A power dispatch automation operation knowledge graph construction method and system

Abstract

This invention proposes a method for constructing a knowledge graph for power dispatch automation operations, comprising the following steps: Step S1, constructing a power dispatch automation operation ontology structure; Step S2, processing the data of the docx text to obtain a first triple in the form of (entity, relation category label, entity) and the domain text to be extracted from the formatted data; Step S3, obtaining the character vector and position vector of each statement in the domain text to be extracted, adding the character vector and the position vector to obtain a new vector representation of the text, and inputting it into an entity relation extraction module that integrates entity category information to obtain a second triple in the form of (entity, relation category label, entity); Step S4, constructing a power dispatch automation operation knowledge graph based on the ontology structure, the first entity triple, and the second entity triple; This invention can fully explore the domain data of power dispatch automation operations and effectively improve the utilization rate of operational knowledge in the power dispatch automation domain.

Description

Technical Field

[0001] This invention relates to the field of power grid operation and maintenance technology, and in particular to a method and system for constructing a knowledge graph for automated power dispatching operations. Background Technology

[0002] my country began its foray into power system dispatch automation in the 1970s, accumulating a vast amount of structured and unstructured data. Currently, traditional manual experience is insufficient for accurately and quickly analyzing faults in large-scale power dispatch automation systems. Knowledge graph technology can effectively address the challenge of utilizing operational knowledge in power dispatch automation. Constructing a comprehensive ontology structure can more fully express the faults, operation and maintenance tasks, and safety measures of the power dispatch automation system, enabling better utilization of historical data. Entity relation joint extraction methods extract triple information from unformatted text, better linking equipment, systems, personnel, faults, and work content. Summary of the Invention

[0003] This invention proposes a method and system for constructing a knowledge graph of power dispatch automation operations. By preprocessing historical texts of dispatch automation operations, and using ontology definition and entity relationship extraction methods, a knowledge graph of power dispatch automation operations is established. This fully explores data in the field of power dispatch automation operations and can effectively improve the utilization rate of operational knowledge in the field of power dispatch automation.

[0004] The present invention adopts the following technical solution.

[0005] A method for constructing a knowledge graph of automated power dispatching operations, which involves preprocessing historical texts of automated dispatching operations and using ontology definition and entity relation extraction to build a knowledge graph of automated power dispatching operations, including the following steps;

[0006] Step S1: Construct the main structure of power dispatch automation operation;

[0007] Step S2: Process the data of the docx text to obtain the first triple in the form of (entity, relation category label, entity) and the domain text to be extracted from the formatted data;

[0008] Step S3: Obtain the character vector and position vector of each sentence in the domain text to be extracted, add the character vector and the position vector to obtain a new vector representation of the text, and input it into the entity relationship extraction module that integrates entity category information to obtain a second triple in the form of (entity, relationship category label, entity).

[0009] Step S4: Construct a knowledge graph for power dispatch automation operations based on the ontology structure, the first entity triplet, and the second entity triplet.

[0010] In step S1, the specific definition of the power dispatch automation operation body structure is as follows:

[0011] Let the power dispatch automation operation ontology be Power-Dispatch-Automation-OperatingOntology; Power-Dispatch-Automation-Operating Ontology = {Entity, Attribute, Relation}, where:

[0012] Step S1A1: Define Entity as a set of entity-related concepts, used to represent a collection of objective entities in the field of power dispatch automation operations. Entity = (Software, Equipment, Staff, Fault, Work). Software is a software entity, referring to the objectively existing business systems and software involved in power dispatch automation, including six subclasses: business systems, processes, databases, data tables, files, and operating systems. Equipment is a device entity, representing objectively existing physical equipment and sites. Staff is a personnel entity, including personnel, companies, departments, and roles involved in power dispatch automation operations. Fault is a fault entity, including fault causes and abnormal phenomena subclasses. Work is a work entity, including work content, work items, instruction operations, and safety measures.

[0013] Step S1A2: Define Attribute as a set of attribute-related concepts, representing the specific attribute characteristics of an Entity, including version, qualification, and risk; among them, Software and Equipment entities have version attributes, identifying the version number of the software and equipment; Staff entities have qualification attributes, including licensed operation categories, qualification certificates, and training information, identifying whether the role, company, and personnel are qualified to perform the work; risks include equipment risks and instruction risks.

[0014] Step S1A3: Define Relation, which includes all relationships between entities and attribute-related concepts except for hierarchical relationships, denoted as <Concept C1, Relation R, Concept C2>. Concept C1 and Concept C2 are contained in the entity-related concept set Entity and the attribute-related concept set Attribute.

[0015] In step S2, the docx text data is processed to obtain the first triplet in the form of (entity, relation category label, entity) and the text of the entity relation to be extracted from the formatted data. The specific steps include: Step S2A1, defining the list of fields to be extracted: NAME = {name1, name2, ..., name...}n The list of outline levels corresponding to the fields: LEVEL = {level1, level2, ..., level} n}, define the temporary dictionary temp as empty;

[0016] Step S2A2: Use the python-docx tool to parse the docx file and obtain a list of paragraphs in XML data format of the document;

[0017] Step S2A3: Read a paragraph data in XML format. If the outline level exists in LEVEL, proceed to step S2A4; otherwise, continue reading the next paragraph until the outline level of the paragraph being read exists in LEVEL or the end of the document is reached.

[0018] Step S2A4: If the temporary dictionary temp is empty, then the content of this segment is... <name i ,content i Save the key-value pairs in the form of > to a temporary dictionary, where name represents the field name and content represents the paragraph content, and then proceed to step S2A3;

[0019] Step S2A5: If the temporary dictionary contains the field "name" i And the corresponding key-value pair has the value "content". i If the outline level is the lowest, then the content of the current paragraph will be appended to the content. i Save it back to temp and execute step S2A3;

[0020] Step S2A6: If the temporary dictionary contains the field "name" i If the outline level is not the last level, then the contents of the temporary dictionary temp are saved to the formatted text database in JSON format, and the value corresponding to the field outline level being the last level is saved to the domain text database to be extracted. The field of this level and all fields of the next level are removed from the temporary dictionary, and step S2A3 is executed.

[0021] Step S2A7: Obtain the first triplet based on the ontology structure and the content of the formatted text database. The domain text to be extracted is stored in the domain text database.

[0022] In step S2, the method further includes obtaining a first triplet based on the ontology structure and the content of the formatted text database, specifically including the following steps:

[0023] Extract the dictionary temp from the formatted text database; obtain the key1 and corresponding value1 of one element in temp; retrieve the associated relation r and entity key2 set of key1 in the ontology structure; find key2 and corresponding value2 in temp; and save (value1, r, value2) as the first triplet.

[0024] In step S3, the character vector and position vector of each sentence in the domain text to be extracted are obtained. The character vector and the position vector are added to obtain a new vector representation of the text, and then input into the entity relation extraction module that integrates entity category information to obtain a second triple in the form of (entity, relation category label, entity). Specifically, this includes the following steps:

[0025] Step S3A1: Obtain the character vector and position vector of each sentence in the domain text to be extracted, add the character vector and the position vector to obtain a new vector representation of the text, and perform first entity recognition through a multi-head self-attention mechanism layer to obtain multiple first entities and first entity categories;

[0026] Step S3A2: Obtain the vector representation of the first entity category information, input the new vector of the text, the first entity, and the first entity category information into the specific relationship second entity recognition module that fuses the entity category information, and obtain a second triplet in the form of (entity, relationship category label, entity).

[0027] In step S3A1, the character vector and position vector of each sentence in the domain text to be extracted are obtained, and the character vector and position vector are added together to obtain a new vector representation of the text. The first entity recognition is performed through a multi-head self-attention mechanism layer to obtain multiple first entities and first entity categories. Specifically, the steps are as follows: Step S3B1: Obtain the character vector and position vector of each sentence. For each character in each sentence, obtain the character vector and position vector of each character.

[0028] Step S3B2: Add the word vector and the position vector together, and obtain the first entity prediction probability of the sentence through a multi-head self-attention mechanism layer;

[0029] Step S3B3: The predicted probability of the first entity is processed through a multi-layer pointer decoding layer to obtain multiple first entities and first entity categories;

[0030] In step S3A2, the vector representation of the first entity category information is obtained. The new vector of the text, the first entity, and the first entity category information are input into the specific relationship second entity recognition module that fuses the entity category information to obtain a second triple in the form of (entity, relationship category label, entity). Specifically, this includes the following steps:

[0031] Step S4C1: Encode the head entity category using the head entity category encoder to obtain the first entity category vector;

[0032] Step S4C2: Obtain the character vector of the first entity from the new vector of the text, and add the character vectors in the first dimension to obtain the word vector of the first entity;

[0033] Step S4C3: Add the new text vector, the first entity word vector, and the first entity category vector together, and obtain the second entity prediction probability through a bidirectional long short-term memory network.

[0034] Step S4C4: Concatenate the new text vector, the first entity word vector, and the first entity category vector, and obtain the second entity prediction probability through a multi-head self-attention mechanism layer;

[0035] Step S4C5: Add the prediction probability of the second entity 1 and the prediction probability of the second entity 2, process them through a multi-layer pointer decoding layer to obtain the second entity and relation related to the first entity, and form a second triplet.

[0036] In step S3, the method further includes restrictions imposed by the triplet construction module on the first entity, the relation category label, and the second entity, specifically including the following steps:

[0037] Specifically, for the second triplet, its first entity e1, second entity e2, and relation r are obtained; if r belongs to the relationship between the fault entity class and the working entity class, and e1 and e2 overlap in the text, then the triplet is removed.

[0038] For the second set of triples, obtain (h) i ,r i ,t i ), (h j ,*,*) and (h i ,r i ,t j ), where * represents any value; if h j The position of the text contains h i At the position of the text or the same h i and r i h i The position of the text contains h j At the position of the text, (h) is retained. i ,r i ,t j ), remove (h j ,*,*) and (h i ,r i ,tj ).

[0039] In step S4, a knowledge graph for automated power dispatching operations is constructed based on the ontology structure, the first triplet, and the second triplet. Specific steps include:

[0040] Step S4A1: Obtain the first entity and the second entity from the first triplet and the second triplet to form an entity list ent = [ent1, ent2, ..., ent2]. n ];

[0041] Step S4A2: Calculate the entity similarity Similar(ent) for each pair of entities ent. i ,ent j ), for Similar(ent i ,ent j Entities with a value greater than λ are stored in different entity sets, where each set represents a specific entity, and λ is a parameter.

[0042] Step S4A3: For the different entity sets, take the first entity ent1 and use it to replace other entities in the sets that exist in the first triplet and the second triplet;

[0043] Step S4A4: Store the replaced first triplet and second triplet in the graph database.

[0044] In step S4A2, the specific formula for calculating entity similarity is as follows:

[0045]

[0046] In the formula, variable a i and b i The entities ent1 and ent are respectively. j The word vector representation.

[0047] A knowledge graph construction system for automated power dispatching includes a memory, a processor, and computer program instructions stored in the memory and executable by the processor. When the processor executes the computer program instructions, it performs the steps described above.

[0048] The system also includes the following functional modules: a data preprocessing module, a triplet extraction module, and a knowledge graph construction module for automated power dispatching operations.

[0049] The data preprocessing module is used to obtain formatted text data and domain text data to be extracted from historical data, and to form triples from the formatted text data;

[0050] The triplet extraction module is used to identify the first entity and the first entity category of the domain text data to be extracted, and input the specific relation second entity recognition module that integrates entity category information, thereby realizing the extraction of (entity, relation, entity) triples in the domain text to be extracted.

[0051] The knowledge graph construction module for power dispatch automation operations is used to store triples in a graph database according to the ontology structure. First, the entity alignment of triples is achieved according to the method proposed in step S4, and then the aligned triples are formed into a knowledge graph and stored in the graph database.

[0052] Compared with the prior art, the present invention has the following beneficial effects:

[0053] Through continuous exploration and experimentation, this invention structures and processes power grid dispatch automation software systems, hardware equipment, personnel information, historical faults, solutions, and safety measures, and uses an ontology approach to link these contents together. It is applicable to power grid dispatch automation operation scenarios, providing auxiliary decision-making for operation and maintenance personnel, effectively improving their ability to handle personnel and equipment anomalies and faults, enhancing the quality and efficiency of knowledge analysis and processing in power grid dispatch automation software and hardware, reducing the operation and maintenance costs of power grid dispatch automation software and hardware, and improving the efficiency of operation and maintenance personnel.

[0054] Furthermore, this invention transforms massive amounts of software, hardware, and personnel information into knowledge and practical productivity for power grid dispatch automation operations through the construction of a knowledge graph foundation. This enables typical application scenarios of power dispatch automation operation knowledge, such as accurate searching of business systems and hardware equipment information and generation of power dispatch automation work orders.

[0055] This invention provides rapid, convenient, accurate, and effective knowledge support for various roles in power dispatch automation equipment maintenance and technical management. It uncovers deep correlations within historical operational data, enhances operational analysis capabilities, and comprehensively improves the informatization and intelligence level of power dispatch automation operations.

[0056] This invention relates to a method and system for constructing a knowledge graph for power dispatch automation operations. The method includes: constructing a power dispatch automation operation ontology structure; processing docx text data to obtain a first triple in the form of (entity, relation category label, entity) and the domain text to be extracted from the formatted data; obtaining the character vector and position vector of each statement in the domain text to be extracted, adding the character vector and position vector to obtain a new vector representation of the text, and inputting this representation into an entity relation extraction module that integrates entity category information to obtain a second triple in the form of (entity, relation category label, entity); and constructing a power dispatch automation operation knowledge graph based on the ontology structure, the first entity triple, and the second entity triple. This invention, by preprocessing historical text of dispatch automation operations and using ontology definition and entity relation extraction methods to establish a power dispatch automation operation knowledge graph, fully explores the domain data of power dispatch automation operations and can effectively improve the utilization rate of operational knowledge in the power dispatch automation domain.

[0057] This invention links five parts in the field of power dispatch automation: software, equipment, personnel, faults, and work content. It obtains formatted data through data preprocessing and entity relationship extraction modules, which can better analyze the potential relationships in historical data and thus improve the knowledge mining capabilities of historical operation data. Attached Figure Description

[0058] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments:

[0059] Appendix Figure 1 This is a schematic diagram of the principle framework of the method described in this invention. Detailed Implementation

[0060] The technical solution of the present invention will now be described in detail with reference to the accompanying drawings.

[0061] This invention provides an overall framework for constructing a knowledge graph of automated power dispatch operations. Specifically, it includes a power dispatch automation operation ontology structure, a data preprocessing module, an entity relationship extraction module integrating entity category information, and a power dispatch automation operation knowledge graph construction module. Considering the characteristics of the power dispatch automation operation field, the ontology structure is designed from five perspectives: software, equipment, personnel, faults, and work content. Historical docx text data is preprocessed, dividing it into formatted and unformatted data. The unformatted text is formatted using a first entity and category extraction module and a second entity acquisition module, ultimately obtaining the power dispatch automation operation knowledge graph.

[0062] As shown in the figure, a method for constructing a knowledge graph of automated power dispatching operations involves preprocessing historical texts of automated dispatching operations and using ontology definition and entity relationship extraction to build a knowledge graph of automated power dispatching operations. The method includes the following steps:

[0063] Step S1: Construct the main structure of power dispatch automation operation;

[0064] Step S2: Process the data of the docx text to obtain the first triple in the form of (entity, relation category label, entity) and the domain text to be extracted from the formatted data;

[0065] Step S3: Obtain the character vector and position vector of each sentence in the domain text to be extracted, add the character vector and the position vector to obtain a new vector representation of the text, and input it into the entity relationship extraction module that integrates entity category information to obtain a second triple in the form of (entity, relationship category label, entity).

[0066] Step S4: Construct a knowledge graph for power dispatch automation operations based on the ontology structure, the first entity triplet, and the second entity triplet.

[0067] In step S1, the specific definition of the power dispatch automation operation body structure is as follows:

[0068] Let the power dispatch automation operation ontology be Power-Dispatch-Automation-OperatingOntology; Power-Dispatch-Automation-Operating Ontology = {Entity, Attribute, Relation}, where:

[0069] Step S1A1: Define Entity as a set of entity-related concepts, used to represent a collection of objective entities in the field of power dispatch automation operations. Entity = (Software, Equipment, Staff, Fault, Work). Software is a software entity, referring to the objectively existing business systems and software involved in power dispatch automation, including six subclasses: business systems, processes, databases, data tables, files, and operating systems. Equipment is a device entity, representing objectively existing physical equipment and sites. Staff is a personnel entity, including personnel, companies, departments, and roles involved in power dispatch automation operations. Fault is a fault entity, including fault causes and abnormal phenomena subclasses. Work is a work entity, including work content, work items, instruction operations, and safety measures.

[0070] Step S1A2: Define Attribute as a set of attribute-related concepts, representing the specific attribute characteristics of an Entity, including version, qualification, and risk; among them, Software and Equipment entities have version attributes, identifying the version number of the software and equipment; Staff entities have qualification attributes, including licensed operation categories, qualification certificates, and training information, identifying whether the role, company, and personnel are qualified to perform the work; risks include equipment risks and instruction risks.

[0071] Step S1A3: Define Relation, which includes all relationships between entities and attribute-related concepts except for hierarchical relationships, denoted as <Concept C1, Relation R, Concept C2>. Concept C1 and Concept C2 are contained in the entity-related concept set Entity and the attribute-related concept set Attribute.

[0072] In step S2, the docx text data is processed to obtain the first triplet in the form of (entity, relation category label, entity) and the text of the entity relation to be extracted from the formatted data. The specific steps include: Step S2A1, defining the list of fields to be extracted: NAME = {name1, name2, ..., name...} n The list of outline levels corresponding to the fields: LEVEL = {level1, level2, ..., level} n}, define the temporary dictionary temp as empty;

[0073] Step S2A2: Use the python-docx tool to parse the docx file and obtain a list of paragraphs in XML data format of the document;

[0074] Step S2A3: Read a paragraph data in XML format. If the outline level exists in LEVEL, proceed to step S2A4; otherwise, continue reading the next paragraph until the outline level of the paragraph being read exists in LEVEL or the end of the document is reached.

[0075] Step S2A4: If the temporary dictionary temp is empty, then the content of this segment is... <name i ,content i Save the key-value pairs in the form of > to a temporary dictionary, where name represents the field name and content represents the paragraph content, and then proceed to step S2A3;

[0076] Step S2A5: If the temporary dictionary contains the field "name" i And the corresponding key-value pair has the value "content". i If the outline level is the lowest, then the content of the current paragraph will be appended to the content.i Save it back to temp and execute step S2A3;

[0077] Step S2A6: If the temporary dictionary contains the field "name" i If the outline level is not the last level, then the contents of the temporary dictionary temp are saved to the formatted text database in JSON format, and the value corresponding to the field outline level being the last level is saved to the domain text database to be extracted. The field of this level and all fields of the next level are removed from the temporary dictionary, and step S2A3 is executed.

[0078] Step S2A7: Obtain the first triplet based on the ontology structure and the content of the formatted text database. The domain text to be extracted is stored in the domain text database.

[0079] In step S2, the method further includes obtaining a first triplet based on the ontology structure and the content of the formatted text database, specifically including the following steps:

[0080] Extract the dictionary temp from the formatted text database; obtain the key1 and corresponding value1 of one element in temp; retrieve the associated relation r and entity key2 set of key1 in the ontology structure; find key2 and corresponding value2 in temp; and save (value1, r, value2) as the first triplet.

[0081] In step S3, the character vector and position vector of each sentence in the domain text to be extracted are obtained. The character vector and the position vector are added to obtain a new vector representation of the text, and then input into the entity relation extraction module that integrates entity category information to obtain a second triple in the form of (entity, relation category label, entity). Specifically, this includes the following steps:

[0082] Step S3A1: Obtain the character vector and position vector of each sentence in the domain text to be extracted, add the character vector and the position vector to obtain a new vector representation of the text, and perform first entity recognition through a multi-head self-attention mechanism layer to obtain multiple first entities and first entity categories;

[0083] Step S3A2: Obtain the vector representation of the first entity category information, input the new vector of the text, the first entity, and the first entity category information into the specific relationship second entity recognition module that fuses the entity category information, and obtain a second triplet in the form of (entity, relationship category label, entity).

[0084] In step S3A1, the character vector and position vector of each sentence in the domain text to be extracted are obtained, and the character vector and position vector are added together to obtain a new vector representation of the text. The first entity recognition is performed through a multi-head self-attention mechanism layer to obtain multiple first entities and first entity categories. Specifically, the steps are as follows: Step S3B1: Obtain the character vector and position vector of each sentence. For each character in each sentence, obtain the character vector and position vector of each character.

[0085] Step S3B2: Add the word vector and the position vector together, and obtain the first entity prediction probability of the sentence through a multi-head self-attention mechanism layer;

[0086] Step S3B3: The predicted probability of the first entity is processed through a multi-layer pointer decoding layer to obtain multiple first entities and first entity categories;

[0087] In step S3A2, the vector representation of the first entity category information is obtained. The new vector of the text, the first entity, and the first entity category information are input into the specific relationship second entity recognition module that fuses the entity category information to obtain a second triple in the form of (entity, relationship category label, entity). Specifically, this includes the following steps:

[0088] Step S4C1: Encode the head entity category using the head entity category encoder to obtain the first entity category vector;

[0089] Step S4C2: Obtain the character vector of the first entity from the new vector of the text, and add the character vectors in the first dimension to obtain the word vector of the first entity;

[0090] Step S4C3: Add the new text vector, the first entity word vector, and the first entity category vector together, and obtain the second entity prediction probability through a bidirectional long short-term memory network.

[0091] Step S4C4: Concatenate the new text vector, the first entity word vector, and the first entity category vector, and obtain the second entity prediction probability through a multi-head self-attention mechanism layer;

[0092] Step S4C5: Add the prediction probability of the second entity 1 and the prediction probability of the second entity 2, process them through a multi-layer pointer decoding layer to obtain the second entity and relation related to the first entity, and form a second triplet.

[0093] In step S3, the method further includes restrictions imposed by the triplet construction module on the first entity, the relation category label, and the second entity, specifically including the following steps:

[0094] Specifically, for the second triplet, its first entity e1, second entity e2, and relation r are obtained; if r belongs to the relationship between the fault entity class and the working entity class, and e1 and e2 overlap in the text, then the triplet is removed.

[0095] For the second set of triples, obtain (h) i ,r i ,t i ), (h j ,*,*) and (h i ,r i ,t j ), where * represents any value; if h j The position of the text contains h i At the position of the text or the same h i and r i h i The position of the text contains h j At the position of the text, (h) is retained. i ,r i ,t j ), remove (h j ,*,*) and (h i ,r i ,t j ).

[0096] In step S4, a knowledge graph for automated power dispatching operations is constructed based on the ontology structure, the first triplet, and the second triplet. Specific steps include:

[0097] Step S4A1: Obtain the first entity and the second entity from the first triplet and the second triplet to form an entity list ent = [ent1, ent2, ..., ent2]. n ];

[0098] Step S4A2: Calculate the entity similarity Similar(ent) for each pair of entities ent. i ,ent j ), for Similar(ent i ,ent j Entities with a value greater than λ are stored in different entity sets, where each set represents a specific entity, and λ is a parameter.

[0099] Step S4A3: For the different entity sets, take the first entity ent1 and use it to replace other entities in the sets that exist in the first triplet and the second triplet;

[0100] Step S4A4: Store the replaced first triplet and second triplet in the graph database.

[0101] In step S4A2, the specific formula for calculating entity similarity is as follows:

[0102]

[0103] In the formula, variable a i and b i The entities ent1 and ent are respectively. j The word vector representation.

[0104] A knowledge graph construction system for automated power dispatching includes a memory, a processor, and computer program instructions stored in the memory and executable by the processor. When the processor executes the computer program instructions, it performs the steps described above.

[0105] The system also includes the following functional modules: a data preprocessing module, a triplet extraction module, and a knowledge graph construction module for automated power dispatching operations.

[0106] The data preprocessing module is used to obtain formatted text data and domain text data to be extracted from historical data, and to form triples from the formatted text data;

[0107] The triplet extraction module is used to identify the first entity and the first entity category of the domain text data to be extracted, and input the specific relation second entity recognition module that integrates entity category information, thereby realizing the extraction of (entity, relation, entity) triples in the domain text to be extracted.

[0108] The knowledge graph construction module for power dispatch automation operations is used to store triples in a graph database according to the ontology structure. First, the entity alignment of triples is achieved according to the method proposed in step S4, and then the aligned triples are formed into a knowledge graph and stored in the graph database.

[0109] This embodiment also provides a knowledge graph construction system for automated power dispatching operations, including a memory and a processor. The memory stores method instructions for steps S1 to S4, and the processor executes the method instructions stored in the memory during runtime.

[0110] Preferably, the system in this embodiment may further include the following functional modules: a data preprocessing module, a triple extraction module, and a power dispatch automation operation knowledge graph construction module.

[0111] The data preprocessing module is used to obtain formatted text data and domain text data to be extracted from historical data, and to form triples from the formatted text data; the triple extraction module is used to identify the first entity and the first entity category from the domain text data to be extracted, and input the specific relation second entity identification module that integrates entity category information, thereby realizing the extraction of (entity, relation, entity) triples from the domain text to be extracted; the power dispatch automation operation knowledge graph construction module is used to store the triples in the graph database according to the ontology structure. First, the entity alignment of the triples is realized according to the method proposed in step S4, and then the aligned triples are formed into a knowledge graph and stored in the graph database.

[0112] This embodiment, tailored to the characteristics of power dispatch automation operations, designs the ontology structure from five perspectives: software, equipment, personnel, faults, and work content. It preprocesses the input historical data to obtain formatted data triples and unformatted data. Formatted triples are obtained through a first entity and first entity category extraction module, and a second entity recognition module that integrates specific relationships with entity category information. Similar entities are aligned to form a knowledge graph, effectively enhancing the knowledge mining and intelligent analysis service capabilities of historical data in dispatch automation operations, and comprehensively improving the informatization and intelligence level of power dispatch automation operations.

[0113] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0114] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0115] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0116] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0117] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention in any other way. Any person skilled in the art may make changes or modifications to the disclosed technical content to create equivalent embodiments. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for constructing a knowledge graph for automated power dispatching operations, characterized in that: By preprocessing historical texts of automated power dispatch operations, a knowledge graph of automated power dispatch operations is established using ontology definition and entity relation extraction methods. Includes the following steps; Step S1: Construct the main structure of power dispatch automation operation; Step S2: Process the data of the docx text to obtain the first triplet and the domain text to be extracted from the formatted data. The first triplet includes the entity, relation category label, and entity form. Step S3: Obtain the character vector and position vector of each sentence in the domain text to be extracted, add the character vector and the position vector to obtain a new vector representation of the text, and input it into the entity relationship extraction module that integrates entity category information to obtain the second triplet; Step S4: Construct a knowledge graph for automated power dispatching operations based on the ontology structure, the first triplet, and the second triplet; In step S2, the docx text data is processed to obtain the first triplet and the text of the entity relations to be extracted from the formatted data. The specific steps include: Step S2A1: Define the list of fields to be extracted. List of outline levels corresponding to fields Define the temporary dictionary temp as empty; Step S2A2: Use the python-docx tool to parse the docx file and obtain a list of paragraphs in XML data format of the document; Step S2A3: Read a paragraph data in XML format. If the outline level exists in LEVEL, proceed to step S2A4; otherwise, continue reading the next paragraph until the outline level of the paragraph being read exists in LEVEL or the end of the document is reached. Step S2A4: If the temporary dictionary temp is empty, then the content of this segment is... The key-value pairs are saved to a temporary dictionary, where Indicates the field name. express For the corresponding content, proceed to step S2A3; Step S2A5, if the field is saved in the temporary dictionary, and the value of the corresponding key-value pair is , and the outline level is the last level, then the content of the current paragraph is spliced to , saved to temp again, and step S2A3 is executed. Step S2A6: If the temporary dictionary contains fields If the outline level is not the last level, then the contents of the temporary dictionary temp are saved to the formatted text database in JSON format, and the value corresponding to the field outline level being the last level is saved to the domain text database to be extracted. The field of this level and all fields of the next level are removed from the temporary dictionary, and step S2A3 is executed. Step S2A7: Obtain the first triplet based on the ontology structure and the content of the formatted text database; the domain text to be extracted is stored in the domain text database to be extracted. In step S3, the character vector and position vector of each sentence in the domain text to be extracted are obtained. The character vector and the position vector are added to obtain a new vector representation of the text, and then input into the entity relation extraction module that integrates entity category information to obtain the second triplet. Specifically, the second triplet includes the following steps: the second triplet includes an entity, a relation category label, and the form of the entity. Step S3A1: Obtain the character vector and position vector of each sentence in the domain text to be extracted, add the character vector and the position vector to obtain a new vector representation of the text, and perform first entity recognition through a multi-head self-attention mechanism layer to obtain multiple first entities and first entity categories; Step S3A2: Obtain the vector representation of the first triplet entity category information, input the new vector of the text, the first triplet entity, and the first triplet entity category information into the specific relationship second entity recognition module that fuses entity category information to obtain the second triplet.

2. The method for constructing a knowledge graph for automated power dispatching operations according to claim 1, characterized in that: In step S1, the specific definition of the power dispatch automation operation body structure is as follows: Let the power dispatch automation operation ontology be Power-Dispatch-Automation-Operating Ontology; Power-Dispatch-Automation-Operating Ontology = {Entity, Attribute, Relation}, where: Step S1A1: Define Entity as a set of entity-related concepts, used to represent a collection of objective entities in the field of power dispatch automation operations. Entity = (Software, Equipment, Staff, Fault, Work). Software is a software entity, referring to the objectively existing business systems and software involved in power dispatch automation, including six subclasses: business systems, processes, databases, data tables, files, and operating systems. Equipment is an equipment entity, representing objectively existing physical equipment and sites. Staff is a personnel entity, including personnel, companies, departments, and roles involved in power dispatch automation operations. Fault is a fault entity, including fault causes and abnormal phenomena subclasses. Work is a work entity, including work content, work items, instruction operations, and safety measures. Step S1A2: Define Attribute as a set of attribute-related concepts, representing the specific attribute characteristics of an Entity, including version, qualification, and risk; among them, Software and Equipment entities have version attributes, identifying the version number of the software and equipment; Staff entities have qualification attributes, including licensed operation categories, qualification certificates, and training information, identifying whether the role, company, and personnel are qualified to perform the work; risks include equipment risks and instruction risks. Step S1A3: Define Relation, which includes all relationships between entities and attribute-related concepts except for hierarchical relationships, denoted as <Concept C1, Relation R, Concept C2>. Concept C1 and Concept C2 are contained in the entity-related concept set Entity and the attribute-related concept set Attribute. 3.The power dispatch automation job knowledge graph construction method of claim 1, wherein: In step S2, the method further includes obtaining a first triplet based on the ontology structure and the content of the formatted text database, specifically including the following steps: Extract the dictionary `temp` from the formatted text database; retrieve one element from `temp`. and corresponding Searching within the ontology structure Relationship r and entity Set, search in temp and corresponding ,Will( ,r, Save it as the first triplet. 4.The power dispatch automation job knowledge graph construction method of claim 1, wherein: In step S3A1, the character vector and position vector of each sentence in the domain text to be extracted are obtained. The character vector and the position vector are added to obtain a new vector representation of the text. First entity recognition is performed through a multi-head self-attention mechanism layer to obtain multiple first entities and first entity categories. Specifically, the steps include: Step S3B1: Obtain the character vector and position vector of each statement. For each character in each statement, obtain the character vector and position vector of each character. Step S3B2: Add the word vector and the position vector together, and obtain the first entity prediction probability of the sentence through a multi-head self-attention mechanism layer; Step S3B3: The predicted probability of the first entity is processed through a multi-layer pointer decoding layer to obtain multiple first entities and first entity categories; In step S3A2, the vector representation of the first entity category information is obtained. The new vector of the text, the first entity, and the first entity category information are input into the second entity recognition module that fuses the entity category information to obtain the second triplet. This specifically includes the following steps: Step S3C1: Encode the head entity category using the head entity category encoder to obtain the first entity category vector; Step S3C2: Obtain the character vector of the first entity from the new vector of the text, and add the character vectors in the first dimension to obtain the first triplet entity word vector; Step S3C3: Add the new text vector, the first entity word vector, and the first entity category vector together, and obtain the second entity prediction probability through a bidirectional long short-term memory network. Step S3C4: Concatenate the new text vector, the first entity word vector, and the first entity category vector, and obtain the second entity prediction probability through a multi-head self-attention mechanism layer; Step S3C5: Add the prediction probability of the second entity 1 and the prediction probability of the second entity 2, process them through a multi-layer pointer decoding layer to obtain the second entity and relation related to the first entity, and form a second triplet.

5. The power dispatch automation job knowledge graph construction method of claim 1, wherein: In step S3, the method further includes restrictions imposed by the triplet construction module on the first entity, the relation category label, and the second entity, specifically... Includes the following steps: For the second triplet, its first entity is obtained. Second entity Relationship r; if r belongs to the relationship between the fault entity class and the working entity class, and and If there is overlap in the positions of the text, the triplet is removed. 6.The power dispatch automation job knowledge graph construction method of claim 1, wherein: In step S3A2, the specific formula for calculating entity similarity is as follows: ， where the variables and are word vector representations of the entities and respectively.

7. A power dispatch automation job knowledge graph construction system, characterized in that: The method includes a memory, a processor, and computer program instructions stored in the memory and executable by the processor. When the processor executes the computer program instructions, it performs the steps of the method as described in any one of claims 1-6. The system also includes the following functional modules: a data preprocessing module, a triplet extraction module, and a knowledge graph construction module for automated power dispatching operations. The data preprocessing module is used to obtain formatted text data and domain text data to be extracted from historical data, and to form triples from the formatted text data; The triple extraction module is used to identify the first entity and the first entity category of the domain text data to be extracted, and input the specific relation second entity recognition module that integrates entity category information, thereby realizing the extraction of triples in the domain text to be extracted. The knowledge graph construction module for power dispatch automation operations is used to store triples in a graph database according to the ontology structure.

Images