Safety assessment method and system for equipotential entering path of power transmission line

Abstract

The invention belongs to the technical field of data processing, and particularly relates to a safety assessment method and system for a power transmission line entering equipotential path, and the method comprises the steps: firstly obtaining and preprocessing point cloud data of an operation tower, building a tower three-dimensional point cloud model, and building a human body three-dimensional model by referring to a standard human body size and an operation posture; setting a candidate path containing a plurality of sampling points in combination with operation requirements, tower structures and safety specifications; in the dynamic simulation, calculating the minimum distance from the human body to the electrified body and the grounding body at each sampling point, summing to obtain a combined gap, predicting interpolation through a finite element method in combination with a deep learning model, and obtaining the electric field intensity distribution of the human body surface; and finally, if the full-path combination gap exceeds the specified threshold value and the body surface field intensity is lower than the maximum bearing value of the human body, determining that the path is safe. The method improves the evaluation precision and efficiency, and guarantees the safety of hot-line work.

Description

Technical Field

[0001] This invention belongs to the field of data processing technology, specifically relating to a safety assessment method and system for transmission lines entering an equipotential path. Background Technology

[0002] With the continuous expansion of the power system and the increasing load on the power grid, the safe, reliable, and economical operation of transmission lines is of great significance to social production and residents' lives. To minimize power outage time and ensure power supply continuity, live-line working has become an important technical means for the inspection and maintenance of transmission lines.

[0003] However, current safety risk assessments for workers entering the equipotential path of transmission lines mainly rely on personal experience and tower drawings. This approach suffers from high subjectivity and insufficient accuracy, easily leading to safety hazards. Furthermore, the complex shapes and uneven electric field distribution of transmission tower components mean that localized hotspots may cause excessive electric field strength on workers' skin, increasing the risk of misjudgment.

[0004] Therefore, there is an urgent need to propose a scientific, objective, and efficient safety assessment method to achieve quantitative assessment of entry into equipotential paths, thereby improving operational safety and efficiency. Summary of the Invention

[0005] To address this issue, the present invention provides a safety assessment method and system for transmission lines entering an equipotential path, thereby solving the aforementioned technical problems.

[0006] This invention provides a safety assessment method for transmission lines entering an equipotential path, comprising the following steps: The point cloud data of the working tower is acquired, and after preprocessing the point cloud data, a three-dimensional point cloud model of the working tower is constructed. Based on standard human body dimensions and common working postures of live-line workers, a 3D human body model was created using 3D modeling software. Based on the requirements of the operation task, the characteristics of the tower structure, and the safety regulations for live-line work, candidate paths for the human body to enter the equipotential are set in the three-dimensional point cloud model of the tower. Each candidate path contains multiple sampling points. During the dynamic simulation of the candidate path, at each sampling point, the minimum distance from the human body to the charged body and the minimum distance from the human body to the grounded body are calculated, and the two are added together to obtain the combined gap; The distribution of electric field intensity on the human body surface is obtained by interpolation based on finite element method calculation combined with deep learning model prediction. Based on the distribution of electric field intensity on the human body surface, if the combined gap is always greater than a specified threshold and the electric field intensity on the body surface is always less than the maximum field strength that the human body can withstand, then the current candidate path is determined to be safe; otherwise, the current candidate path is determined to be unsafe.

[0007] In another aspect, this application also provides a safety assessment system for transmission lines entering an equipotential path, comprising: The point cloud model acquisition module is used to acquire point cloud data of the working tower, and after preprocessing the point cloud data, construct a three-dimensional point cloud model of the working tower. The human body 3D model acquisition module is used to create a human body 3D model by referencing standard human body size data and common working postures of live-line workers using 3D modeling software. The path acquisition module is used to set candidate paths for a human body to enter the equipotential in the three-dimensional point cloud model of the working tower according to the requirements of the work task, the characteristics of the tower structure and the safety regulations for live working. Each candidate path contains multiple sampling points. The combined gap acquisition module is used to calculate the minimum distance from the human body to the charged body and the minimum distance from the human body to the grounded body at each sampling point during the dynamic simulation of the candidate path, and add the two to obtain the combined gap; The module for obtaining the electric field intensity of the human body surface is used to obtain the distribution of the electric field intensity of the human body surface based on the calculation of the finite element method combined with the prediction and interpolation of the deep learning model. The determination module is used to determine whether the current candidate path is safe if, based on the distribution of electric field intensity on the human body surface, the combined gap is always greater than a specified threshold and the electric field intensity on the body surface is always less than the maximum field strength that the human body can withstand, and otherwise, the current candidate path is determined to be unsafe.

[0008] This application also provides an electronic device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the safety assessment method for power transmission lines entering an equipotential path as described above.

[0009] In another aspect, this application provides a computer-readable storage medium having stored thereon computer program instructions that can be executed by a processor to implement the safety assessment method for transmission lines entering an equipotential path as described above.

[0010] Another aspect of this application provides a computer program product, including a computer program that, when executed by a processor, implements the safety assessment method described above for a transmission line entering an equipotential path.

[0011] This invention constructs a 3D model of a transmission tower and a human body, combining point cloud processing, finite element analysis, and deep learning technologies to dynamically simulate and quantitatively assess the process of workers entering an equipotential path. Specifically, by using the finite element method combined with locally adaptive meshes from graph neural networks and deep learning interpolation, it achieves accurate calculation of field strength at key nodes and efficient completion of the overall field strength, significantly improving accuracy compared to traditional experience-based assessments and avoiding the risk of misjudgment. Spatial indexing accelerates distance search, and deep learning reduces computational load, greatly shortening single-path assessment time and meeting real-time assessment requirements. In summary, this invention's method can identify potential path hazards in advance, effectively ensuring the safety of live-line work, improving work efficiency, and reducing power operation and maintenance costs. Attached Figure Description

[0012] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0013] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 A flowchart of a safety assessment method for a transmission line entering an equipotential path, provided in an embodiment of the present invention.

[0014] Figure 2 This is a schematic diagram of the minimum distance search process provided in an embodiment of the present invention.

[0015] Figure 3 This is a schematic diagram of the finite element method solution process provided in an embodiment of the present invention.

[0016] Figure 4 This is a schematic diagram of a safety assessment system for a power transmission line entering an equipotential path, provided as an embodiment of the present invention.

[0017] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0019] like Figure 1 As shown, this embodiment of the invention discloses a safety assessment method 100 for a transmission line entering an equipotential path, comprising the following steps: S101, acquire point cloud data of the working tower, preprocess the point cloud data, and construct a three-dimensional point cloud model of the working tower; S102. Based on standard human body size data and common working postures of live-line workers, a three-dimensional human body model is created using three-dimensional modeling software. S103, Based on the requirements of the operation task, the characteristics of the tower structure and the safety regulations for live working, a candidate path for the human body to enter the equipotential is set in the three-dimensional point cloud model of the operating tower. S104, During the dynamic simulation of the candidate path, at each sampling point, calculate the minimum distance from the human body to the charged body and the minimum distance from the human body to the grounded body, and add the two to obtain the combined gap; S105, based on the finite element method calculation combined with deep learning model prediction interpolation, obtains the distribution of electric field intensity on the human body surface; S106. Based on the distribution of electric field intensity on the human body surface, if the combined gap is always greater than a specified threshold and the electric field intensity on the body surface is always less than the maximum field strength that the human body can withstand, then the current candidate path is determined to be safe; otherwise, the current candidate path is determined to be unsafe.

[0020] In some embodiments, for step S101, for example, a multi-rotor drone equipped with a laser scanner is used. The laser scanner meets the requirements of high-precision long-range measurement, such as selecting a device with a range of ≥200m and a point cloud accuracy of ≤5mm.

[0021] The drone's flight altitude is adjusted according to the tower height (usually 10-15m above the top of the tower). The flight path adopts a combination of circumferential and longitudinal scanning modes. For example, circumferential scanning (taking a set of data every 30°) covers the circumference of the tower, while longitudinal scanning (taking a set of data every 5m from the bottom to the top of the tower) covers the vertical direction, ensuring that no point cloud data is missed and completely capturing small components such as crossarms, insulators, grounding wires, and bolts.

[0022] At the same time, 3-5 ground control points (using GPS positioning with an accuracy of ≤2mm) need to be set up around the tower for subsequent point cloud data registration.

[0023] In this embodiment, point cloud data preprocessing exemplarily includes sequentially performing denoising, registration, and simplification operations: The denoising process employs a combination of statistical filtering and radius filtering algorithms. Statistical filtering calculates the mean and standard deviation of the distance between each point and its neighbors, eliminating noise points (such as dust reflections or bird interference points) whose mean distance exceeds three times the standard deviation. Radial filtering sets a radius (usually 5 cm) to remove isolated points with fewer than five neighbors, such as occasional single-point noise generated during scanning, ensuring that the retained point cloud data consists entirely of valid reflection points from the tower components.

[0024] The registration process uses ground control points as a reference and employs the ICP (Iterative Closest Point) algorithm to unify point cloud data acquired from multiple perspectives into the same coordinate system (geodetic coordinate system). During registration, coarse registration is performed first (approximate alignment using ground control points, with errors controlled within 10cm), followed by fine registration (iterative optimization using the ICP algorithm to ensure an overlap of ≥95% between adjacent point clouds and a registration error ≤3mm), avoiding model splicing misalignment due to perspective differences (such as breaks in the crossarm in point clouds from different perspectives).

[0025] The simplification process employs a voxel mesh simplification algorithm, which sets the voxel size (adjusted according to the component size, such as 5cm for tower body and 1cm for small components like bolts). The point cloud within the same voxel is merged into a representative point, reducing the amount of data (usually by 60%-80% of the original data) while preserving the geometric features of the components, thus avoiding computational overload during subsequent model building and calculation.

[0026] In this embodiment, the preprocessed point cloud data is imported into professional point cloud processing software such as Cloud Compare and Geomagic DesignX, and a three-dimensional model is generated through surface reconstruction.

[0027] Specifically, for regular components such as tower bodies and crossarms, the Poisson reconstruction algorithm is used (to generate smooth surfaces by fitting the density function of the point cloud); for irregular components such as insulators and bolts, the Alpha shape reconstruction algorithm is used (to accurately restore the concave and convex shapes of components, such as the disc edge of an insulator and the thread structure of a bolt, by adjusting the alpha parameters). Further details will not be elaborated here.

[0028] In some embodiments, for S102, the size data of the main group of live-line workers (adult males aged 18-50) are used as a benchmark. Key dimensions include: height 165-180cm, shoulder width 38-45cm, arm length (from acromion to fingertip) 58-65cm, leg length (from anterior superior iliac spine to heel) 85-95cm, palm length 18-22cm, and finger length (middle finger) 8-10cm, to ensure that the model conforms to the physiological characteristics of the workers.

[0029] In addition, it includes six core postures commonly used in live-line work, and the joint angles and limb positions of each posture must be consistent with actual operations.

[0030] Specifically, the climbing posture is as follows: knees bent at 110°-130°, elbows bent at 90°-110°, and hands gripping the tower ladder (fingers touching the ladder crossbar with a gap of ≤2mm); the transition posture (moving from the grounding body to the equipotential): the body is turned to the side, the torso is at an angle of 30°-45° to the tower axis, one arm hangs naturally, and the other arm is slightly bent (elbow angle 120°-140°); the operating posture (approaching the work point): the body leans forward 15°-20°, both arms are extended forward (elbow angle 160°-170°), and the palms face the work point (fingers spread apart with a gap of 3-5mm).

[0031] The modeling process utilizes a combination of Blender (geometric modeling) and COMSOL Multiphysics (physical parameter assignment). Blender supports high-precision polygon modeling, accurately reproducing the contours of human muscles and the range of motion of joints; COMSOL can easily assign dielectric properties to the model to meet the needs of subsequent electric field calculations.

[0032] Specifically, in Blender, you first create the basic geometry (cylinder, sphere) of the human torso, limbs, and head, and then use the subdivision surface tool to add model details such as the curvature of the torso and the thickness variation of the limbs. The head (simplifying hair into smooth curved surfaces to avoid mesh interference during field strength calculation), hands (distinguishing between palms and 5 fingers, adding chamfers at finger joints with a radius of 1mm to restore the realistic finger shape), and feet (adding 10-15mm thick soles to simulate insulated shoes) are modeled in detail to ensure the accurate geometry of these easily accessible charged parts. The model was imported into COMSOL, and the equivalent physical parameters of the human body were assigned a dielectric constant ε=50 (referencing the "Insulation Test Procedure for High Voltage Electrical Equipment" to balance calculation accuracy and efficiency) and a conductivity σ=0.1S / m (approximate to the conductivity of human skin), so that the model could present a realistic dielectric response in an electric field.

[0033] In some embodiments, for S103, for example, the candidate path setting integrates three core factors to ensure that the path is safe, feasible, and compliant with specifications.

[0034] Specifically, regarding the requirements of the work tasks, the destination of the route, such as the work point of the conductor or the installation location of the insulator, should be determined according to the specific work type, such as conductor maintenance or insulator replacement. The route should be direct to the destination to avoid detours that may increase the work time. Regarding the structural characteristics of the tower, avoid densely packed areas of tower components, such as the connection between crossarms and insulators, where the spacing between components is ≤30cm, and avoid areas with sharp components, such as bolt groups and grounding wire joints. These areas are prone to generating local high field strength and have limited space for human movement. Regarding safety regulations for live-line work, the route must meet the following requirements: the initial position must be near a grounding body, such as the grounding platform at the bottom of a tower; during the movement, the initial distance from the live body must be ≥ 1.2 times the safe distance. For example, the safe distance for a 220kV line is 1.8m, and the initial distance must be ≥ 2.16m.

[0035] Optionally, CAD software such as AutoCAD or SolidWorks can be used to draw the path in the 3D point cloud model of the working tower. For example, the steps are as follows: Determine the starting point (a safe location near the grounding electrode) and the ending point (the work site); Draw a preliminary path along the gaps in the tower structure, such as the space between the ladder and the crossarm, and the space under the insulator string. The path is a continuous curve (avoid right-angle turns, with a turning radius ≥1m, in line with human movement habits). Set sampling points along the path (intervals of 50-100cm, and 20-30cm intervals in critical areas such as near charged bodies) for parameter calculation during subsequent dynamic simulation.

[0036] In some embodiments, for step S104, for example, Python combined with Open3D is used to realize the dynamic simulation of the human body model along the candidate path. The joint angles and spatial positions of the human body model are controlled by programming to simulate the movement process of the worker (the movement speed is set to 0.5m / s to match the actual working speed). Meanwhile, sampling areas are divided according to the risk level of the path. For example, in low-risk areas (distance from live conductors ≥ 3 times the safety distance): sampling points are spaced 100cm apart; in medium-risk areas (distance from live conductors 1.5-3 times the safety distance): sampling points are spaced 50cm apart; and in high-risk areas (distance from live conductors ≤ 1.5 times the safety distance): sampling points are spaced 20-30cm apart. This ensures that the parameter calculations for high-risk areas are complete and that potential safety hazards are captured in a timely manner.

[0037] The combined gap calculation involves obtaining the minimum distance (D1) from the human body to the charged body and the minimum distance (D2) from the human body to the grounded body. The sum of the two is the combined gap (D=D1+D2).

[0038] Optionally, to improve computational efficiency, a spatial index structure is used to accelerate the minimum distance search. Among them, KD-tree and octree are preferred for the selection and construction of spatial index structures.

[0039] Specifically, KD-tree construction includes: dividing the point cloud data of the human body model and charged / grounded bodies into dimensions (along the x, y, and z axes in sequence); selecting the median of each dimension as the split point to construct a binary tree; recursively splitting until each leaf node contains ≤10 points; KD-tree is suitable for point cloud data with low dimensionality (3D) and medium data volume (≤1 million points), has a fast construction speed, and can quickly narrow the search range during querying, making it suitable for calculating the distance between the human body and regular components such as tower bodies and crossarms; For example, the octree construction includes: constructing a cubic bounding box centered on the tower model; dividing the bounding box into 8 sub-cubes and determining whether each sub-cube contains a point cloud; repeatedly dividing the sub-cubes containing point clouds until the side length of the sub-cubes is ≤5mm; octrees are suitable for large point cloud data (>1 million points) and complex component shapes such as insulators and bolts, with finer spatial division, which can accurately capture the local shape of complex components and is suitable for distance calculation between the human body and irregular components such as bolts and insulators.

[0040] like Figure 2 As shown, exemplarily, the minimum distance search process includes, S201, extract the point cloud data of the human body model at the current sampling point (approximately 100,000 to 200,000 points), the point cloud data of the charged body (approximately 50,000 to 100,000 points), and the point cloud data of the grounded body (approximately 200,000 to 300,000 points). S202, construct the spatial index structure (KD-tree and octree) for charged and grounded bodies. S203: For each point of the human body model, query the nearest point to the charged / grounded body through the index structure and record the minimum distance; S204, after traversing all points on the human body, take the minimum value as D1 (minimum distance from the human body to the charged body) and D2 (minimum distance from the human body to the grounded body), and calculate the combined gap D.

[0041] In some embodiments, for S105, an electric field calculation model is constructed, including the three-dimensional point cloud model of the working tower, the three-dimensional human body model, and the surrounding space; the field strength data of key nodes in the electric field calculation model is calculated using the finite element method, wherein the key nodes cover the main characteristic areas of the electric field distribution and the dangerous adjacent areas that the human body may pass through; the field strength data of the key nodes is input into a trained deep learning model based on a graph neural network to predict and complete the field strength distribution of the entire electric field calculation model; based on the completed field strength distribution, the field strength change on the surface of the human body model at different positions is monitored.

[0042] Specifically, the electric field calculation model is a carrier for simulating the electric field distribution of transmission lines. For example, it includes a three-dimensional point cloud model of the working tower, a three-dimensional human body model, and the surrounding space.

[0043] For the spatial boundary of the calculation model, the horizontal direction extends to the area with electric field strength ≤0.1kV / m (100-150m for 220kV lines and 200-250m for 500kV lines) centered on the tower, and the vertical direction extends from 1m below the ground to 20-30m above the top of the tower to ensure coverage of the area significantly affected by the electric field. For the dielectric partitioning of the computational model, the model includes two media: air (dielectric constant ε = 8.85 × 10⁻¹² F / m, conductivity σ = 10⁻¹² F / m). 4 S / m) and human body (dielectric constant ε=50, conductivity σ=0.1S / m), the boundaries of the two media are precisely aligned with the geometric boundaries of the model to avoid electric field calculation errors caused by media overlap and boundary gaps; For the boundary conditions of the calculation model, the charged body (conductor, crossarm) is set as Dirichlet boundary (potential = line rated voltage, such as 220kV for a 220kV line), the grounded body (tower body, ground) is set as Neumann boundary (potential = 0V), and the outermost layer of the model is set as an absorbing boundary (to avoid electric field reflection at the boundary, which conforms to the actual electric field diffusion law).

[0044] For the mesh generation of the computational model, a local adaptive meshing strategy is adopted to control the computational load while ensuring accuracy. Non-critical areas (air medium, tower components away from people): grid size 50-100mm; Key areas (within 500mm of the human body and the surface of sharp components of the tower): mesh size 5-10mm, and different density meshes are connected by transition meshes (size gradually transitions from 10mm to 100mm) to avoid computational convergence difficulties caused by abrupt changes in mesh size.

[0045] In this embodiment, the field strength of key nodes is calculated using the finite element method. Specifically, for example, the principle for selecting key nodes is that the key nodes simultaneously satisfy the characteristics of the covered electric field distribution and the focus of the dangerous area.

[0046] For example, nodes are evenly distributed throughout the electric field calculation model, with nodes selected at intervals of 100mm×100mm×100mm to cover the overall distribution of the electric field. For hazardous areas, densification nodes are selected at intervals of 20mm×20mm×20mm near human hands, feet, and head (hazardous adjacent area 1), and near tower bolts and grounding wire joints (hazardous adjacent area 2) to ensure the capture of local high field strength (hot spots). The total number of critical nodes is controlled, usually selecting 10,000 to 20,000 critical nodes, which is only 1% to 2% of the total number of nodes (1 million to 2 million), greatly reducing the amount of finite element calculation.

[0047] For example, such as Figure 3 As shown, the finite element method solution process includes, S301, import the electric field calculation model into finite element calculation software, such as ANSYS Maxwell or COMSOL Multiphysics; this invention does not limit this. S302, establish the electrostatic field control equation: ∇・(ε∇φ)=0 (φ is the potential), and substitute the medium parameters and boundary conditions; S303 uses the Galerkin weighted residual method to discretize the governing equations, transforming the continuous electric field problem into a linear system of equations. S304, the potential value φ of the key node is obtained by solving the linear equation system using the conjugate gradient method; S305, calculate the field strength E: E=-∇φ (field strength is the negative gradient of potential), and obtain the magnitude and direction of the field strength at the key node.

[0048] Specifically, in the calculation of the electric field of transmission lines, the governing equations describing the electric field are continuous, and there are potential values ​​at an infinite number of points in space, making direct computer calculation impossible. The core of the Galerkin method is to decompose this continuous problem into a system of discrete linear equations that can be processed by a computer. The steps can be simplified to three: First, divide the computational region (including the electric field space of towers and human bodies) into many small units (such as small cubes). The vertices of the units are discrete nodes. By calculating the potential of these nodes, the overall electric field can be pieced together.

[0049] For each small cell, the potential φ within the cell is approximated by a simple function (such as a first-order polynomial). app The coefficients of the current function can be represented by the potentials at the unit vertices, thus allowing us to find the continuous function φ. appThis becomes finding the potential values ​​of a finite number of nodes.

[0050] The bias is eliminated using a weighting function, resulting in a system of equations. Substituting the approximate function into the governing equations produces a bias (margin) R. The Galerkin method selects a weighting function w_i that is consistent with the basis functions of the approximate function, ensuring that the product of the margin R and w_i integrals to zero in space (i.e., the sum of the biases within the weighting function's range is zero). After simplification through integration, each node corresponds to a linear equation containing the potentials of all nodes, ultimately forming a system of linear equations. Solving this system yields the potentials of all nodes.

[0051] For S304, specifically, the Galerkin method transforms the continuous problem into a system of linear equations, in the form Kφ=F (K is the stiffness matrix, determined by the dielectric constant, element geometry, etc.; φ is the nodal potential vector to be solved; F is the load vector, determined by the boundary conditions). The conjugate gradient method is an efficient way to solve this type of symmetric positive definite linear equation system. Its core is to find φ through iterative approximation, simplifying the steps to three: Step 1: Initialize the iteration starting point. First, assume an initial potential vector φ0 (e.g., all zeros). Calculate the initial residual r0 = F - Kφ0 (the residual reflects the deviation between the current φ and the true solution; when r = 0, φ is the true solution). Then, set the initial search direction p0 = r0 (start searching for a better solution along the residual direction).

[0052] Step 2: Iterative optimization to approximate the true solution. In each iteration, first calculate the step size α. k (Determine along p) k How far should the direction be moved to minimize the residual? Update potential φ k ₊1=φ k +α k p k ; then calculate the new residual r k ₊1=r k -α k Kp k If r k ₊1 is small enough (to meet engineering precision requirements, such as less than 10⁻) 6 ), then φ k ₊1 is the solution.

[0053] Step 3: Update the search direction. If the accuracy requirement is not met, calculate the direction coefficient β. k (To avoid repeating the same mistakes), update the search direction p k ₊1=r k ₊1+β k p k Repeat step 2 until the residual meets the target.

[0054] The current method does not require storing the complete matrix, has low computational cost, fast convergence, and can efficiently obtain the potential φ of all key nodes.

[0055] In this embodiment, the finite element method can only obtain the field strength data of key nodes, and the deep learning model based on graph neural network (GNN) is used to predict the field strength of non-key nodes to complete the entire electric field distribution.

[0056] Regarding the network structure of the GNN model, for example, a graph convolutional layer (GCN) and a fully connected layer structure are adopted. The graph convolutional layer is used to learn the spatial relationships between nodes (such as the field strength dependence of adjacent nodes), and the fully connected layer is used to output the field strength prediction value of non-critical nodes. The input layer receives two types of data, including the three-dimensional spatial coordinates (x, y, z) of key nodes and the corresponding field strength value (E). The data is organized in the form of a graph structure, with each key node as a vertex of the graph, and the vertex feature vector is [x, y, z, E] (4-dimensional feature). The edges between nodes are defined according to the spatial distance (nodes with a distance ≤ 50 mm are usually regarded as adjacent nodes and assigned an edge connection relationship). The weight of the edge is determined by the reciprocal of the distance (the closer the distance, the greater the weight, reflecting the strong correlation characteristic of the electric field at close range).

[0057] The output layer contains the field strength prediction values ​​for non-critical nodes. The dimension is consistent with the number of non-critical nodes (usually hundreds of thousands of dimensions), and each dimension corresponds to the field strength prediction result of a non-critical node.

[0058] For example, the graph convolutional layer is the core feature extraction module of the network. It achieves feature learning from local to global through 3-5 layers stacked together. The design of each layer follows the principles of neighborhood information aggregation and feature transformation.

[0059] Specifically, the first layer of graph convolution focuses on the direct spatial relationships between nodes. For each node, it aggregates the feature vectors of all neighboring nodes ([x,y,z,E]) and incorporates neighborhood information into the current node's features through weighted summation (the weights are the inverse of the distance between the edges). The feature transformation is performed by a 32-channel convolution kernel, and a 32-dimensional feature vector is output to capture the basic spatial dependencies between nodes (such as the relationship between distance and field strength attenuation).

[0060] The intermediate convolutional layers learn higher-order spatial relationships. Starting from the second layer, the number of channels in each layer doubles (32→64→128) to gradually improve the feature representation ability. A skip connection mechanism is introduced to concatenate the output features of the previous layer with the input features of the current layer, while retaining low-order relationship information. By introducing nonlinearity through the activation function (Leaky ReLU, slope 0.01), we can capture the nonlinear patterns in the electric field distribution (such as electric field distortion near the human body).

[0061] The final layer of graph convolution outputs globally correlated features, increasing the number of channels to 256 dimensions. The feature vectors integrate the correlation information of local neighborhood and global space. The output includes key patterns of electric field distribution, such as large field strength gradients near conductors, low field strength near grounding bodies, and concentrated field strength around sharp components.

[0062] For example, the fully connected layer receives 256-dimensional global features output by the graph convolutional layer and achieves accurate prediction of field strength from features to non-critical nodes through a three-level mapping.

[0063] Specifically, the first-level fully connected layer is used for feature dimensionality reduction and integration. It employs 1024 neurons to map 256-dimensional features to 1024 dimensions and enhances non-linear expression through the ReLU activation function. A Dropout layer (dropout rate of 0.3) is added to prevent overfitting and ensure the model's generalization ability for different tower scenarios.

[0064] The second fully connected layer is used to focus on the features of non-critical nodes, reducing the number of neurons to 512. It adjusts the features based on the spatial distribution characteristics of non-critical nodes, such as the gradual change of field strength in regions far from critical nodes. By standardizing the feature distribution through the BatchNorm layer, model convergence is accelerated and stability is improved.

[0065] The output layer predicts the field strength value. The number of neurons is the same as the total number of non-critical nodes, and a linear activation function is used (because the field strength is a continuous value and positive and negative direction information needs to be retained). Output the predicted field strength value for each non-critical node. The numerical range is controlled by normalizing the previous data (mapping to the [-1,1] interval), and finally restore the actual field strength value (kV / m) by inverse normalization.

[0066] Thus, through the synergy of graph convolutional layers and fully connected layers, the network structure not only retains the accuracy of the field strength at key nodes calculated by the finite element method, but also achieves efficient completion of the field strength across the entire domain by learning spatial correlation.

[0067] In some embodiments, the training of a GNN model may optionally include the following process: Collect electric field data for different scenarios, such as different voltage levels from 110kV to 1000kV, different tower types (straight-line towers / tension towers), and different weather conditions (sunny / cloudy). Calculate the field strength of all nodes (1-2 million points) in each scenario using the finite element method, extract the field strength of key nodes (input) and the field strength of all nodes (output), and construct a training set (80%), a validation set (10%), and a test set (10%). The Adam optimizer was used with an initial learning rate of 0.001 (decreasing by 10% every 50 iterations). The loss function was mean squared error (MSE). The number of iterations was 1000. Training was stopped when the validation set loss did not decrease for 20 consecutive iterations to avoid overfitting. The model performance was evaluated on the test set, requiring the field strength prediction error to be ≤5% (MSE≤(0.05×Emax)², where Emax is the maximum field strength in the test set), and the ability to accurately reproduce the field strength peaks in local hot spots (such as the field strength near bolts, where the traditional linear interpolation error is 15%-20%, and the GNN model error is ≤5%).

[0068] The field strength of key nodes calculated by the finite element method is input into the trained GNN model to predict the field strength of non-key nodes, thus completing the field strength distribution of the entire electric field calculation model. Field strength data of all nodes (approximately 100,000 to 200,000) on the surface of the human body model are extracted, the magnitude of the field strength of each node is recorded, and the maximum field strength (Emax) and distribution characteristics of the surface field strength are determined (e.g., high field strength areas are concentrated in the fingertips and the top of the head).

[0069] As the human body model moves along the planned path, the electric field calculation model updates the spatial medium occupied by the human body in real time and recalculates the field strength distribution to ensure that the field strength data is synchronized with the human body's position. At the same time, the nodes on the surface of the human body model serve as field strength monitoring points, providing the electric field calculation model with target monitoring areas and avoiding meaningless global field strength calculations.

[0070] In some embodiments, for S106, the threshold for the combined gap is set in conjunction with the line voltage level, such as a threshold of ≥2.5m for 220kV lines and ≥3.5m for 500kV lines. The threshold needs to be 10%-20% higher than the safe distance to reserve a safety margin. The maximum electric field strength that the human body can withstand is set at ≤240kV / m (the upper limit of the safe electric field strength that the human body can withstand for a long time under power frequency electric field). If the worker wears insulating clothing, it can be appropriately increased to ≤300kV / m (insulating clothing can reduce the electric field strength on the body surface by 20%-30%).

[0071] Based on the threshold of the combined gap and the maximum field strength that the human body can withstand, all sampling points on the candidate path are traversed to obtain the combined gap D and the maximum field strength Emax of the body surface for each sampling point. The path is determined to be safe if all sampling points meet the conditions that D > the specified threshold and Emax < the maximum field strength that the human body can withstand. If any sampling point does not meet the conditions (D ≤ threshold, Emax ≥ capability value), the path is determined to be unsafe. Optionally, generate a path combination gap variation curve (horizontal axis is path length, vertical axis is combination gap) and a surface field strength distribution cloud map (different colors represent different field strengths) to intuitively display the path safety status.

[0072] Figure 4 A safety assessment system 400 for transmission lines entering an equipotential path is shown. Device embodiments and... Figure 1 Corresponding to the illustrated method embodiments, the specific methods include: The point cloud model acquisition module 401 is used to acquire point cloud data of the working tower, and after preprocessing the point cloud data, construct a three-dimensional point cloud model of the working tower. The human body 3D model acquisition module 402 is used to establish a human body 3D model by referring to standard human body size data and common working postures of live-line workers using 3D modeling software. The path acquisition module 403 is used to set candidate paths for a human body to enter the equipotential in the three-dimensional point cloud model of the working tower according to the requirements of the work task, the characteristics of the tower structure and the safety regulations for live working. Each candidate path contains multiple sampling points. The combined gap acquisition module 404 is used to calculate the minimum distance from the human body to the charged body and the minimum distance from the human body to the grounded body at each sampling point during the dynamic simulation of the candidate path, and add the two to obtain the combined gap; The human body surface electric field intensity acquisition module 405 is used to obtain the distribution of human body surface electric field intensity based on finite element method calculation combined with deep learning model prediction interpolation. The determination module 406 is used to determine the current candidate path is safe if, based on the distribution of the electric field intensity on the human body surface, the combined gap is always greater than a specified threshold and the electric field intensity on the body surface is always less than the maximum field strength that the human body can withstand, then the current candidate path is determined to be unsafe.

[0073] Specific limitations regarding the safety assessment system for transmission lines entering an equipotential path can be found in the limitations of the safety assessment method for transmission lines entering an equipotential path described above, and will not be repeated here. The various modules in the aforementioned safety assessment system for transmission lines entering an equipotential path can be implemented through software, hardware, or a combination thereof. These modules can be embedded in the processor of an electronic device in hardware format or independently of the processor, or stored in the memory of the electronic device in software format, so that the processor can call the corresponding operations of each module.

[0074] It should be noted that, in order to highlight the innovative aspects of this application, this embodiment does not include modules that are not closely related to solving the technical problems proposed in this application, but this does not mean that there are no other modules in this embodiment.

[0075] like Figure 5 As shown, the electronic device 1 provided in this application may include a memory 12, a processor 13 and a bus, and may also include a computer program stored in the memory 12 and executable on the processor 13, such as a safety assessment program for power transmission lines entering an equipotential path.

[0076] The memory 12 includes at least one type of readable storage medium, such as flash memory, portable hard drive, multimedia card, card-type memory, magnetic storage, disk, optical disk, etc. In some embodiments, the memory 12 can be an internal storage unit of the electronic device 1, such as a portable hard drive. In other embodiments, the memory 12 can be an external storage device of the electronic device 1, such as a plug-in portable hard drive, smart memory card, secure digital card, flash memory card, etc., equipped on the electronic device 1. Furthermore, the memory 12 can include both internal and external storage units of the electronic device 1. The memory 12 can be used not only to store application software installed on the electronic device 1 and various types of data.

[0077] The processor 13 executes the operating system of the electronic device 1 and various installed applications. The processor 13 executes the applications to implement the steps in the above-described safety assessment method for power transmission lines entering an equipotential path.

[0078] For example, the computer program may be divided into one or more modules, which are stored in the memory 12 and executed by the processor 13 to complete this application. The one or more modules may be a series of computer program instruction segments capable of performing specific functions, with the current instruction segment describing the execution process of the computer program in the electronic device 1.

[0079] The integrated unit implemented as a software functional module described above can be stored in a computer-readable storage medium, which can be non-volatile or volatile. The software functional module stored in the storage medium includes several instructions to cause a computer device (which may be a personal computer, computer equipment, or network device, etc.) or processor to execute some of the functions of the various embodiments of this application.

[0080] Embodiments of this application also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps described in the above embodiments.

[0081] The above embodiments are merely illustrative of the principles and effects of this application and are not intended to limit this application. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of this application. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in this application should still be covered by the claims of this application.

Claims

1. A safety assessment method for a transmission line entering an equipotential path, characterized in that, Includes the following steps: The point cloud data of the working tower is acquired, and after preprocessing the point cloud data, a three-dimensional point cloud model of the working tower is constructed. Based on standard human body dimensions and common working postures of live-line workers, a 3D human body model was created using 3D modeling software. Based on the requirements of the operation task, the characteristics of the tower structure, and the safety regulations for live-line work, candidate paths for the human body to enter the equipotential are set in the three-dimensional point cloud model of the tower. Each candidate path contains multiple sampling points. During the dynamic simulation of the candidate path, at each sampling point, the minimum distance from the human body to the charged body and the minimum distance from the human body to the grounded body are calculated, and the two are added together to obtain the combined gap; The distribution of electric field intensity on the human body surface is obtained by interpolation based on finite element method calculation combined with deep learning model prediction. Based on the distribution of electric field intensity on the human body surface, if the combined gap is always greater than the specified threshold and the electric field intensity on the body surface is always less than the maximum field strength that the human body can withstand, then the candidate path is determined to be safe; otherwise, the candidate path is determined to be unsafe.

2. The safety assessment method for a transmission line entering an equipotential path according to claim 1, characterized in that, The method uses finite element method calculation combined with deep learning model interpolation to obtain the electric field intensity distribution on the human body surface. include, Construct an electric field calculation model that includes a 3D point cloud model of the working tower, a 3D human body model, and the surrounding space; The electric field strength data of key nodes in the electric field calculation model are calculated using the finite element method. The key nodes cover the main characteristic areas of the electric field distribution and the dangerous adjacent areas that a person may pass through. The field strength data of key nodes are input into the trained deep learning model based on graph neural network to predict and complete the field strength distribution of the entire electric field calculation model. Based on the completed field strength distribution, the changes in field strength on the surface of the human body at different locations are monitored.

3. The safety assessment method for a transmission line entering an equipotential path according to claim 2, characterized in that, The deep learning model includes graph convolutional layers and fully connected layers. The graph convolutional layers are stacked in multiple layers. The first layer aggregates the feature vectors of adjacent nodes and transforms them through a 32-channel convolutional kernel to output a 32-dimensional feature vector. The number of channels in the middle layer is doubled and skip connections and Leaky ReLU activation function are introduced. The last layer increases the number of channels to 256 dimensions to output globally related features. The fully connected layers integrate features through a first-level fully connected layer with 1024 neurons and a second-level fully connected layer with 512 neurons. Finally, the output layer, which is consistent with the total number of non-key nodes, outputs the field strength prediction value through a linear activation function.

4. The safety assessment method for a transmission line entering an equipotential path according to claim 2, characterized in that, The dangerous adjacent area includes the area near the hands, feet, and head of a person, as well as the area around parts of the tower components that are prone to generating local high field strength.

5. The safety assessment method for a transmission line entering an equipotential path according to claim 1, characterized in that, Spatial index structures are used to accelerate the search for the minimum distance during the calculation of the combined gap. Spatial index structures include KD-tree and octree.

6. The safety assessment method for a transmission line entering an equipotential path according to claim 5, characterized in that, The calculation process of the combined gap includes: S201, extracting the point cloud data of the human body model, the point cloud data of the charged body, and the point cloud data of the grounded body at the current sampling point; S202, constructing a spatial index structure for the charged body and the grounded body; S203, for each point of the human body model, querying the nearest point to the charged body / grounded body through the index structure and recording the minimum distance; S204, after traversing all points of the human body, taking the minimum value D1 as the minimum distance from the human body to the charged body and D2 as the minimum distance from the human body to the grounded body, and calculating the combined gap D=D1+D2.

7. A safety assessment system for transmission lines entering an equipotential path, characterized in that, include: The point cloud model acquisition module is used to acquire point cloud data of the working tower, and after preprocessing the point cloud data, construct a three-dimensional point cloud model of the working tower. The human body 3D model acquisition module is used to create a human body 3D model by referencing standard human body size data and the working posture of live-line workers, and using 3D modeling software. The path acquisition module is used to set candidate paths for a human to enter the equipotential in the three-dimensional point cloud model of the working tower according to the requirements of the work task, the characteristics of the tower structure and the safety regulations for live working. Each candidate path contains multiple sampling points. The combined gap acquisition module is used to calculate the minimum distance from the human body to the charged body and the minimum distance from the human body to the grounded body at each sampling point during the dynamic simulation of the candidate path, and add them together to obtain the combined gap; The module for obtaining the electric field intensity of the human body surface is used to obtain the distribution of the electric field intensity of the human body surface based on the calculation of the finite element method combined with the prediction and interpolation of the deep learning model. The determination module is used to determine whether the current candidate path is safe if the combined gap is always greater than a specified threshold and the electric field strength on the body surface is always less than the maximum field strength that the human body can withstand, based on the distribution of electric field strength on the human body surface. Otherwise, the current candidate path is determined to be unsafe.

8. An electronic device, the electronic device comprising: At least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform any one of the methods of claims 1-6.

9. A computer-readable medium having stored thereon computer program instructions that can be executed by a processor to implement the method of any one of claims 1-6.

10. A computer program product comprising a computer program that, when executed by a processor, implements the method of any one of claims 1-6.

Images