Keywords

1 Introduction

Under the background of the construction of new power system, the pressure of the distribution side consumption and the distributed new energy grid connection service is increasing, the operation control form is evolving to the active distribution network, the topology analysis and model construction is difficult, inefficient, and costly, and the distribution network digital resource sharing and maintenance, multi-dimensional data simulation and analysis capabilities are insufficient [1, 2]. It is urgent to improve the level of intelligent operation, lean operation and maintenance, and efficient operation of the distribution network by digital twinning technology. The distribution network facilities are characterized by diversified types, massive scale and complex structure [3, 4]. The traditional digital twins of distribution network facilities tend to static three-dimensional models, while the conventional construction methods such as point cloud measurement and CAD design have problems of high cost and slow speed. High-cost local area construction modeling is difficult to support large-scale system-level applications [5, 6]. The distribution network digital twin spatial layout technology needs to be studied to solve the above problems.

2 Automatic Layout Technology of Power Grid Diagram

At home and abroad, there is little research on the automatic layout technology of power grid diagram, and it is only limited to the transmission network with relatively simple structure [7, 8]. However, the use of GIS (geographic information system) and PMS (generation management system) systems as the source of diagram data for power grid dispatching services is also at the experimental stage in the industry, and there is no mature and unified standard to follow [9, 10]. In recent years, GIS-based topology analysis can not only realize traditional topology analysis, but also realize power point tracking analysis, power supply radius analysis, optimal repair path analysis, dynamic coloring, etc. [11, 12]. However, at present, the automatic layout technology of power grid diagrams and models lack three-dimensional space support and point cloud correction, which leads to the inability to achieve accurate spatial analysis function of distribution network.

3 Layout of Electrical Topological Relationship in Digital Twin Space

Under the dual-carbon background, the new power system needs to support the two-way interaction between power supply and load, and needs to build a digital twin-space covering the electrical topology to support the electrical analysis, spatial analysis, comprehensive analysis and other functions of the power grid. The digital twin space needs to be built on the basis of the distribution network electrical topology map, combined with geographical map, 3D model and point cloud model to build a digital twin space with mutual integration of models. The line type of the distribution network electrical topology diagram system is divided into overhead line, ground cable line and overhead ground cable hybrid line. There are many multi-circuit station building equipment in the underground cable line, such as switching post, ring network cabinet, distribution room, box transformer, cable branch box, high voltage user, etc. These station building equipment are independent elements on the electrical topology diagram, and there are detailed electrical wiring diagrams inside these independent elements.

In the traditional electrical topology diagram system, each power equipment is associated with a point on the topology description, and the relevant power equipment includes distribution transformer, switch, grounding switch, etc. The relationship between equipment is connected by lines. The line includes the power line and the connecting line inside the station building. In order to build a digital twin space that integrates geographic model map, electrical topology map, 3D model and point cloud model, it is necessary to automatically generate a 3D model based on the electrical topology map, geographic resource map and point cloud model. The automatic generation of 3D models requires searching the electrical topology map and geographical resource map, and constructing the distribution network single-line diagram and ring network diagram topology structure based on the digital twin space.

The search algorithm is shown in Fig. 3.1. The search process for a distribution network line is based on the node corresponding to the outgoing line switch (for switching station) or distribution line outgoing line (for substation) in the electrical topology map or geographical resource map. On the electrical topology map, search all relevant lines, equipment and topology data from the load side (including distributed power supply) along the dotted line, and stop the search of this line until it meets the end of the line or the contact switch, Search the electrical topology data of all selected distribution line circuits. During the search process, the data information of tie switch is recorded in order to generate a digital twin space for multiple tie associated feeders. According to the electrical topology data of the distribution line that has been searched, the single-line diagram and ring network diagram tree structure of the line are established.

Fig. 3.1
A 3-step flow diagram is as follows. Outlet switch or outlet as the starting point. Search for all relevant equipment on load side 1. Stop searching when encountering the end of the line or contact switch.

Flow chart of distribution network route search

Full size image

After establishing the tree structure of single-line diagram and ring network diagram, it is necessary to set the coordinates of each device in the tree object in 3D space. To set the coordinates of each equipment in three-dimensional space, it is necessary to determine whether the equipment is a trunk line or a branch line in the electrical topology diagram, and whether it is an equipment outside the station building or an equipment inside the station building. Therefore, it is necessary to integrate the equipment in the established tree model. The final result of integration is to establish the level of each element in the equipment list. The integration basis is the search order of the equipment in the in-depth search from the distribution line outlet to the load side (including distributed power supply) based on the topological relationship.

The layout of the 3D model needs to solve two problems: the routing mode of critical path nodes and the layout of branch line equipment. Determine the latest digital twin space layout according to all equipment on the critical path in combination with the geographical resource map and point cloud scanning map. After determining the routing mode, the coordinates of all equipment and lines on the critical path of the geographical resource map in the digital twin space can be determined. According to the direction from left to right, from top to bottom, and according to the topological relationship between the devices, the device nodes on the critical path are determined according to the needs of digital twinning analysis. If the trunk line contains resources in the station building, then when the specific layout is carried out, a station building area is cut out from the trunk line by combining the point cloud model, and placed below or above the trunk line, and connected with the digital twin space of the station building. When drawing the branch line part, set the 3D position of the branch line equipment based on the spatial layout relationship of the point cloud model. Before drawing the branch line equipment, it is necessary to clarify the digital twin space position of the trunk line. Poll the branch equipment to be drawn in the single-line diagram, calculate the maximum number of nodes of the branch, and confirm that the branch is drawn above or below the trunk line. Set the coordinates of each device on the branch in order from parent to child.

4 Automatic Generation Algorithm of Distribution Network 3D Model Drawing

The electrical topology constructed based on the layout of electrical topology relationships in digital twin spaces, combined with point cloud models, 3D models, and geographic models, is automatically generated into a distribution network 3D model diagram based on feature matching, layout, and routing algorithms. The processing logic of the 3D model graph generation algorithm is shown in Fig. 3.2.

Fig. 3.2
A 3-step flow diagram is as follows. Distribution network line and station building data loading. Build an automatic mapping model. Layout and wiring.

Algorithm diagram of 3D model drawing generation

Full size image

Distribution network line and station building data loading: first, the initial loading process of the distribution network line, station building data, point cloud model, 3D model, and geographic model is used to build a memory object model for subsequent automatic mapping. In this process, the topology fusion and simplification of the distribution network equipment 3D resources, line model, and station building wiring diagram and other models are required to support efficient mapping processing in the subsequent automatic generation process. Specifically, match 3D model resources based on the line model and equipment types and models in the station building wiring diagram, and establish a mapping relationship between equipment and 3D models on the topology diagram.

Establish automatic mapping model: realize the creation of 3D model space and all 3D equipment entity objects of the distribution network based on the data structure in the middle memory, and establish the association relationship with the source objects of the distribution network electrical topology diagram and internal wiring diagram. Based on the logical relationship of distribution business, point cloud model, geographic model and electrical topology model matching model are established, mainly to achieve feature matching. Neural network is used for model training, and finally the distribution network digital twin mapping model is constructed.

Layout and wiring: based on the geographical model, electrical topology model and point cloud model, the distribution network digital twin mapping model matching algorithm is adopted to realize the initial layout and wiring of the 3D model diagram through the layout and wiring algorithm. After completion, the 3D model diagram is saved to realize the complete initialization and generation process of the 3D model diagram.

5 3D Correction Technology of Distribution Network Equipment Based on Point Cloud Mode

The original 3D point cloud only contains geometric information, which can only meet the simple measurement and analysis functions, and is difficult to meet the deeper level of intelligent services and applications such as distribution network planning and spatial analysis. The semantic 3D model of the entity not only contains necessary geometric information, but also has rich semantic information and topological association information. The automatic extraction of distribution network equipment is a necessary and important step to automatically reconstruct the semantic rich entity model from the point cloud. Based on the constraints of distribution network topology prior knowledge, this paper extracts the semantics of the distribution network point cloud model, and further extracts the geometric information of the identified semantic objects. The partition and extraction of space is the process of dividing space into semantic regions. Each partition space represents a distribution network device. The partition and extraction of space can determine the ownership of each component, and determine the semantic relationship between distribution network devices by judging the connectivity of each space.

The method used in this paper is mainly to build a point cloud identification model library for fixed distribution network equipment, automatically identify the shape of the cable in the distribution network room and the ground through the point cloud model, correct the outdoor overhead line tower based on the 3D model and the point cloud model, identify the content of the object in the station room and the ground cable based on the distribution network topology prior knowledge, and correct the 3D model position based on the distribution network business knowledge, 3D model and the point cloud model, as shown in Fig. 3.3:

Fig. 3.3
A 3-step flow diagram is as follows. Point cloud recognition model library. Correction of outdoor overhead line poles and towers. Equipment correction in the station building.

Point cloud modification algorithm diagram

Full size image
  1. 1.

    Take all the overhead line, underground cable and overhead cable point cloud data as samples, use the 3D depth learning network to extract depth features from each point cloud data, cluster all the sample depth features through K-mean (K-mean algorithm) algorithm, and obtain corresponding cluster centers. At the same time, regard them as words, and combine these words together to form an overhead line, underground cable and overhead cable point cloud dictionary.

  2. 2.

    Point cloud correction of indoor and underground cables and outdoor overhead lines: The point cloud model of overhead lines, underground cables and overhead cables is the point cloud information scanned by UAV, and the frame of the point cloud model of the station building and underground cables is defined in advance, and then the approximate point cloud information of the station building, underground cables and overhead lines is determined by combining the 3D model; The main model of the overhead line tower is determined according to the three-dimensional model, and then the point cloud model is used to correct the conductor position, circuit model, and tower type. Through this step, the specific three-dimensional model of the station building, ground cable shape, and the overhead line tower is determined.

  3. 3.

    Object correction of indoor and underground cables: this paper uses the algorithm based on region growth to segment and extract the plane of point cloud. The algorithm first sorts the points according to the curvature of each point in the point set, and takes the points corresponding to the minimum curvature as the initial seed points of region growth; Define the K neighborhood search range, search for points within the neighborhood of the seed point, if the angle between the normal direction of the point in the neighborhood and the normal direction of the current seed point is less than the angle threshold θt. Then add the neighborhood point to the current plane area and check the curvature value of all points in the neighborhood. If the curvature value of the neighborhood point is less than the curvature threshold Ct, the neighborhood point will be grown as a new seed point; Repeat the above steps until the seed point set is empty, and complete the region growth algorithm.

The indoor plane point sets are extracted by the region growth algorithm, and the plane parameters need to be estimated according to the plane point sets. The commonly used plane fitting algorithms include the least square fitting algorithm and the random sampling consistent plane fitting algorithm.

The following prior constraints are mainly involved in the process of plane marking of indoor components (ceiling, floor, wall):

  1. (1)

    The normal direction is divided by calculating the angle between the normal direction of the fitting plane and the gravity direction. Generally, the normal direction of the indoor ceiling and floor plane is collinear with the gravity direction, and the normal direction of the wall plane is perpendicular to the gravity direction;

  2. (2)

    Area and size: in the case of the same floor height in the multi-indoor space scene, the floor point cloud plane is usually the largest plane, and the area, height and width of the ceiling plane and the wall plane also meet certain constraints;

  3. (3)

    Plane position: the interior ceiling and floor plane are generally located at the highest and lowest plane of each room, and the wall plane is perpendicular to the ceiling and floor plane, forming a closed space.

  4. (4)

    The transformer is located in the middle of the room, and the main body is composed of several cuboids. There are several types of cylindrical bushings on the top, which are connected with conductors;

  5. (5)

    The switch body is composed of several cuboids and connected with wires.

With the help of the above priori knowledge, the extracted plane is marked and identified, and the point cloud of transformer, wire, switch, reactive compensation box, fuse and DC screen is proposed. The point cloud space needs to be further divided to determine the attribution of each component.

6 Distribution Network Digital Twin Spatial Layout Component

Integrate geographic, physical, management and business information, and develop distribution network digital twin spatial layout components. Support the spatial layout and scene integration of twin objects based on semantic understanding, realize the spatial layout and rapid reconstruction of distribution network digital twins based on electrical topology, and support the interaction between distribution network equipment and topology network.

The specific functions are as follows: distribution network data acquisition tool, which carries out the collection and regulation of distribution network account and measurement data, including basic data, operation monitoring data, maintenance test data, automation data, information communication data, environmental data, etc. Two-dimensional and three-dimension model association, using the configuration mapping association model of two-dimensional topology and three-dimensional model, realizes the interaction and deduction simulation between the distribution network equipment twins and the topology network. The space constraint rule base mainly includes the space constraint of the transformer, fixed switch cabinet, channel width, single row assembled capacitor bank, double row assembled capacitor bank, and a complete set of capacitor cabinet. The object topology linkage rules, the mapping linkage rules of 2D entities and 3D models, realize the linkage operation and deduction simulation of 2D topology wiring diagram elements and 3D twin models. Geospatial engines, build 3D GIS engine, and use open source framework to realize grid 2D topology, 3D rendering and spatial interaction. The 3D twin engine adopts a lightweight cross platform H5 architecture, supporting browser based 3D models and a fast rendering visualization engine for large scenes. The holographic topology engine realizes the construction technology and interactive mapping of standard entity libraries, graphic editing and topology generation, real-time verification of line topology and station topology business rules, association of topology nodes and ledger data, interactive analysis and access of power points, open capacity calculation, economic analysis of wiring, and load characteristic analysis. The topology constraint rule library mainly includes two types: business specification and data specification. The business protocol library includes islands, no power supply, connectivity, etc.; The data specification library includes constraint rules such as null value constraints, duplicate constraints, including relationship constraints, excluding relationship constraints, threshold constraints, temporal data anomalies, etc.,

The overall architecture of the system is mainly divided into three parts: browser client, cloud storage, and high-performance service cluster.

The browser client local cache and cloud cache constitute a multi-level cache system. Browser client is an important user-oriented interactive window for data visualization, mainly composed of user interface, Cesium rendering engine and local data cache. The user interface is implemented in HTML, JavaScript and other programming languages, and Cesium is used for distribution network digital twin space visualization and interaction. The user triggers the client to send a request to the server by clicking or checking the layer resource and other interactive operations, and renders the data responded by the server in the browser. In addition, the local data cache is mainly the data cache pool set based on the browser's cache mechanism, which is an important part of the multi-level cache system. According to the user's interaction, the browser first requests the local cache data, and then sends the data request to the server if the data does not exist in the local cache. After receiving the server response data, add it to the local cache to reduce the response time of multiple requests for data.

Cloud storage and browser local storage form a multi-level cache system. Multi-level cache is an effective method to improve the efficiency of large-scale data loading.

The high-performance server cluster mainly has the following two functions.

  1. (a)

    Data preprocessing: The main function is to convert data formats that are not supported or recommended by the rendering engine into data formats that the rendering engine can load and render.

  2. (b)

    Data storage: including the file system and database system. The file system mainly stores electrical topology map, point cloud model, 3D model resources, etc.; The database system mainly stores the detailed information corresponding to 3D model resources.

7 Conclusion

Under the background of a new power system construction, the distribution network facilities have the characteristics of diversified types, massive scale and complex structure. The traditional digital twin of distribution network facilities tends to static three-dimensional models, while the conventional construction methods such as point cloud measurement and CAD design have the problems of high cost and slow speed, and the high-cost local area facility modeling is difficult to support large-scale system-level applications. This paper studies the layout of the electrical topological relationship of the digital twin space. Based on the geographical resource map and the electrical topological map, the distribution network topological nodes are extracted using the single-line diagram and the ring network diagram tree structure. According to the hierarchical mark of the equipment, the single-line diagram and the ring network diagram structure, combined with the point cloud model, the distribution network digital twin space is rapidly constructed using the automatic generation algorithm of the distribution network three-dimensional model diagram. In this paper, based on the point cloud model, we carry out the research on the three-dimensional correction technology of distribution network equipment, build the point cloud recognition model library of fixed equipment, automatically identify and correct the outdoor overhead line tower, station building and other equipment through the point cloud model, identify the contents of the objects in the station building and underground cable line according to the prior knowledge, and correct the three-dimensional model position based on the three-dimensional model and the point cloud model. This paper develops the spatial layout component of the distribution network digital twins that integrates geographic, physical, management and business information, and realizes the spatial layout and rapid reconstruction of the distribution network digital twins based on electrical topology.