Modeling method and device for power transformation equipment standard, electronic equipment and storage medium

Abstract

The embodiment of the invention provides a modeling method and device for a power transformation equipment standard, electronic equipment and a storage medium. The method comprises the following steps: extracting terminologies and context features related to a power transformation equipment standard from an original document by adopting a natural language processing model, generating entities according to the terminologies and the context features and determining a relationship between the entities, and then calculating similarities of the entities at different granularities and different dimensions, the knowledge graph can be constructed according to the entities, the relationship between the entities and the similarity. Through a natural language processing model, an unstructured standard text can be automatically converted into structured information which can be understood by a machine, in combination with multi-granularity and multi-dimensional similarity calculation, the structured degree and semantic coherence of the standard documents can be remarkably improved, potential correlation among the standard documents can be mined, and the method is suitable for popularization and application. The integrity and connectivity of the knowledge graph are improved, efficient and accurate modeling of the power transformation equipment standard is realized, and the management difficulty of the power transformation equipment standard is reduced.

Description

Technical Field

[0001] This application relates to the field of power data processing, and in particular to a modeling method, apparatus, electronic device and storage medium for a standard of power equipment. Background Technology

[0002] In the power industry, substation equipment is a core facility ensuring the safe and stable operation of the power grid. Its standardized management is crucial for equipment selection, operation and maintenance, and fault diagnosis. Substation equipment standards cover multiple levels, including industry standards and enterprise standards. The number of documents is large, the content is complex, and the structure is loose, making unified management and application difficult.

[0003] Traditional standards management methods rely primarily on keyword matching and human experience, requiring manual organization of equipment standards, industry clauses, and other content from numerous standard documents to create models and build a unified standards management system. This approach is inefficient, and as the data scale increases, manual modeling can easily overlook useful information such as implicit relationships between standards. Summary of the Invention

[0004] This application provides a modeling method, apparatus, electronic device, and storage medium for power equipment standards, in order to solve the problem of low efficiency in the management and control of power customer service indicators.

[0005] In a first aspect, embodiments of this application provide a modeling method for a standard power equipment system, including:

[0006] A natural language processing model was used to extract technical terms and contextual features related to power equipment standards from the original document.

[0007] Based on technical terms and contextual features, generate entities and determine the relationships between entities;

[0008] Calculate the first similarity of different entities at multiple granularities, and calculate the second similarity of different entities at multiple dimensions;

[0009] A knowledge graph is constructed based on entities, relationships, first similarity, and second similarity as the modeling result.

[0010] In some possible implementations, the first similarity between different entities at multiple granularities and the second similarity between different entities at multiple dimensions are calculated, including:

[0011] A multi-granularity similarity calculation framework is constructed to calculate the semantic similarity between entities at the term level, clause level, and document level, respectively.

[0012] Design a multi-dimensional similarity representation system to calculate the similarity between entities from the functional, structural, attribute, and application dimensions;

[0013] The weighting coefficients of similarity in each dimension are dynamically adjusted based on the entity's professional attributes and context.

[0014] In some possible implementations, a multi-granularity similarity calculation framework is constructed to calculate semantic similarity at the term, clause, and document levels, respectively, including:

[0015] In the term-level similarity calculation, a deep learning-based semantic vector representation method is adopted to map each term into a semantic vector in a high-dimensional semantic space, and the cosine similarity of the semantic vectors is calculated.

[0016] In clause-level similarity calculation, clauses are treated as a set of terms, and a weighted matching method is used to calculate the similarity between clauses.

[0017] In document-level similarity calculation, clauses in two different original documents are matched into clause pairs. Weight coefficients are determined based on the position of the clauses in the original documents. The similarity of all clause pairs in the two original documents is then weighted and summed according to the weight coefficients to obtain the document-level similarity.

[0018] In some possible implementations, after constructing the knowledge graph, the following are also included:

[0019] Obtain application requirements for power equipment standards, and perform any of the following target operations using a knowledge graph based on those requirements:

[0020] For standard retrieval needs, entities with similarity greater than a preset threshold are retrieved based on the knowledge graph, and similarity explanation information is provided, including matching points, similarity scores, and matching criteria.

[0021] For compliance inspection needs, the parameters of the equipment to be inspected are mapped to entities in the knowledge graph, and the matching criteria are determined based on the relationships between the entities. The compliance inspection results are then output based on the comparison between the matching criteria and the equipment parameters.

[0022] For discrepancy analysis, the entities corresponding to different versions of power equipment standards in the knowledge graph are identified, and the attributes and relationships of the identified entities are compared to generate discrepancy analysis results.

[0023] In some possible implementations, after constructing the knowledge graph, the following are also included:

[0024] Retrieve new documents containing standards for power equipment;

[0025] Calculate the similarity between the new document and the original document;

[0026] If the similarity is greater than a preset threshold, the differences between the new document and the original document are extracted, and the knowledge graph is updated based on the differences.

[0027] In some possible implementations, the knowledge graph is updated based on the differences in content, including:

[0028] Transform the discrepancies into new entities and relationships in the knowledge graph;

[0029] Based on the mechanism of influence propagation, the impact of newly added entities and relationships after transformation on the knowledge graph is evaluated, and the knowledge graph is updated according to the newly added entities and relationships whose impact exceeds a preset threshold.

[0030] In some possible implementations, after constructing the knowledge graph, the following are also included:

[0031] Obtain operating data of power equipment;

[0032] Establish a mapping relationship between operational data and standard knowledge in the knowledge graph to form a multimodal fusion knowledge graph;

[0033] Based on graph-based reasoning rules, cross-modal knowledge reasoning is performed on a multimodal fusion knowledge graph, and the reasoning results provide standard basis and / or decision suggestions for the operation status of substation equipment.

[0034] Secondly, embodiments of this application provide a modeling apparatus for a standard power equipment system, comprising:

[0035] The extraction module is used to extract technical terms and contextual features related to power equipment standards from the original document using a natural language processing model.

[0036] The entity generation module is used to generate entities and determine the relationships between entities based on technical terms and contextual features.

[0037] The calculation module is used to calculate the first similarity of different entities at multiple granularities, and to calculate the second similarity of different entities at multiple dimensions.

[0038] The knowledge graph construction module is used to construct a knowledge graph as the modeling result based on entities, relations, first similarity, and second similarity.

[0039] Thirdly, embodiments of this application provide an electronic device, including: a memory and a processor;

[0040] The memory stores instructions that the computer executes;

[0041] The processor executes computer execution instructions stored in memory, causing the processor to perform the first aspect and / or various possible implementations of the first aspect as described above.

[0042] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the first aspect and / or various possible implementations of the first aspect.

[0043] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the first aspect and / or various possible implementations of the first aspect.

[0044] The modeling method, apparatus, electronic device, and storage medium for substation equipment standards provided in this application employ a natural language processing model to extract relevant technical terms and contextual features from original documents. Entities are generated based on these technical terms and contextual features, and relationships between entities are determined. Then, the similarity of entities at different granularities and dimensions is calculated. A knowledge graph can be constructed based on entities, relationships between entities, and similarity. The natural language processing model can automatically transform unstructured standard text into machine-understandable structured information. Combined with multi-granularity and multi-dimensional similarity calculations, the structure and semantic coherence of standard documents can be significantly improved. Furthermore, potential connections between standard documents can be mined, enhancing the completeness and connectivity of the knowledge graph. This enables efficient and accurate modeling of substation equipment standards, helping to reduce the management difficulty of substation equipment standards. Attached Figure Description

[0045] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0046] Figure 1 A flowchart illustrating a modeling method for a standard power equipment provided in this application embodiment;

[0047] Figure 2 A schematic diagram of an architecture for semantic similarity calculation provided in an embodiment of this application;

[0048] Figure 3 A schematic diagram illustrating a knowledge graph update mechanism provided in an embodiment of this application;

[0049] Figure 4 A schematic diagram of an inference process that integrates multi-source data is provided in an embodiment of this application;

[0050] Figure 5 A schematic diagram illustrating the application of a modeling system for a standard power equipment provided in this application embodiment;

[0051] Figure 6 A schematic diagram of the structure of a modeling device for a standard power equipment provided in this application embodiment;

[0052] Figure 7 This is a schematic diagram of the structure of an electronic device provided in this application.

[0053] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0054] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0055] Standards for power equipment are crucial technical documents in the power industry, playing a vital role in ensuring the safe and stable operation of the power grid. However, the management of power equipment standards currently faces the following challenges: On the one hand, the number of standards is vast and continues to grow, with complex hierarchical and referencing relationships existing between standards at different levels (enterprise standards, industry standards, and higher-level regional standards); on the other hand, standard texts exist in natural language, lacking structured and semantic representation, resulting in low standard retrieval efficiency and difficulty in discovering implicit connections between standards.

[0056] Traditional standards management methods primarily rely on keyword matching and human experience, which are ill-suited to the complexity and dynamism of substation equipment standards management. In recent years, with the development of artificial intelligence technology, deep learning technology can be used to replace manual labor in specific tasks, but it still has the following shortcomings: first, it lacks semantic understanding capabilities for specialized terminology in the substation equipment field; second, it struggles to automatically discover implicit relationships between standards. This means that modeling and management of substation equipment standards still cannot escape reliance on human intervention. However, manual methods are prone to errors and omissions when dealing with a vast number of standard documents, making it difficult to fully uncover effective information such as implicit relationships between standards.

[0057] Based on this, this application proposes a technical concept that, based on deep learning and knowledge graph technology, constructs a multi-granularity, multi-dimensional semantic similarity calculation framework. This framework transforms technical terms and other content in standard documents into entities and relationships in a knowledge graph, and calculates multi-granularity and multi-dimensional similarity to represent implicit relationships between standards. By constructing a knowledge graph, intelligent management of power equipment standards can be achieved. This technical concept extracts technical terms and contextual features from standard documents using a natural language processing model, transforming unstructured standard text into machine-understandable structured information. Combining multi-granularity (terminology level, clause level, document level) and multi-dimensional (function, structure, attribute, application) similarity calculation methods, a semantic association network is constructed. This network can uncover potential relationships between standard documents, address the fragmentation problem of standard knowledge, and improve the completeness and connectivity of the knowledge graph.

[0058] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0059] Figure 1 A flowchart illustrating a standard modeling method for power equipment provided in this application is shown below. Figure 1 As shown, the method includes:

[0060] Step S101: Use a natural language processing model to extract professional terms and contextual features related to power equipment standards from the original document.

[0061] Natural language processing models can refer to deep learning-based text analysis models used to extract technical terms and contextual features from the natural language in input documents. Technical terms can refer to technical keywords in power equipment standards (such as "short-circuit impedance"). Contextual features can refer to semantic information such as adjacent words and syntactic structure of the technical terms within the text.

[0062] For example, original documents related to power equipment standards (such as standard documents developed by enterprises) can be obtained, and the original documents can be input into a trained natural language processing model to extract professional terms and contextual features from the documents.

[0063] The natural language processing model's processing steps may include: segmenting the standard document for power equipment, dividing the document into chapters, clauses, and rules according to the hierarchical structure of the standard document; using a professional domain dictionary and context rules to identify professional terms for power equipment from the segmented document; and for each identified professional term, using a sliding window technique to extract the context of the five words before and after it.

[0064] Step S102: Generate entities and determine the relationships between entities based on technical terms and contextual features.

[0065] For example, for each technical term, entity relationships can be obtained in its context to construct preliminary semantic connections. The acquisition of entity relationships adopts a method based on dependency parsing and semantic role labeling, which can identify relationship types between terms such as inclusion, belonging, reference, and testing.

[0066] Step S103: Calculate the first similarity of different entities at multiple granularities, and calculate the second similarity of different entities at multiple dimensions.

[0067] Multiple granularities can include different levels such as terminology, clause, and document. Multiple dimensions can include functional, structural, attribute, and application dimensions.

[0068] The terminology level can refer to the technical terms extracted in step S101. The clause level can refer to a sentence containing one or more technical terms, or to a set of terms generated from multiple technical terms. The document level can refer to the original document containing technical terms and clauses.

[0069] For example, a multi-granularity similarity calculation framework can be constructed, which uses semantic vector representation and other methods to map entities representing professional terms into vectors in a high-dimensional space, calculates vector similarity as the similarity between different professional terms, and calculates the similarity at the clause level and document level by weighted combination based on the structural relationship between terms, clauses and documents, thereby achieving accurate calculation of similarity between standard documents and automatic discovery of implicit associations.

[0070] In addition to multi-granularity similarity calculation frameworks, multi-dimensional similarity relationship representation systems can be constructed to evaluate the degree of similarity between entities from functional, structural, attribute, and application dimensions. This multi-dimensional representation can more comprehensively capture the similarity relationships between entities and support more accurate similarity assessment.

[0071] Step S104: Construct a knowledge graph as the modeling result based on entities, relationships, first similarity, and second similarity.

[0072] In this embodiment of the application, a knowledge graph can be constructed based on a semantic network with a graph structure, using the above-mentioned entities and their relationships. During the construction process, the calculated first similarity and second similarity can be associated with the corresponding entities, so that the knowledge graph not only contains the basic information and semantic relationships of professional terms, but also contains the implicit similarity relationships between terms.

[0073] For example, a two-layer entity hierarchy can be constructed. The first layer is the standard system entity, which includes three parts: power equipment, standard system, and version. The second layer is the standard entity, which includes technical standards, industry standards, and enterprise management standards. Then, a relationship representation mechanism between entities can be established, where the relationship between the standard system entity and the standard layer is represented as a standard, the relationship between the standard system entity and the version layer is represented as a reference, and the relationship between the referencing standard and the standard layer is represented as being referenced. Next, a standard entry structure can be designed, dividing the standard entries into two parts: standard clauses and test methods, forming a complete semantic network.

[0074] In the above embodiments, a natural language processing (NLP) model is used to extract technical terms and contextual features related to power equipment standards from the original documents. Entities are generated based on the technical terms and contextual features, and the relationships between entities are determined. Then, the similarity of entities at different granularities and dimensions is calculated. A knowledge graph can be constructed based on entities, relationships between entities, and similarity. The NLP model can automatically transform unstructured standard text into machine-understandable structured information. Combined with multi-granularity and multi-dimensional similarity calculations, the structure and semantic coherence of standard documents can be significantly improved. Furthermore, it can uncover potential connections between standard documents, enhance the completeness and connectivity of the knowledge graph, and achieve efficient and accurate modeling of power equipment standards, thus helping to reduce the management difficulty of power equipment standards.

[0075] The modeling method for power equipment standards proposed in this application can be used to construct a three-layer processing architecture comprising a document layer, a knowledge layer, and an application layer. In this method, the three layers work together to achieve intelligent management of power equipment standards.

[0076] The document layer can perform semantic parsing of the original document using a deep learning-based natural language processing model. The knowledge layer can be used to build and update the knowledge graph. The application layer can provide interactive functions to users, such as retrieving relevant data from the knowledge graph based on user-input search criteria and returning it to the user.

[0077] Specifically, the document layer can first segment the standard documents for power equipment, dividing them into chapters, clauses, and detailed rules according to the hierarchical structure of the standard documents. This step uses a combination of rule-based and statistical methods to identify structural markers (such as chapter numbers, clause numbers, etc.) in the standard documents and perform reasonable segmentation.

[0078] Next, the document layer can utilize a domain-specific dictionary and context rules to identify specialized terminology related to power equipment from the segmented document. For example, the domain-specific dictionary can contain over 10,000 commonly used terms in the power equipment field and their variants, covering various aspects such as equipment type, component names, technical parameters, and testing methods. Meanwhile, the context rules are designed based on language habits within the power equipment field, effectively identifying new terms in the text not included in the dictionary.

[0079] For each identified term, the document layer uses a sliding window technique to extract the context of ±5 surrounding words and sets a weighting decreasing mechanism. In practical applications, the size of this window can be adjusted according to the text characteristics and can be set between 3 and 7 words. The weighting decreasing mechanism follows the distance decay principle, with words closer to the central term having higher weights. The specific calculation formula can be expressed as the following formula (1).

[0080] (1)

[0081] in: Let be the weight of the word at position i. This is the weighting base (usually set to 1.0). Here, is the decay coefficient (in practice, a value between 0.2 and 0.5 is more suitable), c is the position index of the central term, |ic| represents the absolute value of the distance between the i-th position and the central position, and e is the base of the natural logarithm, approximately equal to 2.718. This weighting setting can better capture the direct contextual semantics of terms. The weight decays exponentially with the increase of distance from the central term, reflecting the degree of influence of contextual words on term understanding.

[0082] Finally, for each technical term, the document layer can retrieve entity relationships within its context and construct preliminary semantic connections. The retrieval of entity relationships employs a method based on dependency parsing and semantic role labeling, which can identify relationship types between terms such as inclusion, belonging, referencing, and testing.

[0083] The knowledge layer can further process the structured information (such as entities and their relationships) provided by the document layer to build and maintain a standard knowledge graph for power equipment.

[0084] Based on the structured information provided by the document layer, the knowledge layer can set up two entity layers. The first layer is the standard system entity, which includes three parts: power equipment, standard system, and version. The second layer is the standard entity, which includes technical standards, industry standards, and enterprise management standards.

[0085] Regarding the representation of relationships between entities, the knowledge layer can establish a clear relationship representation mechanism: the relationship between a standard system entity and the standard layer is represented as a standard; the relationship between a standard system entity and the version layer is represented as a reference; and the relationship between a referencing standard and the standard layer is represented as being referenced. Furthermore, specific relationships also exist between different types of standards; for example, technical standards may have a reference relationship, and enterprise standards may have a reference / compliance check relationship.

[0086] The knowledge layer also designed the internal structure of the standard entries, dividing them into two parts: standard clauses and test methods. The standard clauses describe the technical requirements that the equipment should meet, while the test methods describe the specific methods for verifying these requirements. This fine-grained structural division provides a foundation for subsequent semantic analysis and applications.

[0087] In practical implementations of knowledge graphs, the knowledge layer can use graph databases (such as Neo4j) to store the knowledge graph, leveraging their efficient graph traversal and query capabilities to support complex knowledge retrieval and reasoning. Nodes in a graph database represent entities, and edges represent relationships. Both nodes and edges can be appended with attributes to store detailed information about entities and relationships. The semantic network storage structure built upon graph database technology can comprehensively represent standard entities and their complex relationships, laying the foundation for the unified management and application of standard knowledge.

[0088] In one embodiment, calculating the first similarity of different entities at multiple granularities and calculating the second similarity of different entities at multiple dimensions may include:

[0089] A multi-granularity similarity calculation framework is constructed to calculate semantic similarity at the term, clause, and document levels. A multi-dimensional similarity relationship representation system is designed to evaluate the degree of similarity between entities from the dimensions of function, structure, attribute, and application. The weight coefficients of each dimension of similarity are dynamically adjusted according to the professional attributes and context of the entities. The semantic similarity matrix between entities of standard clauses and standard entries is obtained, and the association matrix of standard clauses and standard entries is constructed to form a general standard terminology library.

[0090] For example, a multi-granularity similarity calculation framework is constructed to calculate semantic similarity at the term level, clause level, and document level, respectively, including:

[0091] In term-level similarity calculation, a deep learning-based semantic vector representation method is used to map each term into a semantic vector in a high-dimensional semantic space, and the cosine similarity of the semantic vectors is calculated. In clause-level similarity calculation, clauses are treated as a set of terms, and a weighted matching method is used to calculate the similarity between clauses. In document-level similarity calculation, clauses in two different original documents are matched into clause pairs, and a weight coefficient is determined based on the position of the clause in the original document. The similarity of all clause pairs in the two original documents is then weighted and summed according to the weight coefficient to obtain the document-level similarity.

[0092] Figure 2 This is a schematic diagram of an architecture for semantic similarity calculation provided in an embodiment of this application. For example... Figure 2 As shown, the knowledge layer can construct a multi-granularity similarity calculation framework to calculate semantic similarity at the term level, clause level, and document level, respectively.

[0093] In the term-level similarity calculation, the knowledge layer adopts a deep learning-based semantic vector representation method to map each term to a high-dimensional semantic space. The similarity between terms is obtained by calculating the cosine similarity of the semantic vectors, and the calculation process can be carried out using the following formula (2).

[0094] (2)

[0095] In the above formula, sim(t1,t2) is the similarity between terms t1 and t2, and its value ranges from [-1,1]. The larger the value, the higher the similarity. and The semantic vectors for terms t1 and t2 are vector representations of the terms after mapping them to a high-dimensional semantic space (usually 300-768 dimensions). This represents the vector dot product operation; The magnitude of a vector is represented by its length, i.e. Where d is the vector dimension, v i Let be the value of the i-th dimension of the vector. This similarity calculation method can capture the semantic association between terms. Even if the surface forms of the terms are different, as long as they are semantically similar, a high similarity can be obtained.

[0096] In clause-level similarity calculation, clauses can be regarded as a set of terms, and the weighted matching method is used to calculate the similarity between clauses. The calculation process can use the following formula (3).

[0097] (3)

[0098] In the above formula, sim(C1,C2) is the similarity between clause C1 and clause C2, with a value range of [0,1]; t1 and t2 are terms in clauses C1 and C2, respectively; sim(t1,t2) is the similarity between terms t1 and t2, which is obtained by the term-level similarity calculation formula. and The importance weights of two terms are usually determined based on the terms' position in the clause (heading, body, etc.) and frequency, with weight values ​​ranging from [0,1]. This represents the maximum weighted similarity between term t1 and all terms in clause C2, i.e., finding the term t2 that is most similar to t1 and calculating the weighted similarity; This indicates a summation operation on all terms in Clause C1. The formula calculates the average of the best-match similarity between two clauses, effectively capturing semantic similarity at the clause level.

[0099] In document-level similarity calculation, the knowledge layer considers the structural characteristics of standard documents and can adopt a hierarchical weighting method. The calculation process can use the following formula (4).

[0100] (4)

[0101] In the above formula, sim(D1,D2) represents the similarity between documents D1 and D2, with a value range of [0,1]; C 1i and C 2j These refer to the i-th clause in document D1 and the j-th clause in document D2, respectively; sim(C 1i C 2j Clause C 1i and C 2j The similarity is obtained by the clause-level similarity calculation formula; and The number of clauses in documents D1 and D2, respectively; w 1i and w 2j The importance weight of a clause is usually related to its level (e.g., chapter, section, article) and position (e.g., beginning, middle, end) in the document. The weight value ranges from [0,1] and satisfies the following conditions: and ; This represents a weighted sum of the similarities across all clause pairs. The formula calculates the weighted similarity sum of all clause pairs in two documents, providing a comprehensive assessment of document-level similarity.

[0102] like Figure 2As shown, in addition to the multi-granularity computation framework, the knowledge layer also designs a multi-dimensional similarity relationship representation system to evaluate the degree of similarity between entities from the dimensions of function, structure, attribute, and application. This multi-dimensional representation can more comprehensively capture the similarity relationships between entities and support more accurate similarity assessment.

[0103] Based on the entity's professional attributes and context, the knowledge layer can dynamically adjust the weighting coefficients of similarity across different dimensions. For example, for entities related to technical parameters, the weight of attribute dimensions will be relatively high; for entities related to experimental methods, the weight of functional dimensions will be even higher. This adaptive adjustment mechanism improves the accuracy and flexibility of similarity calculation.

[0104] Finally, the knowledge layer obtains the semantic similarity matrix between entities of standard clauses and standard entries, constructs the association matrix of standard clauses and standard entries, and forms a general standard terminology database. This terminology database not only contains basic information about terms, but also includes similarity relationships and semantic connections between terms, providing important support for the construction and application of knowledge graphs.

[0105] In one embodiment, after constructing the knowledge graph, the following is also included:

[0106] Retrieve new documents containing standards for power equipment; calculate the similarity between the new document and the original document; if the similarity is greater than a preset threshold, extract the differences between the new document and the original document, and update the knowledge graph based on the differences.

[0107] In this embodiment of the application, the document layer can also be used to automatically identify and process newly added standard documents, and update the existing knowledge graph (i.e., the knowledge graph generated based on the original documents) based on the newly added documents.

[0108] For example, the document layer can acquire new standard documents, extract document feature fingerprints, and quickly compare them with existing original documents in the standard library; the document feature fingerprint is a highly generalized representation of the document content, generated by hash calculation of key document content, which can quickly determine the similarity of documents.

[0109] After obtaining the new standard document, the document layer can calculate the similarity matrix between the new standard document and the original document, and set a similarity threshold θ. The similarity calculation adopts a multi-feature fusion method, which comprehensively considers the structural features, content features, and semantic features of the standard document. The specific calculation formula is as follows.

[0110]

[0111] In the above formula, sim(D1,D2) represents the combined similarity between documents D1 and D2. strStructural similarity represents the degree of similarity between two documents at the structural level (such as chapter organization, paragraph layout, etc.). co Content similarity, representing the degree of similarity between two documents in terms of text content, Sim se Semantic similarity represents the degree of similarity between two documents at the semantic level. γ1, γ2, and γ3 are the weight coefficients for each similarity score, satisfying γ1 + γ2 + γ3 = 1. In practical applications, based on the characteristics of standard documents, γ1 = 0.2, γ2 = 0.3, and γ3 = 0.5 can be set to emphasize the importance of semantic similarity. These weight coefficients can be adjusted according to specific application scenarios and needs to optimize the accuracy of similarity calculation.

[0112] Next, the document layer determines whether the similarity value between the new standard document and the existing i-th original standard document is greater than the similarity threshold θ. The similarity threshold θ is a key parameter that determines the system's sensitivity to standard similarity. In a preferred embodiment of the invention, the value of θ ranges from 0.7 to 0.85, and is typically set to 0.75. When the similarity value is greater than θ, it indicates that the new standard document is highly similar to the original document, which may be an update or revision of the original document. In this case, the document layer extracts the differences between the new standard document and the original document, forms a new entity, and adds it to the existing standard for updating.

[0113] If the similarity between the new standard document and a single original document is not high enough, the document layer will further determine whether the average similarity between the new standard document and the top n original documents (n≥i) is greater than the similarity threshold θ. The parameter n is usually set to 3-5, representing the existing standards most similar to the new standard. If the average similarity is less than θ, it is added to the new standard library as a new standard. This dual-judgment mechanism can effectively distinguish between standard updates and new standards, improving the system's intelligent processing capabilities.

[0114] In one embodiment, updating the knowledge graph based on the differing content includes:

[0115] The differential content is transformed into new entities and relationships in the knowledge graph; based on the association influence propagation mechanism, the influence of the transformed new entities and relationships on the knowledge graph is evaluated, and the knowledge graph is updated according to the new entities and relationships with an influence greater than a preset threshold.

[0116] For example, the knowledge layer can transform the new standard documents processed by the document layer into new entities and relationships in the knowledge graph; design an association influence propagation mechanism to assess the scope of influence of the newly added entities and relationships on the existing knowledge graph; and perform update operations on the entities and relationships within the scope of influence to ensure the consistency of the knowledge graph.

[0117] Figure 3This is a schematic diagram illustrating a knowledge graph update mechanism provided in an embodiment of this application. Figure 3 As shown, the knowledge layer can transform newly added standard documents processed by the document layer into new entities and relations in the knowledge graph. The transformation process maintains the standardized representation of entities and relations, ensuring compatibility with existing knowledge graphs.

[0118] The knowledge layer can assess the impact of newly added entities and relationships on the existing knowledge graph based on the mechanism of influence propagation. The assessment of the impact scope is based on graph structure analysis, calculating the connectivity and influence between the newly added nodes and existing nodes. The specific formula for calculating the influence can be expressed as follows (5).

[0119] (5)

[0120] In the above formula, For the newly added entity e new For existing entity e existing The degree of influence ranges from (0, +∞), with a larger value indicating a greater influence. The strength of the association between two entities, with a value range of [0,1], is determined by the semantic similarity and relationship type between the entities; The shortest path distance between two entities in the graph, i.e., from e new to e existing The minimum number of edges required, with a minimum value of 1 (direct connection). When two entities have a strong association and are close to each other, the influence value is large, indicating that the new entity has a significant impact on the existing entity. In practical applications, entities with an influence value greater than a preset threshold (e.g., 0.3) can be included in the scope of influence.

[0121] For entities and relationships within the scope of influence, the knowledge layer can perform update operations to ensure the consistency of the knowledge graph. Update operations include adding new entities, establishing new relationships, modifying existing entity attributes, and adjusting relationship weights, with the specific operations dynamically determined based on the entity type and the degree of influence.

[0122] like Figure 3 As shown, the knowledge layer can also extract implicit knowledge needs based on user interaction data at the application layer, and optimize and adjust the knowledge graph structure accordingly. User interaction behavior is an important feedback signal that reflects the practicality and completeness of the knowledge graph. By analyzing user retrieval patterns, click behavior, and usage frequency, the module can identify weaknesses and optimization directions in the knowledge graph.

[0123] By processing newly added standard documents at the document and knowledge layers, new standards can be automatically identified and intelligently processed, enabling automatic updates to the knowledge graph. This dynamic update mechanism allows the knowledge graph to continuously optimize as standards evolve and user needs change, maintaining the timeliness and practicality of knowledge.

[0124] In one embodiment, after constructing the knowledge graph, the following is also included:

[0125] Obtain application requirements for power equipment standards, and perform any of the following target operations using a knowledge graph based on those requirements:

[0126] For standard retrieval needs, entities with similarity greater than a preset threshold are retrieved based on the knowledge graph, and similarity explanation information is provided, including matching points, similarity scores, and matching criteria.

[0127] For compliance inspection needs, the parameters of the equipment to be inspected are mapped to entities in the knowledge graph, and the matching criteria are determined based on the relationships between the entities. The compliance inspection results are then output based on the comparison between the matching criteria and the equipment parameters.

[0128] For discrepancy analysis, the entities corresponding to different versions of power equipment standards in the knowledge graph are identified, and the attributes and relationships of the identified entities are compared to generate discrepancy analysis results.

[0129] In this embodiment of the application, the application layer can provide users with interactive functions such as standard retrieval, compliance checks, and version difference analysis.

[0130] For example, the standard search function may include: receiving search conditions input by the user and converting them into structured queries; calculating the similarity between the search conditions and standard entities in the knowledge graph based on a multi-dimensional semantic similarity algorithm; setting a similarity threshold and filtering standard entities with similarity exceeding the threshold as candidate results; sorting the search results according to the comprehensive score of multi-dimensional similarity and providing similarity explanation information.

[0131] First, the application layer can receive user-input search criteria and transform them into structured queries. This transformation process involves technologies such as keyword extraction, query intent recognition, and query expansion, enabling an effective understanding of the user's search needs.

[0132] Next, the application layer can calculate the similarity between query conditions and standard entities in the knowledge graph based on a multi-dimensional semantic similarity algorithm. The calculation process considers matching across multiple dimensions, enabling the discovery of semantically related but differently expressed standard entities.

[0133] The application layer can pre-set a similarity threshold (typically 0.65-0.75) to filter standard entities with similarity exceeding the threshold as candidate results. The threshold setting needs to balance retrieval precision and recall, and can be adjusted according to the actual application scenario. In the retrieval of standard entities for power equipment, the threshold can be set to 0.7, ensuring retrieval accuracy without overlooking potentially relevant standard entities.

[0134] Finally, the application layer ranks the search results based on a comprehensive similarity score across multiple dimensions and provides explanatory information on similarity. This explanatory information includes matching points, similarity scores, and matching criteria, helping users understand the relevance of the search results and improving the search experience.

[0135] For example, the compliance inspection function may include: acquiring the equipment parameter data to be inspected and mapping it to the technical parameter entity in the knowledge graph; identifying the valid standard requirements applicable to the equipment based on the association relationship of the knowledge graph; comparing the equipment parameters with the standard requirements to determine whether they comply with the specifications; and generating a compliance inspection report, including compliance items, non-compliance items, and gap analysis.

[0136] First, the application layer can obtain the device parameter data to be inspected and map it to the technical parameter entity in the knowledge graph. The mapping process needs to resolve issues such as inconsistent parameter naming, unit conversion, and differences in numerical representation to ensure accurate correspondence between device parameters and standard parameters.

[0137] Next, the application layer identifies the applicable standard requirements for the device based on the relationships within the knowledge graph. This identification process utilizes a graph traversal algorithm, starting from the device type node, to find all relevant standard clause nodes, and then determines the final applicable standard requirements based on the standard's effective date, scope of application, and priority.

[0138] Then, the application layer compares the device parameters with the standard requirements to determine whether they comply with the specifications. The comparison method varies depending on the parameter type: for numerical parameters, it checks whether they are within the range specified by the standard; for enumerated parameters, it checks whether they conform to the values ​​allowed by the standard; for composite parameters, it performs a comprehensive evaluation based on multiple conditions.

[0139] Finally, the application layer can generate a compliance check report, including compliant items, non-compliant items, and gap analysis. The report presents the check results in an intuitive way, helping users quickly understand the compliance status of the equipment and providing a basis for equipment improvement and optimization.

[0140] For the version difference analysis function, the application layer can compare changes in entity relationships between different standard versions to visualize the differences in standard content. In specific implementation, the application layer first identifies the corresponding entities in the two versions, then compares the changes in entity attributes and relationships, and finally displays the differences in a graphical manner.

[0141] The results of the difference analysis can include: new content (entities and relationships added in the new version), deleted content (entities and relationships removed in the new version), modified content (entities whose attributes or relationships have changed), and unchanged content. This intuitive display of differences helps users quickly understand the content and impact of standard updates, supporting more efficient standard management and application.

[0142] In the above embodiments, the application layer can provide diversified intelligent application services, including accurate standard retrieval, automated compliance checks, and intuitive version difference analysis, which greatly improves the efficiency and quality of standard management and application.

[0143] In one embodiment, after constructing the knowledge graph, the following is also included:

[0144] Acquire operational data of substation equipment; establish a mapping relationship between operational data and standard knowledge in the knowledge graph to form a multimodal fusion knowledge graph; perform cross-modal knowledge reasoning on the multimodal fusion knowledge graph based on graph reasoning rules, and provide standard basis and decision-making suggestions for the operational status of substation equipment based on the reasoning results.

[0145] Figure 4 This is a schematic diagram of an inference process that integrates multi-source data, provided as an embodiment of this application. Figure 4 As shown, the application layer can also perform multi-source data fusion and knowledge reasoning based on the constructed knowledge graph, realizing the deep integration and collaborative application of standard knowledge and equipment operation data.

[0146] First, the application layer can acquire operational data from power equipment, extract feature fingerprints, and construct a representation of the equipment's operating status. Operational data typically comes from SCADA (Supervisory Control and Data Acquisition) systems, smart meters, monitoring devices, etc., and includes parameters such as voltage, current, temperature, and vibration. The feature fingerprint extraction process involves data cleaning, feature calculation, and anomaly detection, ultimately forming a compact representation of the equipment's operating status.

[0147] Next, a mapping relationship is established between standard knowledge and equipment operation data, forming a multimodal fusion knowledge graph. The mapping relationship is mainly based on parameter correspondence and semantic association; for example, mapping the A-phase voltage in the operation data to the rated voltage parameter in the standard. The fusion process needs to handle issues such as inconsistent data formats, differences in sampling frequencies, and different naming conventions to ensure the effective fusion of data from different sources.

[0148] In the process of multimodal fusion, it is necessary to ensure the compatibility of data types. For matrix data, its dimensions need to be clearly defined to ensure the validity of matrix operations. For example, the operating state of a device can be represented as a state vector. Where n is the number of state parameters; the standard requirements can be represented as a requirement matrix. Where m is the number of requirements and k is the number of parameters for each requirement. During the fusion process, a mapping matrix is ​​needed. The state vector is converted into a form compatible with the required matrix for comparison and analysis.

[0149] Then, graph-based reasoning rules are designed to perform cross-modal knowledge reasoning on a fused knowledge graph. These rules, formulated based on domain knowledge and statistical learning methods, are capable of inferring device status and potential problems from a combination of standard requirements and operational data. Specific reasoning methods include rule-based reasoning, statistical reasoning, and graph neural network-based reasoning, with appropriate inference strategies selected according to different scenarios.

[0150] Finally, based on the reasoning results, standard guidelines and decision-making suggestions are provided for the equipment's operational status. These suggestions may include compliance assessments, potential risk warnings, maintenance recommendations, and optimization plans, helping maintenance personnel make more informed decisions and improving the intelligence level of equipment management.

[0151] In the above embodiments, the application layer can also integrate standard knowledge with equipment operation data. Through multi-source data fusion and knowledge reasoning technology, static standard knowledge is combined with dynamic equipment operation data to provide standard basis and decision suggestions for equipment operation status.

[0152] Reference Figure 5 This application also provides a modeling system for substation equipment standards. The system may include three main parts: document layer 1, knowledge layer 2, and application layer 3. Each part contains multiple functional modules, which together realize the intelligent management of substation equipment standards.

[0153] Document layer 1 includes a file storage management module 11, a text extraction module 12, and an incremental learning module 13. The file storage management module 11 manages substation equipment standard documents, including document storage, version control, and basic management functions. The text extraction module 12 uses deep learning-based natural language processing technology to extract technical terms and their contextual features from the substation equipment standard documents, generating a standard specification semantic structure system. The incremental learning module 13 automatically identifies new standard documents and determines whether to update or add new standard documents through similarity evaluation.

[0154] The modules in the document layer can work together to achieve intelligent management and processing of standard documents. The file storage management module 11 provides the original standard documents to the text extraction module 12. After processing the documents, the text extraction module 12 passes the structured information to the knowledge layer 2. The incremental learning module 13 monitors changes in the standard library and triggers the update process.

[0155] Knowledge Layer 2 includes a semantic modeling module 21, a semantic similarity algorithm module 22, a knowledge graph management module 23, and an inference engine module 24. The semantic modeling module 21 constructs a semantic network storage structure based on a graph database, storing standard entities and their relationships as a knowledge graph. The semantic similarity algorithm module 22 uses a deep learning-based multi-dimensional semantic similarity algorithm to calculate the similarity matrix between standard documents, automatically discovering implicit relationships between standards. The knowledge graph management module 23 dynamically updates the knowledge graph based on newly added standards and user interaction data, achieving continuous knowledge optimization. The inference engine module 24 performs inference tasks on the knowledge graph, supporting complex queries and knowledge discovery.

[0156] Knowledge Layer 2 is the core of the system, responsible for knowledge representation, processing, and optimization. The modules collaborate closely: the semantic modeling module 21 constructs the basic structure of the knowledge graph; the semantic similarity algorithm module 22 discovers semantic relationships between entities; the knowledge graph management module 23 maintains and updates the knowledge graph; and the reasoning engine module 24 provides reasoning capabilities based on the knowledge graph.

[0157] Application Layer 3 includes a standard retrieval module 31, a compliance check module 32, a version difference analysis module 33, and a user interaction module 34. The standard retrieval module 31 calculates the text similarity of standard descriptions, providing accurate standard retrieval services. The compliance check module 32 provides a standard effectiveness query service, implementing compliance check functionality. The version difference analysis module 33 compares changes in entity relationships between different standard versions, visualizing differences in standard content. The user interaction module 34 processes user input and presents system results, providing a user-friendly interface.

[0158] Application layer 3 is user-facing, providing specific application functions. User interaction module 34 receives user input, distributes it to the corresponding functional modules for processing, and then presents the processing results in a user-friendly manner. Simultaneously, user interaction data is collected and fed back to knowledge layer 2 to optimize the knowledge graph and algorithm model.

[0159] The system architecture of this application is clearly designed, with reasonable module division and coordinated cooperation among modules to achieve intelligent management of power equipment standards. The system has good scalability and adaptability, and can meet the application scenarios of different scales and needs, providing strong support for the standardization work of the power industry.

[0160] The implementation process and effects of this application will be illustrated below with specific examples.

[0161] Example 1, Standard Management of Main Transformers in Substations:

[0162] A power grid company needs to implement intelligent management of standards related to main transformers in substations. The company possesses a large number of main transformer standard documents, including national regulations, industry standards, and enterprise standards. These standards have complex citation relationships and implicit connections, and new standards are constantly being released and old standards are being updated.

[0163] First, the system acquires all relevant standard documents for main transformers and constructs a three-layer processing architecture. At the document layer, the system preprocesses and analyzes the standard documents. Using deep learning-based natural language processing, key terms such as main transformer, rated voltage, short-circuit impedance, and temperature rise test are extracted from the standard documents, along with their contextual features and entity relationships. Through incremental learning algorithms, the system can automatically identify newly released main transformer standards and determine their relationship to existing standards.

[0164] At the knowledge layer, the system constructs a standard knowledge graph for main transformers. This graph contains approximately 5,000 entity nodes and 8,000 relational edges, covering various technical parameters, test methods, and quality requirements of main transformers. Through a multi-dimensional semantic similarity algorithm, the system discovered several implicit relationships. For example, different descriptions of short-circuit impedance in different standards actually point to the same concept, or certain test methods have slight differences in different standards but are essentially the same. These findings greatly enrich the semantic connotation of the knowledge graph.

[0165] At the application layer, the system provides a variety of functions. For example, when engineers need to query the temperature rise limit requirements for main transformers, the system can quickly find relevant standard clauses through the standard search service and present them in a sorted order of relevance. When it is necessary to check whether a certain model of main transformer complies with the latest standard, the compliance check function can automatically compare equipment parameters with standard requirements and generate a detailed compliance report. When national standards are updated, the version difference analysis function can intuitively display the changes between the old and new versions, helping relevant personnel quickly understand the key points of the update.

[0166] Using the methods and systems described in this application, the power grid company's standards management efficiency has been significantly improved: standards retrieval time has been reduced by 65%, standards update processing time has been shortened by 80%, and the accuracy of compliance checks has increased by 35%. At the same time, the system has discovered several previously overlooked standard correlations and potential conflicts, providing important references for standards optimization.

[0167] Example 2: Intelligent application of preventive testing standards for power equipment:

[0168] A power maintenance company is responsible for preventative testing of a large number of substation equipment, requiring them to perform tests and assess equipment condition according to various testing standards. These standards are scattered across multiple documents and are frequently updated, posing a significant challenge to maintenance personnel.

[0169] The system first acquires all relevant standard documents for preventive testing and processes them intelligently. At the document level, the system extracts key information such as test name, test method, and judgment criteria, and identifies the relationships between this information. For example, the system can identify the correlation between insulation resistance testing and absorption ratio measurement, as well as the different requirements for the same test for different equipment types.

[0170] At the knowledge layer, the system constructs a knowledge graph of preventative testing standards. This graph not only contains static knowledge of testing standards but also establishes connections with equipment operating data. Through multi-source data fusion and knowledge reasoning technology, the system can determine the health status of the equipment and provide targeted testing recommendations based on historical testing data and operating status.

[0171] At the application layer, the system provides intelligent test management functions. When maintenance personnel need to conduct preventive tests on a device, the system can automatically recommend applicable test items and standards; after the test results are entered, the system can automatically determine whether the results are qualified and provide a detailed analysis report; when the standards are updated, the system can automatically identify the changes and remind relevant personnel to make adjustments.

[0172] The methods and systems described in this application significantly improve the efficiency and quality of the maintenance unit's work: test preparation time is reduced by 50%, the accuracy of test result judgment is increased by 40%, and the response time for test standard updates is shortened by 70%. More importantly, the system can predict potential equipment problems based on historical equipment data and test results, enabling predictive maintenance and effectively reducing equipment failure rates.

[0173] In summary, this application constructs a complete modeling and management solution for power equipment standards, transforming static standard documents into a dynamic knowledge network. This enables intelligent understanding, correlation discovery, dynamic updating, and knowledge reasoning of standards, bringing revolutionary changes to the standardization work in the power industry and promoting the development of power equipment standard management towards intelligence and refinement.

[0174] Figure 6 A schematic diagram of a modeling device for a standard power equipment provided in this application is shown below. Figure 6 As shown, the substation equipment standard modeling device 600 provided in this embodiment includes:

[0175] Extraction module 601 is used to extract professional terms and contextual features related to power equipment standards from the original document using a natural language processing model;

[0176] The entity generation module 602 is used to generate entities and determine the relationships between entities based on technical terms and contextual features.

[0177] The calculation module 603 is used to calculate the first similarity of different entities at multiple granularities and the second similarity of different entities at multiple dimensions.

[0178] The knowledge graph construction module 604 is used to construct a knowledge graph as a modeling result based on entities, relations, first similarity, and second similarity.

[0179] In some possible implementations, the calculation module 603 can also be used to: construct a multi-granularity similarity calculation framework to calculate semantic similarity at the term, clause, and document levels respectively; design a multi-dimensional similarity relationship representation system to evaluate the degree of similarity between entities from the functional, structural, attribute, and application dimensions; dynamically adjust the weight coefficients of each dimension of similarity according to the professional attributes and context of the entities; obtain the semantic similarity matrix between entities of standard clauses and standard entries, construct the association matrix of standard clauses and standard entries, and form a general standard terminology library.

[0180] In some possible implementations, the calculation module 603 can also be used to: in term-level similarity calculation, employ a deep learning-based semantic vector representation method to map each term into a semantic vector in a high-dimensional semantic space, and calculate the cosine similarity of the semantic vectors; in clause-level similarity calculation, treat clauses as a set of terms, and use a weighted matching method to calculate the similarity between clauses; in document-level similarity calculation, match clauses in two different original documents into clause pairs, determine weight coefficients based on the position of the clauses in the original documents, and perform a weighted summation of the similarities of all clause pairs in the two original documents according to the weight coefficients to obtain the document-level similarity.

[0181] In some possible implementations, the knowledge graph construction module 604 can also be used to: obtain application requirements for substation equipment standards, and perform any of the following target operations through the knowledge graph based on the application requirements: for standard retrieval requirements, retrieve entities with similarity greater than a preset threshold based on the knowledge graph, and provide similarity explanation information, including matching points, similarity scores, and matching criteria; for compliance inspection requirements, map the equipment parameters to be inspected to entities in the knowledge graph, determine the matching standards based on the association relationships of the entities, and output compliance inspection results based on the comparison results between the matching standards and the equipment parameters; for difference analysis requirements, identify the entities corresponding to different versions of substation equipment standards in the knowledge graph, and compare the attributes and relationships of the identified entities to generate difference analysis results.

[0182] In some possible implementations, the knowledge graph construction module 604 can also be used to: obtain new documents containing substation equipment standards; calculate the similarity between the new documents and the original documents; if the similarity is greater than a preset threshold, extract the differences between the new documents and the original documents, and update the knowledge graph based on the differences.

[0183] In some possible implementations, the graph construction module 604 can also be used to: transform differential content into new entities and relationships in the knowledge graph; evaluate the impact of the transformed new entities and relationships on the knowledge graph based on the association influence propagation mechanism, and update the knowledge graph based on new entities and relationships with an impact greater than a preset threshold.

[0184] In some possible implementations, the graph construction module 604 can also be used to: acquire the operating data of the substation equipment; establish a mapping relationship between the operating data and the standard knowledge in the knowledge graph to form a multimodal fusion knowledge graph; perform cross-modal knowledge reasoning on the multimodal fusion knowledge graph based on the graph reasoning rules, and provide standard basis and decision suggestions for the operating status of the substation equipment based on the reasoning results.

[0185] The modeling device for the standard of power equipment provided in this embodiment can execute the method provided in the above method embodiment. Its implementation principle and technical effect are similar, and will not be described in detail here.

[0186] Figure 7 This is a schematic diagram of the structure of an electronic device provided in this application. Figure 7 As shown, the electronic device 70 provided in this embodiment includes at least one processor 701 and a memory 702. Optionally, the device 70 further includes a communication component 703. The processor 701, memory 702, and communication component 703 are connected via a bus 704.

[0187] In a specific implementation, at least one processor 701 executes computer execution instructions stored in memory 702, causing at least one processor 701 to perform the above-described method.

[0188] The specific implementation process of processor 701 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.

[0189] In the above embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0190] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.

[0191] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0192] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0193] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described method.

[0194] The aforementioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.

[0195] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and the readable storage medium can exist as discrete components in the device.

[0196] The division of units is merely a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0197] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0198] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0199] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0200] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0201] Finally, it should be noted that other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein, and is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. A standard modeling method for power equipment, characterized in that, include: A natural language processing model was used to extract technical terms and contextual features related to power equipment standards from the original document. Based on the technical terms and the contextual features, entities are generated and the relationships between the entities are determined. Calculate the first similarity of different entities at multiple granularities, and calculate the second similarity of different entities at multiple dimensions; A knowledge graph is constructed as the modeling result based on the entity, the relationship, the first similarity, and the second similarity.

2. The method according to claim 1, characterized in that, The calculation of the first similarity of different entities at multiple granularities and the calculation of the second similarity of different entities at multiple dimensions include: A multi-granularity similarity calculation framework is constructed to calculate the semantic similarity between entities at the term level, clause level, and document level, respectively. Design a multi-dimensional similarity representation system to calculate the similarity between entities from the functional, structural, attribute, and application dimensions; The weighting coefficients of similarity in each dimension are dynamically adjusted based on the entity's professional attributes and context. Obtain the semantic similarity matrix between entities of standard clauses and standard entries, and construct the association matrix between standard clauses and standard entries.

3. The method according to claim 2, characterized in that, The constructed multi-granularity similarity calculation framework calculates semantic similarity at the term level, clause level, and document level, including: In the term-level similarity calculation, a semantic vector representation method based on deep learning is adopted to map each term into a semantic vector in a high-dimensional semantic space, and the cosine similarity of the semantic vectors is calculated. In clause-level similarity calculation, clauses are treated as a set of terms, and a weighted matching method is used to calculate the similarity between clauses. In document-level similarity calculation, clauses in two different original documents are matched into clause pairs. A weight coefficient is determined based on the position of the clause in the original document. The similarity of all clause pairs in the two original documents is then weighted and summed according to the weight coefficient to obtain the document-level similarity.

4. The method according to any one of claims 1 to 3, characterized in that, After constructing the knowledge graph, the following is also included: Obtain the application requirements for the aforementioned power equipment standard, and perform any of the following operations based on the application requirements using the knowledge graph: For standard retrieval requirements, entities with similarity greater than a preset threshold are retrieved based on the knowledge graph, and similarity explanation information is provided, including matching points, similarity scores, and matching criteria. For compliance inspection requirements, the parameters of the equipment to be inspected are mapped to entities in the knowledge graph, and the matching criteria are determined according to the association relationship of the entities. The compliance inspection results are output according to the comparison results between the matching criteria and the equipment parameters. For discrepancy analysis, the entities corresponding to different versions of the power equipment standards in the knowledge graph are identified, and the attributes and relationships of the identified entities are compared to generate discrepancy analysis results.

5. The method according to any one of claims 1 to 3, characterized in that, After constructing the knowledge graph, the following is also included: Retrieve new documents containing standards for power equipment; Calculate the similarity between the newly added document and the original document; If the similarity is greater than a preset threshold, the differences between the new document and the original document are extracted, and the knowledge graph is updated based on the differences.

6. The method according to claim 5, characterized in that, The step of updating the knowledge graph based on the differences includes: The discrepancies are then transformed into new entities and relationships in the knowledge graph. Based on the mechanism of association and influence propagation, the influence of the newly added entities and relationships after transformation on the knowledge graph is evaluated, and the knowledge graph is updated according to the newly added entities and relationships whose influence is greater than a preset threshold.

7. The method according to any one of claims 1 to 3, characterized in that, After constructing the knowledge graph, the following is also included: Obtain operating data of power equipment; Establish a mapping relationship between the operational data and the standard knowledge in the knowledge graph to form a multimodal fusion knowledge graph; Based on graph-based reasoning rules, cross-modal knowledge reasoning is performed on the multimodal fusion knowledge graph, and standard basis and / or decision suggestions are provided for the operation status of the substation based on the reasoning results.

8. A modeling device for a standard of power equipment, characterized in that, include: The extraction module is used to extract technical terms and contextual features related to power equipment standards from the original document using a natural language processing model. An entity generation module is used to generate entities and determine the relationships between the entities based on the technical terms and the context features. The calculation module is used to calculate the first similarity of different entities at multiple granularities, and to calculate the second similarity of different entities at multiple dimensions. The knowledge graph construction module is used to construct a knowledge graph as a modeling result based on the entity, the relationship, the first similarity and the second similarity.

9. An electronic device, characterized in that, include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory to implement the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1 to 7.

Images