Construction and interaction method and system of electrical equipment defect knowledge graph, and medium

Abstract

The invention discloses a power equipment defect knowledge graph construction and interaction method and system and a medium, and the method comprises the steps: obtaining a tabular industry knowledge text from a power field standard, extracting nodes and edges in a knowledge graph, and extracting node attributes and edge attributes; generating'unindicated default nodes' as well as front related description attributes and rear related description attributes of the'unindicated default nodes' according to default missing information elements; constructing a power equipment defect knowledge graph according to nodes, edges, node attributes and edge attributes of the knowledge graph; and querying the power equipment defect knowledge graph through the big language model by using the query and understanding cues to perform knowledge questions and answers. The invention aims to provide a professional, accurate and detailed knowledge question and answer new method for providing knowledge expression, knowledge storage and knowledge application for the field of professional power equipment aiming at the problem that field-related common knowledge default exists in a power standard-related text of the power equipment based on the power standard text content.

Description

Technical Field

[0001] The present invention relates to the field of power equipment knowledge services, and particularly to a method, system and medium for constructing and interacting with a knowledge graph of power equipment defects. Background Art

[0002] Power equipment knowledge services collect technical standards, regulations and other materials in the equipment specialty, and use technologies such as natural language processing and knowledge graphs to digitally process equipment knowledge resources and make knowledge-based associations, so as to provide precise retrieval, intelligent question answering and other services for equipment knowledge resources. At present, the research on the construction and application of power equipment knowledge based on the text content of power standards is still in its infancy, and there are problems such as low intelligence in equipment knowledge processing, low accuracy in retrieval and question answering, and insufficient deep integration with business scenarios, and a set of general computing paradigms and engineering technology route standards have not been formed. In particular, there is a problem of "lack of domain-related common sense defaults" in the text of power standards regarding power equipment, which affects the accuracy of constructing and applying power equipment knowledge based on the text content of power standards. Summary of the Invention

[0003] The technical problem to be solved by the present invention: In view of the above problems of the prior art, a method, system and medium for constructing and interacting with a knowledge graph of power equipment defects are provided. The present invention aims to provide a new method for knowledge expression, knowledge storage and knowledge application for providing professional, accurate and detail-specific knowledge questions and answers in the professional power equipment field in view of the problem of "lack of domain-related common sense defaults" in the text of power standards regarding power equipment.

[0004] To solve the above technical problems, the technical solution adopted by the present invention is as follows: A method for constructing and interacting with a knowledge graph of power equipment defects includes the following steps: S1, obtaining a tabular industry knowledge text from power field standards, where one row in the tabular industry knowledge text is a knowledge point, and some or all of the information element columns of some knowledge points have missing information elements that are ignored because they can be deduced based on domain common sense and are in the form of blanks; S2, extracting nodes and edges, node attributes and edge attributes in the knowledge graph according to the tabular industry knowledge text, and generating "unspecified default nodes" for the missing information elements in the form of blanks to be used for placeholder in the knowledge graph, and generating pre-related description attributes and post-related description attributes for the "unspecified default nodes", where the pre-related description attributes are used to describe the domain common sense description of the association between the "unspecified default node" and the previous node, and the post-related description attributes are used to describe the domain common sense description of the association between the "unspecified default node" and the subsequent node; S3. Construct a power equipment defect knowledge graph based on the nodes and edges of the knowledge graph, node attributes, and edge attributes. S4. Use query and understanding prompts to query the power equipment defect knowledge graph through a large language model for knowledge Q&A.

[0005] Optionally, the tabularized industry knowledge text in step S1 includes the following information element columns: "equipment type", "equipment category", "component", "component category", "location", "defect description", "classification basis", "defect classification", "corresponding status quantity", "judgment basis", and "evaluation guidelines". The information elements that are missing by default due to being ignored because they can be deduced based on domain common sense refer to the information elements in "equipment type", "equipment category", "component", "component category", "location" that are missing by default due to being ignored because they can be deduced based on domain common sense, and there is a preset logical relationship between the information element columns.

[0006] Optionally, step S2 includes: S2.1. Traverse and obtain a row from the tabularized industry knowledge text as the current knowledge point. If the traversal and acquisition are successful, jump to step S2.2; otherwise, jump to step S3. S2.2. Divide the current knowledge point into nodes based on the information element columns. The content of the information element columns is used as the attributes of the nodes, generate edges between adjacent information element columns, and the logical relationship content between adjacent information element columns is used as the attributes of the edges. S2.3. Determine whether there are information elements that are missing by default and have a blank form in the current knowledge point. If there are information elements that are missing by default and have a blank form, generate an "unspecified default node" for this information element to be used as a placeholder in the knowledge graph, and generate pre-related description attributes and post-related description attributes for the "unspecified default node". The pre-related description attributes are used to describe the domain common sense description of the association between the "unspecified default node" and the previous node, and the post-related description attributes are used to describe the domain common sense description of the association between the "unspecified default node" and the next node. Jump to step S2.1.

[0007] Optionally, when generating pre-related description attributes and post-related description attributes for the "unspecified default node" in step S2.3, generating pre-related description attributes for the "unspecified default node" includes: S2.3.1A. Determine the previous node of the "unspecified default node", where the previous node refers to the previous node of the "unspecified default node" in the information element columns of "equipment type", "equipment category", "component", "component category", "location". S2.3.2A. Determine the information elements of the previous node of the "unspecified default node", the knowledge points of the row where the "unspecified default node" is located, and the knowledge points of the rows before and after the row where the "unspecified default node" is located in the tabularized industry knowledge text, and form a combined text. Use the GraphRAG technology to generate a relevant descriptive text from the combined text; S2.3.3A. Input the combined text and the relevant descriptive text to the large language model through the preset prompt words, and use the output text of the large language model as the relevant descriptive attributes before generating for the "unspecified default node".

[0008] Optionally, when generating the relevant descriptive attributes before and after for the "unspecified default node" in step S2.3, the relevant descriptive attributes after generating for the "unspecified default node" include: S2.3.1B. Determine the next node of the "unspecified default node", where the next node refers to the next node of the "unspecified default node" in the information element columns of "equipment type", "equipment category", "component", "component category", and "location"; S2.3.2B. Determine the information elements of the next node of the "unspecified default node", the knowledge points of the row where the "unspecified default node" is located, and the knowledge points of the rows before and after the row where the "unspecified default node" is located in the tabularized industry knowledge text, and form a combined text. Use the GraphRAG technology to generate a relevant descriptive text from the combined text; S2.3.3B. Input the combined text and the relevant descriptive text to the large language model through the preset prompt words, and use the output text of the large language model as the relevant descriptive attributes after generating for the "unspecified default node".

[0009] Optionally, when using the query and understanding prompt words to query the power equipment defect knowledge graph through the large language model for knowledge answering in step S4, it includes generating a question for asking the judgment basis for a specified power equipment to have a specified defect using the query and understanding prompt words, inputting the question for asking the judgment basis for a specified power equipment to have a specified defect to the large language model, and querying the power equipment defect knowledge graph through the large language model to obtain the answer of the judgment basis for a specified power equipment to have a specified defect.

[0010] Optionally, when performing knowledge Q&A by querying the power equipment defect knowledge graph through the large language model using the query and understanding prompts in step S4, it includes generating questions for defect description and defect classification of a specified power equipment with defects by combining the query and understanding prompts with the corresponding status quantities and judgment bases of the power equipment, inputting the questions for defect description and defect classification of a specified power equipment with defects into the large language model, and querying the power equipment defect knowledge graph through the large language model to obtain answers for defect description and defect classification of a specified power equipment with defects.

[0011] In addition, the present invention also provides a construction and interaction system for a power equipment defect knowledge graph, including a microprocessor and a memory connected to each other, and the microprocessor is programmed or configured to execute the construction and interaction method of the power equipment defect knowledge graph.

[0012] In addition, the present invention also provides a computer-readable storage medium, in which a computer program or instruction is stored, and the computer program or instruction is programmed or configured to execute the construction and interaction method of the power equipment defect knowledge graph through a processor.

[0013] In addition, the present invention also provides a computer program product, including a computer program or instruction, and the computer program or instruction is programmed or configured to execute the construction and interaction method of the power equipment defect knowledge graph through a processor.

[0014] Compared with the prior art, the present invention mainly has the following advantages: The present invention aims at the problem of "lack of domain-related common sense defaults" in the text content of power standards for power equipment, and provides a new method for knowledge expression, knowledge storage and knowledge application to provide professional, accurate and detail-specific knowledge Q&A for the professional power equipment field. The present invention can realize professional and accurate answers to specific detail questions of power equipment, can support power equipment operation and maintenance personnel to give professional answers to specific on-site operation and maintenance questions, and assist in the operation and maintenance of power equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a schematic diagram of the basic process of the method in the embodiment of the present invention.

[0016] Figure 2 It is a schematic diagram of the structure of the tabular industry knowledge text in the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] In order to enable those skilled in the art to better understand the technical solutions of the present invention, the technical solutions of the present invention will be further described in detail below with reference to the accompanying drawings in the embodiments of the present invention.

[0018] Regarding the problem of "lack of domain-related common sense defaults" in the text about power standards of power equipment for text content based on power standards, based on the parsed structured power equipment knowledge text with "domain-related common sense defaults" from national or enterprise power standards, construct a knowledge graph and provide a supporting query method based on a large model to support power equipment of different models, different performances, and different application scenarios to obtain equipment defect situations according to the equipment situation description. As Figure 1 As shown, this embodiment provides a method for constructing and interacting with a power equipment defect knowledge graph, including the following steps: S1, Obtain a tabular industry knowledge text from power domain standards. One row in the tabular industry knowledge text represents a knowledge point, and some or all of the information elements in some of the knowledge points have default missing information elements that are ignored because they can be deduced based on domain common sense and are in the form of emptiness. S2, Extract nodes and edges, node attributes and edge attributes in the knowledge graph according to the tabular industry knowledge text, and generate "unspecified default nodes" for the default missing information elements in the form of emptiness for placeholder in the knowledge graph, and generate pre-related description attributes and post-related description attributes for the "unspecified default nodes". The pre-related description attributes are used to describe the domain common sense description of the association between the "unspecified default node" and the previous node, and the post-related description attributes are used to describe the domain common sense description of the association between the "unspecified default node" and the subsequent node. S3, Construct a power equipment defect knowledge graph according to the nodes and edges, node attributes and edge attributes of the knowledge graph. S4, Use query and understanding prompts (Prompt) to query the power equipment defect knowledge graph through a large language model for knowledge Q&A.

[0019] In this embodiment, the tabular industry knowledge text in step S1 is a parsed complete tabular industry knowledge text, specifically the tabular form content with complete and clear rows and columns after parsing and cleaning based on the "Q / GDW 02 1 2001-2011 Power Equipment (Facilities) Defect Qualification Technical Standard" and "Q / GDW 1906-2013 Classification Standard for Transmission and Substation Primary Equipment Defects" of power domain standards. One row in this tabular industry knowledge text represents a specific knowledge point, and several columns form several information elements of this knowledge point. There are logical relationships (such as attribution relationships and other logical association relationships) between the information elements. However, some rows (knowledge points) have certain columns (a certain information element) omitted, but this omission is considered "derivable from domain common sense" within the industry and is "omissible" for people. But how to make the computer handle "domain-related common sense defaults" well is the problem to be solved in this embodiment. As Figure 2As shown in the figure, in step S1 of this embodiment, the tabularized industry knowledge text includes the following information element columns: "equipment type", "equipment category", "component", "component category", "location", "defect description", "classification basis", "defect classification", "corresponding status quantity", "judgment basis", and "evaluation guidelines". Among them, the "corresponding status quantity" is the status quantity involved in the "judgment basis". For example, for the defect of unqualified winding resistance (the "defect description" is unqualified winding resistance), the "corresponding status quantity" is the winding DC resistance. Combining the winding DC resistance with the "judgment basis" and the "classification basis" can determine the corresponding "defect classification". The information elements that are missing by default because they can be deduced based on common sense in the field refer to the information elements that are missing by default because they can be deduced based on common sense in the field in "equipment type", "equipment category", "component", "component category", and "location", and there is a preset logical relationship between the information element columns.

[0020] In this embodiment, step S2 includes: S2.1, traverse and obtain a row from the tabularized industry knowledge text as the current knowledge point. If the traversal and acquisition are successful, jump to step S2.2; otherwise, jump to step S3. S2.2, divide the current knowledge point into nodes with information element columns as the unit, and the content of the information element column is used as the attribute of the node. An edge is generated between adjacent information element columns, and the logical relationship content between adjacent information element columns is used as the attribute of the edge. S2.3, determine whether there is an information element that is missing by default and has an empty form in the current knowledge point. If there is an information element that is missing by default and has an empty form, generate an "unspecified default node" for this information element to be used as a placeholder in the knowledge graph, and generate a pre-related description attribute and a post-related description attribute for the "unspecified default node". The pre-related description attribute is used to describe the common sense description of the association field between the "unspecified default node" and the previous node, and the post-related description attribute is used to describe the common sense description of the association field between the "unspecified default node" and the subsequent node. Jump to step S2.1.

[0021] When generating the pre-related description attribute and the post-related description attribute for the "unspecified default node" in step S2.3 of this embodiment, generating the pre-related description attribute for the "unspecified default node" includes: S2.3.1A, determine the previous node of the "unspecified default node", where the previous node refers to the previous node of the "unspecified default node" in the information element columns of "equipment type", "equipment category", "component", "component category", and "location". S2.3.2A. Determine the information elements of the previous node of the "unspecified default node", the knowledge points of the row where the "unspecified default node" is located, and the knowledge points of the rows before and after the row where the "unspecified default node" is located in the tabular industry knowledge text, and combine them to form a combined text. Then, use the GraphRAG technology to generate a relevant descriptive text from the combined text. Among them, GraphRAG is a well-known existing retrieval-augmented generation (RAG) technology based on a knowledge graph. By combining a large language model and a knowledge graph, it aims to improve the performance of processing complex information questions and answers, especially in the query-focused summarization task. The core advantage of GraphRAG is its ability to utilize the structural information between entities to achieve more accurate retrieval, capture relational knowledge, and generate more accurate and context-aware responses. For details, see the literature: Microsoft Research Team. (2023). From Local to Global: A Graph RAG Approach to Query-Focused Summarization. Journal of Artificial Intelligence Research, 10(4), 123-145. https: / / doi.org / 10.1016 / j.ijar.2023.03.001; S2.3.3A. Input the combined text and the relevant descriptive text into the large language model through a preset prompt, and use the output text of the large language model as the relevant descriptive attributes generated before the "unspecified default node".

[0022] When generating the relevant descriptive attributes before and after the "unspecified default node" in step S2.3 of this embodiment, the relevant descriptive attributes generated after the "unspecified default node" include: S2.3.1B. Determine the next node of the "unspecified default node", where the next node refers to the next node of the "unspecified default node" in the information element columns of "device type", "device category", "component", "component category", and "location". S2.3.2B. Determine the information elements of the next node of the "unspecified default node", the knowledge points of the row where the "unspecified default node" is located, and the knowledge points of the rows before and after the row where the "unspecified default node" is located in the tabular industry knowledge text, and combine them to form a combined text. Then, use the GraphRAG technology to generate a relevant descriptive text from the combined text. S2.3.3B. Input the combined text and the relevant descriptive text into the large language model through a preset prompt, and use the output text of the large language model as the relevant descriptive attributes generated after the "unspecified default node".

[0023] For example, as an alternative embodiment, the prompt words in steps S2.3.3A and S2.3.3B of this embodiment are as follows: " # Answer rules - In the provided document content, there is no explanation of why the problem occurred, only the phenomenon of the problem, the location where it occurred, the device type, the device category, the component, the component category, the defect description, the classification basis, the defect classification, the corresponding status quantity, the judgment basis, and the keywords of the evaluation guidelines need to be exactly matched.

[0024] - There will be some fields called "self-unclassified" in the provided document content. Filter these fields.

[0025] - For the content mentioned in the knowledge base that is relevant to the question, it is necessary to elaborate on it completely and in detail, and when matching keywords, it needs to be exactly matched without changing the corresponding relationship in the knowledge base.

[0026] - When the content mentioned in the knowledge base is not relevant to the question, it is necessary to answer that no relevant content was found.

[0027] - It is necessary to generate a more specific description of the XXXX attribute that conforms to the content and logic described in the text for the XXXX node, which needs to be specific to the attribute attribution relationship, attribute representation, attribute details, and attribute category.

[0028] Among them, the "XXXX node" is the "unspecified default node", and the "XXXXX attribute" is the relevant description attribute before or the relevant description attribute after.

[0029] As an alternative embodiment, when using the query and understanding prompt words to query the power equipment defect knowledge graph through a large language model for knowledge Q&A in step S4 of this embodiment, it includes using the query and understanding prompt words to generate a question for asking about the judgment basis for a specified defect occurring in a specified power equipment, inputting the question for asking about the judgment basis for a specified defect occurring in a specified power equipment into the large language model, and querying the power equipment defect knowledge graph through the large language model to obtain the answer of the judgment basis for a specified defect occurring in a specified power equipment. Specifically, the query and understanding prompt words designed in this embodiment are: " # Answer rules - In the provided document content, there is no explanation of why the problem occurred, only the phenomenon of the problem, the location where it occurred, the device type, the device category, the component, the component category, the defect description, the classification basis, the defect classification, the corresponding status quantity, the judgment basis, and the keywords of the evaluation guidelines need to be exactly matched.

[0030] - There will be some fields called "self-unclassified" in the provided document content. Filter these fields.

[0031] - For the content related to the question mentioned in the knowledge base, it is necessary to elaborate on it completely and in detail, and when matching keywords, it must be an exact match without changing the corresponding relationships in the knowledge base.

[0032] - When the content mentioned in the knowledge base does not match the question, it is necessary to answer that no relevant content has been found.

[0033] - Interpret the text content of the node attributes fully.

[0034] - Please answer the question based on the content in the knowledge base. Note: Your answer must strictly follow the content in the knowledge base. If the content in the knowledge base is irrelevant to the question, then you should answer that no relevant content has been found.

[0035] - Do not answer for typos.

[0036] - If the provided document does not explain why the problem occurred, directly answer: No answer to this question has been found in the knowledge base.

[0037] - Be sure to add at the end of the answer which document this answer is from.

[0038] Please answer the question according to the answer rules for the content in the knowledge base, answering the question of "XXXXXX".

[0039] Among them, the question of "XXXXXX" is the question of asking about the judgment basis for a specified defect occurring in a specified power equipment.

[0040] To verify the construction and interaction method of the power equipment defect knowledge graph in this embodiment, when using the query and understanding prompt words to query the power equipment defect knowledge graph through the large language model for knowledge Q&A in step S4 of this embodiment, the question of asking about the judgment basis for a specified defect occurring in a specified power equipment is: "What is the judgment basis for oil leakage in the bushing of an oil-immersed transformer?"

[0041] In this embodiment, by querying the power equipment defect knowledge graph through a large language model, the answer for obtaining the judgment basis for a specified power equipment having a specified defect is: "The following are the results found in 2 standards: (1) Oil leakage on the surface of the bushing, forming oil droplets, defect classification: critical, this content is from: Q_GDW_2013 - Classification Basis for Defects of Primary Transmission and Distribution Equipment. (2) The judgment basis for oil leakage of the oil-immersed transformer bushing is that oil leakage appears on the surface of the bushing, forming oil droplets, and the defect classification is critical. This content comes from Q_GDW_02_1_2001 - 2011 - Technical Standard for Qualifying Defects of Power Equipment (Facilities)". In contrast, without being processed by the method of this embodiment, using the large model method to get the answer for the same question is: "The defect classification of oil leakage of the oil-immersed transformer bushing is divided into critical defects, serious defects and general defects according to the severity. According to the description in 'State Grid (Operation and Maintenance 4) 632 - 2005(F) 110(6)KV~500K (Technical Supervision Regulations for Oil-Immersed Transformers (Reactors))', the defects of transformers (reactors) usually refer to situations such as poor insulation or abnormal operating conditions, damage of components, etc., such as oil leakage. It is divided into critical defects, serious defects and general defects according to the severity. Therefore, as a defect of the transformer, the defect classification of oil leakage of the oil-immersed transformer bushing needs to be evaluated and classified according to specific circumstances. The defect classification of oil leakage of the oil-immersed transformer bushing is divided into critical defects, serious defects and general defects according to the severity. According to the description in 'State Grid (Operation and Maintenance 4) 632 - 2005(F) 10(66)KV~50K Technical Supervision Regulations for Oil-Immersed Transformers (Reactors))', the defects of transformers (reactors) usually refer to situations such as poor insulation or abnormal operating conditions, damage of components, etc., such as oil leakage. It is divided into critical defects, serious defects and general defects according to the severity. Therefore, as a defect of the transformer, the defect classification of oil leakage of the oil-immersed transformer bushing needs to be evaluated and classified according to specific circumstances".Through comparative analysis, it can be seen that the knowledge graph processed by the method of this embodiment has the ability of "default common sense processing". When dealing with the question of "What is the basis for judging oil leakage in the bushing of an oil-immersed transformer?", a specific default common sense will be encountered, that is, "the knowledge node of the component type of the bushing belongs to the default, which is defined as an unspecified default node in this embodiment, and this default node has the pre-related description attribute" and "post-related description attribute" texts. Among them, the text of the pre-related description attribute is "There is a component in the oil-immersed transformer called the bushing. According to the currently defined knowledge structure, the bushing should have a component type. Based on basic electrical knowledge, the bushing is a core component of the oil-immersed transformer. Without type information, it belongs to a fixed dependency relationship", and the text of the post-related description attribute is "The bushing is a core component of the oil-immersed transformer. Without type information, the bushing is a relatively independent and complete component. From the perspective of defect analysis, there is no further division of parts under the bushing, and the bushing belongs to the oil-immersed transformer". Then, through the fourth step query of this embodiment, it will successfully pass through this unspecified default node, and further query that it is the bushing itself that has the oil leakage situation and there is no further division of parts for the bushing. In contrast, when querying the knowledge graph not processed by the method of this embodiment and encountering that "the component type of the bushing is not specified", according to the logic of the knowledge structure, the query cannot continue, and a common sense answer without specific details will be returned as shown in the above comparison.

[0042] In addition, as another optional implementation manner, when using the query and understanding prompt words to query the power equipment defect knowledge graph through a large language model for knowledge Q&A in step S4 of this embodiment, it includes generating a question for the defect description and defect classification of a specified power equipment with a defect by combining the query and understanding prompt words with the corresponding status quantity and judgment basis of the power equipment, inputting the question for the defect description and defect classification of a specified power equipment with a defect into the large language model, and querying the power equipment defect knowledge graph through the large language model to obtain the answer for the defect description and defect classification of a specified power equipment with a defect.

[0043] In summary, this embodiment provides a method for constructing and interacting with a knowledge graph of power equipment defects for the phenomenon of domain-related common sense default. The input of the method in this embodiment is the parsed and complete tabular industry knowledge text. Each line of this tabular industry knowledge text represents a specific knowledge point, and several columns form several information elements of this knowledge point. There is a logical relationship between the information elements; design a special "unspecified default node" in the knowledge graph to represent the "default knowledge element" and "occupy a position" at the corresponding position in the knowledge graph; design the attribute types of the edges connecting the "unspecified default nodes", and add "pre-related description attributes" and "post-related description attributes" to the description of the edge attribute types, and use targeted prompt words to design the implementation of text generation for "pre-related description attributes" and "post-related description attributes" based on RAG and GraphRAG; design query and understanding prompt words, use these prompt words with the large model to query the knowledge graph, and answer questions according to the query results. The method in this embodiment can utilize the semantic understanding ability of the existing large model and the basic operation methods of the knowledge graph consensus, and for the "domain common sense default", set an unspecified default node, pre-related description attributes, post-related description attributes, and an attribute text generation method including domain-specific domain common sense default at the corresponding position in the knowledge graph, and then match the corresponding query method to overall solve the "domain common sense default" problem. The method in this embodiment can achieve professional and accurate answers to specific detail questions of power equipment, support power equipment maintenance personnel to give professional answers to specific problems on the maintenance site, and assist power equipment maintenance.

[0044] In addition, this embodiment also provides a system for constructing and interacting with a knowledge graph of power equipment defects, including a microprocessor and a memory connected to each other. The microprocessor is programmed or configured to execute the method for constructing and interacting with the knowledge graph of power equipment defects.

[0045] In addition, this embodiment also provides a computer-readable storage medium, in which a computer program or instruction is stored. The computer program or instruction is programmed or configured to execute the method for constructing and interacting with the knowledge graph of power equipment defects through a processor.

[0046] In addition, this embodiment also provides a computer program product, including a computer program or instruction. The computer program or instruction is programmed or configured to execute the method for constructing and interacting with the knowledge graph of power equipment defects through a processor.

[0047] Those skilled in the art should understand that the technical solutions provided by the embodiments of the present application can be in the form of methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-readable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The present application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in the process Figure 1 one process or multiple processes and / or blocks Figure 1 These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including instruction means, and the instruction means implements the functions specified in the process Figure 1 one process or multiple processes and / or blocks Figure 1 These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process. Therefore, the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in the process Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks

[0048] The above is only the preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions within the idea of the present invention belong to the protection scope of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. A method for constructing and interacting with a knowledge graph of power equipment defects, characterized in that: The steps include: S1, obtaining a tabular industry knowledge text derived from the electric power field standard, wherein one line in the tabular industry knowledge text is one knowledge point, and some information element columns of some or all knowledge points contain default missing information elements that are ignored and left empty because they can be derived based on domain common sense; S2, extracting nodes and edges, node attributes and edge attributes in the knowledge graph according to the tabular industry knowledge text, and generating "unspecified default nodes" for the default missing information elements in the form of empty to occupy the place in the knowledge graph, and generating the previous related description attribute and the next related description attribute for the "unspecified default node", wherein the previous related description attribute is used to describe the domain common sense description of the association between the "unspecified default node" and the previous node, and the next related description attribute is used to describe the domain common sense description of the association between the "unspecified default node" and the next node; S3, constructing a knowledge graph of power equipment defects based on the nodes and edges, node attributes and edge attributes of the knowledge graph; S4, uses query and understanding prompt words to query the power equipment defect knowledge graph through a large language model to perform knowledge question answering.

2. The method for constructing and interacting with the knowledge graph of power equipment defects according to claim 1 is characterized in that: The tabular industry knowledge text in step S1 includes the following information element columns: "equipment type", "equipment category", "component", "component category", "location", "defect description", "classification basis" and "defect classification", "corresponding state quantity", "judgment basis" and "evaluation guidelines", and the information elements that are ignored and missing due to being deduced based on domain common sense refer to the information elements in "equipment type", "equipment category", "component", "component category" and "location" that are ignored and missing due to being deduced based on domain common sense, and the information element columns contain preset logical relationships.

3. The method for constructing and interacting with the knowledge graph of power equipment defects according to claim 2 is characterized in that: Step S2 includes: S2.1, traverse and obtain a row from the tabular industry knowledge text as the current knowledge point. If the traversal is successful, jump to step S2.2; otherwise, jump to step S3; S2.2, divide the current knowledge point into nodes based on information element columns, use the content of the information element columns as the attributes of the nodes, generate edges between adjacent information element columns, and use the logical relationship between adjacent information element columns as the attributes of the edges; S2.3, determine whether there is a default missing information element in the form of an empty space in the current knowledge point. If there is a default missing information element in the form of an empty space, generate an "unspecified default node" for the information element to occupy a place in the knowledge graph, and generate a front-related description attribute and a back-related description attribute for the "unspecified default node", wherein the front-related description attribute is used to describe the domain common sense description of the association between the "unspecified default node" and the previous node, and the back-related description attribute is used to describe the domain common sense description of the association between the "unspecified default node" and the next node; jump to step S2.

1.

4. The method for constructing and interacting with the knowledge graph of power equipment defects according to claim 3 is characterized in that: When generating the front-related description attribute and the back-related description attribute for the "unspecified default node" in step S2.3, generating the front-related description attribute for the "unspecified default node" includes: S2.3.1A, determining the previous node of the "unspecified default node", wherein the previous node refers to the previous node of the "unspecified default node" in the information element columns of "equipment type", "equipment category", "component", "component category", and "location"; S2.3.2A, the tabular industry knowledge text is used to determine the information element of the previous node of the "unspecified default node", the knowledge point of the row where the "unspecified default node" is located, and the knowledge points of the rows before and after the row where the "unspecified default node" is located to form a combined text, and the combined text is used to generate a relevant description text using GraphRAG technology; S2.3.3A, the combined text and the related description text are input into the large language model through the preset prompt words, and the output text of the large language model is used as the related description attribute before generating for the "unspecified default node".

5. The method for constructing and interacting with the knowledge graph of power equipment defects according to claim 3 is characterized in that: When generating the front-related description attribute and the back-related description attribute for the "unspecified default node" in step S2.3, generating the back-related description attribute for the "unspecified default node" includes: S2.3.1B, determine the next node of the "unspecified default node", wherein the next node refers to the next node of the "unspecified default node" in the information element columns of "equipment type", "equipment category", "component", "component category", and "location"; S2.3.2B, the tabular industry knowledge text is used to determine the information element of the node after the "unspecified default node", the knowledge point of the row where the "unspecified default node" is located, and the knowledge points of the rows before and after the row where the "unspecified default node" is located to form a combined text, and the combined text is used to generate a relevant description text using the GraphRAG technology; S2.3.3B, the combined text and the related description text are input into the large language model through the preset prompt words, and the output text of the large language model is used as the related description attribute generated for the "unspecified default node".

6. The construction and interaction method of the power equipment defect knowledge graph according to claim 4 or 5 is characterized in that: In step S4, when querying and understanding prompt words through the large language model to query the power equipment defect knowledge graph to perform knowledge question and answer, it includes using the query and understanding prompt words to generate questions for inquiring about the basis for judging the occurrence of specified defects in the specified power equipment, inputting the questions for inquiring about the basis for judging the occurrence of specified defects in the specified power equipment into the large language model, and querying the power equipment defect knowledge graph through the large language model to obtain the answer to the basis for judging the occurrence of the specified defects in the specified power equipment.

7. The method for constructing and interacting with the knowledge graph of power equipment defects according to claim 4 or 5, characterized in that: In step S4, when querying and understanding prompt words through a large language model to query the power equipment defect knowledge graph for knowledge question and answering, it includes using the query and understanding prompt words in combination with the corresponding state quantity and judgment basis of the power equipment to generate questions for inquiring about the defect description and defect classification of the specified power equipment, inputting the questions for inquiring about the defect description and defect classification of the specified power equipment into the large language model, and querying the power equipment defect knowledge graph through the large language model to obtain answers to the questions about the defect description and defect classification of the specified power equipment.

8. A construction and interactive system for a knowledge graph of power equipment defects, comprising a microprocessor and a memory connected to each other, characterized in that: The microprocessor is programmed or configured to execute the method for constructing and interacting with the power equipment defect knowledge graph as described in any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program or instruction stored therein, characterized in that: The computer program or instruction is programmed or configured to execute the method for constructing and interacting with the power equipment defect knowledge graph described in any one of claims 1 to 7 through a processor.

10. A computer program product comprising a computer program or instructions, characterized in that The computer program or instruction is programmed or configured to execute the method for constructing and interacting with the power equipment defect knowledge graph described in any one of claims 1 to 7 through a processor.

Images