Substation secondary circuit three-dimensional drawing generation method and system for intelligent operation inspection

Abstract

The application is suitable for the field of digital substation technology, and provides a substation secondary circuit three-dimensional drawing generation method and system for intelligent operation and inspection. The method comprises the following steps: obtaining a two-dimensional drawing, an SCD system configuration file and point cloud data of a substation secondary system; constructing a three-dimensional structure model of the secondary circuit based on the two-dimensional drawing and the point cloud data, and establishing a mapping relationship between the SCD system configuration file and physical devices; and generating a three-dimensional visual drawing of the secondary circuit according to the three-dimensional structure model. The application can fuse the two-dimensional drawing geometric parameters and the point cloud data to construct a high-precision three-dimensional visual drawing, and can stereoscopically present the screen cabinet structure, device panel and cable path, so that the operation and maintenance personnel can directly "see through" the screen cabinet, check the device layout, terminal row arrangement mode and cable laying path, and facilitate the operation and maintenance personnel to operate and maintain the secondary circuit.

Description

Technical Field

[0001] This application relates to the field of substation digital technology, specifically to a method and system for generating three-dimensional drawings of substation secondary circuits for intelligent operation and maintenance. Background Technology

[0002] The secondary circuit of a substation refers to the electrical circuit system composed of relay protection devices, automatic devices, measuring instruments, control switches, signaling equipment, and connecting cables. It undertakes the functions of monitoring, controlling, protecting, and transmitting signals for primary equipment and is the "nervous system" that ensures the safe and stable operation of the substation.

[0003] Currently, the operation and maintenance of substation secondary circuits heavily relies on drawings. During the design phase, CAD 2D drawings are created, including information such as cabinet layout, panel openings, terminal block wiring, and cable lists. During the operation and maintenance phase, paper or electronic versions of these drawings are archived for reference during on-site operations. Maintenance personnel must bring these drawings to the site to verify the status of pressure plates, inspect terminal wiring, and troubleshoot faulty circuits.

[0004] Two-dimensional drawings cannot depict the internal three-dimensional structure of the cabinet, the details of the device panels, and the cable routing. Moreover, information updates are lagging, and drawings are updated slowly after equipment modifications, sometimes resulting in discrepancies between the drawings and the secondary circuit structure. In addition, two-dimensional drawings, SCD logic configurations, and ledger records are stored in a scattered manner, requiring information from multiple systems for fault analysis, leading to low operation and maintenance efficiency. Summary of the Invention

[0005] In view of this, the embodiments of this application provide a method and system for generating three-dimensional drawings of substation secondary circuits for intelligent operation and maintenance. It can integrate the geometric parameters of two-dimensional drawings with point cloud data to construct high-precision three-dimensional visualization drawings, which present the cabinet structure, device panels, cable paths, etc. in three dimensions. Maintenance personnel can directly "see through" the inside of the cabinet and view the device layout, terminal block arrangement, and cable laying path, which facilitates the maintenance personnel to perform operation and maintenance on the secondary circuits.

[0006] To achieve the above objectives, this application adopts the following technical solution:

[0007] In a first aspect, this application provides a method for generating three-dimensional drawings of substation secondary circuits for intelligent operation and maintenance, comprising: acquiring two-dimensional drawings of the substation secondary system, SCD system configuration files, and point cloud data, wherein the two-dimensional drawings include secondary cabinet layout diagrams, device panel opening diagrams, terminal block wiring diagrams, cable lists, and protection logic diagrams, and the SCD system configuration files include IED device lists, logic node configurations, virtual terminal connection relationships, and GOOSE / SV control block information; constructing a three-dimensional structural model of the secondary circuit based on the two-dimensional drawings and the point cloud data, and establishing a mapping relationship between the SCD system configuration files and the physical devices of the three-dimensional structural model; and generating three-dimensional visualization drawings of the substation secondary circuits based on the three-dimensional structural model.

[0008] Based on the first aspect, in some embodiments, constructing a three-dimensional structural model of the secondary circuit based on the two-dimensional drawings and the point cloud data includes: parsing the two-dimensional drawings to obtain structural parameter information of the secondary circuit, which includes geometric parameters and spatial layout information of the secondary cabinet, internal equipment and terminal blocks, as well as cable list and protection logic information; establishing a drawing parameter database based on the structural parameter information of the secondary circuit, which includes: the size and positional relationship of the secondary cabinet and internal equipment, the number and type of terminal blocks, the opening layout and pressure plate type of the internal equipment, cable list and protection logic; performing three-dimensional modeling based on the drawing parameter database to generate a preliminary three-dimensional structural model; extracting key point cloud features from the point cloud data, which includes planar point cloud features, edge line point cloud features and corner point cloud features; registering the key point cloud features with the preliminary three-dimensional structural model, and correcting the preliminary three-dimensional structural model to obtain the three-dimensional structural model.

[0009] Based on the first aspect, in some embodiments, the step of extracting key point cloud features from the point cloud data includes: inputting the original point cloud into a trained point cloud recognition model to determine the category of each point cloud and obtain a classified point cloud dataset; based on the classified point cloud dataset, using a clustering algorithm to segment connected point clouds of the same category into independent point cloud objects; determining the plane of each point cloud object to obtain planar point cloud features; determining the intersection line of adjacent planes or the boundary contour line of the plane based on the plane of each point cloud object to obtain edge line point cloud features; and calculating the key corner point positions based on the planar point cloud features and the edge line point cloud features to obtain corner point cloud features.

[0010] Based on the first aspect, in some embodiments, registering the key point cloud features with the preliminary three-dimensional structural model and correcting the preliminary three-dimensional structural model to obtain the three-dimensional structural model includes: extracting model feature points of the same type as the key point cloud features from the preliminary three-dimensional structural model; calculating a coarse registration matrix based on the model feature points and the key point cloud features; transforming the preliminary three-dimensional structural model to a point cloud coordinate system using the coarse registration matrix, wherein the coarse registration matrix includes a rotation matrix and a translation vector; transforming the preliminary three-dimensional structural model to a point cloud coordinate system using the rotation matrix and the translation vector; extracting a point cloud subset P of the cabinet from the point cloud data; sampling the surface of the cabinet of the preliminary three-dimensional structural model after coordinate system transformation to obtain a sampling point set Q; calculating the nearest neighbor of each point in the sampling point set Q in the point cloud subset P, and constructing a set of corresponding point pairs; calculating the target rotation matrix and target translation vector corresponding to the minimum sum of distances between all corresponding point pairs; updating the position of the preliminary three-dimensional structural model in the point cloud coordinate system based on the target rotation matrix and target translation vector to obtain the three-dimensional structural model.

[0011] Based on the first aspect, in some embodiments, determining the plane of each point cloud object to obtain planar point cloud features includes: for each point cloud object, randomly selecting three points from the point cloud object to construct a plane, calculating the distance from other points to the plane, and determining the number of interior points whose distance to the plane is less than a set threshold; repeating the previous step to construct multiple planes, each plane corresponding to a number of interior points; taking the plane with the most interior points as the main plane of the point cloud object, and taking the normal vector of the main plane, the coordinates of the plane's center point, and the set of interior points constituting the main plane as planar point cloud features.

[0012] Based on the first aspect, in some embodiments, the step of determining the intersection line of adjacent planes or the boundary contour line of the plane based on the plane of each point cloud object to obtain edge line point cloud features includes: calculating the curvature value of each point in the planar point cloud features; determining points with curvature values ​​greater than a threshold as potential edge points; grouping all determined edge points according to connectivity, fitting each group of point clouds into a straight line segment; and using the start coordinates, end coordinates, and direction vector of each straight line segment as edge line point cloud features.

[0013] Based on the first aspect, in some embodiments, the step of calculating the key corner point positions based on planar point cloud features and edge line point cloud features to obtain corner point point cloud features includes: determining preliminary corner points based on the planar point cloud features and the edge line point cloud features; generating local point cloud sets with each preliminary corner point as the center and points within a preset radius of the preliminary corner point; performing principal component analysis on each local point cloud set to determine the main direction of the point cloud distribution; and selecting the point with the smallest weighted sum of squared distances from all points in the local point cloud set as the corner point, wherein each point in the local point cloud set corresponds to a weight, and the weight is inversely proportional to the distance from the point to the main direction of the point cloud distribution.

[0014] Based on the first aspect, in some embodiments, the method for generating three-dimensional drawings of substation secondary circuits for intelligent operation and maintenance further includes: after the secondary circuit is modified, obtaining updated two-dimensional drawings, the updated two-dimensional drawings containing the geometric parameters and installation positions of the newly built equipment; scanning the modified area with a laser scanner to obtain point cloud data of the modified area; and updating the three-dimensional structure model based on the updated two-dimensional drawings and the point cloud data of the modified area.

[0015] Based on the first aspect, in some embodiments, the method for generating three-dimensional drawings of substation secondary circuits for intelligent operation and maintenance further includes: responding to a user's operation command on a target device in a three-dimensional visualization drawing, and verifying the user's operation permission; when the user has operation permission on the target device, performing operation logic verification and operation sequence verification on the operation of the target device; when the operation on the target device conforms to the operation logic and operation sequence, sending the operation command to the target device so that the target device executes the operation command; obtaining the execution result after the target device executes the operation command, and mapping the execution result to the three-dimensional visualization drawing.

[0016] Secondly, embodiments of this application provide a three-dimensional drawing generation system for secondary circuits of substations for intelligent operation and maintenance. The system includes a terminal device, which includes a memory and a processor. The memory stores a computer program that can run on the processor. When the processor executes the computer program, it implements the steps of the three-dimensional drawing generation method for secondary circuits of substations for intelligent operation and maintenance as described in any of the first aspects.

[0017] The beneficial effects of the embodiments of this application compared with the prior art include:

[0018] In this embodiment, a high-precision three-dimensional visualization drawing is constructed by integrating the geometric parameters of two-dimensional drawings with point cloud data. This drawing can fully present the three-dimensional structure of the cabinet, the details of the device panel, and the spatial routing of cables. Maintenance personnel can directly "see through" the inside of the cabinet in the three-dimensional visualization drawing, such as viewing the layout of indicator lights on the device panel, the arrangement of terminal blocks, and the actual laying path of cables in the cable tray, greatly reducing the cognitive burden.

[0019] Furthermore, when secondary circuits are modified, traditional methods require redrawing the entire set of drawings, which is time-consuming and prone to errors. This application uses a handheld scanner to acquire point clouds of the modified area, registers and corrects them with the updated drawings, and updates only the model of the changed area, updating the 3D visualization drawings and displaying the updated 3D visualization drawings to maintenance personnel. Maintenance personnel can then view the 3D visualization drawings to ensure consistency between the drawings and the actual site conditions, avoiding misoperations caused by discrepancies between the drawings and reality.

[0020] When a secondary circuit fails, traditional fault analysis requires repeatedly switching between an SCD viewer, CAD drawings, and a ledger system to determine the fault information, which is inefficient. This application integrates scattered information into a unified 3D visualization drawing. For example, when a user clicks on a protection device, a panel pops up displaying the IED name, a list of logical nodes, and associated virtual terminals; clicking on a terminal block highlights the virtual connection signal flow and cable number corresponding to that terminal; clicking on a cable displays the start and end points and core wire allocation in the cable list. When a secondary circuit fails, the system can automatically highlight the faulty circuit and pop up fault information. All relevant information is presented centrally on the same interface, greatly reducing fault handling time. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a flowchart illustrating the method for generating three-dimensional drawings of substation secondary circuits for intelligent operation and maintenance provided in this application embodiment.

[0023] Figure 2 This is a structural block diagram of the terminal device provided in the embodiments of this application. Detailed Implementation

[0024] The present application will be described more clearly below with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the function of the present application, but do not limit the present application in any way. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present application. These all fall within the protection scope of the present application.

[0025] To make the objectives, technical solutions, and advantages of this application clearer, the following description will be provided in conjunction with the accompanying drawings and specific embodiments.

[0026] SeeFigure 1 The method for generating three-dimensional drawings of substation secondary circuits for intelligent operation and maintenance disclosed in this application may include the following steps:

[0027] Step 101: Obtain the two-dimensional drawings of the substation secondary system, the SCD system configuration file, and the point cloud data.

[0028] The aforementioned two-dimensional drawings may include secondary cabinet layout diagrams, device panel opening diagrams, terminal block wiring diagrams, cable lists, and protection logic diagrams. The aforementioned SCD system configuration file may include IED device lists, logic node configurations, virtual terminal connection relationships, and GOOSE / SV control block information.

[0029] Specifically, the secondary cabinet layout drawing can include the cabinet's floor plan, arrangement order, cabinet number, external dimensions, and spacing within the power distribution room, providing a spatial layout benchmark for the construction of the 3D structural model. The drawing can be imported into modeling software to extract the width, depth, and height dimensions of each cabinet, as well as its installation location coordinates and rotation angle.

[0030] The device panel opening diagram can include the panel outline of each protection device and control device in the secondary cabinet, as well as the opening positions and dimensions of indicator lights, buttons, and LCD screens. Layer separation technology can be used to extract the panel outline and opening layout from the device panel opening diagram. The panel outline includes panel width, height, thickness, and mounting hole positions, while the opening layout includes the type of opening (e.g., round holes for indicator lights, square holes for buttons, display windows), position coordinates, and geometric parameters.

[0031] Terminal block wiring diagrams can include the arrangement of each terminal block, terminal numbers, terminal types, and the wiring relationships between terminals and external cables. OCR technology can be used to extract terminal number text and create a terminal block configuration table, including terminal block number, starting terminal number, ending terminal number, circuit number for each terminal, and connection destination.

[0032] The cable inventory can contain a complete list of cables for the entire station, including cable number, cable type, starting equipment, starting terminal, ending equipment, ending terminal, cable length, and core wire allocation. The cable inventory can be stored in a structured format, creating a physical connection table.

[0033] Protection logic diagrams can include internal logic block diagrams of protection devices, showing the input-output relationships of each protection function, logic gate combinations, and delay settings. Logic diagrams can be transformed into behavioral rule descriptions for subsequent behavioral simulation in digital twin models.

[0034] In this step, an SCD file can be exported from the substation control layer. This file conforms to the IEC 61850 standard and uses XML format to describe the substation's secondary system configuration. First, the IED (Intelligent Electronic Device) section of the SCD file can be parsed to extract the name, model, manufacturer, version number, and bay of each intelligent electronic device. Then, the communication section is parsed to obtain the access point configuration, GOOSE (Generic Object Oriented Substation Event) control block, and SV (Sampled Value) control block information for each IED, including the multicast MAC (Media Access Control) address, APPID (Application Identifier), dataset path, and transmission parameters. Finally, the input section is parsed to extract the virtual terminal connection relationships, establish a signal mapping table between the transmitting and receiving IEDs, and record the signal type and functional description of each virtual connection.

[0035] Specifically, the IED device list includes the name, model, manufacturer, version number, and interval of each intelligent electronic device, creating a device list. The logical node configuration includes the logical nodes within each IED, including overcurrent protection PIOC, distance protection PDIS, circuit breaker XCBR, etc., recording the data objects and data attributes of each logical node.

[0036] Virtual terminal connections are used to describe the logical signal exchange between IED devices, including transmitter information (e.g., transmitter IED name, logical node, data object, and data attributes), receiver information (e.g., receiver IED name, logical node, data object, and data attributes), and signal descriptions (signal type and function description).

[0037] The GOOSE / SV control block information includes multicast MAC address, APPID, dataset path, transmission interval, etc., and is used for visualizing signal flow in the 3D structural model.

[0038] In this step, a laser scanner can be used to scan the secondary circuit of the substation to obtain point cloud data of the secondary circuit.

[0039] Specifically, a measurement control network can be set up in the secondary loop, and the absolute coordinates of the control points can be measured using a total station as the reference for the point cloud coordinates. Spherical targets are deployed, with multiple targets placed in overlapping areas of adjacent stations to ensure accurate stitching of multi-station cloud data. A high-precision 3D laser scanner is used to scan each station, covering all cabinets and cable layers. Close-up, detailed scanning of the terminal block area is performed, scanning from multiple angles to ensure no obstructions; color scanning is enabled on the device panels to acquire RGB information; and high-density scanning is performed on the pressure plate area. After each station scan, the data is previewed to check for voids or noise, confirming complete target scanning. Stations with substandard quality are immediately re-scanned. Multi-station cloud data is automatically matched and stitched based on spherical targets, and isolated noise points are removed through statistical filtering. The point cloud is converted to an absolute coordinate system based on the control point coordinates to obtain high-precision point cloud data.

[0040] Step 102: Construct a three-dimensional structural model of the secondary loop based on two-dimensional drawings and point cloud data, and establish a mapping relationship between the SCD system configuration file and the physical devices of the three-dimensional structural model.

[0041] In this step, a preliminary three-dimensional structural model can be generated from two-dimensional drawings. Then, the preliminary three-dimensional structural model can be corrected using point cloud data. The SCD system configuration file is parsed, and a mapping relationship is established between the parsed information and the corrected three-dimensional structural model. This achieves a precise correspondence between physical entities and logical configurations, enabling the three-dimensional model to have both geometric accuracy and behavioral semantics, and supporting virtual-real interaction.

[0042] In some embodiments, the above-mentioned construction of a three-dimensional structural model of the secondary circuit based on two-dimensional drawings and point cloud data may include: parsing the two-dimensional drawings to obtain structural parameter information of the secondary circuit, which includes geometric parameters and spatial layout information of the secondary cabinet, internal equipment and terminal blocks, as well as cable list and protection logic information; establishing a drawing parameter database based on the structural parameter information of the secondary circuit, which includes: the size and positional relationship of the secondary cabinet and internal equipment, the number and type of terminal blocks, the opening layout and pressure plate type of the internal equipment, cable list and protection logic; performing three-dimensional modeling based on the drawing parameter database to generate a preliminary three-dimensional structural model; extracting key point cloud features from the point cloud data, which includes planar point cloud features, edge line point cloud features and corner point cloud features; registering the key point cloud features with the preliminary three-dimensional structural model, and correcting the preliminary three-dimensional structural model to obtain a three-dimensional structural model.

[0043] The following explains the process of "analyzing two-dimensional drawings to obtain the structural parameter information of secondary circuits" and "establishing a drawing parameter database based on the structural parameter information of secondary circuits".

[0044] First, perform geometric parameter analysis on each layer of the 2D drawings. For cabinets, measure their width, depth, and height, and record the position coordinates of internal columns and mounting rails. For internal equipment such as protection devices and control devices, read the panel dimensions from the opening diagram, extract the type and coordinates of each opening, including indicator light round holes, button square holes, and display window, and record the diameter or side length of each opening. For terminal blocks, count the total number of terminal blocks, record the starting and ending terminal numbers of each terminal block, and classify the quantity by terminal type. For pressure plates, read the pressure plate type and function identifier, and record the installation position.

[0045] Secondly, spatial layout information is extracted from the two-dimensional drawings. The cabinet layout diagram provides the planar coordinates of each cabinet within the room, the cabinet structure diagram provides the installation height and horizontal position of the devices on the cabinet, and the terminal block layout diagram provides the specific location of the terminal blocks at the rear of the cabinet. All spatial coordinates are unified into the same coordinate system to establish a complete spatial layout dataset.

[0046] Next, the extracted geometric parameters and spatial layout information are stored according to a unified data structure to establish a drawing parameter database. The drawing parameter database can adopt a relational structure, containing multiple interrelated data tables: cabinet parameter table, device parameter table, terminal block parameter table, pressure plate parameter table, and connection relationship table.

[0047] For example, the cabinet parameter table records the unique ID, cabinet number, width, depth, height, installation coordinates X, Y, Z, and rotation angle for each cabinet. The device parameter table records the unique ID, device model, cabinet ID, installation height, horizontal position, panel width, panel height, and a JSON-formatted description of the opening layout for each device. The terminal block parameter table records the unique ID, cabinet ID, installation location, total number of terminals, number of current terminals, number of voltage terminals, number of signal terminals, number of power terminals, starting terminal number, and ending terminal number for each terminal block. The pressure plate parameter table records the unique ID, device ID, pressure plate type, function identifier, and installation coordinates for each pressure plate. The connection relationship table records the correspondence between terminal blocks and devices, the circuit number and signal name corresponding to each terminal. The database uses the device ID as the primary key and foreign keys to link the tables, forming a complete device parameter system.

[0048] The following describes the process of "creating a preliminary three-dimensional structural model based on the drawing parameter database".

[0049] First, a parametric template for the cabinet is created, with width, depth, and height set as variables. A cabinet model with specific dimensions is generated based on the parameter values ​​from the drawing parameter database. The cabinet model includes the cabinet frame, front and rear doors, internal columns, and mounting rails. The door hinge positions and opening directions are set according to the drawings.

[0050] Next, models of the device, terminal blocks, and pressure plates are created. Specifically, the corresponding template is called according to the device model, and the panel base is generated according to the panel's external dimensions. The JSON data of the opening layout is traversed to precisely place components such as indicator lights, buttons, and displays on the panel. Indicator lights are represented by cylinders, buttons by cubes, and displays by cubes with transparent materials. The housing is generated by extruding according to the device depth, creating plug-in modules. The total number and type distribution of terminals in the terminal block parameter table are read, and individual terminals are created iteratively at fixed intervals. Current terminals are given a red material, voltage terminals yellow, signal terminals blue, and power terminals black. Terminal numbers are engraved on the side of each terminal. The generated terminal blocks are placed at the rear of the cabinet according to the installation coordinates. The pressure plate model includes a base and a rotatable handle, with the rotation axis and angle range set.

[0051] Finally, all the generated models are assembled into the corresponding cabinets according to the spatial layout information to generate a preliminary three-dimensional structural model.

[0052] In some embodiments, the extraction of key point cloud features from point cloud data may include: inputting the original point cloud into a trained point cloud recognition model to determine the category of each point cloud and obtain a classified point cloud dataset; based on the classified point cloud dataset, using a clustering algorithm to segment connected point clouds of the same category into independent point cloud objects; determining the planes of each point cloud object to obtain planar point cloud features; determining the intersection lines of adjacent planes or the boundary contour lines of the planes based on the planes of each point cloud object to obtain edge line point cloud features; and calculating the key corner point positions based on the planar point cloud features and the edge line point cloud features to obtain corner point cloud features.

[0053] Specifically, point cloud recognition models can employ deep learning-based semantic segmentation networks, such as PointNet++ or RandLA-Net. These networks can directly process unordered point clouds and extract features from each point. The network structure consists of two parts: an encoder and a decoder. The encoder extracts local and global features of the point cloud step by step through multilayer perceptrons and downsampling operations; the decoder recovers the feature dimension of each point through upsampling and feature propagation; and finally, the probability of each point belonging to a category is output through fully connected layers and a softmax function.

[0054] Training this model requires a large amount of labeled substation point cloud data as the training set. The training dataset contains point cloud samples of various categories: cabinet body, device panel, terminal block, pressure plate, cable, ground, wall, etc. Each sample's point cloud is manually labeled, and each point has the correct category label. During training, the model continuously optimizes the network parameters through the backpropagation algorithm, minimizing the cross-entropy loss between the predicted category and the true label.

[0055] The original point cloud is input into the trained model for forward inference to obtain the category label for each point. The output is a classified point cloud dataset, in which each point, in addition to the original 3D coordinate information, also has a category label field, such as "cabinet" for the main body of the cabinet, "panel" for the device panel, and "terminal" for the terminal block.

[0056] Based on the classified point cloud dataset, a clustering algorithm is used to segment connected point clouds of the same category into independent point cloud objects. For each category, instance segmentation is performed separately. Taking the cabinet main category as an example, all points labeled "cabinet" constitute a point cloud set, but it may contain multiple different cabinets. Therefore, a clustering algorithm is needed to separate the point clouds belonging to different cabinets.

[0057] In 3D space, point clouds belonging to the same object are usually close together, while point clouds of different objects have gaps between them. Based on this, a spatial index structure for the point cloud is first constructed. Then, an unvisited point is randomly selected as a seed point, and its neighboring points are searched within a set search radius. The neighboring points are added to the current cluster. The search for neighboring points continues for newly added points until no new points are added. This process is repeated until all points have been visited, and each connected region forms an independent point cloud object, which is assigned a unique object ID.

[0058] During clustering, the search radius parameter needs to be set appropriately. For the main body of the cabinet, the search radius can be set to 10-20 cm to ensure that the point clouds of the same cabinet are connected, while the gaps between different cabinets can be effectively separated. For smaller objects such as device panels and terminal blocks, the search radius should be reduced accordingly. After segmentation, a point cloud dataset with instance IDs is obtained, and each point cloud object corresponds to a specific physical device.

[0059] For example, the process of determining the planes of each point cloud object and obtaining planar point cloud features can include: for each point cloud object, randomly selecting three points from the point cloud object to construct a plane, calculating the distances of other points to the plane, and determining the number of interior points whose distances to the plane are less than a set threshold; repeating the previous step to construct multiple planes, each plane corresponding to a certain number of interior points; taking the plane with the most interior points as the principal plane of the point cloud object, and using the normal vector of the principal plane, the coordinates of the plane's center point, and the set of interior points constituting the principal plane as planar point cloud features.

[0060] For each point cloud object, the RANSAC algorithm can be used for multiple random samplings. Each time, three non-collinear points are randomly selected from the object's point cloud, and these three points determine a plane equation. For cabinet objects, the sampling points may fall on the front, side, or top surface, and the plane direction obtained from each sampling is different; for device panels, since the point cloud is concentrated on a single plane, most sampling points fall on the same plane; for terminal blocks, due to their small size and thinness, the sampling points are easily affected by noise, and a small distance threshold needs to be set.

[0061] For each sampled plane, traverse all point clouds of the object and calculate the spatial distance from each point to the plane. Set a distance threshold based on the object type and point cloud accuracy. For large objects like cabinets, the distance threshold can be set to 2mm-3mm; for device panels, it can be set to 1mm-2mm; and for terminal blocks, it can be set to 0.5-1mm. Points whose spatial distance to the plane is less than the distance threshold are recorded as interior points, and the number of interior points is recorded.

[0062] Compare the number of interior points of all candidate planes, and determine the plane with the most interior points as the main plane of the object. For cabinets, the plane with the most interior points is usually the front with the largest area; for device panels, the main plane is the panel plane; for terminal blocks, the main plane is the narrow strip plane on its front.

[0063] For example, the above method of determining the intersection line or boundary contour line of adjacent planes based on the plane of each point cloud object to obtain edge line point cloud features may include: calculating the curvature value of each point in the planar point cloud features; identifying points with curvature values ​​greater than a threshold as potential edge points; grouping all identified edge points according to connectivity, fitting each group of point clouds into a straight line segment; and using the start coordinates, end coordinates, and direction vector of each straight line segment as edge line point cloud features.

[0064] For each point in the planar point cloud feature, its local curvature value is calculated. Principal component analysis is used: with the current point as the center, neighboring points within a preset radius r (e.g., 5-10 mm) are searched, a covariance matrix is ​​constructed, and the eigenvalues ​​λ1, λ2, and λ3 (λ1≥λ2≥λ3) are solved. The curvature calculation formula is σ = λ3 / (λ1+λ2+λ3), which reflects the degree of curvature of the local surface of the point cloud. The curvature of points inside the plane is close to 0, while the curvature at the edges and corners increases significantly.

[0065] A threshold T (typically 0.1-0.2) is set based on the curvature distribution. Points with curvature values ​​greater than T are marked as potential edge points. These points are located at object edges, abrupt changes in surface curvature, or planar boundaries. For isolated high-curvature points caused by noise, preliminary filtering is performed using neighborhood consistency to retain geometrically meaningful edge candidate points.

[0066] Euclidean clustering is used to group adjacent edge points into the same group. A spatial index is constructed, and a cluster radius (typically 2-5 mm) is set. Starting from any unvisited edge point, neighboring points are searched, recursively growing to form connected regions. The same group of point clouds corresponds to the same continuous edge line, and different edge lines are naturally segmented into independent point sets. A straight line segment is fitted to each group of edge point clouds. Two points are randomly selected to determine the straight line model, and the distances from other points to this line are calculated, counting the number of interior points. The straight line model with the most interior points is iteratively selected as the optimal fit. The endpoints of the straight line segment are determined based on the minimum and maximum projection parameters of the interior point set, and the direction vector is calculated.

[0067] For example, the above calculation of key corner point positions based on planar point cloud features and edge line point cloud features to obtain corner point point cloud features may include: determining preliminary corner points based on planar point cloud features and edge line point cloud features; generating local point clouds around each preliminary corner point with points within a preset radius of that preliminary corner point; performing principal component analysis on each local point cloud to determine the main direction of the point cloud distribution; and selecting the point with the smallest weighted sum of squared distances from all points in the local point cloud as the corner point, where each point in the local point cloud corresponds to a weight that is inversely proportional to the distance from the point to the main direction of the point cloud distribution.

[0068] Specifically, for a screen cabinet, eight corner points are calculated using the intersection lines of the front plane, left and right sides, and top and bottom surfaces. Taking a screen cabinet as an example, the front plane P is obtained. f Right side P r Top surface P t Calculate P f With P r Intersection line L1, P f With P t Intersection line L2, P r With P t The intersection of line L3 and the three lines is the upper right corner point. Similarly, calculate the remaining seven corner points, initially labeled P. init For the device panel, the four corner points are calculated using the panel plane and the four boundary edge lines: the intersection of the top edge and the left edge is the top left corner, and so on. For the terminal block, the endpoints of the top and bottom edge lines are used directly as the initial corner points.

[0069] With each initial corner point P init Centered on a point cloud, a search radius R is set, and all points within this radius are extracted from the original point cloud to form a local point cloud set. Statistical filtering is then used to remove isolated noise points from the local point cloud set, retaining the clean corner point regions.

[0070] Construct the covariance matrix for the local point cloud and solve for the eigenvalues ​​λ1, λ2, and λ3 (λ1≥λ2≥λ3) and their eigenvectors v1, v2, and v3. The v1 corresponding to the largest eigenvalue λ1 represents the principal direction of the edge near the corner point, i.e., the intersection direction of the three edges.

[0071] Establish a weighted objective function min Σw i ||P opt - P i || 2 Weight w i According to point P i The distance to the principal direction edge is determined by: w i = exp(-d i 2 / σ 2 ), where d i Let be the distance from point i to the edge along the direction v1; the closer the distance, the greater the weight. The precise corner coordinates P are obtained by solving the weighted objective function. opt This makes the corner points closer to the feature points of the edge rather than simply the centroid.

[0072] In some embodiments, the above-mentioned registration of key point cloud features with a preliminary 3D structural model and the correction of the preliminary 3D structural model to obtain a 3D structural model may include: extracting model feature points of the same type as the key point cloud features from the preliminary 3D structural model; calculating a coarse registration matrix based on the model feature points and key point cloud features, and using the coarse registration matrix to transform the preliminary 3D structural model to a point cloud coordinate system, the coarse registration matrix containing a rotation matrix and a translation vector, and transforming the preliminary 3D structural model to a point cloud coordinate system using the rotation matrix and translation vector; extracting a point cloud subset P of the cabinet from the point cloud data, and sampling the surface of the cabinet of the preliminary 3D structural model after coordinate system transformation to obtain a sampling point set Q; calculating the nearest neighbor of each point in the sampling point set Q in the point cloud subset P, and constructing a set of corresponding point pairs; calculating the target rotation matrix and target translation vector corresponding to the minimum sum of distances between all corresponding point pairs; updating the position of the preliminary 3D structural model in the point cloud coordinate system based on the target rotation matrix and target translation vector, and obtaining the 3D structural model.

[0073] Specifically, model feature points of the same type as the key point cloud features can be extracted from the preliminary 3D structural model. For a cabinet object, the 3D coordinates of eight corner points are extracted from the model; for a device panel, the coordinates of four corner points are extracted from the model; and for a terminal block, the coordinates of both end points are extracted from the model. These model feature points constitute the initial spatial reference of the model.

[0074] Select 3-5 evenly distributed and distinctive pairs of corresponding points, such as the lower left, lower right, and upper left corners of the cabinet. Each pair of corresponding points contains the coordinates p of a model feature point. i And the coordinates q of a key corner point in a point cloud i Based on the selected point pairs, the coarse registration matrix is ​​calculated using the singular value decomposition method: first, the center points of the two point sets are calculated. and Then construct the covariance matrix. Perform singular value decomposition on H. Calculate the rotation matrix Translation vector The resulting rotation matrix R and translation vector T are the coarse registration matrix. This matrix is ​​used to transform the preliminary 3D structural model to the point cloud coordinate system, achieving initial alignment between the model and the point cloud.

[0075] Extract the point cloud subset P corresponding to each cabinet from the preprocessed point cloud data. Based on the semantic segmentation results, select all point clouds belonging to the target cabinet, ensuring that the point cloud subset completely covers all surfaces of the cabinet, including the front, sides, top, and bottom. Downsample the point cloud subset to maintain uniform density, facilitating subsequent nearest neighbor search.

[0076] For the preliminary 3D structural model after coordinate system transformation, the 3D model surface of the target cabinet is extracted, and a sampling point set Q is generated using a uniform sampling algorithm. The sampling density is set to be comparable to the point cloud subset P, and the sampling points are evenly distributed on all outer surfaces of the model, including the front of the cabinet, the device panel, and the terminal block area, to ensure that the sampling point set can fully represent the geometric features of the model.

[0077] For each point q in the sampling point set Q i Search for its nearest neighbor in the point cloud subset P, and calculate q. i The point with the smallest Euclidean distance to each candidate point in P is selected as the nearest neighbor point p. i Construct corresponding point pairs (q) i , p i ).

[0078] To improve registration accuracy and robustness, the constructed corresponding point pairs are screened. The initial distance of all point pairs is calculated, and pairs with excessively large or obviously incorrect distances are discarded, retaining those with distances less than a set threshold (e.g., 10 cm). Simultaneously, a normal vector consistency check is performed, and q is calculated. i The model normal vector at p i The point cloud normal vectors at a given location are used to eliminate point pairs where the angle between the normal vectors is greater than a threshold, ensuring that corresponding point pairs have similar geometric properties.

[0079] Based on the filtered set of corresponding point pairs, an iterative nearest-point algorithm is used for fine registration. In each iteration, the objective function E = ∑||R×q is solved. i + T - p i The objective function is solved using the minimum rotation matrix R and translation vector T. The singular value decomposition method is used to solve the objective function: first, the center points of all corresponding point pairs are calculated. 'and Construct the covariance matrix H' = ∑(qi - ')(p i - ')ᵀ, Singular value decomposition of H' yields R k and T k Applying the solution to R k and T k Update the position of all sampling points Q: Q = R k ×Q + T k Reconstruct the corresponding point pairs, calculate the new nearest neighbors, and solve for the rotation matrix and translation vector again. Repeat the iteration until the root mean square error change between two consecutive iterations is less than a set threshold (e.g., 0.1 mm) or the maximum number of iterations is reached.

[0080] After the iteration is complete, the final rotation matrix R is obtained. final Translation vector T final The coarse registration matrix and the rotation matrix R obtained through iterative optimization are then compared. final Translation vector T final Combining the transformations, we obtain the complete transformation matrix from the original model coordinate system to the point cloud coordinate system: R total = R final ×R init T total = R final ×T init + T final R init Let T be the initial rotation matrix. init This is the initial translation vector. This transformation matrix is ​​applied to update the position and orientation of all devices in the initial 3D structural model, generating the corrected 3D structural model.

[0081] The following explains the process of "establishing the mapping relationship between the SCD system configuration file and the physical devices of the three-dimensional structural model".

[0082] Parse the SCD file to obtain the IED device list, including the name, model, and manufacturer of each intelligent electronic device. Add an "IEDName" attribute field to the intelligent devices such as protection devices, monitoring and control devices, and intelligent terminals in the 3D structural model, and establish a device mapping table by mapping the IED names in the SCD to the physical devices one by one.

[0083] Analyze the configuration of each IED's internal logical nodes, such as overcurrent protection PIOC (Phase Instantaneous Overcurrent) and circuit breakers. Associate the logical nodes with specific components of the device: map the device panel indicator lights to LPHD (Logical Physical Device) logical nodes, map the LCD screen display content to protection logical nodes such as PDIS (Phase Distance Protection), and map the device backplane interfaces to communication-related logical nodes.

[0084] The virtual terminal connection relationships are analyzed to obtain the transmitting and receiving information for each signal. A unique identifier is established for each physical terminal in the 3D structural model, associating the virtual connection with the specific terminal. For example, the trip output virtual terminal of the protection device is mapped to the corresponding terminal of the cabinet terminal block.

[0085] Parse GOOSE / SV control block information, including APPID, MAC address, and dataset path. Add control block attributes to devices that support communication, associating the control block with the device's fiber optic interface or network port.

[0086] Through the above mapping relationship, each physical device in the 3D structural model can be associated with complete logical configuration information, supporting querying IED parameters, virtual terminal connections and communication configurations from the model.

[0087] Step 103: Generate a 3D visualization drawing of the substation's secondary circuit based on the 3D structural model.

[0088] First, the 3D visualization drawings can be divided into four levels of views according to spatial scale: station-level view, cabinet-level view, device-level view, and circuit-level view. The station-level view displays the overall layout of the substation's secondary circuits, specifically showing the layout of all cabinets and cable corridors. The cabinet-level view displays the cabinet layout, specifically showing the three-dimensional arrangement of internal devices, terminal blocks, and pressure plates. The device-level view displays the details of the devices within the cabinets, specifically showing panel indicator lights, buttons, LCD screen details, as well as the device housing, heat dissipation holes, communication interfaces, and other structural elements. The circuit-level view displays cable paths and signal flows. Views at each level can be seamlessly switched through zooming and sectioning. When maintenance personnel enter the secondary circuit, the system displays the corresponding view based on the distance from the user to the secondary cabinet. As the distance to the secondary cabinet decreases, the system sequentially displays the station-level view, cabinet-level view, and device-level view.

[0089] When maintenance personnel need to operate on terminals or check cable routes, the system displays a loop-level view. For example, when maintenance personnel need to verify specific terminal wiring, measure terminal voltage and current, or locate loose terminals, they need to approach the terminal block within 1 meter. The system automatically switches to a loop-level view, displaying the terminal number, loop number, and cable connection relationship. When it is necessary to find the cable route, verify the core wire allocation, or troubleshoot cable breaks, approaching the cable or clicking on the cable model will display the complete cable path, intermediate joint locations, and opposite terminals.

[0090] Secondly, corresponding annotation information is automatically generated based on the spatial location of the 3D structural model. For example, the name and model number are labeled next to equipment, the terminal number and circuit number are labeled next to terminals, and the cable number and core wire assignment are labeled next to cables. When annotating a certain device, terminal, or cable, multiple candidate annotation positions are generated at equal intervals in a 360° direction around the device, terminal, or cable as the center. Rays are emitted from each candidate position to the device, terminal, or cable, and it is detected whether the rays intersect with other models, recording any occluded candidate points. Unoccluded candidate points are scored according to rules such as closest distance and best viewing angle, and the optimal position is selected for annotation. In addition, the annotation density can be dynamically controlled according to the view level. The station-level view can only display the cabinet number, the cabinet-level view can display the device name and terminal block number, the device-level view can display all terminal annotations, and the circuit-level view can display the terminal number and circuit number. The station-level view covers the entire station area. If all terminal annotations are displayed, the annotations will overlap densely and become completely unrecognizable. Displaying only the cabinet number keeps the view simple and facilitates quick location of the target cabinet. The device-level view focuses on a single cabinet, with a short viewing distance and small field of view, but there is enough space to display terminal labels. Maintenance personnel need to check the detailed terminal information to verify wiring and troubleshoot when inspecting specific devices.

[0091] For example, when maintenance personnel wearing MR (Mixed Reality) glasses enter the relay protection room, they are initially in a station-level view (distance from the secondary cabinets > 10 meters). The MR glasses display the outlines and numbers of all cabinets. When the distance to the secondary cabinets decreases to 8 meters, the system automatically switches to a cabinet-level view: the cabinet front unfolds, and the internal devices, terminal blocks, pressure plates, and other equipment are presented in 3D. Device names and terminal block numbers are displayed floatingly, allowing maintenance personnel to quickly locate the target device based on the view displayed in the MR glasses.

[0092] Next, the virtual terminals obtained from SCD parsing are visualized and configured in a 3D drawing. Specifically, different colored lines can represent signal types, such as red for tripping, yellow for alarms, and green for normal operation; dynamic arrows can represent signal flow, with the arrow's movement speed reflecting signal priority; and users can click on the connection system to bring up an information panel displaying configuration information such as the sender, receiver, and APPID.

[0093] Next, the connection relationships in the cable inventory are generated into a three-dimensional cable path. Specifically, the path can be automatically planned along the cable tray based on the coordinates of the starting and ending terminals, generating a curved tubular model; the cable sheath color can be distinguished by voltage level, and the core wires are colored according to a standard color spectrum; labels are generated at both ends of the cable to display the cable number and core wire allocation.

[0094] In addition, a real-time data interface needs to be reserved to map SCADA (Supervisory Control And Data Acquisition) data into the 3D structural model. Specifically, indicator lights can change color according to real-time status, the LCD screen displays the current measurement value, and the pressure plate rotates into position according to the actual status. When the equipment malfunctions, it will automatically highlight and display an alarm message.

[0095] Finally, LOD (Level of Detail) technology can be used to generate multi-precision versions of 3D visualization drawings. When the user is far away from the secondary display cabinet, a simplified model of the 3D visualization drawing is used; when the user is close to the secondary display cabinet, details are loaded into the simplified model of the 3D visualization drawing.

[0096] In some embodiments, the above-mentioned method for generating three-dimensional drawings of substation secondary circuits for intelligent operation and maintenance may further include: after the secondary circuit is modified, obtaining updated two-dimensional drawings, the updated two-dimensional drawings containing the geometric parameters and installation location of the newly built equipment; scanning the modified area with a laser scanner to obtain point cloud data of the modified area; and updating the three-dimensional structural model based on the updated two-dimensional drawings and the point cloud data of the modified area.

[0097] For example, when a secondary circuit is modified, updated 2D drawings can be obtained from the design department, including the geometric parameters and installation locations of the new equipment. Simultaneously, a handheld laser scanner is used to perform a localized scan of the modified area, acquiring point cloud data of the changed parts, covering areas such as newly built cabinets, new devices, and modified terminal blocks. By comparing the old and new 2D drawings, a list of changed equipment is extracted, including the model, quantity, and installation location of new equipment, a list of removed equipment, and changes in the connection relationships of reconnected cables.

[0098] Based on the geometric parameters in the updated 2D drawings, a 3D model of the new equipment is generated. For new protection devices, a device model with indicator lights and buttons is generated based on the panel opening diagram; for new terminal blocks, numbered terminal individuals are automatically generated based on the terminal configuration table; for new cables, a 3D path is generated based on the cable list.

[0099] Register the newly scanned point cloud of the modification area with the original overall point cloud to ensure coordinate system consistency. Extract key features (planes, edge lines, corner points) from the point cloud of the modification area, calculate the precise installation position of the new equipment, correct the deviation between the design position on the drawing and the actual installation, and ensure that the model is aligned with the reality.

[0100] The update operation is performed in the 3D structural model: dismantled equipment is marked as "dismantled" in the model and displayed in grayscale, preserving historical data; new equipment is inserted into the model at the corrected and precise location; equipment requiring reconnection has its terminal attributes and cable paths adjusted according to the new connection relationships. The update process only processes the changed areas, keeping the rest of the model unchanged to ensure update efficiency.

[0101] Finally, the mapping relationship between the SCD configuration file and the model is updated synchronously: IED name and logical node configuration are added for newly created devices; the virtual terminal connection table is updated to associate the new virtual terminals with the corresponding physical terminals; and the correspondence between GOOSE / SV control blocks and communication interfaces is adjusted.

[0102] In some embodiments, the above-described method for generating three-dimensional drawings of substation secondary circuits for intelligent operation and maintenance may further include: responding to a user's operation command on a target device in a three-dimensional visualization drawing and verifying the user's operation permission; when the user has operation permission on the target device, performing operation logic verification and operation sequence verification on the operation of the target device; when the operation on the target device conforms to the operation logic and operation sequence, sending the operation command to the target device so that the target device executes the operation command; obtaining the execution result after the target device executes the operation command, and mapping the execution result to the three-dimensional visualization drawing.

[0103] For example, the system monitors user interactions in real time within a 3D visualization drawing. When it detects a user operating a target device (such as a protection device control panel, air switch, circuit breaker, etc.), it captures the operation command, including the object ID, operation type, and operation parameters. It then queries the user permission database to verify if the current user's role grants them permission to operate on the device. If the user lacks permission, the operation is immediately terminated, and a "No permission" message is sent to the user. If the user has permission, the operation undergoes logical validation.

[0104] The system acquires the current status of the target equipment (e.g., the current engagement / disengagement status of the pressure plate, the current open / closed status of the circuit breaker) and verifies the rationality of the operation logic: for example, performing an engagement operation on an already engaged pressure plate is considered a logical error; performing a closing operation on an operational circuit breaker is considered a dangerous operation. Simultaneously, it checks the standard operating procedures in conjunction with the current maintenance work order to verify the correctness of the operation sequence: for example, operating the equipment before obtaining a work permit, or operating the circuit breaker before disconnecting the isolating switch, is considered a sequence violation. If any operation logic fails verification, the operation is terminated and an error message is pushed out.

[0105] After the operation logic verification is passed, for remotely operable devices (such as protection devices, circuit breakers, etc.), the operation instructions are converted into a protocol format that the target device can recognize, and control commands are sent to the physical device. After receiving the commands, the physical device executes the corresponding operations, and the execution results returned by the physical device are monitored. For devices that require on-site operation (such as pressure plates, some air switches, etc.), operation instructions are displayed to maintenance personnel, and changes in device status are detected through sensors. The execution results and the detected device status are mapped in real time to a 3D visualization drawing to achieve synchronization between virtual and real status.

[0106] For example, the above-mentioned method for generating three-dimensional drawings of substation secondary circuits for intelligent operation and maintenance may further include: obtaining the real-time location information of the user during maintenance, and mapping the user's operation to a three-dimensional visualization drawing based on the user's real-time location information.

[0107] Specifically, multiple UWB (Ultra-Wide Band) base stations and cameras can be set up within the secondary loop site. UWB tags and IMU (Inertial Measurement Unit) sensors are placed on the user's wearable device. The user's real-time location information is determined through the multiple UWB base stations and UWB tags, and the user's body posture is determined through the IMU sensors. Visual synchronous positioning and mapping technology is used through the multiple cameras to identify and track the position of the tool and the user's hand in real time. The user's real-time location information and body posture, as well as the positions of the tool and the user's hand, are mapped onto a 3D visualization map. Based on the user's real-time location information, body posture, tool position, and user's hand position, the user's operation information is identified.

[0108] First, at least four benchmark points can be set up within the secondary circuit training area, and their three-dimensional coordinates can be measured using a high-precision total station to establish an absolute reference system in space. Visual feature markers should be placed at key operating points of the secondary control cabinet, such as cabinet door handles, device panel buttons, pressure plate handles, terminal block terminals, and air switch levers, using QR codes or ArUco markings for easy identification by the vision system. The training area should be divided into grid cells, with a local coordinate system established for each cell, and rapid coarse positioning achieved by combining UWB positioning.

[0109] The coordinate system of the 3D visualization drawing is converted to the same engineering coordinate system as the physical space coordinate system to ensure strict correspondence between position and scale. The locations of feature points corresponding to the physical space are marked in the 3D visualization drawing, including each reference calibration point and each visual marker point, establishing a one-to-one mapping relationship and generating a coordinate transformation matrix. The operable areas of each device are predefined in the 3D visualization drawing, such as pressure plate operation areas, button pressing areas, terminal measurement areas, as well as spatial constraints such as safety distance ranges and tool placement positions.

[0110] Secondly, multiple UWB base stations can be deployed around the site to form a positioning network. Users wear helmets or vests with integrated UWB tags, which transmit pulse signals at a preset frequency. After receiving the signals, the base stations use a time difference of arrival (TDOA) algorithm to calculate the tag's three-dimensional coordinates, achieving real-time positioning of the user's head. Combined with site grid division, this enables rapid positioning of the user's location within the designated cabinet area.

[0111] Nine-axis IMU sensors can be worn on key areas such as the user's helmet, wrists, and ankles to measure the attitude angles, acceleration, and angular velocity of each part in real time. By fusing data from accelerometers, gyroscopes, and magnetometers using extended Kalman filtering, the three-dimensional attitude quaternions of each sensor are calculated. Based on the attitudes of key points such as hands, feet, and head, the angles of the entire skeletal joints are inferred to construct a complete user posture.

[0112] Additionally, multiple high-definition cameras can be deployed on-site to cover all secondary display cabinet operation areas. Visual SLAM (Simultaneous Localization and Mapping) technology is used to identify and track the user's hand and tool positions in real time, detect key hand points, and identify finger joint positions; the six-DOF pose of the tool is tracked through QR code recognition or feature point matching.

[0113] A miniature eye tracker is integrated into the user's helmet or glasses. It captures the pupil and corneal reflection points through an infrared camera, calculates the user's gaze direction vector, and then combines it with the head orientation measured by the head IMU to calculate the user's actual gaze point in the world coordinate system, thereby determining the specific device component that the user is observing.

[0114] Next, the quaternions of poses calculated by the IMU for each part are directly applied to the joint rotations of the corresponding virtual character in the 3D visualization drawing, driving the virtual human to achieve movements completely synchronized with the user. The calculated user gaze direction vector is drawn as a visual gaze ray in the 3D visualization drawing, emanating from the virtual character's eyes and highlighted at its intersection with the device model, intuitively showing the user's focus of attention. In addition, during multi-user collaborative operation and maintenance, each user's gaze can be represented by a different color, making it easier to understand the distribution of each user's attention.

[0115] A mapping table between physical and virtual tools is established, mapping the position, posture, and motion state of physical tools to virtual tools in real time. Virtual tool models are then bound to corresponding positions on a virtual hand, maintaining the same relative positional relationship as the real tools. When the user changes tools, the corresponding virtual tool model is automatically switched through visual recognition.

[0116] Finally, precise collision bodies are constructed for all operable devices in the 3D visualization drawing, including pressure plate handles, device buttons, terminal blocks, and circuit breaker levers. Collision detection is triggered when the distance between the user's hand or tool and the collision body is less than a threshold. The operation type is identified based on the collision location and the movement state of the hand / tool: a hand approaching the pressure plate handle and pinching and rotating is identified as a pressure plate engagement / disengagement operation; a finger approaching a button and pressing it is identified as a button operation; a multimeter probe approaching a terminal and making contact is identified as a measurement operation; and a screwdriver approaching a terminal and rotating it is identified as a wiring operation.

[0117] Force sensors or six-dimensional force sensors can be integrated into key tools to measure the user's operating force in real time, such as the torque of tightening screws or the force applied to pressing buttons. The force level is visually indicated in a 3D visualization drawing; for example, bolts change color according to torque (green indicates acceptable torque, red indicates too tight or too loose), and buttons show varying degrees of indentation based on the pressure applied. After operation, the system automatically evaluates whether the operating force conforms to standard procedures, such as whether the bolt tightening torque is within the specified range, and records the evaluation results.

[0118] When a user initiates a specific operation, such as tightening a bolt, the 3D visualization can play a pre-set, high-quality animation that is synchronized with the user's actual actions in time. The animation enhances the visual representation of the operation process, such as the number of bolt rotations and the displacement of the nut, while concealing any motion jitter that may be caused by sensor accuracy limitations, thus improving the visual experience.

[0119] In addition, it can record the timing data of all user operations in real time, including operation time, operating device, operation type, operation location, operation force, operation duration, as well as the user's movement trajectory and changes in eye focus. The user's operations are compared with preset standard operating procedures to determine the completeness rate of steps, the correctness rate of operation sequence, and the compliance rate of operation force. For example, for pressure plate operations, it checks whether the pressure and release are performed in the prescribed order; for terminal measurements, it checks whether the probe contact positions are correct; for wiring operations, it checks whether the torque is within the standard range.

[0120] In some embodiments, this application can also map the faulty physical equipment onto a 3D visualization drawing, shortening the fault location time. Specifically, SCADA data and protection action signals can be analyzed. After an anomaly is detected, correlation analysis is performed based on the virtual loop topology in the 3D visualization drawing, traversing the signal path to locate the fault source equipment and specific loop, and determining the 3D coordinates of the fault location. After receiving the fault location information, the corresponding cable model and fault terminal model can be displayed prominently in the 3D visualization drawing. At the same time, the view is focused on the fault area, and an information panel pops up to display the fault type and handling suggestions, achieving a rapid response from fault detection to visual location.

[0121] In this embodiment, a high-precision three-dimensional visualization drawing is constructed by integrating the geometric parameters of two-dimensional drawings with point cloud data. This drawing can fully present the three-dimensional structure of the cabinet, the details of the device panel, and the spatial routing of cables. Maintenance personnel can directly "see through" the inside of the cabinet in the three-dimensional visualization drawing, such as viewing the layout of indicator lights on the device panel, the arrangement of terminal blocks, and the actual laying path of cables in the cable tray, greatly reducing the cognitive burden.

[0122] Furthermore, when secondary circuits are modified, traditional methods require redrawing the entire set of drawings, which is time-consuming and prone to errors. This application uses a handheld scanner to acquire point clouds of the modified area, registers and corrects them with the updated drawings, and updates only the model of the changed area, updating the 3D visualization drawings and displaying the updated 3D visualization drawings to maintenance personnel. Maintenance personnel can then view the 3D visualization drawings to ensure consistency between the drawings and the actual site conditions, avoiding misoperations caused by discrepancies between the drawings and reality.

[0123] When a secondary circuit fails, traditional fault analysis requires repeatedly switching between an SCD viewer, CAD drawings, and a ledger system to determine the fault information, which is inefficient. This application integrates scattered information into a unified 3D visualization drawing. For example, when a user clicks on a protection device, a panel pops up displaying the IED name, a list of logical nodes, and associated virtual terminals; clicking on a terminal block highlights the virtual connection signal flow and cable number corresponding to that terminal; clicking on a cable displays the start and end points and core wire allocation in the cable list. When a secondary circuit fails, the system can automatically highlight the faulty circuit and pop up fault information. All relevant information is presented centrally on the same interface, greatly reducing fault handling time.

[0124] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0125] In addition, this application also provides a three-dimensional drawing generation system for substation secondary circuits for intelligent operation and maintenance. This system includes terminal equipment, such as... Figure 2 As shown, the terminal device includes a processor 210 and a memory 220. The memory 220 stores computer programs that can run on the processor 210, such as a substation secondary circuit 3D drawing generation program for intelligent operation and maintenance. When the processor 210 executes the above-mentioned computer program, it implements the steps in the above-described embodiment of the substation secondary circuit 3D drawing generation method for intelligent operation and maintenance, for example... Figure 1 Steps 101 to 103 are shown.

[0126] The terminal device can be a desktop computer, laptop, or cloud server, etc. The terminal device may include, but is not limited to, processor 210 and memory 220. Those skilled in the art will understand that... Figure 2 This is merely an example of a terminal device and does not constitute a limitation on the terminal device. It may include more or fewer components than shown in the figure, or combine certain components, or different components. For example, a terminal device may also include input / output devices, network access devices, buses, etc.

[0127] The processor 210 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0128] The memory 220 can be an internal storage unit of the terminal device, such as the hard drive or RAM of the terminal device. The memory 220 can also be an external storage device of the terminal device, such as a plug-in hard drive, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card. Furthermore, the memory 220 can include both internal and external storage units of the terminal device. The memory 220 is used to store computer programs and other programs and data required by the terminal device. The memory 220 can also be used to temporarily store data that has been output or will be output.

[0129] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for generating three-dimensional drawings of secondary circuits in substations for intelligent operation and maintenance, characterized in that, include: The substation secondary system acquires two-dimensional drawings, SCD system configuration files, and point cloud data. The two-dimensional drawings include secondary cabinet layout diagrams, device panel opening diagrams, terminal block wiring diagrams, cable lists, and protection logic diagrams. The SCD system configuration files include IED device lists, logic node configurations, virtual terminal connection relationships, and GOOSE / SV control block information. Based on the two-dimensional drawings and the point cloud data, a three-dimensional structural model of the secondary loop is constructed, and a mapping relationship is established between the SCD system configuration file and the physical devices of the three-dimensional structural model. Based on the three-dimensional structural model, generate three-dimensional visualization drawings of the substation's secondary circuit.

2. The method for generating three-dimensional drawings of substation secondary circuits for intelligent operation and maintenance according to claim 1, characterized in that, The construction of a three-dimensional structural model of a secondary loop based on the two-dimensional drawing and the point cloud data includes: The structural parameter information of the secondary circuit is obtained by analyzing the two-dimensional drawings. This structural parameter information includes the geometric parameters and spatial layout information of the secondary cabinet, internal equipment and terminal blocks, as well as the cable list and protection logic information. A drawing parameter database is established based on the structural parameter information of the secondary circuit. The drawing parameter database includes: the size and positional relationship of the secondary cabinet and its internal equipment, the number and type of terminal blocks, the opening layout and pressure plate type of the internal equipment, the cable list and protection logic. Based on the drawing parameter database, perform 3D modeling to generate a preliminary 3D structural model; Key point cloud features are extracted from the point cloud data, including planar point cloud features, edge line point cloud features, and corner point cloud features. The key point cloud features are registered with the preliminary three-dimensional structure model, and the preliminary three-dimensional structure model is corrected to obtain the three-dimensional structure model.

3. The method for generating three-dimensional drawings of substation secondary circuits for intelligent operation and maintenance according to claim 2, characterized in that, The extraction of key point cloud features from the point cloud data includes: The original point cloud is input into the trained point cloud recognition model to determine the category of each point cloud, thus obtaining the classified point cloud dataset. Based on the classified point cloud dataset, a clustering algorithm is used to segment connected point clouds of the same category into independent point cloud objects; Define the plane of each point cloud object to obtain the planar point cloud features; Based on the planes of each point cloud object, determine the intersection line of adjacent planes or the boundary contour line of the plane to obtain the edge line point cloud features; The key corner point positions are calculated based on the planar point cloud features and the edge line point cloud features, thus obtaining the corner point point cloud features.

4. The method for generating three-dimensional drawings of substation secondary circuits for intelligent operation and maintenance according to claim 3, characterized in that, The step of registering the key point cloud features with the preliminary 3D structure model and correcting the preliminary 3D structure model to obtain the 3D structure model includes: Extract model feature points of the same type as the key point cloud features from the preliminary three-dimensional structural model; A coarse registration matrix is ​​calculated based on the model feature points and the key point cloud features. The coarse registration matrix is ​​then used to transform the preliminary three-dimensional structure model to the point cloud coordinate system. The coarse registration matrix includes a rotation matrix and a translation vector. The preliminary three-dimensional structure model is then transformed to the point cloud coordinate system using the rotation matrix and the translation vector. Extract the point cloud subset P of the cabinet from the point cloud data, and sample the surface of the cabinet of the preliminary three-dimensional structural model after coordinate system transformation to obtain the sampling point set Q; Calculate the nearest neighbor of each point in the sampling point set Q in the point cloud subset P, and construct the corresponding set of point pairs; Calculate the target rotation matrix and target translation vector that minimize the sum of distances between all corresponding point pairs; The position of the preliminary three-dimensional structural model in the point cloud coordinate system is updated based on the target rotation matrix and the target translation vector to obtain the three-dimensional structural model.

5. The method for generating three-dimensional drawings of substation secondary circuits for intelligent operation and maintenance according to claim 1, characterized in that, The process of determining the plane of each point cloud object to obtain planar point cloud features includes: For each point cloud object, three points are randomly selected from the point cloud object to construct a plane, the distances from other points to the plane are calculated, and the number of interior points whose distances to the plane are less than a set threshold are determined. Repeat the previous step to construct multiple planes, each plane corresponding to a certain number of interior points; The plane with the most interior points is taken as the principal plane of the point cloud object, and the normal vector of the principal plane, the coordinates of the plane's center point, and the set of interior points that constitute the principal plane are taken as the features of the planar point cloud.

6. The method for generating three-dimensional drawings of substation secondary circuits for intelligent operation and maintenance according to claim 5, characterized in that, The step of determining the intersection line of adjacent planes or the boundary contour line of the plane based on the plane of each point cloud object to obtain the edge line point cloud features includes: Calculate the curvature value of each point in the planar point cloud feature; Points with curvature values ​​greater than a threshold are identified as potential edge points; Group all identified edge points according to connectivity, and fit each group of point clouds into a straight line segment; The starting point coordinates, ending point coordinates, and direction vector of each straight line segment are used as edge line point cloud features.

7. The method for generating three-dimensional drawings of substation secondary circuits for intelligent operation and maintenance according to claim 6, characterized in that, The calculation of key corner point positions based on planar point cloud features and edge line point cloud features to obtain corner point point cloud features includes: Preliminary corner points are determined based on the planar point cloud features and the edge line point cloud features; Using each initial corner point as the center, generate a local point cloud from the points within a preset radius of that initial corner point; Principal component analysis is performed on each local point cloud to determine the main direction of the point cloud distribution; The point with the smallest weighted sum of squared distances from all points in the local point cloud set is taken as the corner point. Each point in the local point cloud set corresponds to a weight, and the weight is inversely proportional to the distance from the point to the main direction of the point cloud distribution.

8. The method for generating three-dimensional drawings of substation secondary circuits for intelligent operation and maintenance according to claim 1, characterized in that, The method for generating three-dimensional drawings of substation secondary circuits for intelligent operation and maintenance also includes: After the secondary circuit is modified, the updated two-dimensional drawings are obtained. The updated two-dimensional drawings contain the geometric parameters and installation location of the new equipment. The area to be modified is scanned using a laser scanner to obtain point cloud data of the area. The three-dimensional structural model is updated based on the updated two-dimensional drawings and the point cloud data of the modified area.

9. The method for generating three-dimensional drawings of substation secondary circuits for intelligent operation and maintenance according to claim 1, characterized in that, The method for generating three-dimensional drawings of substation secondary circuits for intelligent operation and maintenance also includes: Respond to user commands for operating the target device in the 3D visualization drawing and verify the user's operating permissions; When a user has operating permissions for the target device, the operation logic and operation sequence will be verified. When the operation on the target device conforms to the operation logic and operation sequence, the operation instruction is sent to the target device so that the target device executes the operation instruction; Obtain the execution result of the target device after executing the operation command, and map the execution result to a 3D visualization drawing.

10. A three-dimensional drawing generation system for secondary circuits of substations for intelligent operation and maintenance, the system comprising a terminal device, the terminal device including a memory and a processor, the memory storing a computer program executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method for generating three-dimensional drawings of substation secondary circuits for intelligent operation and maintenance as described in any one of claims 1 to 9.

Images