby obtaining the risk value of the adjacent power operation. Specifically: Construct a KD-tree, and perform a range search based on the KD-tree to obtain the set of nearest neighbors; Calculate the final weight of each nearest neighbor in the nearest neighbor set; The weighted average of the Euclidean distances of all nearest neighbors is calculated based on the final weights to obtain the weighted distance; the minimum value of the weighted distance is taken as the risk value of nearby electrical work.

9. A risk prevention and control method for near-electricity work at power transmission sites according to claim 8, characterized in that: The calculation of the final weight of each nearest neighbor in the nearest neighbor set is specifically as follows: ； in, Indicates the first The final weight of the nearest neighbor points, The number of nearest neighbors. For nearest neighbor index, For variable precision parameters.

10. A risk control system for near-electric work at power transmission sites, comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the steps of the method according to any one of claims 1-9.

Images