Layered and partitioned visual modeling method, system and equipment for power distribution network and medium

Abstract

The invention provides a hierarchical and partitioned visual modeling method, system and equipment for a power distribution network and a medium. The method comprises the following steps: acquiring multi-source data of the power distribution network and unifying a space-time reference; dynamically partitioning an area where the power distribution network is located, and constructing a digital twinborn hierarchical model of each partition by using power distribution network multi-source data of a unified space-time reference and a power distribution network component library for each partition; and splicing the digital twinning layering models of the partitions, and rendering a working condition scene of the power distribution network by adopting a deduction technology to construct the digital twinning layering model of the power distribution network. According to the method, the multi-source data is unified to the same space-time coordinate system, the reference difference of the multi-source data is eliminated, and the positioning precision is improved; a hierarchical model is constructed by using multi-source data of a unified space-time reference, and the problem of separation of equipment geographic coordinates and an electrical connection relation is solved; a standardized component library is established, so that the modeling flexibility and the multiplexing capability in different scenes are improved; a partition model seamless splicing mechanism is constructed, and the limitation that local updating of a large-scale power distribution network model needs overall reconstruction is broken through.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of power system digitization, and particularly relates to a layered and partitioned visual modeling method, system, device and medium for a distribution network. BACKGROUND

[0002] Nowadays, digital twin technology has become a core support means for the intelligent construction of modern distribution networks. Building a high-precision and high-timeliness three-dimensional visual twin model of the distribution network can realize real-time mapping of the power grid operation state, full life cycle management of equipment, and rapid fault positioning, which has great significance for improving the reliability and operation efficiency of the power grid. However, the distribution network has the significant characteristics of a large number of devices, complex spatial distribution, and dynamic changes in topology, which puts forward the requirements of timeliness, spatial topology coupling precision, dynamic expansion capability, and multi-source data fusion for the rapid construction and dynamic updating of the digital twin model.

[0003] The current digital twin modeling technology has significant technical bottlenecks: first, the modeling method based on artificial surveying and mapping. It relies on professional surveying and mapping personnel to collect coordinate data on site using total station, RTK and other devices, and manually draws the topology graph through CAD software. This method can guarantee the basic accuracy, but it is low in efficiency, lagging in updating and poor in new energy adaptability. Second, the automatic modeling technology based on a single data source. It uses unmanned aerial vehicle oblique photography or LiDAR point cloud data to automatically generate a three-dimensional model, and typical representatives are commercial software such as ContextCapture and Pix4D. It has limitations such as lack of topology identification, failure in complex scenes, and difficulty in dynamic updating. Third, the multi-source data fusion modeling method. In recent years, research has tried to fuse geographic information system (GIS) account, data acquisition and monitoring control system (SCADA) data and image information to build a model, and realize the spatial layout of equipment through coordinate mapping, but there are still problems such as data fragmentation, image recognition bottlenecks and model expansion rigidity. In summary, the existing technology cannot meet the core requirements of "high precision, high efficiency and high flexibility" of the digital twin of the distribution network. SUMMARY

[0004] In order to solve the core defects of low efficiency, insufficient precision and poor dynamic expansion capability in traditional digital twin modeling technology of distribution network, the present application proposes a layered and partitioned visual modeling method for distribution network, which comprises:

[0005] acquiring multi-source data of the distribution network and unifying the space-time reference;

[0006] dynamically partitioning the area where the distribution network is located, and constructing a digital twin layered model for each partition using the unified space-time reference of the multi-source data of the distribution network and the pre-constructed component library of the distribution network for each partition;

[0007] The partitioned digital twin layered models are spliced, and a power distribution network working condition scene is rendered by using a deduction technology to construct a power distribution network digital twin layered model.

[0008] Preferably, the acquiring of the power distribution network multi-source data and the unification of the space-time reference comprises:

[0009] The power distribution network multi-source data is acquired and preprocessed.

[0010] The timestamp of the preprocessed power distribution network multi-source data is acquired, and the timestamp is recorded in the header file of the power distribution network multi-source data by using a global navigation satellite system to unify the time reference.

[0011] The latitude and longitude of the location of the power distribution network are acquired by using a global navigation satellite system, and the latitude and longitude of the location of the power distribution network are used as a reference to convert the unified time reference of the power distribution network multi-source data into a global coordinate system by using a seven-parameter Bursa model to unify the space reference.

[0012] Preferably, the power distribution network multi-source data comprises global image data of the power distribution network acquired by a drone, videos and / or pictures of local equipment acquired by a patrol robot and / or manual shooting, electrical account data, laser point cloud data, and infrared thermal imaging data acquired by a data acquisition and monitoring control system; and the preprocessing comprises denoising, distortion correction, and format unification.

[0013] Preferably, the power distribution network located area is dynamically partitioned, and the power distribution network multi-source data of each partition is used to construct a digital twin layered model of each partition by using the unified space-time reference and a pre-constructed power distribution network component library, comprising:

[0014] The power distribution network located area is dynamically partitioned based on the equipment, topology, and function of the power distribution network.

[0015] Based on the component library and the power distribution network multi-source data of each partition, a topology-space joint analysis algorithm is used to couple the physical space, topological connection, and electrical property of the equipment in each partition to obtain a digital twin layered model of each partition.

[0016] Preferably, the topology-space joint analysis algorithm is constructed by combining a neural network, a YOLOv8 model, an attention mechanism, and a density-based clustering algorithm.

[0017] Preferably, the topology-space joint analysis algorithm is constructed by combining a neural network, a YOLOv8 model, an attention mechanism, and a density-based clustering algorithm.

[0018] For each partition, a density-based clustering algorithm is used to perform semantic segmentation on the laser point cloud of each partition, and based on the semantic segmentation result, a device three-dimensional model is retrieved from the component library to perform physical space coupling of the device, obtaining a digital twin physical layer model of each partition;

[0019] Based on the video and / or picture of each partition, a YOLOv8 model combined with an attention mechanism is used to detect and identify the devices in each partition, and based on the detection and identification result, a neural network is used to couple the topology connection of the device, obtaining a digital twin topology layer model of each partition;

[0020] The electrical account and the devices in the digital twin physical layer model of each partition are dynamically associated with the electrical properties, obtaining a digital twin attribute layer model of each partition;

[0021] The digital twin physical layer model, topology layer model and attribute layer model of each partition are coupled to construct a digital twin hierarchical model of each partition.

[0022] Preferably, the density-based clustering algorithm is used to perform semantic segmentation on the laser point cloud of each partition, and based on the semantic segmentation result, a device three-dimensional model is retrieved from the component library to perform physical space coupling of the device, obtaining a digital twin physical layer model of each partition, comprising:

[0023] The density-based clustering algorithm is used to perform semantic segmentation on the laser point cloud of each partition, obtaining the bounding box of each cluster, the center point of each cluster, the device coordinates, the device name and the device type in each partition;

[0024] Based on the bounding box of each cluster, the center point of each cluster, the device coordinates, the device name and the device type in each partition, a convex hull model of each partition is obtained by convex hull modeling;

[0025] Based on the device type in each partition, a device three-dimensional model is retrieved from the component library;

[0026] Based on the device name in each partition, the device three-dimensional model and the convex hull model of each partition are coupled to obtain a digital twin physical layer model of each partition.

[0027] Preferably, based on the video and / or picture of each partition, a YOLOv8 model combined with an attention mechanism is used to detect and identify the devices in each partition, and based on the detection and identification result, a neural network is used to generate the topology connection of the device to be coupled, obtaining a digital twin topology layer model of each partition, comprising:

[0028] The YOLOv8 model is used to detect the device type in the video and / or picture of each partition, obtaining the device type, device coordinates and confidence in each partition;

[0029] Based on the device type, device coordinates and confidence in each partition, the connection relationship and connection point between devices in each partition are identified using an attention mechanism;

[0030] The devices corresponding to the connection points in each partition are taken as nodes, and the connection relationship between devices in each partition is taken as edges to construct a graph data structure for each partition;

[0031] Based on the graph data structure for each partition, a neural network model is used to infer the topology structure graph of each partition, and the topology structure graph of each partition is taken as the digital twin topology layer model of each partition.

[0032] Preferably, the electrical account and the digital twin physical layer model of each partition are dynamically associated with the electrical attributes of the devices to obtain the digital twin attribute layer model of each partition, which comprises:

[0033] The electrical account of each partition is taken as an electrical attribute, and the electrical attribute is bound to the devices in the digital twin physical layer model of each partition;

[0034] Through the establishment of a real-time updating mechanism, the electrical attribute and the digital twin physical layer model of each partition are dynamically associated to construct the digital twin attribute layer model of each partition.

[0035] Preferably, after the electrical account and the digital twin physical layer model of each partition are dynamically associated with the electrical attributes of the devices to obtain the digital twin attribute layer model of each partition, the digital twin physical layer model, topology layer model and attribute layer model of each partition are coupled to construct the digital twin hierarchical model of each partition, which further comprises:

[0036] The digital twin attribute layer model of each partition is connected to a visualization component through a data interface.

[0037] Preferably, after the digital twin hierarchical model of each partition is obtained by coupling the physical space, topology connection and electrical attribute of the devices in each partition using the topology-space joint analysis algorithm based on the component library and the multi-source data of the power distribution network of each partition, it further comprises:

[0038] A spatial hash index is constructed, and the spatial hash index is used to monitor the changed devices in the power distribution network and to detect the occluded or damaged devices in each partition using a deep learning model;

[0039] When any partition of the power distribution network is found to have changed devices, the boundary devices of the adjacent partition of the partition where the change occurs are identified, and the digital twin hierarchical model of the partition where the change occurs is reconstructed based on the changed devices and the boundary devices of the adjacent partition of the partition where the change occurs;

[0040] When it is monitored that there is equipment occlusion or damage in any subarea of the power distribution network, the occluded or damaged equipment is identified, and the feature of the occluded or damaged equipment is completed using a generative adversarial network to reconstruct the digital twin model of the occluded or damaged equipment.

[0041] Preferably, when it is monitored that there is equipment change in any subarea of the power distribution network, the boundary equipment of the subarea adjacent to the subarea where the change occurs is identified, and the digital twin hierarchical model of the subarea where the change occurs is reconstructed based on the changed equipment and the boundary equipment of the subarea adjacent to the subarea where the change occurs, including:

[0042] When it is monitored that there is equipment change in any subarea of the power distribution network, the boundary equipment of the subarea adjacent to the subarea where the change occurs is located through spatial hash index, and a boundary subgraph is constructed based on the boundary equipment of the subarea adjacent to the subarea where the change occurs.

[0043] Based on the changed equipment, a topological-spatial joint analysis algorithm is used to reconstruct the digital twin hierarchical model of the subarea where the change occurs, and a VF2 algorithm is used to match the topological structure of the reconstructed digital twin hierarchical model and the boundary subgraph to unify the coordinates and establish a mapping relationship.

[0044] Preferably, the construction of the power distribution network component library includes:

[0045] According to the type of power distribution network equipment, parameterized modeling templates of various types of equipment are created;

[0046] The parameterized modeling templates of various types of equipment are configured with parameters, and a mapping relationship between the parameters and the model attributes is established;

[0047] Based on the mapping relationship between the parameters and the model attributes, installation constraint rules for various types of equipment are defined;

[0048] Based on the parameterized modeling templates of various types of equipment, the mapping relationship between the parameters and the model attributes, and the installation constraint rules for various types of equipment, the power distribution network component library is constructed.

[0049] Preferably, after the power distribution network multi-source data is obtained and unified with the space-time reference, the area where the power distribution network is located is dynamically partitioned, and before the digital twin hierarchical model of each subarea is constructed using the power distribution network multi-source data unified with the space-time reference and the pre-constructed power distribution network component library, it further includes:

[0050] The power distribution network multi-source data is enhanced based on simultaneous localization and mapping technology.

[0051] Preferably, after the power distribution network multi-source data is acquired and the time and space reference is unified, the area where the power distribution network is located is dynamically partitioned, and before the power distribution network multi-source data with the unified time and space reference and the pre-constructed power distribution network component library are used to construct the digital twin hierarchical model of each partition, the method further comprises the following steps:

[0052] A global optimization problem of the power distribution network multi-source data is constructed, and a general graph optimization library is used to solve the optimization problem to eliminate the reference differences between the power distribution network multi-source data.

[0053] Based on the same inventive concept, the application also provides a power distribution network hierarchical partition visualization modeling system, comprising a time and space unification module, a hierarchical modeling module and a partition splicing module.

[0054] The time and space unification module is used to acquire power distribution network multi-source data and unify the time and space reference.

[0055] The hierarchical modeling module is used to dynamically partition the area where the power distribution network is located, and to construct the digital twin hierarchical model of each partition by using the power distribution network multi-source data with the unified time and space reference and the pre-constructed power distribution network component library for each partition.

[0056] The partition splicing module is used to splice the digital twin hierarchical model of each partition, and render the power distribution network working condition scene by using deduction technology to construct the power distribution network digital twin hierarchical model.

[0057] Preferably, the time and space unification module comprises a preprocessing unit, a time unification unit and a space unification unit.

[0058] The preprocessing unit is used to acquire the power distribution network multi-source data and perform preprocessing.

[0059] The time unification unit is used to acquire the time stamp of the preprocessed power distribution network multi-source data, and use a global navigation satellite system to record the time stamp in the header file of the power distribution network multi-source data to unify the time reference.

[0060] The space unification unit is used to acquire the longitude and latitude of the location of the power distribution network by using a global navigation satellite system, and to convert the synchronous positioning and map construction coordinates of the power distribution network multi-source data after the unified time reference to a global coordinate system by using a seven-parameter Bursa model with the longitude and latitude of the location of the power distribution network as the reference to unify the space reference.

[0061] The power distribution network multi-source data comprises global image data of the power distribution network taken by a drone, videos and / or pictures of local equipment taken by a patrol robot and / or manually, electrical account data, laser point cloud data and infrared thermal imaging data acquired through a data acquisition and monitoring control system; the preprocessing comprises denoising, distortion correction and format unification.

[0062] Preferably, the hierarchical modeling module comprises a dynamic partitioning submodule and a modeling submodule.

[0063] The dynamic partitioning submodule is configured to perform dynamic partitioning on a region where a power distribution network is located based on equipment, topology, and functions of the power distribution network.

[0064] The modeling submodule is configured to obtain a digital twin hierarchical model of each partition by coupling physical space, topological connection, and electrical properties of equipment in each partition based on the component library and multi-source data of the power distribution network of each partition using a topological-space joint analysis algorithm.

[0065] The topological-space joint analysis algorithm is constructed by combining a neural network, a YOLOv8 model, an attention mechanism, and a density-based clustering algorithm.

[0066] Preferably, the modeling submodule comprises a physical layer modeling unit, a topological layer modeling unit, an attribute layer modeling unit, and a model coupling unit.

[0067] The physical layer modeling unit is configured to perform semantic segmentation on the laser point cloud of each partition using a density-based clustering algorithm, and retrieve a three-dimensional model of equipment from the component library based on the semantic segmentation result to couple the physical space of equipment, thereby obtaining a digital twin physical layer model of each partition.

[0068] The topological layer modeling unit is configured to detect and identify equipment in each partition using a YOLOv8 model combined with an attention mechanism based on the video and / or picture of each partition, and couple the topological connection of equipment using a neural network based on the detection and identification result, thereby obtaining a digital twin topological layer model of each partition.

[0069] The attribute layer modeling unit is configured to dynamically associate the electrical account of each partition with equipment in the digital twin physical layer model of each partition, thereby obtaining a digital twin attribute layer model of each partition.

[0070] The model coupling unit is configured to couple the digital twin physical layer model, the topological layer model, and the attribute layer model of each partition, thereby constructing the digital twin hierarchical model of each partition.

[0071] Preferably, the physical layer modeling unit comprises a semantic segmentation subunit, a convex hull modeling subunit, a retrieval subunit, and a physical layer construction subunit.

[0072] The semantic segmentation subunit is configured to perform semantic segmentation on the laser point cloud of each partition using a density-based clustering algorithm, thereby obtaining a bounding box of each cluster, a center point of each cluster, equipment coordinates, equipment name, and equipment type in each partition.

[0073] The convex hull modeling subunit is configured to perform convex hull modeling based on the bounding box of each cluster, the center point of each cluster, the device coordinates, the device name, and the device type in each subzone to obtain a subzone convex hull model;

[0074] The calling subunit is configured to call a device three-dimensional model from the component library based on the device type in each subzone;

[0075] The physical layer construction subunit is configured to couple the device three-dimensional model and the subzone convex hull model based on the device name in each subzone to obtain a digital twin physical layer model of each subzone.

[0076] Preferably, the topology layer modeling unit comprises a detection subunit, an identification subunit, a graph data subunit, and a topology layer construction subunit.

[0077] The detection subunit is configured to perform device type detection on the video and / or picture of each subzone by using a YOLOv8 model to obtain the device type, the device coordinates, and the confidence in each subzone.

[0078] The identification subunit is configured to identify the connection relationship and the connection point between devices in each subzone by using an attention mechanism based on the device type, the device coordinates, and the confidence in each subzone.

[0079] The graph data subunit is configured to construct a graph data structure of each subzone by taking the device corresponding to the connection point in each subzone as a node and taking the connection relationship between devices in each subzone as an edge.

[0080] The topology layer construction subunit is configured to infer a topology structure graph of each subzone by using a neural network model based on the graph data structure of each subzone, and take the power distribution network topology structure graph of each subzone as a digital twin topology layer model of each subzone.

[0081] Preferably, the attribute layer model unit comprises a binding subunit and an attribute layer construction subunit.

[0082] The binding subunit is configured to take the electrical account book of each subzone as an electrical attribute, and bind the electrical attribute on the device in the digital twin physical layer model of each subzone.

[0083] The attribute layer construction subunit is configured to dynamically associate and construct a digital twin attribute layer model of each subzone by establishing a real-time updating mechanism and associating the electrical attribute with the digital twin physical layer model of each subzone.

[0084] Preferably, the modeling sub-module further comprises a visualization unit.

[0085] The visualization unit is configured to connect the digital-twin attribute layer model of each subzone with a visualization component through a data interface before the model coupling unit is invoked after the attribute layer model unit is invoked.

[0086] Preferably, the hierarchical modeling module further comprises a monitoring submodule, a change reconstruction submodule, and an occlusion reconstruction submodule.

[0087] The monitoring submodule is configured to construct a spatial hash index after the modeling submodule is invoked, and to monitor the changed equipment in the power distribution network by using the spatial hash index, and to detect the occluded or damaged equipment by using a deep learning model.

[0088] The change reconstruction submodule is configured to, when it is monitored that there is equipment change in any subzone of the power distribution network, identify the boundary equipment of the adjacent subzone of the subzone where the change occurs, and to reconstruct the digital-twin hierarchical model of the subzone where the change occurs based on the changed equipment and the boundary equipment of the adjacent subzone of the subzone where the change occurs.

[0089] The occlusion reconstruction submodule is configured to, when it is monitored that there is equipment occlusion or damage in any subzone of the power distribution network, identify the features of the occluded or damaged equipment, and to complete the features of the occluded or damaged equipment by using a generative adversarial network to reconstruct the digital-twin model of the occluded or damaged equipment.

[0090] Preferably, the change reconstruction submodule comprises a boundary subgraph unit and a change reconstruction unit.

[0091] The boundary subgraph unit is configured to, when it is monitored that there is equipment change in any subzone of the power distribution network, locate the boundary equipment of the adjacent subzone of the subzone where the change occurs by using the spatial hash index, and to construct a boundary subgraph based on the boundary equipment of the adjacent subzone of the subzone where the change occurs.

[0092] The change reconstruction unit is configured to, based on the changed equipment, reconstruct the digital-twin hierarchical model of the subzone where the change occurs by using a topology-space joint analysis algorithm, and to match the topology structure of the reconstructed digital-twin hierarchical model and the boundary subgraph by using a VF2 algorithm to perform unified coordinate and establish a mapping relationship.

[0093] Preferably, the system further comprises a component library construction module.

[0094] The component library construction module is configured to create parameterized modeling templates for various types of equipment according to the power distribution network equipment types; configure parameters for the parameterized modeling templates for various types of equipment, and establish a mapping relationship between the parameters and model attributes; define installation constraint rules for various types of equipment based on the mapping relationship between the parameters and model attributes; and construct a power distribution network component library based on the parameterized modeling templates for various types of equipment, the mapping relationship between the parameters and model attributes, and the installation constraint rules for various types of equipment.

[0095] Preferably, the system further comprises a data enhancement module.

[0096] The data enhancement module is configured to enhance the power distribution network multi-source data based on a simultaneous localization and mapping technique after the spatio-temporal unification module is invoked and before the hierarchical modeling module is invoked.

[0097] Preferably, the system further comprises a difference elimination module.

[0098] The difference elimination module is configured to construct a global optimization problem of the power distribution network multi-source data after the spatio-temporal unification module is invoked and before the hierarchical modeling module is invoked, and solve the optimization problem using a general graph optimization library to eliminate the benchmark difference among the power distribution network multi-source data.

[0099] Based on the same inventive concept, the present application further provides an electronic device comprising at least one processor and a memory; the memory and the processor are connected through a bus;

[0100] The memory is configured to store one or more programs.

[0101] When the one or more programs are executed by the at least one processor, a power distribution network hierarchical partition visualization modeling method as described above is implemented.

[0102] Based on the same inventive concept, the present application further provides a readable storage medium having an execution program stored thereon, and the execution program, when executed, implements a power distribution network hierarchical partition visualization modeling method as described above.

[0103] Compared with the prior art, the present application has the following beneficial effects:

[0104] The application provides a power distribution network layered partition visual modeling method, system, device and medium, the method comprises: acquiring power distribution network multi-source data and unifying space-time reference; dynamically partitioning the area where the power distribution network is located, and constructing a digital twin layered model for each partition by using the power distribution network multi-source data with a unified space-time reference and a pre-constructed power distribution network component library for each partition; splicing the digital twin layered model of each partition, and rendering the power distribution network working condition scene by using deduction technology to construct a power distribution network digital twin layered model. The application unifies multi-source data to the same space-time coordinate system, eliminates the reference difference of multi-source data, and improves the positioning accuracy; the layered model is constructed by using multi-source data with a unified space-time reference, which solves the problem of the split of device geographic coordinates and electrical connection relationship; a standardized component library is established, the modeling flexibility and reusability in different scenarios are improved, and the modeling efficiency is significantly improved; a seamless splicing mechanism of partition model is constructed, which breaks through the limitation that local update of large-scale power distribution network model needs to be reconstructed as a whole, greatly shortens the modeling cycle, and meets the demand of high-frequency update and rapid response of power grid. BRIEF DESCRIPTION OF DRAWINGS

[0105] Figure 1 A flowchart of a power distribution network layered partition visual modeling method of the application;

[0106] Figure 2 A standardized component library construction flowchart of a power distribution network layered partition visual modeling method of the application;

[0107] Figure 3 A physical layer modeling flowchart of a power distribution network layered partition visual modeling method of the application;

[0108] Figure 4 A topology layer modeling flowchart of a power distribution network layered partition visual modeling method of the application;

[0109] Figure 5 An attribute layer modeling flowchart of a power distribution network layered partition visual modeling method of the application;

[0110] Figure 6 A dynamic partition splicing and incremental update flowchart of a power distribution network layered partition visual modeling method of the application;

[0111] Figure 7 An abnormal data compensation and multi-source closed-loop verification flowchart of a power distribution network layered partition visual modeling method of the application;

[0112] Figure 8 A flowchart of an embodiment of a power distribution network layered partition visual modeling method of the application;

[0113] Figure 9A basic structure schematic diagram of a layered and partitioned visual modeling system of a power distribution network of the present application;

[0114] Figure 10 An electronic device structure schematic diagram of the present application. DETAILED DESCRIPTION

[0115] The existing power distribution network digital twin modeling technology cannot meet the core requirements of high precision, high efficiency and high flexibility of power distribution network digital twin, and the following key technologies need to be broken through:

[0116] (1) Solve the problem of data benchmark unification of unmanned aerial vehicle aerial photography, field image and electrical account;

[0117] (2) Realize the synchronous and accurate identification of device connection relationship and geographical coordinates;

[0118] (3) Local model rapid splicing and global topology consistency maintenance;

[0119] (4) New type of power equipment parameterized modeling system, establish standard component library of photovoltaic, energy storage and other equipment.

[0120] The present application proposes an innovative solution to the above industry pain points, through systematic innovation, to solve the problem of efficiency, accuracy and expansion of the power distribution network digital twin field for a long time, and to provide core digital support for building a new type of power system with high flexibility and high reliability. In order to better understand the present application, the contents of the present application will be further described in conjunction with the drawings and examples of the specification.

[0121] Example 1:

[0122] A layered and partitioned visual modeling method of a power distribution network, the flowchart thereof is shown in Figure 1 , which comprises:

[0123] Step 1: Obtain power distribution network multi-source data and unify the space-time benchmark;

[0124] Step 2: dynamically partition the area where the power distribution network is located, and use the power distribution network multi-source data unified by the space-time benchmark and the pre-constructed power distribution network component library to construct digital twin layered models of each partition;

[0125] Step 3: splice the digital twin layered models of each partition, render the power distribution network working condition scene by using deduction technology, and construct the power distribution network digital twin layered model.

[0126] Step 1 specifically comprises:

[0127] 1.1, obtain the power distribution network multi-source data and pre-process.

[0128] The global image data of the power distribution network is obtained by unmanned aerial vehicle aerial photography, the local device video and / or picture is collected by inspection robot and / or artificial shooting, the SCADA electrical account, laser point cloud and infrared thermal imaging data are synchronously obtained. The multi-source data is preprocessed such as denoising, distortion correction and format unification to construct the standardized input data set.

[0129] 1.2, embedding timestamp to realize time synchronization of multi-source data.

[0130] The timestamp of all data collection is obtained based on the Beidou timing module of the unmanned aerial vehicle, field shooting device and laser radar collection terminal, and is recorded in the data header file. The GNSS (Global Navigation Satellite System) cycle count and cycle seconds are recorded in the data header file to establish a unified time reference.

[0131] 1.3, defining global coordinate system and constructing coordinate conversion model. The WGS84 geodetic coordinate system is selected as the global reference, and the reference point longitude and latitude of the power distribution network area are obtained by GNSS. The local coordinate system (such as the device self-contained SLAM (Simultaneous Localization and Mapping) coordinate system) of the laser point cloud and field image is converted to the global coordinate system to unify the spatial reference. The seven-parameter Bursa-Wolf model is used to calculate the coordinate conversion parameters:

[0132]

[0133] In the formula, X 全局 is the target global coordinate system coordinate, X 局部 is the source local coordinate system coordinate (i.e. the three-dimensional coordinate vector of the device self-contained coordinate system (such as laser radar, camera SLAM)), ΔX is the translation parameter vector (indicating the offset of the local coordinate system origin in the global coordinate system), k is the scale factor (compensating for the scale difference between different coordinate systems (usually 10 -6 orders of magnitude)), R(θ) is a composite rotation matrix composed of three Euler angles (∈ z , ∈ y , ∈ x ), R z (∈ z ) is the rotation matrix of Euler angle ∈ z , R y (∈ y ) is the rotation matrix of Euler angle ∈ y , and R x (∈ x ) is the rotation matrix of Euler angle ∈ x .

[0134] After unifying the time and space reference, before step 2 is executed, it also includes:

[0135] 1.4 Local Data Augmentation Based on SLAM. The LOAM (Lidar Odometry and Mapping) algorithm is used to extract point cloud edges and planar features. Inter-frame matching is achieved through ICP (Iterative Closest Point) to construct a high-precision local 3D map. The VINS-Fusion algorithm is used for on-site shooting equipment, with the IMU (Inertial Measurement Unit) providing high-frequency pose estimation. Camera image feature point matching corrects drift errors, outputting a 6-DOF pose.

[0136] 1.5 Joint Optimization of Multi-Source Data. A global optimization problem is constructed, and g2o (a general graph optimization library) is used to solve large-scale sparse matrices, eliminating cross-source data concatenation residuals. The optimization formula is expressed as follows:

[0137]

[0138] In the formula, T i and T j Let Z represent the 6-DOF poses (translation + rotation) of the i-th and j-th nodes, respectively, where T is the set of all nodes, and Z is the set of all nodes. ij For the actual measured values ​​from node i to node j in the observation model (such as laser ranging, visual feature matching distance), the observation model h(T) i ,T j The theoretical distance between nodes predicted by pose is h = ||T i [1:3]-T j [1:3]|| 2 The robustness coefficient ρ is a function of the residual r: (This is the threshold value, typically taken as 1.345).

[0139] The above process uses GNSS high-precision positioning data and SLAM algorithm to unify UAV aerial photography (global), on-site images (local), and laser point clouds (details) into the same spatiotemporal coordinate system, eliminating the differences in reference from multiple data sources and improving positioning accuracy.

[0140] 1.6. Construct a power distribution network component library. This step is as follows: Figure 2 As shown, it includes:

[0141] (1) Define equipment categories and templates (equipment categories): Based on the types of distribution network equipment, define 15 types of standardized equipment (photovoltaic, energy storage, charging piles, etc.) and create parametric modeling templates for each type of equipment.

[0142] (2) Design parametric modeling structure (parameter definition): Define configurable parameters for each type of equipment, including geometric dimension parameters (length, width, height, etc.), electrical attribute interfaces (voltage, power, etc.), and functional characteristic parameters; establish the mapping relationship between parameters and model attributes.

[0143] (3) Based on the mapping relationship between the parameters and the model attributes, define the installation constraint rules of the equipment, including spatial constraints (minimum distance, direction requirements), electrical constraints (voltage matching, capacity limitation), environmental constraints (waterproof, temperature range), safety constraints (safety distance, protection requirements).

[0144] (4) Based on the parameterized modeling templates of various types of equipment, the mapping relationship between the parameters and the model attributes, and the installation constraint rules of various types of equipment, a power distribution network component library is constructed, and a visual equipment panel is provided to support dragging equipment into the scene and automatically attaching to the appropriate position. Automatically detect the installation space, apply the constraint rules to optimize the equipment position, intelligently avoid existing equipment, and generate the optimal placement scheme.

[0145] Step 2 specifically includes:

[0146] 2.1, Dynamic partitioning. Based on the equipment, topology and function of the power distribution network, the area where the power distribution network is located is dynamically partitioned; to break through the limitation of local update of large-scale power distribution network model requiring overall reconstruction, significantly reduce the operation and maintenance complexity and resource consumption, the invention constructs a seamless partitioning mechanism for the partitioned model. For the use occasion of partitioning, it can be performed before the construction of the global model, or after the construction of the global model is completed and a partition is changed. The following first introduces the method of partitioning modeling before the construction of the global model.

[0147] 2.2, Based on the component library and the power distribution network multi-source data of each partition, a topology-space joint analysis algorithm is used to couple the physical space, topology connection and electrical properties of the equipment in each partition to obtain a digital twin hierarchical model of each partition.

[0148] 2.2.1, Physical layer modeling

[0149] For each partition, a density-based clustering algorithm is used to perform semantic segmentation on the laser point cloud of each partition, and based on the semantic segmentation result, a three-dimensional model of the equipment is retrieved from the component library to couple the physical space of the equipment, and a digital twin physical layer model of each partition is obtained. The flowchart of this step is shown in Figure 3

[0150] After the initialization coordinate conversion and preprocessing of step 1, a density-based clustering algorithm (DBSCAN) is first used to perform semantic segmentation on the laser point cloud of each partition to obtain the bounding box of each cluster, the center point of each cluster, the equipment coordinates, the equipment name and the equipment type in each partition;

[0151] Based on the bounding box of each cluster, the center point of each cluster, the equipment coordinates, the equipment name and the equipment type in each partition, a convex hull model of each partition is obtained by convex hull modeling;

[0152] ​Retrieving a device three-dimensional model from the component library based on the device type in each subzone;

[0153] Coupling the device three-dimensional model and the subzone convex hull model based on the device name in each subzone to obtain a digital twin physical layer model of each subzone.

[0154] 2.2.2, Topology layer modeling

[0155] Based on the video and / or picture of each subzone, using a YOLOv8 model combined with an attention mechanism to detect and identify the devices in each subzone, and based on the detection and identification results, using a neural network to couple the topology connection of the devices to obtain a digital twin topology layer model of each subzone; the flowchart of this step is shown in Figure 4

[0156] After the multi-source data collection and preprocessing of step 1, using a YOLOv8 model to detect the device type of the video and / or picture (including switch cabinet, transformer, fuse, etc. Key equipment) in each subzone, to obtain the device type, device coordinates and confidence in each subzone;

[0157] Based on the device type, device coordinates and confidence in each subzone, using an attention mechanism to identify the connection relationship and connection point between devices in each subzone;

[0158] Taking the device corresponding to the connection point in each subzone as a node, and taking the connection relationship between devices in each subzone as an edge to construct a graph data structure of each subzone;

[0159] Based on the graph data structure of each subzone, using a neural network model to infer to obtain a topology structure graph of each subzone, and taking the topology structure graph of each subzone distribution network as a digital twin topology layer model of each subzone. Through reinforcement learning, the model is continuously optimized and compared with the actual distribution network topology for verification.

[0160] 2.2.3, Attribute layer modeling

[0161] Dynamically associating the electrical account and the devices in the digital twin physical layer model of each subzone with electrical attributes to obtain a digital twin attribute layer model of each subzone; the flowchart of this step is shown in Figure 5

[0162] After the multi-source data collection and preprocessing of step 1, taking the electrical account (electrical parameters such as capacity, impedance, operating state, etc.) of each subzone as an electrical attribute, and binding the electrical attribute on the devices in the digital twin physical layer model of each subzone;

[0163] ​​The real-time updating mechanism is established to ensure that the parameters are synchronized with the SCADA data, and a dynamic association database is constructed to store the data associated with the equipment attributes and the three-dimensional model. The electrical properties and the digital twin physical layer model of each partition are dynamically associated to construct the digital twin attribute layer model of each partition. The equipment attributes are dynamically associated with the three-dimensional model, and the three-dimensional model changes in real time with the equipment state (such as color, animation, etc.).

[0164] In addition, the attribute layer also provides a data interface and a visualization component.

[0165] 2.2.4, hierarchical coupling

[0166] The digital twin physical layer model, the topology layer model and the attribute layer model of each partition are coupled to construct the hierarchical digital twin model of each partition.

[0167] 2.3, before step 3 is performed, in order to support elastic expansion and dynamic updating, the present application also monitors each partition. Specifically:

[0168] A spatial hash index is constructed, and the spatial hash index is used to monitor the changed equipment in the distribution network, and a deep learning model is used to detect the obscured or damaged equipment in the partition;

[0169] 2.3.1, dynamic partition splicing and incremental updating, this step is as shown in Figure 6 When it is monitored that any partition of the distribution network has changed equipment, the spatial hash index is used to quickly locate the boundary equipment (switch, connection point, etc.) of the adjacent partition of the partition where the change is located, and a boundary subgraph is constructed based on the boundary equipment of the adjacent partition of the partition where the change is located for subsequent matching. A1, A2, …, E5 in the figure represent the partitions.

[0170] Based on the changed equipment, a topology-space joint analysis algorithm is used to reconstruct the hierarchical digital twin model of the partition where the change exists, and a VF2 algorithm is used to match the topology structure of the reconstructed hierarchical digital twin model and the boundary subgraph to unify the coordinates and establish a mapping relationship M: Aboundary→Bboundary(A zone boundary to B zone boundary).

[0171] Wherein, the unified coordinates only need to recalculate the changed area (≤5% of the global model), and the coordinate transformation formula: T(x) = s·R·x + t is applied to ensure that the changed area is accurately aligned with the global model. Where T(x) is the changed coordinate of the original coordinate x, s is the scaling factor, R is the rotation matrix, and t is the translation vector.

[0172] Then, the boundary connection points and the electrical connectivity are checked, and the transactional update realizes seamless splicing to generate a complete merged partition topology.

[0173] 2.3.2, when monitoring any sub-area of the power distribution network exists equipment shielding or damage, the shielding or damaged equipment feature recognition, and the use of generative adversarial network to complete the shielding or damaged equipment features, to reconstruct the shielding or damaged equipment digital twin model. This step is shown in Figure 7

[0174] (1) shielding / damage scene detection: identify the presence of complex scenes of equipment shielding or damage, locate the missing feature area.

[0175] (2) GAN feature completion: apply generative adversarial network (GAN) to complete the missing features, use deep convolutional generative adversarial network (DCGAN) + conditional generative network architecture, generator and discriminator against training.

[0176] (3) complete model reconstruction: based on the completion of the feature reconstruction of the complete device model, three-dimensional point cloud reconstruction and texture mapping, improve the feature completion rate.

[0177] (4) unmanned aerial vehicle inspection data collection: regular unmanned aerial vehicle automatic inspection, multi-sensor data collection (visible light, infrared, laser radar), weekly comprehensive inspection + daily key area inspection.

[0178] (5) multi-source data comparison and analysis: unmanned aerial vehicle data and account system comparison, combined with SCADA real-time data, analyze the consistency of equipment location, running state and parameters.

[0179] (6) model deviation dynamic correction: adaptive weighted fusion algorithm to correct the deviation, improve data consistency and improve model accuracy. Continuous optimization and closed loop feedback: the correction result is fed back to the GAN training, the unmanned aerial vehicle inspection path is optimized, the account system reference data is updated, and the system robustness is improved.

[0180] Step 3 integrates a dynamic rendering engine to support smooth interaction of massive node models; embeds modules such as power flow calculation and fault deduction, users can directly simulate scenarios such as line load and short circuit fault in the three-dimensional model. And model verification and iterative optimization function. Verify the model accuracy and efficiency through actual power grid data, continuously optimize algorithm parameters; can establish version management algorithm according to demand, support model history backtracking and difference analysis.

[0181] The method also has another flow chart, that is, after building a global model, then partitioning, as Figure 8 ​As shown. First, build the power distribution network component library, then collect multi-source data and unify the space-time reference; According to the multi-source data, the components are retrieved from the component library to build a global hierarchical model; After the global model is built, the area where the power distribution network is located is monitored dynamically, and when it is monitored that there is a change in a partition, the model of the partition is rebuilt, and the model of the partition is spliced to the original global model, and whether there is equipment shielding or damage in the partition is detected, and when it is detected that there is equipment shielding or damage, the model of the equipment is rebuilt and replaces the original model in the global model; Finally, the model is functionalized and iteratively optimized.

[0182] The method provided by the application has the following effects:

[0183] (1) Revolutionary improvement in modeling efficiency: Through intelligent fusion and automatic analysis of multi-source data, the traditional modeling mode dominated by manual work is completely changed; Through partition modeling and component library, a high-frequency update support system is established to meet the rapid response demand of dynamic changes of power grid, and the overall modeling efficiency is revolutionarily improved, and the operation and maintenance response speed is improved by orders of magnitude.

[0184] (2) Fundamental improvement in spatial topology accuracy: By unifying the space-time reference of multi-source data and constructing the physical layer and the topology layer, the problem of separation of device geographic coordinates and electrical connection relationship is solved, and the accurate mapping relationship between physical space and logical topology is formed; A new paradigm of space-electricity coupling is established by using a topology-space joint analysis algorithm to support high-reliability fault diagnosis and state analysis.

[0185] (3) Major breakthrough in dynamic expansion capability: innovative seamless partition splicing mechanism, breaking through the limitation of traditional modeling scale; Local update mechanism avoids overall reconstruction, significantly optimizes the utilization of computing resources.

[0186] (4) Comprehensive improvement of new device integration capability: Establish a complete parameterized component library of new devices to support the rapid access of various new energy devices, and realize the rapid deployment of devices by dragging, which greatly improves the scene adaptation capability.

[0187] (5) Essential enhancement of robustness in complex scenarios: effectively solve the modeling difficulties under extreme conditions such as vegetation shielding and device rust, establish a modeling guarantee mechanism in complex environments, ensure the integrity and usability of the model, and realize intelligent identification and automatic repair of model deviation.

[0188] Embodiment 2:

[0189] The application based on the same inventive concept also provides a power distribution network hierarchical partition visual modeling system, a basic structure diagram thereof is as shown in Figure 9 , comprising: a space-time unification module, a hierarchical modeling module and a partition splicing module;

[0190] The space-time unification module is configured to acquire power distribution network multi-source data and unify space-time reference;

[0191] The hierarchical modeling module is configured to dynamically partition a region where the power distribution network is located, and construct a digital twin hierarchical model of each partition by using the power distribution network multi-source data unified with the space-time reference and a pre-constructed power distribution network component library;

[0192] The partition splicing module is configured to splice the digital twin hierarchical models of the partitions, render a power distribution network working condition scene by using a deduction technique, and construct a power distribution network digital twin hierarchical model.

[0193] Preferably, the space-time unification module comprises a preprocessing unit, a time unification unit, and a space unification unit.

[0194] The preprocessing unit is configured to acquire the power distribution network multi-source data and perform preprocessing.

[0195] The time unification unit is configured to acquire a time stamp of the preprocessed power distribution network multi-source data, and record the time stamp in a header file of the power distribution network multi-source data by using a global navigation satellite system, so as to unify a time reference.

[0196] The space unification unit is configured to acquire longitude and latitude of a region where the power distribution network is located by using a global navigation satellite system, and convert synchronous positioning and map construction coordinates of the power distribution network multi-source data unified with the time reference to a global coordinate system by using a seven-parameter Bursa model, so as to unify a space reference.

[0197] The power distribution network multi-source data comprises global image data of the power distribution network acquired by a drone, videos and / or pictures of local equipment acquired by a patrol robot and / or manually, electrical account data, laser point cloud data, and infrared thermal imaging data acquired by a data acquisition and monitoring control system; and the preprocessing comprises denoising, distortion correction, and format unification.

[0198] Preferably, the hierarchical modeling module comprises a dynamic partitioning submodule and a modeling submodule.

[0199] The dynamic partitioning submodule is configured to dynamically partition a region where the power distribution network is located based on equipment, topology, and functions of the power distribution network.

[0200] The modeling submodule is configured to couple physical space, topological connection, and electrical properties of equipment in each partition by using a topological-space joint analysis algorithm based on the component library and the power distribution network multi-source data of each partition, so as to obtain a digital twin hierarchical model of each partition.

[0201] The topological-space joint analysis algorithm is constructed by combining a neural network, a YOLOv8 model, an attention mechanism, and a density-based clustering algorithm.

[0202] Preferably, the modeling submodule comprises a physical layer modeling unit, a topological layer modeling unit, an attribute layer modeling unit, and a model coupling unit.

[0203] The physical layer modeling unit is configured to perform semantic segmentation on the laser point cloud of each partition by using a density-based clustering algorithm, and retrieve a device three-dimensional model from the component library based on the semantic segmentation result to perform physical space coupling of the device, thereby obtaining a digital twin physical layer model of each partition.

[0204] The topological layer modeling unit is configured to perform detection and identification of devices in each partition by using a YOLOv8 model combined with an attention mechanism based on the video and / or picture of each partition, and perform coupling of the topological connection of the devices by using a neural network based on the detection and identification result, thereby obtaining a digital twin topological layer model of each partition.

[0205] The attribute layer modeling unit is configured to dynamically associate the electrical account of each partition with the devices in the digital twin physical layer model, thereby obtaining a digital twin attribute layer model of each partition.

[0206] The model coupling unit is configured to couple the digital twin physical layer model, the topological layer model, and the attribute layer model of each partition, thereby constructing a digital twin layered model of each partition.

[0207] Preferably, the physical layer modeling unit comprises a semantic segmentation subunit, a convex hull modeling subunit, a retrieval subunit, and a physical layer construction subunit.

[0208] The semantic segmentation subunit is configured to perform semantic segmentation on the laser point cloud of each partition by using a density-based clustering algorithm, thereby obtaining a bounding box of each cluster, a center point of each cluster, device coordinates, a device name, and a device type in each partition.

[0209] The convex hull modeling subunit is configured to perform convex hull modeling based on the bounding box of each cluster, the center point of each cluster, the device coordinates, the device name, and the device type in each partition, thereby obtaining a convex hull model of each partition.

[0210] The retrieval subunit is configured to retrieve a device three-dimensional model from the component library based on the device type in each partition.

[0211] The physical layer construction subunit is configured to couple the device three-dimensional model and the convex hull model of each partition based on the device name in each partition, thereby obtaining a digital twin physical layer model of each partition.

[0212] Preferably, the topology layer modeling unit comprises a detection subunit, an identification subunit, a graph data subunit, and a topology layer construction subunit.

[0213] The detection subunit is configured to perform device type detection on the video and / or picture of each partition by using a YOLOv8 model to obtain device types, device coordinates, and confidence levels in each partition.

[0214] The identification subunit is configured to identify the connection relationship and connection points between devices in each partition based on the device types, device coordinates, and confidence levels in each partition by using an attention mechanism.

[0215] The graph data subunit is configured to construct a graph data structure for each partition by taking the devices corresponding to the connection points in each partition as nodes and taking the connection relationship between devices in each partition as edges.

[0216] The topology layer construction subunit is configured to obtain a topology structure graph for each partition by using a neural network model based on the graph data structure for each partition, and take the topology structure graph of the power distribution network in each partition as a digital twin topology layer model for each partition.

[0217] Preferably, the attribute layer model unit comprises a binding subunit and an attribute layer construction subunit.

[0218] The binding subunit is configured to take the electrical account of each partition as an electrical attribute, and bind the electrical attribute on the devices in the digital twin physical layer model of each partition.

[0219] The attribute layer construction subunit is configured to dynamically associate and construct a digital twin attribute layer model for each partition by establishing a real-time updating mechanism and associating the electrical attribute with the digital twin physical layer model of each partition.

[0220] Preferably, the modeling sub-module further comprises a visualization unit.

[0221] The visualization unit is configured to connect the digital twin attribute layer model of each partition to a visualization component through a data interface after the attribute layer model unit is invoked and before the model coupling unit is invoked.

[0222] Preferably, the hierarchical modeling module further comprises a monitoring sub-module, a change reconstruction sub-module, and an occlusion reconstruction sub-module.

[0223] The monitoring sub-module is configured to construct a spatial hash index after the modeling sub-module is invoked, and use the spatial hash index to monitor the changed devices in the power distribution network and use a deep learning model to detect the occluded or damaged devices in each partition.

[0224] The change reconstruction submodule is used to identify the boundary equipment of the adjacent partitions when a change in equipment is detected in any partition of the distribution network, and to reconstruct the partition with the change based on the changed equipment and the boundary equipment of the adjacent partitions.

[0225] The occlusion reconstruction submodule is used to identify the features of the occluded or damaged equipment when any part of the distribution network is detected to be occluded or damaged, and to use a generative adversarial network to complete the features of the occluded or damaged equipment in order to reconstruct a digital twin model of the occluded or damaged equipment.

[0226] Preferably, the change and reconstruction submodule includes: a boundary subgraph unit and a change and reconstruction unit;

[0227] The boundary subgraph unit is used to locate the boundary devices of the adjacent partitions of the distribution network when a device change is detected in any partition of the distribution network, by using a spatial hash index, and to construct a boundary subgraph based on the boundary devices of the adjacent partitions of the changed partition.

[0228] The change reconstruction unit is used to reconstruct a digital twin hierarchical model of the partition with changes based on the changed device, using a topology-spatial joint analytical algorithm, and to perform topological structure matching between the reconstructed digital twin hierarchical model and the boundary subgraph using the VF2 algorithm, so as to unify coordinates and establish mapping relationships.

[0229] Preferably, the system further includes: a component library construction module;

[0230] The component library construction module is used to create parametric modeling templates for various types of equipment based on the type of distribution network equipment; configure parameters for the parametric modeling templates of various types of equipment and establish a mapping relationship between parameters and model attributes; define installation constraint rules for various types of equipment based on the mapping relationship between parameters and model attributes; and construct a distribution network component library based on the parametric modeling templates of various types of equipment, the mapping relationship between parameters and model attributes, and the installation constraint rules for various types of equipment.

[0231] Preferably, the system further includes: a data enhancement module;

[0232] The data enhancement module is used to enhance the multi-source data of the power distribution network based on synchronous positioning and map building technology after calling the spatiotemporal unification module and before calling the hierarchical modeling module.

[0233] Preferably, the system further includes: a difference elimination module;

[0234] The difference elimination module is configured to construct a global optimization problem of the power distribution network multi-source data before calling the spatio-temporal unification module and before calling the hierarchical modeling module, and to use a general graph optimization library to solve the optimization problem to eliminate the benchmark difference between the power distribution network multi-source data.

[0235] The system provided by the embodiment has the following effects:

[0236] (1) Improve modeling efficiency. Break through the traditional artificial dependence mode, and greatly shorten the modeling cycle through intelligent fusion and automatic analysis of multi-source data, to meet the demand of high-frequency update and rapid response of power grid.

[0237] (2) Strengthen the accuracy of spatial topology coupling. Solve the problem of splitting of device geographic coordinates and electrical connection relationship, realize sub-meter level spatial positioning and accurate matching of topology, and support high-precision fault diagnosis and state analysis.

[0238] (3) Support elastic expansion and dynamic update. Build a partition model seamless splicing mechanism, break through the limitation of whole reconstruction for local update of large-scale power distribution network model, and significantly reduce the operation and maintenance complexity and resource consumption.

[0239] (4) Adapt to rapid integration of new power equipment. Establish a standardized component library, compatible with distributed photovoltaic, energy storage and other new equipment, improve the modeling flexibility and reusability in new energy scenarios.

[0240] (5) Enhance the robustness of complex scenarios. Improve the modeling reliability in harsh environments such as vegetation obstruction and equipment corrosion, and guarantee the integrity and availability of the whole-terrain power distribution network model.

[0241] (6) Ensure the consistency of multi-level data. Eliminate multi-source data conflicts through hierarchical verification mechanism, realize accurate cooperation of physical space, topological connection and electrical properties, and ensure the global consistency of the model.

[0242] Through systematic innovation, the present application solves the problem of collaborative optimization of efficiency, accuracy and expansibility in the field of power distribution network digital twin, and provides core digital support for building a new type of power system with high flexibility and high reliability.

[0243] Embodiment 3:

[0244] As Figure 10As shown, the present application also provides an electronic device, which can be a computer device, a single-chip microcomputer device, a smart mobile device, etc. The electronic device in the embodiment can include a processor, a memory, a transceiver component, etc. The memory, the processor and the transceiver component are connected through a bus; the memory can be used to store an execution program, and the exemplary execution program can include instructions; the processor is used to execute the instructions stored in the memory. The memory can also be used to store data, which can be called and / or modified when the instructions are executed.

[0245] The processor can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc., which are the computing core and control core of the terminal, and are suitable for implementing one or more instructions, and are specifically suitable for loading and executing one or more instructions in the storage medium to implement a corresponding method flow or a corresponding function, so as to implement the steps of the power distribution network layered partition visualization modeling method in the above embodiment.

[0246] The present application unifies multi-source data to the same space-time coordinate system, eliminates the reference difference of multi-source data, and improves the positioning accuracy; the layered model is constructed by using the multi-source data of the unified space-time reference, the problem of the split of the geographical coordinates of the equipment and the electrical connection relationship is solved; the standardized component library is established, the modeling flexibility and the reuse ability in different scenes are improved, and the modeling efficiency is significantly improved; the seamless splicing mechanism of the partition model is constructed, the limitation that the local update of the large-scale power distribution network model needs to be reconstructed as a whole is broken through, the modeling cycle is greatly shortened, and the demand for high-frequency update and rapid response of the power grid is met.

[0247] Embodiment 4:

[0248] Based on the same inventive concept, the application also provides a readable storage medium, specifically an electronic device readable storage medium (Memory), which is a memory device in the electronic device and is used for storing programs and data. It can be understood that the storage medium herein can include the built-in storage medium in the electronic device, and of course can also include the expansion storage medium supported by the electronic device. The storage medium provides a storage space, and the storage space stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space, and the instructions can be one or more execution programs (including program codes). It should be noted that the storage medium herein can be a high-speed RAM memory or a non-volatile memory such as at least one disk memory. Loading and executing one or more instructions stored in the storage medium by the processor can realize the steps of the power distribution network layered partition visualization modeling method in the above embodiment.

[0249] The application unifies multi-source data to the same space-time coordinate system, eliminates the reference difference of multi-source data, and improves the positioning accuracy; the layered model is constructed by using the multi-source data of the unified space-time reference, the problem of the split of the geographical coordinates of the equipment and the electrical connection relationship is solved; the standardized component library is established, the modeling flexibility and reusability in different scenarios are improved, and the modeling efficiency is significantly improved; the seamless splicing mechanism of the partition model is constructed, the limitation that the local update of the large-scale power distribution network model needs to be reconstructed as a whole is broken through, the modeling cycle is greatly shortened, and the high-frequency update and rapid response demand of the power grid are met.

[0250] Those skilled in the art should understand that the embodiments of the application can be provided as a method, a system, or a computer program product. Therefore, the application can adopt a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the application can adopt the form of a computer program product implemented on one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer usable program codes.

[0251] The application is described with reference to flowcharts and / or block diagrams of the method, device (system), and computer program product according to the embodiments of the application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of the flows and / or blocks in the flowcharts and / or block diagrams can be realized by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices produce a machine that implements the functions described in the flowcharts and / or block diagrams. Figure 1one or more processes and / or blocks Figure 1 an apparatus for performing the functions recited in one or more processes and / or blocks.

[0252] These computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer readable memory produce an article of manufacture including instructions which implement the Figure 1 one or more processes and / or blocks Figure 1 an apparatus for performing the functions recited in one or more processes and / or blocks.

[0253] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the Figure 1 one or more processes and / or blocks Figure 1 an apparatus for performing the functions recited in one or more processes and / or blocks.

[0254] The above merely provides an embodiment of the present application, but is not intended to limit the present application, and any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall fall within the scope of the present application.

Claims

1. A hierarchical and zonal visualization modeling method for power distribution networks, characterized in that, include: Acquire multi-source data from the power distribution network and unify the spatiotemporal reference; The distribution network area is dynamically partitioned, and a digital twin hierarchical model is constructed for each partition using the multi-source data of the distribution network with a unified spatiotemporal reference and a pre-built distribution network component library. The digital twin hierarchical models of each zone are spliced ​​together, and the power distribution network operating conditions are rendered using simulation technology to construct a digital twin hierarchical model of the power distribution network.

2. The method as described in claim 1, characterized in that, The acquisition of multi-source data from the distribution network and the unification of spatiotemporal references include: Acquire multi-source data from the power distribution network and perform preprocessing; The timestamps of the preprocessed multi-source data of the distribution network are obtained, and the timestamps are recorded in the header file of the multi-source data of the distribution network using the Global Navigation Satellite System to unify the time reference. The latitude and longitude of the distribution network location are obtained using the Global Navigation Satellite System. Based on the latitude and longitude of the distribution network location, a seven-parameter Bursa model is used to transform the synchronous positioning and map construction coordinates corresponding to the multi-source data of the distribution network after unifying the time reference to the global coordinate system, so as to unify the spatial reference. The multi-source data of the power distribution network includes global image data of the power distribution network taken by drones, videos and / or pictures of local equipment taken by inspection robots and / or manuals, electrical ledgers, laser point clouds and infrared thermal imaging data obtained by data acquisition and monitoring control systems; the preprocessing includes noise reduction, distortion correction and format unification.

3. The method as described in claim 2, characterized in that, The process of dynamically partitioning the distribution network area and constructing a hierarchical digital twin model for each partition using multi-source data of the distribution network based on a unified spatiotemporal reference and a pre-built distribution network component library includes: Dynamically partition the area where the distribution network is located based on the equipment, topology, and functions of the distribution network. Based on the component library and the multi-source data of the distribution network in each zone, the topology-space joint analysis algorithm is used to couple the physical space, topology connection and electrical attributes of the equipment in each zone to obtain the digital twin hierarchical model of each zone; The topology-space joint analytical algorithm is constructed by combining neural networks, the YOLOv8 model, attention mechanisms, and density-based clustering algorithms.

4. The method as described in claim 3, characterized in that, Based on the component library and the multi-source data of the distribution network in each zone, a topology-spatial joint analytical algorithm is used to couple the physical space, topological connections, and electrical attributes of the equipment in each zone to obtain a hierarchical digital twin model for each zone, including: For each partition, a density-based clustering algorithm is used to perform semantic segmentation on the laser point cloud of each partition, and the physical space coupling of the device is performed by retrieving the 3D model of the device from the component library based on the semantic segmentation results, so as to obtain the digital twin physical layer model of each partition. Based on the videos and / or images of each partition, the YOLOv8 model combined with the attention mechanism is used to detect and identify the devices in each partition. Based on the detection and identification results, the topological connections of the devices are coupled using a neural network to obtain the digital twin topological layer model of each partition. The electrical attributes of the devices in the electrical ledger and the digital twin physical layer model of each partition are dynamically associated to obtain the digital twin attribute layer model of each partition. The digital twin physical layer model, topology layer model, and attribute layer model of each partition are coupled to construct a layered digital twin model for each partition.

5. The method as described in claim 4, characterized in that, The process employs a density-based clustering algorithm to perform semantic segmentation on the laser point cloud of each partition, and retrieves the device's 3D model from the component library based on the semantic segmentation results to perform physical spatial coupling of the device, thereby obtaining a digital twin physical layer model for each partition, including: A density-based clustering algorithm is used to perform semantic segmentation on the laser point cloud of each partition, and the bounding box, center point, device coordinates, device name and device type of each cluster in each partition are obtained. Convex hull models for each partition are obtained by performing convex hull modeling based on the bounding boxes of each cluster, the center points of each cluster, the device coordinates, the device names, and the device types in each partition. Based on the device type in each partition, retrieve the 3D model of the device from the component library; Based on the device names in each partition, the 3D model of the device and the convex hull model of each partition are coupled to obtain the digital twin physical layer model of each partition.

6. The method as described in claim 4, characterized in that, The video and / or image data for each partition are used to detect and identify devices in each partition using a YOLOv8 model combined with an attention mechanism. Based on the detection and identification results, a neural network is used to generate and couple the topological connections of the devices, resulting in a digital twin topology layer model for each partition, including: The YOLOv8 model is used to detect the device type of the videos and / or images in each partition, and the device type, device coordinates and confidence score in each partition are obtained. Based on the device type, device coordinates, and confidence level in each partition, an attention mechanism is used to identify the connection relationships and connection points between devices in each partition; The data structure of each partition graph is constructed using the devices corresponding to the connection points in each partition as nodes and the connection relationships between devices in each partition as edges. Based on the data structure of each partition map, a neural network model is used to infer the topology structure of each partition, and the distribution network topology structure of each partition is used as the digital twin topology layer model of each partition.

7. The method as described in claim 4, characterized in that, The step of dynamically associating the electrical attributes of the devices in the electrical ledgers and digital twin physical layer models of each partition to obtain the digital twin attribute layer models of each partition includes: The electrical ledgers of each partition are used as electrical attributes, and the electrical attributes are bound to the devices in the digital twin physical layer model of each partition; By establishing a real-time update mechanism, electrical attributes and the digital twin physical layer models of each partition are dynamically correlated to construct the digital twin attribute layer models of each partition.

8. The method as described in claim 4, characterized in that, Before dynamically associating the electrical attributes of the devices in the electrical ledger and the digital twin physical layer model of each partition to obtain the digital twin attribute layer model of each partition, and then coupling the digital twin physical layer model, topology layer model, and attribute layer model of each partition to construct the digital twin layered model of each partition, the process further includes: The digital twin attribute layer model of each partition is connected to the visualization component through a data interface.

9. The method as described in claim 3, characterized in that, After obtaining the hierarchical digital twin model of each partition by coupling the physical space, topological connections, and electrical attributes of the equipment in each partition using a topology-spatial joint analytical algorithm based on the component library and the multi-source data of the distribution network in each partition, the process further includes: Construct a spatial hash index and use the spatial hash index to perform partitioned monitoring of changed equipment in the distribution network, and use a deep learning model to perform partitioned detection of obstructed or damaged equipment; When a change in equipment is detected in any part of the distribution network, the boundary equipment of the adjacent parts of the affected parts is identified, and a digital twin hierarchical model is reconstructed for the affected parts based on the changed equipment and the boundary equipment of the adjacent parts of the affected parts. When equipment blockage or damage is detected in any section of the distribution network, the features of the blocked or damaged equipment are identified, and the features of the blocked or damaged equipment are completed using a generative adversarial network to reconstruct a digital twin model of the blocked or damaged equipment.

10. The method as described in claim 9, characterized in that, When a device change is detected in any section of the distribution network, the boundary devices of the adjacent sections are identified, and a digital twin hierarchical model is reconstructed for the changed section based on the changed device and the boundary devices of the adjacent sections, including: When a device change is detected in any partition of the distribution network, the boundary device of the partition adjacent to the partition where the change occurred is located by spatial hash index, and a boundary subgraph is constructed based on the boundary device of the partition adjacent to the partition where the change occurred. Based on the changed equipment, a topology-spatial joint analytical algorithm is used to reconstruct a digital twin hierarchical model for the changed partitions, and the VF2 algorithm is used to perform topological structure matching between the reconstructed digital twin hierarchical model and the boundary subgraph in order to unify coordinates and establish mapping relationships.

11. The method as described in claim 1, characterized in that, The construction of the power distribution network component library includes: Based on the type of power distribution network equipment, create parametric modeling templates for various types of equipment; Configure parameters for parametric modeling templates for various types of equipment and establish a mapping relationship between parameters and model attributes; Based on the mapping relationship between the parameters and model attributes, installation constraint rules for various types of equipment are defined; Based on the parametric modeling templates of various equipment, the mapping relationship between parameters and model attributes, and the installation constraint rules of various equipment, a power distribution network component library is constructed.

12. The method as described in claim 1, characterized in that, After acquiring multi-source data of the distribution network and unifying the spatiotemporal reference, the region where the distribution network is located is dynamically partitioned, and before constructing a hierarchical digital twin model for each partition using the multi-source data of the distribution network with the unified spatiotemporal reference and a pre-built distribution network component library, the process further includes: The multi-source data of the power distribution network is enhanced based on synchronous positioning and mapping technology.

13. The method as described in claim 1, characterized in that, After acquiring multi-source data of the distribution network and unifying the spatiotemporal reference, the region where the distribution network is located is dynamically partitioned, and before constructing a hierarchical digital twin model for each partition using the multi-source data of the distribution network with the unified spatiotemporal reference and a pre-built distribution network component library, the process further includes: A global optimization problem for the multi-source data of the distribution network is constructed, and the optimization problem is solved using a general graph optimization library to eliminate the benchmark differences among the multi-source data of the distribution network.

14. A hierarchical and zoning visualization modeling system for power distribution networks, characterized in that, include: Spatiotemporal unification module, hierarchical modeling module, and partition splicing module; The spatiotemporal unification module is used to acquire multi-source data of the distribution network and unify the spatiotemporal reference. The hierarchical modeling module is used to dynamically partition the area where the distribution network is located, and to construct a digital twin hierarchical model for each partition using the multi-source data of the distribution network with a unified spatiotemporal reference and a pre-built distribution network component library. The partition splicing module is used to splice the hierarchical digital twin models of each partition and use inference technology to render the power distribution network operating conditions to construct a hierarchical digital twin model of the power distribution network.

15. The system as described in claim 14, characterized in that, The spatiotemporal unification module includes: a preprocessing unit, a time unification unit, and a space unification unit; The preprocessing unit is used to acquire and preprocess the multi-source data of the power distribution network. The time unification unit is used to obtain the timestamps of the preprocessed multi-source data of the distribution network, and to record the timestamps in the header file of the multi-source data of the distribution network using the Global Navigation Satellite System, so as to unify the time reference. The unified spatial unit is used to obtain the latitude and longitude of the distribution network location using the global navigation satellite system, and uses the latitude and longitude of the distribution network location as a reference to transform the synchronous positioning and map construction coordinates corresponding to the multi-source data of the distribution network after unifying the time reference to the global coordinate system using a seven-parameter Bursa model, so as to unify the spatial reference. The multi-source data of the power distribution network includes global image data of the power distribution network taken by drones, videos and / or pictures of local equipment taken by inspection robots and / or manuals, electrical ledgers, laser point clouds and infrared thermal imaging data obtained by data acquisition and monitoring control systems; the preprocessing includes noise reduction, distortion correction and format unification.

16. The system as described in claim 15, characterized in that, The hierarchical modeling module includes: a dynamic partitioning submodule and a modeling submodule; The dynamic partitioning submodule is used to dynamically partition the area where the distribution network is located based on the equipment, topology and functions of the distribution network. The modeling submodule is used to couple the physical space, topological connection and electrical attributes of the equipment in each partition with the component library and the multi-source data of the distribution network in each partition to obtain the digital twin hierarchical model of each partition. The topology-space joint analytical algorithm is constructed by combining neural networks, the YOLOv8 model, attention mechanisms, and density-based clustering algorithms.

17. The system as claimed in claim 16, characterized in that, The modeling submodule includes: a physical layer modeling unit, a topology layer modeling unit, an attribute layer modeling unit, and a model coupling unit; The physical layer modeling unit is used to perform semantic segmentation of the laser point cloud of each partition using a density-based clustering algorithm, and to retrieve the device 3D model from the component library based on the semantic segmentation results to perform physical space coupling of the device, thereby obtaining the digital twin physical layer model of each partition. The topology layer modeling unit is used to detect and identify devices in each partition based on the video and / or image of each partition, using the YOLOv8 model combined with an attention mechanism, and to couple the topological connections of the devices using a neural network based on the detection and identification results, so as to obtain a digital twin topology layer model of each partition. The attribute layer model unit is used to dynamically associate the electrical attributes of the electrical ledgers and the devices in the digital twin physical layer model of each partition to obtain the digital twin attribute layer model of each partition. The model coupling unit is used to couple the digital twin physical layer model, topology layer model and attribute layer model of each partition to construct the digital twin layered model of each partition.

18. The system as claimed in claim 17, characterized in that, The physical layer modeling unit includes: a semantic segmentation subunit, a convex hull modeling subunit, a retrieval subunit, and a physical layer construction subunit; The semantic segmentation subunit is used to perform semantic segmentation on the laser point cloud of each partition using a density-based clustering algorithm to obtain the bounding box of each cluster, the center point of each cluster, the device coordinates, the device name and the device type in each partition. The convex hull modeling subunit is used to perform convex hull modeling based on the bounding box of each cluster in each partition, the center point of each cluster, the device coordinates, the device name and the device type to obtain the convex hull model of each partition. The retrieval subunit is used to retrieve the 3D model of the device from the component library based on the device type in each partition; The physical layer construction subunit is used to couple the 3D model of the device and the convex hull model of each partition based on the device name in each partition to obtain the digital twin physical layer model of each partition.

19. The system as claimed in claim 17, characterized in that, The topology layer modeling unit includes: a detection subunit, an identification subunit, a graph data subunit, and a topology layer construction subunit; The detection subunit is used to perform device type detection on the videos and / or images in each partition using the YOLOv8 model, and obtain the device type, device coordinates and confidence level in each partition; The identification subunit is used to identify the connection relationships and connection points between devices in each partition based on the device type, device coordinates and confidence level in each partition, using an attention mechanism. The graph data subunit is used to construct the graph data structure of each partition with the devices corresponding to the connection points in each partition as nodes and the connection relationships between the devices in each partition as edges. The topology layer construction sub-unit is used to obtain the topology structure diagram of each partition based on the data structure of each partition diagram using a neural network model, and to use the distribution network topology structure diagram of each partition as the digital twin topology layer model of each partition.

20. The system as claimed in claim 17, characterized in that, The attribute layer model unit includes: a binding subunit and an attribute layer construction subunit; The binding subunit is used to use the electrical ledger of each partition as the electrical attribute, and bind the electrical attribute to the device in the digital twin physical layer model of each partition; The attribute layer construction subunit is used to construct the digital twin attribute layer model of each partition by dynamically associating electrical attributes with the digital twin physical layer model of each partition through a real-time update mechanism.

21. The system as claimed in claim 16, characterized in that, The hierarchical modeling module also includes: a monitoring submodule, a change reconstruction submodule, and an occlusion reconstruction submodule; The monitoring submodule is used to construct a spatial hash index after calling the modeling submodule, and to use the spatial hash index to perform partitioned monitoring of changed equipment in the distribution network, and to use a deep learning model to perform partitioned detection of obstructed or damaged equipment. The change reconstruction submodule is used to identify the boundary equipment of the adjacent partitions when a change in equipment is detected in any partition of the distribution network, and to reconstruct the partition with the change based on the changed equipment and the boundary equipment of the adjacent partitions. The occlusion reconstruction submodule is used to identify the features of the occluded or damaged equipment when any part of the distribution network is detected to be occluded or damaged, and to use a generative adversarial network to complete the features of the occluded or damaged equipment in order to reconstruct a digital twin model of the occluded or damaged equipment.

22. The system as claimed in claim 21, characterized in that, The change and reconstruction submodule includes: a boundary subgraph unit and a change and reconstruction unit; The boundary subgraph unit is used to locate the boundary devices of the adjacent partitions of the distribution network when a device change is detected in any partition of the distribution network, by using a spatial hash index, and to construct a boundary subgraph based on the boundary devices of the adjacent partitions of the changed partition. The change reconstruction unit is used to reconstruct a digital twin hierarchical model of the partition with changes based on the changed device, using a topology-spatial joint analytical algorithm, and to perform topological structure matching between the reconstructed digital twin hierarchical model and the boundary subgraph using the VF2 algorithm, so as to unify coordinates and establish mapping relationships.

23. An electronic device, characterized in that, include: At least one processor and memory; The memory and processor are connected via a bus; The memory is used to store one or more programs; When the one or more programs are executed by the at least one processor, a hierarchical and zonal visualization modeling method for power distribution networks as described in any one of claims 1 to 13 is implemented.

24. A readable storage medium, characterized in that, It contains an executable program, which, when executed, implements a distribution network hierarchical and zoning visualization modeling method as described in any one of claims 1 to 13.

Images