Construction Method and System of Distribution Network Visualization Platform Based on SG-CIM

Abstract

This invention discloses a method and system for constructing a distribution network visualization platform based on SG-CIM, belonging to the technical field of distribution network visualization platform construction. The method for constructing this SG-CIM-based distribution network visualization platform involves acquiring power text information of the distribution network and inputting it into a pre-trained semantic recognition model for parsing and analysis. This yields equipment attribute triples and equipment relationship triples for each power device. Topology reconstruction processing is performed on the equipment relationship triples to generate a power device topology diagram. Graph modeling is then performed to generate a semantic knowledge graph. Finally, quality verification analysis is conducted to obtain a semantic structure coordination index. This invention constructs a distribution network visualization platform using the semantic structure coordination index, thereby ensuring the consistency of the semantic expression and topological connection logic of each power device in the distribution network from the source. This improves the intelligent parsing capability, operational visualization capability, and automated verification capability of the distribution network structure.

Description

Technical Field

[0001] This invention relates to the field of power distribution network visualization platform construction technology, specifically to a method and system for constructing a power distribution network visualization platform based on SG-CIM. Background Technology

[0002] SG-CIM, as the unified standard for power grid information modeling in my country, originates from the IEC 61970 / 61968 international standard and has been localized and extended to suit the business characteristics of the State Grid Corporation of China. It is the authoritative foundation for building a semantic knowledge graph for distribution networks. Through the standardized definition of equipment classes, attributes, and connection relationships, SG-CIM provides power companies with a unified data semantic description system, effectively supporting equipment lifecycle management, operational status awareness, and data interoperability. However, despite the standardization and consistency of SG-CIM in model description, how to transform its abstract semantic model into an intuitive and interactive graphical visualization, and achieve the integrated display of equipment topology, attribute information, and operational status, remains one of the current technical challenges in the industry. At present, most distribution network visualization platforms only support static structure display and lack the ability to integrate, associate, and reason about attribute semantics, resulting in severe information fragmentation, unclear interaction levels, and low fault tolerance and poor repairability for incomplete data and ambiguous relationships.

[0003] Existing technologies, such as the patent application with publication number CN113191687B, disclose a method and system for visualizing panoramic information of a resilient distribution network. This method monitors system data in real time, calculates and displays dynamic resilience indicators of the distribution network, provides refined early warnings based on indicator thresholds, and marks risky components. It also simulates extreme weather events offline, calculates and displays the comprehensive resilience score of the distribution network, ranks distribution network components by importance, and marks the most important components. This invention reflects the resilience changes of the distribution network through both offline assessment and online monitoring, and further provides an intuitive display through a panoramic information visualization system. This allows for guidance on component reinforcement schemes and resource allocation strategies at the planning level, and assists staff in emergency dispatch and rapid recovery at the operational level.

[0004] Based on the above findings, the limitations of existing technologies include at least the following issues: Existing technologies lack a deep understanding and fusion expression capability of the semantic hierarchy of power equipment, resulting in structure diagrams that only present connection relationships while lacking equipment semantic content. This leads to the widespread existence of structural silos and semantic fragments, severely restricting the intelligent parsing capability of distribution network data and the depth of expression of visualization maps. For example, connection relationships lack semantic rule support, making it difficult to automatically judge the rationality of connections, such as whether there are issues like incorrect connections across voltage levels, abnormal trunk-branch connections, or incorrect loop dependency relationships. These errors in connection logic will not be apparent in traditional topology diagrams, but they can easily create hidden dangers in actual operation, forming potential semantic conflict paths, thereby affecting the structural security and operation and maintenance efficiency of the entire distribution network. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method and system for constructing a power distribution network visualization platform based on SG-CIM. This solves the problem that existing technologies lack the ability to perform semantic fusion and connection logic verification of equipment, resulting in structural islands and semantic conflict paths in the power distribution network structure diagram, making it difficult to support intelligent parsing and safe operation and maintenance.

[0006] To achieve the above objectives, this invention provides the following technical solution: a method for constructing a distribution network visualization platform based on SG-CIM, comprising the following steps: acquiring the power text information of the distribution network to be constructed and inputting it into a pre-trained semantic recognition model for parsing and analysis to obtain a structured semantic dataset for each power device in the distribution network to be constructed, including device attribute triples and device relationship triples; performing topology reconstruction processing on the device relationship triples of each power device in the distribution network to be constructed to generate a power device topology diagram of the distribution network to be constructed; performing graph modeling processing on the power device topology diagram based on the device attribute triples of each power device in the distribution network to be constructed to generate a semantic knowledge graph of the distribution network to be constructed; and performing quality verification analysis on the semantic knowledge graph of the distribution network to be constructed to obtain a semantic structure coordination index of the semantic knowledge graph of the distribution network to be constructed, and constructing a distribution network visualization platform based on the semantic structure coordination index.

[0007] Furthermore, the semantic recognition model is specifically a semantic fusion network, which includes an input layer, a semantic encoding layer, a ternary construction layer, a graph structure modeling layer, and a ternary output layer.

[0008] Further, the specific steps for obtaining the structured semantic dataset of each power device in the distribution network to be constructed are as follows: In the input layer of the semantic fusion network, the power text information of each power device in the distribution network to be constructed is received and preprocessed; in the semantic encoding layer of the semantic fusion network, multi-layer contextual semantic modeling is performed on the preprocessed power text information of the distribution network to be constructed to obtain a high-dimensional semantic embedding vector set of the distribution network to be constructed; in the triple construction layer of the semantic fusion network, semantic role recognition is performed on the high-dimensional semantic embedding vector set of the distribution network to be constructed to obtain an initial triple structure set of each power device in the distribution network to be constructed; in the graph structure modeling layer of the semantic fusion network, graph semantic propagation is performed on the initial triple structure set of each power device in the distribution network to be constructed to obtain a set of structure-enhanced triples of each power device in the distribution network to be constructed; in the triple output layer of the semantic fusion network, the set of structure-enhanced triples of each power device in the distribution network to be constructed is filtered and classified to obtain a triple set of device attributes and a triple set of device relationships of each power device in the distribution network to be constructed.

[0009] Furthermore, the specific steps for generating the power equipment topology diagram of the distribution network to be constructed are as follows: read the equipment relationship triplet of each power equipment in the distribution network to be constructed, and perform comprehensive analysis to obtain several groups of candidate connected power equipment in the distribution network to be constructed; perform topology connectivity analysis on each group of candidate connected power equipment in the distribution network to be constructed to obtain several groups of actual connected power equipment in the distribution network to be constructed, and generate the power equipment topology diagram of the distribution network to be constructed.

[0010] Further, the specific steps for obtaining several groups of actually connected power equipment in the distribution network to be constructed are as follows: Obtain the connection index set for each group of candidate connected power equipment in the distribution network to be constructed, including node consistency index, structural direction difference index, and semantic confidence index; perform a comprehensive analysis on the connection index set for each group of candidate connected power equipment in the distribution network to be constructed to obtain the topology connection confidence index for each group of candidate connected power equipment in the distribution network to be constructed; compare the topology connection confidence index of each group of candidate connected power equipment in the distribution network to be constructed with a preset topology connection confidence index threshold to obtain several groups of actually connected power equipment in the distribution network to be constructed.

[0011] Furthermore, the specific formula for calculating the topology connectivity confidence index of a certain group of candidate connected power devices in the distribution network to be constructed is as follows: ;in, Let be the topology connectivity confidence index for a set of candidate connected power devices in a power distribution network to be constructed. The node consistency index of a group of candidate connected power devices in a power distribution network to be constructed. The node consistency adjustment coefficient is stored in the database. The structural orientation difference index of a group of candidate connected power equipment in the power distribution network to be constructed. These are the structural orientation adjustment coefficients stored in the database. The collaborative penalty adjustment coefficient is stored in the database. The semantic confidence index of a group of candidate connected power devices for a power distribution network to be constructed. The semantic suppression coefficients are stored in the database. These are the semantic adjustment coefficients stored in the database.

[0012] Furthermore, the specific steps for generating the semantic knowledge graph of the distribution network to be constructed are as follows: read the equipment attribute triples of each power device in the distribution network to be constructed, and perform matching processing in combination with the power device topology diagram to obtain the node attribute set of the power device topology diagram of the distribution network to be constructed; perform semantic integration processing on the node attribute set of the power device topology diagram of the distribution network to be constructed to obtain the graph node set of the power device topology diagram of the distribution network to be constructed, and perform comprehensive analysis to generate the semantic knowledge graph of the distribution network to be constructed.

[0013] Furthermore, the specific steps for obtaining the semantic structure coordination index of the semantic knowledge graph of the distribution network to be constructed are as follows: A comprehensive analysis is performed on the semantic knowledge graph of the distribution network to be constructed to obtain a set of graph quality evaluation indices, including the structural connectivity integrity index, semantic consistency index, and topological semantic mapping index; and a comprehensive analysis is performed on the set of graph quality evaluation indices of the semantic knowledge graph of the distribution network to be constructed to obtain the semantic structure coordination index of the semantic knowledge graph of the distribution network to be constructed.

[0014] Furthermore, the specific formula for calculating the semantic structure coordination index of the semantic knowledge graph of the distribution network to be constructed is as follows: ;

[0015] in, This is the semantic structure coordination index for the semantic knowledge graph of the distribution network to be constructed. These are the superimposed adjustment coefficients stored in the database. The structural connectivity completeness index is used to construct the semantic knowledge graph of the power distribution network. The structural connectivity integrity adjustment coefficient is stored in the database. This is the semantic consistency index for the semantic knowledge graph of the distribution network to be constructed. These are the semantic consistency adjustment coefficients stored in the database. This is the topological semantic mapping index for the semantic knowledge graph of the distribution network to be constructed. These are the semantic mapping adjustment coefficients stored in the database. This refers to the coordination baseline adjustment coefficient stored in the database.

[0016] The system for constructing a distribution network visualization platform based on SG-CIM includes: a data acquisition and parsing module, used to acquire the power text information of the distribution network to be constructed and input it into a pre-trained semantic recognition model for parsing and analysis, to obtain a structured semantic dataset for each power device in the distribution network to be constructed, including device attribute triples and device relationship triples; a device topology reconstruction module, used to perform topology reconstruction processing on the device relationship triples of each power device in the distribution network to be constructed, to generate a power device topology diagram of the distribution network to be constructed; a power grid graph modeling module, used to perform graph modeling processing on the power device topology diagram based on the device attribute triples of each power device in the distribution network to be constructed, to generate a semantic knowledge graph of the distribution network to be constructed; and a distribution visualization platform construction module, used to perform quality verification analysis on the semantic knowledge graph of the distribution network to be constructed, to obtain the semantic structure coordination index of the semantic knowledge graph of the distribution network to be constructed, and to construct the distribution network visualization platform based on the semantic structure coordination index.

[0017] The present invention has the following beneficial effects:

[0018] (1) The construction method of the distribution network visualization platform based on SG-CIM constructs equipment attribute triples and equipment relationship triples based on the SG-CIM standard, and performs deep recognition in conjunction with semantic fusion network. At the same time, it combines graph neural network to enhance structural consistency, thereby ensuring the consistency of semantic expression and topological connection logic of each power device in the distribution network from the source. Then, in the process of constructing topological structure graph and semantic knowledge graph, attribute mounting and semantic integration processing are performed on each device node to realize joint modeling of node context attributes and connection relationship. Then, the connection rationality and semantic consistency are explicitly identified through semantic structure coordination index, so that the distribution network topology no longer has isolated nodes, semantic conflict links or mismatched connection relationships, and significantly improves the intelligent analysis capability, operation visualization capability and automated verification capability of distribution network structure, while providing a solid semantic foundation for subsequent structural evolution analysis.

[0019] (2) The construction method of the distribution network visualization platform based on SG-CIM is to construct a distribution network topology connection confidence index, integrate the node consistency index, structural direction difference index and semantic confidence index, and combine coordinated adjustment and nonlinear function combination to construct a comprehensive evaluation index system for the quality of power equipment connection relationship. In this way, by quantitatively analyzing and constraining the topology connectivity, spatial geometric direction matching degree and semantic recognition confidence of candidate connection equipment pairs, it can automatically identify problematic connection relationships with structural anomalies, directional conflicts or semantic inconsistencies, and generate a power equipment topology structure diagram accordingly. On this basis, the topology diagram not only presents electrical connection relationships, but also explicitly embeds connection edge weights, thereby supporting subsequent functions such as topology correction suggestion generation and semantic conflict path awareness, and comprehensively improving the physical rationality and semantic credibility of the distribution network structure expression.

[0020] (3) The construction method of the distribution network visualization platform based on SG-CIM, through the construction of the semantic structure coordination index system, and the introduction of three evaluation standards, namely the structural connectivity integrity index, the semantic consistency index and the topological semantic mapping index, realizes the multi-dimensional quantitative judgment of the overall quality of the distribution network semantic knowledge graph. Each index is based on the entity-level analysis results of graph nodes, attributes and relationships, which can measure the degree of equipment topology coverage, the degree of equipment attribute standardization and the degree of semantic matching of connection relationship, thereby driving the semantic repair suggestion generation and graph evolution judgment mechanism, thus ensuring the adaptive ability, evolution controllability and semantic interpretability of the knowledge graph in the running state.

[0021] (4) The construction system of the distribution network visualization platform based on SG-CIM forms a closed loop of data parsing, topology reconstruction, graph modeling and visualization presentation through modular construction architecture. This enables the transformation of the distribution network from original semantic text to structural graph and then to the visualization platform to be highly automated and controllable. Moreover, the functions of each module in the system are clear and the boundaries are clear. For example, the semantic recognition model realizes the unified extraction of equipment attributes and relationships in the data acquisition and parsing module, while the equipment topology reconstruction module ensures the accurate restoration of structural connection logic. Furthermore, the power grid graph modeling module further completes the deep integration of attribute semantics and structural topology. Finally, the distribution visualization platform construction module completes the semantic consistency verification and view-level presentation mode control, thereby realizing a vertical closed loop link from the bottom semantic understanding to the upper visual interaction. This effectively improves the efficiency of distribution network modeling, the flexibility of platform deployment and the semantic governance capability of the overall system, and provides scalable and highly adaptable technical support for the construction of large-scale smart distribution networks.

[0022] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0023] Figure 1 This is a flowchart illustrating the construction method of the distribution network visualization platform based on SG-CIM according to the present invention.

[0024] Figure 2 This is a schematic diagram of the equipment attribute triplet sequence of the distribution network to be constructed in the construction method of the distribution network visualization platform based on SG-CIM of the present invention.

[0025] Figure 3 This is a schematic diagram of the equipment relationship triplet sequence of the distribution network to be constructed in the construction method of the distribution network visualization platform based on SG-CIM of the present invention.

[0026] Figure 4 This is a flowchart illustrating the specific steps involved in constructing a distribution network visualization platform based on SG-CIM according to the present invention to obtain several groups of actually connected power equipment in the distribution network to be constructed.

[0027] Figure 5 This is a schematic diagram of the candidate sequence connection index set of the distribution network to be constructed in the construction method of the distribution network visualization platform based on SG-CIM of the present invention.

[0028] Figure 6 This is a system block diagram of the SG-CIM-based power distribution network visualization platform of the present invention. Detailed Implementation

[0029] Please see Figure 1-3This invention provides a technical solution: a method for constructing a distribution network visualization platform based on SG-CIM, comprising the following steps: acquiring the power text information of the distribution network to be constructed (i.e., text information about equipment topology and attribute semantics conforming to the SG-CIM standard), and inputting it into a pre-trained semantic recognition model for parsing and analysis to obtain a structured semantic dataset for each power device in the distribution network to be constructed, including device attribute triples (representing the basic parameter information of the power device, having a structure format of device-attribute-attribute value, i.e., device name, attribute name, attribute value, with the subject being the specific device), and so on. For example, a predicate for a line or circuit breaker is a static attribute or configuration parameter of the device, and the object is the specific value of that attribute. A device relationship triple (describing the connection, dependency, or control relationship between power devices, with a device-relationship-device structure, i.e., device A, connection relationship, device B, where the subject and object are two different devices or facilities, and the predicate indicates a certain electrical connection, control, attribution, or dependency relationship between them) is used. Topology reconstruction processing is performed on the device relationship triples of each power device in the distribution network to be constructed, generating a power device topology diagram of the distribution network to be constructed. Based on the topology diagram of the distribution network to be constructed... The equipment attribute triples of each power device in the power grid are used to perform graph modeling on the power device topology diagram, generating a semantic knowledge graph of the distribution network to be constructed. A quality verification analysis is then performed on the semantic knowledge graph of the distribution network to be constructed to obtain its semantic structure coordination index. Based on this index, a distribution network visualization platform is constructed. The specific steps are as follows: The semantic structure coordination index of the distribution network to be constructed is compared with a preset semantic structure coordination index threshold. If the semantic structure coordination index of the distribution network to be constructed is lower than or equal to a certain threshold... At the preset semantic structure coordination index threshold, the system switches to a low-confidence visualization mode, loading only the basic topology and core node information (i.e., only loading backbone equipment, such as substations, main lines, buses, and the connection relationships of primary nodes, without presenting secondary branches and auxiliary equipment). At the same time, it explicitly labels semantically incomplete or unclearly connected equipment and generates semantic repair suggestions (generating a list of structural optimization suggestions, including recommended supplementary triplet items, connection relationships to be confirmed, and suspicious attribute conflict information, to assist users in manually repairing the semantic knowledge graph or triggering retraining), providing a basis for subsequent knowledge graph optimization.If the semantic structure coordination index of the semantic knowledge graph of the distribution network to be constructed is higher than the preset semantic structure coordination index threshold, then a high-confidence semantic-driven visualization mode is entered. This includes semantic fusion graph rendering (based on the topology, the attribute triples and connection relationships of each device node are fused and displayed, supporting the overlay of multi-dimensional information layers, such as attribute labels, type icons, parameter heatmaps, etc.), full-graph interactive navigation (users can click on device nodes to dynamically expand detailed device attributes, view their upstream and downstream device links, device control relationships, and operating status semantics), support for semantic path filtering based on attribute conditions (such as voltage level, load level, and device type), display of subnet structures that meet specific conditions, and structural evolution and comparative analysis (multiple versions of the semantic graph can be loaded to visually compare the process of topology changes, attribute updates, and connection logic evolution).

[0030] Specifically, the semantic recognition model is a semantic fusion network (combining graph neural networks and BERT), which includes an input layer, a semantic encoding layer, a ternary construction layer, a graph structure modeling layer, and a ternary output layer.

[0031] The input layer is used to parse the semantic text of electricity, extract the token sequence and standardize the units, providing structured input for semantic modeling.

[0032] The semantic encoding layer is used to perform contextual semantic modeling on the token sequence, generating a high-dimensional vector representation that expresses device attributes and relationships.

[0033] The ternary construction layer is used to identify entities, attributes, and relational terms, and to construct a preliminary set of attribute and relational triples for each power device.

[0034] The graph structure modeling layer is used to construct a graph structure from triples, leveraging graph neural networks to enhance the consistency between device semantics and topology.

[0035] The ternary output layer is used to filter and classify structure-enhanced triples, outputting standardized sets of attribute and relation triples according to the device.

[0036] The specific steps to obtain the structured semantic dataset of each power device in the distribution network to be constructed are as follows: In the input layer of the semantic fusion network, the power text information of each power device in the distribution network to be constructed is received and preprocessed, that is, the structured text is parsed for semantic tags, removing format tags and nested structures in XML, and only retaining entity values ​​and attribute fields in nodes to ensure that the extracted core text content that can be semantically modeled. Then, token-level word segmentation is performed. The cleaned text content is processed using a word segmentation algorithm based on rule-driven and language model combination, such as WordPiece + device dictionary recognition, to decompose compound word structures into token sequences with semantic meaning, while retaining key entity words and connecting verbs. All numerical tokens and their physical units are normalized to unify different expressions into a standard form. Then, entity location information is bound to ensure that entity values ​​and unit information can be distinguished in subsequent semantic modeling. Then, based on the equipment categories and semantic rules defined in the CIM model, each token is labeled with its semantic role (such as equipment name, attribute name, attribute value, connecting words, etc.), forming a semantically meaningful set. The token sequence structure of color tags; in the semantic encoding layer of the semantic fusion network, the preprocessed power text information of the distribution network to be constructed is subjected to multi-layer contextual semantic modeling (the input token sequence is converted into a fixed-dimensional word vector matrix, and positional encoding is introduced to preserve text order information. Then, the word vector matrix is ​​input into a multi-layer Transformer encoder structure. Each layer of the encoder includes a multi-head self-attention mechanism and a feedforward neural network module. The semantic relevance weight between different tokens is dynamically calculated through the self-attention mechanism to realize the modeling of device semantic fragments across sentence structures, nested expressions and long-distance dependencies. In each layer of the encoder structure, residual connections and layer normalization mechanisms are introduced to enhance the depth of semantic information transmission and improve the stability of network training. Finally, the high-dimensional semantic embedding vector set of each token after fusion in the context is output, which represents the semantic unit represented by the token, such as the device number, topological relationship, parameter value in the contextual semantic meaning of the overall distribution network, and is used to guide the triplet construction layer to complete semantic extraction and entity localization).In the triplet construction layer of the semantic fusion network, semantic role recognition processing is performed on the high-dimensional semantic embedding vector set of the distribution network to be constructed. (A sequence labeling mechanism is used to classify the semantic role of the embedding vector of each semantic token. Using the BIO labeling system, combined with a fully connected layer and a Softmax classifier, semantic labels are determined for each token, resulting in semantic role categories including but not limited to: device name, attribute name, attribute value, unit, and connection relation words. The classification process can be further combined with a conditional random field layer to enhance the recognition ability of semantic continuity. Through the above labeling operation, the semantic role recognition of each token in the text of the distribution network to be constructed is completed, generating a set of token sequences with semantic labels as the semantic input basis for subsequent triplet construction. Then, based on the completed semantic role labeling, the triplet construction operation is performed, based on context window matching.) By establishing rules and semantic positional dependencies, the attribution logic between devices and attributes, as well as the connection semantics between devices, is identified. This leads to the construction of preliminary candidate sets of attribute triples and relation triples. Specifically, if a token sequence is detected to simultaneously contain a device name, attribute name, and attribute value or unit, it is constructed as an attribute triple; if a connection relation word is detected between two device entities, it is constructed as a relation triple. To improve construction accuracy, semantic consistency verification can be performed on candidate triples based on similarity scoring or attention weighting mechanisms between semantic embeddings to remove redundant or conflicting triple structures. Finally, the constructed triples are clustered and integrated according to the unique identifier of the power equipment to obtain the initial triple structure set of each power equipment in the distribution network to be constructed (including the initial attribute triple set and the initial relation triple set).In the graph structure modeling layer of the semantic fusion network, the initial triplet set of each power device to be constructed in the distribution network undergoes graph semantic propagation processing (converting the initial triplet set of each power device into a graph structure, where the subject, predicate, and object in the triplet set are mapped to nodes and edges in the graph structure, respectively, with the subject and object as nodes in the graph and the predicate as directed edges connecting the subject and object, forming an ontology attribute relationship graph and a device connection relationship graph centered on the power device). Through the above mapping process, the device attribute triplets and device relationship triplets can be uniformly represented as a semantic graph structure, fully preserving the attribute features and topological connections of the power devices. The relationship is then established, and a graph neural network is used to model the constructed semantic graph structure. A semantic vector representation is initialized for each node in the graph. Combining its positional relationship within the graph with the semantic labels of its adjacent edges, the graph neural network stored in the database performs semantic feature aggregation and multi-hop dependency modeling on the nodes. During this process, the graph neural network integrates the semantic information of each node with the information of its upstream and downstream nodes through a layer-by-layer propagation mechanism, thereby enhancing the structure of the device's semantic representation. Through multiple rounds of information propagation, each node can perceive the semantic context within its graph structure, thus enhancing the global consistency and topology of the power equipment's semantic expression. Ultimately, the system outputs a set of structurally enhanced triplet structures for each power device. These triplet structures, while retaining the original attributes and connection semantics, incorporate topological context information from the graph structure, providing a set of triplets with more accurate semantic expression and a more rational structural organization for subsequent semantic knowledge graph construction. This yields a set of structurally enhanced triplet structures for each power device in the distribution network to be constructed. In the triplet output layer of the semantic fusion network, the set of structurally enhanced triplet structures for each power device in the distribution network to be constructed is filtered and categorized (a triplet filtering operation is performed, and for each triplet in the set of structurally enhanced triplet structures, its generation is combined with its...). The contextual information and attention weight distribution of the graph structure propagation process are used to score the semantic confidence of triples. Abnormal triples with semantic repetition, semantic conflict or scores below a preset threshold are removed. Only triples with complete semantics, reasonable structure and high semantic relevance are retained. Then, the retained high-quality triples are classified into attribute class and relation class according to the semantic type of their predicate part. If the predicate represents the internal parameters, operating characteristics or static configuration of a device, such as hasVoltage, hasLength, hasStatus, hasPhase, etc., then the triple is classified as a device attribute triple.If the predicate represents a structural relationship such as connection, affiliation, or control between the device and other devices, such as connectsTo, locatedIn, feedsInto, controlledBy, etc., then this triple is classified as a device relation triple. Then, based on the subject of each triple (i.e., the unique identifier of the power device), all attribute triples and relation triples belonging to the same device are aggregated separately to form a structured semantic set specific to that device. Finally, the aggregated triple sets for each power device are standardized to ensure that the representation of all triples follows a unified format (including standardized device numbers, unified naming of attribute fields, and normalized numerical units), resulting in the device attribute triples and device relation triples for each power device in the distribution network to be constructed.

[0037] Furthermore, the pre-training process of the semantic fusion network is as follows:

[0038] Obtain labeled text information, including a large amount of high-quality training corpus of semantic text of power equipment. The corpus should conform to the SG-CIM standard, and each piece of corpus should have a clear description of equipment attributes and connection relationships. The corresponding standard triple tag set should be manually or automatically labeled.

[0039] The labeled text information is divided into data segments to construct separate text training and text validation sets, ensuring that the data distribution is representative and structurally balanced. The text training set is used for parameter learning, and the text validation set is used for performance evaluation during the training process.

[0040] The semantic fusion network is initialized, which means initializing all model parameters of the semantic fusion network, including the output layer, semantic coding layer, graph neural network layer and output layer. The initialization method can be Xavier uniform initialization or BERT pre-trained vector transfer initialization strategy.

[0041] The training is conducted on a text training set, with a set number of training loops (e.g., 100). In each training loop, the following steps are performed sequentially: forward propagation (the current electricity semantic text is input into the semantic fusion network model, passing through the input layer, semantic encoding layer, triplet candidate construction layer, and graph structure modeling layer, ultimately generating a set of device triplets at the triplet output layer); loss calculation (the set of device triplets output by the model is precisely compared with the corresponding labeled triplets in the training set, and the semantic recognition bias is calculated based on the triplet-level cross-entropy loss function or label matching accuracy index); backpropagation (according to the gradient result of the loss function, the error information is backpropagated to each model layer to update the embedding parameters, Transformer encoder weights, and graph neural network parameters); and parameter optimization (optimization algorithms, such as Adam or AdamW, are used to perform iterative parameter updates to improve the model's prediction ability in the next training batch).

[0042] After each training cycle, an evaluation analysis is performed based on the text validation set. This involves fixing the currently trained model parameters and inputting the electrical semantic text from the validation set that was not used in the training into the semantic fusion network to perform a complete forward propagation process, resulting in predicted device attribute triples and device relationship triples. Subsequently, the prediction results are compared item by item with the manually labeled triples in the validation set, and performance metrics such as precision, recall, and F1 score at the triple level are calculated to evaluate the model's semantic recognition ability and structural reconstruction accuracy on unseen data.

[0043] If the validation set evaluation metrics do not show significant improvement over several consecutive rounds, or if the model performance has reached the preset performance threshold, the early stopping mechanism is triggered to prevent the model from overfitting on the training set and improve its generalization ability in actual distribution network semantic extraction tasks. The model is then saved as a trained semantic fusion network.

[0044] This implementation scheme introduces a multi-level semantic recognition path consisting of an input layer, an encoding layer, a triplet construction layer, a graph modeling layer, and an output layer. This allows for the automatic extraction and normalization of key semantic elements such as equipment names, attributes, units, and relational terms from complex power structured text, thus avoiding problems such as information fragmentation and ambiguous resolution in manual modeling. Furthermore, the encoding layer introduces a multi-layer Transformer structure and positional encoding to effectively integrate long-distance semantic dependencies, ensuring clear attribution logic in the triplet construction stage. This enables precise entity location and attribute pairing, laying an accurate foundation for downstream graph structure modeling. Secondly, the initial triples are projected into the graph structure, and through the graph neural network propagation mechanism, multi-hop dependency modeling of node semantic representations is achieved. This preserves the local expression of equipment attributes and integrates the semantic background of upstream and downstream power equipment, resulting in a semantically consistent and topologically reasonable structural expression. This improves semantic accuracy and network structure consistency. Finally, a rigorous pre-training process is introduced to enhance the model's transferability and stability in real-world scenarios, ensuring engineering-level reliability throughout the entire process of constructing the distribution network semantic knowledge graph.

[0045] Specifically, the steps for generating the power equipment topology diagram of the distribution network to be constructed are as follows: Read the equipment relationship triplet of each power device in the distribution network to be constructed and perform comprehensive analysis (for the predicate part of each equipment relationship triplet, classify its semantics based on a predefined semantic vocabulary, dividing the predicate into semantic categories such as electrical connection, structural affiliation, and control command, retaining only triplets belonging to the electrical connection category as candidate triplets; then, verify the validity of the subject and object devices in each triplet, determining whether the two devices exist in the same topology domain, such as the same substation or feeder; check whether it has complete equipment identification and Terminal information; exclude devices marked as failed or isolated; finally, retain only those triplets with electrical connection semantic category, complete equipment identification, consistent topology domain, and not marked as failed, and combine the subject and object devices to form a candidate connected power equipment set), obtaining several sets of candidate connected power equipment for the distribution network to be constructed; for each set of candidate connected power equipment for the distribution network to be constructed... To prepare for the topology connectivity analysis, several sets of actual connected power devices in the distribution network to be constructed are obtained, and a power device topology diagram of the distribution network to be constructed is generated. This involves extracting the power device identification information appearing in all actual connected power device pairs, removing duplicates, and establishing a set of nodes for the topology diagram, with each node corresponding to a power device entity. Next, based on the triplet of each set of device relationships, a set of edges for the topology diagram is established. Each edge represents a direct connection between two power devices, and the edge only contains the starting device, the ending device, the connection relationship type (such as connectTo, controlBy, etc.), and the connection confidence index generated by the topology analysis process as the edge weight. Subsequently, the set of nodes and the set of edges are combined to form a graph structure model, generating a power device topology diagram with power devices as nodes and connection relationships as edges, used to realistically reproduce the structural connection relationships between devices in the distribution network. Finally, the integrity of the graph structure is checked, and disconnected devices or structurally abnormal edges are removed. The final topology diagram is used as the basic structural output for subsequent graph construction and visualization processing.

[0046] like Figure 4As shown, the specific steps to obtain several groups of actual connected power equipment in the distribution network to be constructed are as follows: Obtain the connection index set for each group of candidate connected power equipment in the distribution network to be constructed, including the node consistency index (measuring the strength of electrical node connectivity between candidate connected equipment), the structural direction difference index (measuring the geometric rationality and physical feasibility of the connection direction of candidate connected equipment in spatial structure), and the semantic confidence index; perform a comprehensive analysis on the connection index set for each group of candidate connected power equipment in the distribution network to be constructed to obtain the topology connection confidence index for each group of candidate connected power equipment in the distribution network to be constructed; compare the topology connection confidence index of each group of candidate connected power equipment in the distribution network to be constructed with a preset topology connection confidence index threshold (i.e., each group of candidate connected power equipment with a topology connection confidence index higher than the preset topology connection confidence index threshold is marked as a group of actual connected power equipment, while each group of candidate connected power equipment with a topology connection confidence index lower than or equal to the preset topology connection confidence index threshold is not marked), thus obtaining several groups of actual connected power equipment in the distribution network to be constructed.

[0047] The specific steps for obtaining the node consistency index are as follows: Read the directly connected power devices (such as devices that have a direct connection relationship with power device A in the same group of candidate connected power devices) of each group of candidate connected power devices in the distribution network to be constructed, and form corresponding adjacent device sets (such as adjacent device set A and adjacent device set B, which correspond to the two power devices in the same group of candidate connected power devices respectively). Perform set intersection and union operations on adjacent device set A and adjacent device set B, calculate the degree of overlap between the two connected devices, that is, use the Jaccard similarity formula, and then use the sigmoid function to map the result to between 0 and 1, and use it as the node consistency index.

[0048] The specific steps for obtaining the structural orientation difference index are as follows: Read the spatial coordinate information of the two terminals of each candidate connected power equipment, such as the three-dimensional coordinates of terminal A and terminal B respectively. Based on the difference between the three-dimensional coordinates of terminal A and terminal B, construct the actual connection direction vector to represent the actual connection direction between the equipment. Then, extract the theoretical connection direction vectors of terminal A and terminal B. This vector can be obtained based on the equipment structural standard or the preset geometric connection direction. Calculate the cosine of the angle between the actual connection direction vector and the theoretical connection direction vectors of terminal A and terminal B respectively. Then, weight the calculation results and use the sigmoid function to map the result to between 0 and 1. The obtained result is the structural orientation difference index.

[0049] The specific steps for obtaining the semantic confidence index are as follows: read the semantic confidence of each triple in the triple output layer of the semantic fusion network, then use the sigmoid function to map the result to a range of 0-1, and use its semantic confidence as the semantic confidence index of the candidate connected power device.

[0050] The specific formula for calculating the topology connectivity confidence index of a group of candidate connected power devices in the distribution network to be constructed is as follows: ;in, Let be the topology connectivity confidence index for a set of candidate connected power devices in a power distribution network to be constructed. The node consistency index of a group of candidate connected power devices in a power distribution network to be constructed. The node consistency adjustment coefficient is stored in the database. The structural orientation difference index of a group of candidate connected power equipment in the power distribution network to be constructed. These are the structural orientation adjustment coefficients stored in the database. The collaborative penalty adjustment coefficient is stored in the database. The semantic confidence index of a group of candidate connected power devices for a power distribution network to be constructed. This is the semantic suppression coefficient stored in the database, and in this implementation example, it takes the value of 0.650. These are the semantic adjustment coefficients stored in the database.

[0051] It needs to be explained that the specific form of the tanh function is as follows: ,in, It is a natural constant, and in this example it can be taken as 2.71, with a domain of (−∞, +∞) and a range of (−1, +1).

[0052] , , , The following steps can be taken to obtain the following: Based on historical data, combined with the node consistency index, structural direction difference index, and semantic confidence index, statistical regression analysis is performed to quantify the specific impact of each factor on the topology connectivity confidence index, thereby fitting the initial weight values. Secondly, sensitivity analysis is used to adjust the value range of each coefficient and observe its impact on the topology connectivity confidence assessment results to ensure the stability and rationality of the model. Based on the structural characteristics and actual situation, the initially fitted coefficients are corrected and optimized, and finally, the coefficient values ​​applicable to the structure are determined.

[0053] The following is a specific implementation example for calculating the topology connectivity confidence index of a group of candidate connected power devices in a distribution network to be constructed. The available data includes the node consistency index, structural orientation difference index, and semantic confidence index of 5 groups of candidate connected power devices (randomly selected) in the distribution network to be constructed, as detailed in Table 1 and... Figure 5 As shown:

[0054] Table 1. Candidate Sequence Connection Index Set for Distribution Networks to be Constructed

[0055]

[0056] The node consistency adjustment coefficient stored in the database Approximately 0.249;

[0057] The structural orientation adjustment coefficient stored in the database Approximately 0.316;

[0058] Cooperative penalty adjustment coefficients stored in the database Approximately 0.367;

[0059] Semantic suppression coefficients stored in the database The value is: 0.650;

[0060] Semantic adjustment coefficients stored in the database Approximately 0.438;

[0061] Substituting the data from Table 1 and the coefficients mentioned above into the specific formula for calculating the topology connectivity confidence index of a group of candidate connected power devices in the distribution network to be constructed, we obtain:

[0062] The topology connectivity confidence index of the first group of candidate connected power equipment in the distribution network to be constructed (i.e., candidate group 1) = ((0.783) 0.249 × (1-0.176) 0.316 ) / (1+|0.783-(1-0.176)| 0.367 ))×(1+(tanh(0.742-0.650)) 0.438 ≈0.914;

[0063] The topology connectivity confidence index of the second group of candidate connected power equipment (i.e., candidate group 2) of the distribution network to be constructed is = ((0.728) 0.249 ×(1-0.139) 0.316 ) / (1+|0.728-(1-0.139)| 0.367))×(1+(tanh(0.794-0.650)) 0.438 ≈0.850;

[0064] The topology connectivity confidence index of the third group of candidate connected power equipment (i.e., candidate group 3) of the distribution network to be constructed = ((0.823) 0.249 ×(1-0.218) 0.316 ) / (1+|0.823-(1-0.218)| 0.367 ))×(1+(tanh(0.697-0.650)) 0.438 ≈0.849;

[0065] The topology connectivity confidence index of the fourth group of candidate connected power equipment (i.e., candidate group 4) of the distribution network to be constructed is = ((0.684) 0.249 × (1-0.145) 0.316 ) / (1+|0.684-(1-0.145)| 0.367 ))×(1+(tanh(0.711-0.650)) 0.438 ≈0.730;

[0066] The topology connectivity confidence index of the fifth group of candidate connected power equipment (i.e., candidate group 5) in the distribution network to be constructed = ((0.751) 0.249 × (1-0.105) 0.316 ) / (1+|0.751-(1-0.105)| 0.367 ))×(1+(tanh(0.816-0.650)) 0.438 ≈0.876.

[0067] This implementation scheme introduces a topology connection confidence index construction mechanism, thereby achieving accurate identification of distribution network equipment connection relationships and generating a highly reliable topology diagram. This significantly improves the accuracy and security of distribution network topology modeling. For example, semantic classification and validity verification are performed on equipment relationship triples, retaining only candidate connection pairs with electrical connection semantics, topological domain consistency, and valid state. This ensures that the connection relationships involved in the modeling have physical rationality and structural legality. Subsequently, by constructing a three-dimensional multi-dimensional index—node consistency index, structural direction difference index, and semantic confidence index—and comprehensively considering the overlap of adjacent network structures, the spatial direction rationality of equipment connections, and the accuracy of semantic extraction during index fusion, it ensures that the selected actual connected power equipment pairs have highly reliable structural and semantic consistency. As a result, the final generated power equipment topology diagram not only excludes isolated equipment and pseudo-connections but also has complete edge weight expression and topological completeness. This provides a clear, reliable, and structurally organic basic framework for subsequent graph modeling and visualization platforms, and improves the accuracy, robustness, and engineering applicability of the distribution network semantic modeling system.

[0068] Specifically, the steps for generating the semantic knowledge graph of the distribution network to be constructed are as follows: Read the equipment attribute triples of each power device in the distribution network to be constructed, and perform matching processing in conjunction with the power device topology graph (i.e., traverse all graph nodes in the topology graph, extract their node identifier information as a matching reference; secondly, for the subject field in each attribute triple, use it as the unique identifier information of the device, and compare and search one by one, using a complete equivalence matching method to determine whether a corresponding node exists in the node set of the topology graph; if a completely matching node exists, then the attribute triple is...). The predicate and object of a ternary triad are attached to the node as key-value pairs, forming the node's attribute mounting structure. The subject field undergoes format standardization processing, such as case uniformity, prefix normalization, and underscore / hyphen replacement. Unmatched subject fields are recorded as attribute islands for subsequent consistency checks. After matching, the node attribute set is output, where each graph node carries zero or more attribute key-value pairs, forming the semantic basis for the device in the semantic graph. This yields the node attribute set of the power equipment topology graph for the distribution network to be constructed. The node attribute set of the power equipment topology diagram of the distribution network is semantically integrated (i.e., multiple attribute triples are combined into a structured attribute dictionary, attribute names are semantically normalized (e.g., rated voltage and voltage level are unified as voltageLevel), and attribute value units are standardized and converted to form a consistent and semantically standardized node attribute expression form, ensuring the uniformity and parsability of subsequent graph visualization, querying, and semantic reasoning). This yields the graph node set of the power equipment topology diagram of the distribution network to be constructed, which is then comprehensively analyzed (i.e., the graph node set with completed attribute mounting and semantic integration is fused and modeled with the edge set in the original topology diagram to form a graph structure with semantic capabilities, including keeping the original graph nodes and edges unchanged, adding attribute information to nodes to construct semantic nodes, and if there is cross-node association information in the attributes, such as the relationship of attachment to equipment or equipment status, it can be expanded into semantic edges to further enhance the graph association expression. The final generated graph has a complete ternary structure expression capability of entity nodes, attribute information, and connection relationships), generating a semantic knowledge graph of the distribution network to be constructed.

[0069] This implementation scheme constructs a distribution network semantic graph structure that integrates connection logic and device semantics, thereby achieving a key leap from a topology graph to a complete semantic knowledge graph. This enhances the completeness of power equipment information expression and structural parsing capabilities. Secondly, it accurately matches device attribute triples with topology graph nodes and performs unified format standardization processing, achieving deep binding between device entities and semantic information. This prevents attribute data from being detached from the structure, ensuring that each node has a clear semantic carrying capacity. Finally, it constructs a structured attribute dictionary and integrates it into graph nodes to form a semantic node set, supporting semantic visualization, conditional querying, and rule reasoning of the graph. The semantic nodes are integrated with the original edge structure for modeling, thus maintaining the physical connection logic of the original topology graph. At the same time, semantic extension edges enhance the expressive ability of potential semantic connections between devices, enabling the generated distribution network semantic knowledge graph to possess the triple characteristics of structural reproducibility, semantic parsability, and data traceability. This provides a solid and accurate semantic foundation and data skeleton for subsequent graph analysis, anomaly detection, and decision support.

[0070] Specifically, the steps to obtain the semantic structure coordination index of the semantic knowledge graph of the distribution network to be constructed are as follows: A comprehensive analysis is performed on the semantic knowledge graph of the distribution network to be constructed to obtain a set of graph quality evaluation indices, including a structural connectivity integrity index (used to measure the topological integrity in the distribution network semantic knowledge graph), a semantic consistency index (measuring the standardization and semantic uniformity of attribute descriptions for similar types of power equipment), and a topological semantic mapping index (measuring the degree of reasonable matching between connection structures and equipment attribute semantics in the graph); and a comprehensive analysis is performed on the set of graph quality evaluation indices of the semantic knowledge graph of the distribution network to be constructed to obtain the semantic structure coordination index of the semantic knowledge graph of the distribution network to be constructed.

[0071] The structural connectivity integrity index can be obtained through the following steps: traverse all nodes in the semantic knowledge graph of the distribution network to be constructed, count the number of times each node appears as the subject (head entity) or object (tail entity) of a triple. If a node does not appear as either the subject or object in any triple, it is marked as an isolated node, and the total number of isolated nodes is recorded as the isolated value. Then, traverse all connection triples. If any device (subject or object) in a triple does not appear in the node set, the triple is marked as a dangling edge, and the total number of dangling edges is recorded as the dangling value. Count the total number of graph device nodes as the total number of nodes, and the total number of connection relationships as the total number of edges. Perform a comprehensive analysis, i.e., 1 - ((isolated value + dangling value) / (total number of nodes + total number of edges)). The result is the structural connectivity integrity index.

[0072] The semantic consistency index can be obtained through the following steps: Classify the devices in the semantic knowledge graph of the distribution network to be constructed according to device type (e.g., circuit breakers, cables, distribution transformers, etc.). For each type of device, extract its corresponding attribute triplet set in the format (device ID-attribute name-attribute value). Perform attribute name conflict detection (check for multiple names of the attribute name, such as rated voltage, U-rated, Un, etc.), unit misuse detection (check for mixed units in the attribute value, such as 10kV, 10000V, 10kV, etc.), and value inconsistency detection (check for negative values, null values, extreme values, etc., such as power = -20kW, etc.). Count the total number of conflicting triplets as the conflict quantity value, and count the total number of attribute triplets for all devices in that type, recording it as the total quantity value for that type of device. Perform a ratio analysis, i.e., 1 - (conflict quantity value / total quantity value). Based on the ratio analysis results, perform weighted processing, and the result is the semantic consistency index.

[0073] The topology semantic mapping index can be obtained through the following steps: Extract the device relationship triples (i.e., connection triples, such as device A - connection relationship - device B) from the semantic knowledge graph of the distribution network to be constructed, and extract the attribute context, such as extracting the attribute triples of device A and device B respectively, and obtaining the voltage level (e.g., 10kV, 35kV), the voltage layer (e.g., medium voltage, low voltage), and the purpose of the equipment (e.g., feeder, distribution transformer). Then, based on the preset topology semantic rules (e.g., feeder equipment of the same voltage level is allowed to be directly connected, upstream voltage of transformer > downstream voltage, low voltage load is not allowed to be directly connected to the main transformer, etc.), perform semantic matching judgment. For each connection relationship, judge whether it meets the above rules according to the attributes of its two ends. If the rules are violated, it is marked as a semantic conflict connection, and the total number of semantic conflict connections is counted. Then, the total number of connection triples is counted and a ratio analysis is performed, i.e., 1 - (semantic conflict connection) / total number of connection triples. The result is the topology semantic mapping index.

[0074] The specific formula for calculating the semantic structure coordination index of the semantic knowledge graph of the distribution network to be constructed is as follows: ;in, This is the semantic structure coordination index for the semantic knowledge graph of the distribution network to be constructed. These are the superimposed adjustment coefficients stored in the database. The structural connectivity completeness index is used to construct the semantic knowledge graph of the power distribution network. The structural connectivity integrity adjustment coefficient is stored in the database. This is the semantic consistency index for the semantic knowledge graph of the distribution network to be constructed. These are the semantic consistency adjustment coefficients stored in the database. This is the topological semantic mapping index for the semantic knowledge graph of the distribution network to be constructed. These are the semantic mapping adjustment coefficients stored in the database. This refers to the coordination baseline adjustment coefficient stored in the database.

[0075] It needs to be explained that in the formula This item is used to adjust the tolerance for local quality deviations (i.e., the suppressive strength of the structural connectivity integrity index, semantic consistency index, and topological semantic mapping index on the overall evaluation results, so as to achieve control over the tolerance of the minimum quality indicators), thereby balancing the robustness and sensitivity of the evaluation.

[0076] , , , , The following steps can be used to obtain the initial influence weights of each variable (structural connectivity integrity index, semantic consistency index, and topological semantic mapping index) on the semantic structure coordination index based on historical data. Then, the range of coefficient values ​​is adjusted using sensitivity analysis to evaluate the stability and applicability of these parameters to the formula output. Next, the weights are further fitted through model optimization (such as machine learning algorithms or multi-objective optimization) to ensure that the formula can accurately reflect the actual coordination state of the semantic structure.

[0077] This implementation plan introduces three quality dimensions—structural connectivity integrity, semantic consistency, and topological semantic mapping—to construct a multi-index evaluation system for quantitatively assessing the coordination status of the distribution network semantic knowledge graph. This enables comprehensive control over the overall structure and semantic fusion quality of the graph. For example, the structural connectivity integrity index can identify potential isolated nodes and dangling edges in the semantic graph to ensure the closure of the topology and network availability. The semantic consistency index performs conflict detection and unified standardization processing on the attribute descriptions of similar power equipment, thereby significantly improving the standardization, uniformity, and reasonability of the graph at the semantic level. Then, the topological semantic mapping index verifies the rule constraints between connection relationships and attribute contexts, thereby ensuring that the connection logic of the distribution network is reasonable and interpretable at the semantic level. Finally, by introducing a lower limit-dominant mechanism and parameter weight adjustment terms through a comprehensive formula, the tolerance level of the overall evaluation result is dynamically adjusted according to the minimum index, making the final semantic structure coordination index both sensitive and robust. This provides accurate and controllable quantitative basis for dynamic verification, repair suggestions, and visualization mode switching of the graph.

[0078] Please see Figure 6This invention provides a technical solution: a system for constructing a distribution network visualization platform based on SG-CIM, comprising: a data acquisition and parsing module, used to acquire power text information of the distribution network to be constructed and input it into a pre-trained semantic recognition model for parsing and analysis to obtain a structured semantic dataset of each power device in the distribution network to be constructed, including device attribute triples and device relationship triples; a device topology reconstruction module, used to perform topology reconstruction processing on the device relationship triples of each power device in the distribution network to be constructed to generate a power device topology diagram of the distribution network to be constructed; a power grid graph modeling module, used to perform graph modeling processing on the power device topology diagram based on the device attribute triples of each power device in the distribution network to be constructed to generate a semantic knowledge graph of the distribution network to be constructed; and a distribution visualization platform construction module, used to perform quality verification analysis on the semantic knowledge graph of the distribution network to be constructed to obtain a semantic structure coordination index of the semantic knowledge graph of the distribution network to be constructed, and construct a distribution network visualization platform based on the semantic structure coordination index.

[0079] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0080] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A construction method of a power distribution network visualization platform based on SG-CIM, characterized in that, The method comprises the following steps: Obtain the power text information of the to-be-constructed power distribution network, and input the power text information into a pre-trained semantic recognition model for analysis to obtain a structured semantic data set of each power device of the to-be-constructed power distribution network, including device attribute triples and device relationship triples; Perform topological reconstruction processing on the device relationship triples of each power device of the to-be-constructed power distribution network to generate a power device topological structure diagram of the to-be-constructed power distribution network; Perform graph modeling processing on the power device topological structure diagram based on the device attribute triples of each power device of the to-be-constructed power distribution network to generate a semantic knowledge graph of the to-be-constructed power distribution network; Perform quality check analysis on the semantic knowledge graph of the to-be-constructed power distribution network to obtain a semantic structure coordination index of the semantic knowledge graph of the to-be-constructed power distribution network, and construct a power distribution network visualization platform based on the semantic structure coordination index; The specific steps of obtaining the semantic structure coordination index of the semantic knowledge graph of the to-be-constructed power distribution network are as follows: Perform comprehensive analysis on the semantic knowledge graph of the to-be-constructed power distribution network to obtain a graph quality evaluation index set of the semantic knowledge graph of the to-be-constructed power distribution network, including a structure connectivity completeness index, a semantic consistency index, and a topological semantic mapping index; Perform comprehensive analysis on the graph quality evaluation index set of the semantic knowledge graph of the to-be-constructed power distribution network to obtain a semantic structure coordination index of the semantic knowledge graph of the to-be-constructed power distribution network, and the specific formula is as follows: ； wherein, , , , are, in sequence, a semantic structure coordination index, a structure connectivity completeness index, a semantic consistency index, and a topological semantic mapping index of a semantic knowledge graph of the power distribution network to be constructed, , , , , are, in sequence, a superposition adjustment coefficient, a structure connectivity completeness adjustment coefficient, a semantic consistency adjustment coefficient, a semantic mapping adjustment coefficient, and a coordination bottom line adjustment coefficient stored in the database.

2. The method for constructing the SG-CIM-based power distribution network visualization platform according to claim 1, characterized in that, The semantic recognition model is specifically a semantic fusion network, and the semantic fusion network comprises an input layer, a semantic encoding layer, a triple construction layer, a graph structure modeling layer, and a triple output layer.

3. The method for constructing the SG-CIM-based power distribution network visualization platform according to claim 2, characterized in that, The specific steps of obtaining the structured semantic data set of each power device of the to-be-constructed power distribution network are as follows: In the input layer of the semantic fusion network, receive the power text information of each power device of the to-be-constructed power distribution network and perform preprocessing; In the semantic encoding layer of the semantic fusion network, perform multi-layer context semantic modeling processing on the preprocessed power text information of the to-be-constructed power distribution network to obtain a high-dimensional semantic embedding vector set of the to-be-constructed power distribution network; In the triple construction layer of the semantic fusion network, perform semantic role recognition processing on the high-dimensional semantic embedding vector set of the to-be-constructed power distribution network to obtain an initial triple structure set of each power device of the to-be-constructed power distribution network; In the graph structure modeling layer of the semantic fusion network, perform graph semantic propagation processing on the initial triple structure set of each power device of the to-be-constructed power distribution network to obtain a structure-enhanced triple set of each power device of the to-be-constructed power distribution network; In the triple output layer of the semantic fusion network, perform screening and classification processing on the structure-enhanced triple set of each power device of the to-be-constructed power distribution network to obtain device attribute triples and device relationship triples of each power device of the to-be-constructed power distribution network.

4. The method for constructing the SG-CIM based power distribution network visualization platform according to claim 1, characterized in that, The specific steps of generating the power device topological structure diagram of the to-be-constructed power distribution network are as follows: Read the device relationship triples of each power device of the to-be-constructed power distribution network and perform comprehensive analysis to obtain a plurality of groups of candidate connected power devices of the to-be-constructed power distribution network; The topology connection analysis is performed on each group of candidate connected power equipment of the to-be-constructed power distribution network, to obtain a plurality of groups of actually connected power equipment of the to-be-constructed power distribution network, and to generate a power equipment topology structure diagram of the to-be-constructed power distribution network.

5. The method for constructing the SG-CIM-based power distribution network visualization platform according to claim 4, characterized in that, The specific steps of obtaining the plurality of groups of actually connected power equipment of the to-be-constructed power distribution network are as follows: The connection index set of each group of candidate connected power equipment of the to-be-constructed power distribution network is obtained, including node consistency index, structure direction difference index, and semantic confidence index; The topology connection confidence index of each group of candidate connected power equipment of the to-be-constructed power distribution network is obtained by comprehensively analyzing the connection index set of each group of candidate connected power equipment of the to-be-constructed power distribution network; The topology connection confidence index of each group of candidate connected power equipment of the to-be-constructed power distribution network is respectively compared with the preset topology connection confidence index threshold, to obtain a plurality of groups of actually connected power equipment of the to-be-constructed power distribution network.

6. The method for constructing the SG-CIM-based power distribution network visualization platform according to claim 5, characterized in that, The specific formula for calculating the topology connection confidence index of a group of candidate connected power equipment of the to-be-constructed power distribution network is as follows: ； wherein, , , , are, in sequence, a topology connection confidence index, a node consistency index, a structure direction difference index, a semantic confidence index of a certain group of candidate connected power equipment of the power distribution network to be constructed, , , , are, in sequence, a node consistency adjustment coefficient, a structure direction adjustment coefficient, a coordination penalty adjustment coefficient, a semantic adjustment coefficient stored in the database, is a semantic suppression coefficient stored in the database.

7. The method for constructing the SG-CIM based power distribution network visualization platform according to claim 1, wherein, The specific steps of generating the semantic knowledge graph of the to-be-constructed power distribution network are as follows: The device attribute triple of each power equipment of the to-be-constructed power distribution network is read, and is matched and processed in combination with the power equipment topology structure diagram, to obtain a node attribute set of the power equipment topology structure diagram of the to-be-constructed power distribution network; The node attribute set of the power equipment topology structure diagram of the to-be-constructed power distribution network is subjected to semantic integration processing, to obtain a graph node set of the power equipment topology structure diagram of the to-be-constructed power distribution network, and is subjected to comprehensive analysis, to generate the semantic knowledge graph of the to-be-constructed power distribution network.

8. A system for constructing a power distribution network visualization platform based on SG-CIM, applying the method for constructing a power distribution network visualization platform based on SG-CIM according to any one of claims 1-7, characterized in that, It comprises: A data acquisition and analysis module is configured to acquire power text information of the to-be-constructed power distribution network, and input the power text information into a pre-trained semantic recognition model for analysis, to obtain a structured semantic data set of each power equipment of the to-be-constructed power distribution network, including device attribute triple and device relationship triple; A device topology reconstruction module is configured to perform topology reconstruction processing on the device relationship triple of each power equipment of the to-be-constructed power distribution network, to generate a power equipment topology structure diagram of the to-be-constructed power distribution network; A power grid graph modeling module is configured to perform graph modeling processing on the power equipment topology structure diagram based on the device attribute triple of each power equipment of the to-be-constructed power distribution network, to generate a semantic knowledge graph of the to-be-constructed power distribution network; A power distribution visual platform construction module is configured to perform quality verification analysis on the semantic knowledge graph of the to-be-constructed power distribution network, to obtain a semantic structure coordination index of the semantic knowledge graph of the to-be-constructed power distribution network, and to construct a power distribution visual platform based on the semantic structure coordination index.

Images