User side power transformation and distribution station operation and maintenance service scheme intelligent generation system and method

Abstract

The invention relates to a user side power transformation and distribution station operation and maintenance service scheme intelligent generation system and method, and belongs to the technical field of electric power operation and maintenance. According to the intelligent generation system and method for the operation and maintenance service scheme of the user-side power transformation and distribution station, firstly, professional documents in the field of operation and maintenance of the power transformation and distribution station are subjected to vectorization processing, and a vector database is generated; a keyword is obtained based on the interactive human-computer interface; then, according to the keyword, re-forming a vectorized query problem; operation and maintenance service scheme knowledge points corresponding to the vectorization query problem are retrieved in a vector database; and a corresponding operation and maintenance service scheme is generated and output, so that the efficiency and the accuracy of operation and maintenance service are effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of power operation and maintenance technology, in particular to the field of substation operation and maintenance technology, and specifically refers to a system and method for intelligently generating an operation and maintenance service solution for a user-side substation. Background Art

[0002] User-side substations generally refer to user-side substations and distribution stations that are not under the jurisdiction of power companies, also known as "special substations". This type of substation includes various voltage levels, from 220kV to 10kV, with different structures and numerous equipment, including transformers, busbars, circuit breakers, disconnectors, capacitors, reactors, power cables, grounding devices, mutual inductors, etc.

[0003] The operation and maintenance of substations includes operation inspection and daily maintenance inspection. With the rapid development of the power system, the operation and maintenance of substations has become more and more complicated. Both daily maintenance and equipment inspection require the formulation of detailed operation and maintenance plans, which must be approved by professionals before implementation.

[0004] The preparation of operation and maintenance plans relies on the experience and judgment of staff, which is inefficient and prone to errors.

[0005] Therefore, how to provide an intelligent system to assist operation and maintenance personnel in quickly writing operation and maintenance repair plans and improving the efficiency and accuracy of operation and maintenance services has become an urgent problem to be solved in this field. Summary of the invention

[0006] The purpose of the present invention is to overcome the shortcomings of the above-mentioned prior art and provide a user-side substation operation and maintenance service plan intelligent generation system and method, which realizes vectorized storage of professional documents in the field of substation operation and maintenance, extracts the query intention of operation and maintenance personnel based on an interactive human-computer interface, retrieves relevant field knowledge, generates and outputs corresponding operation and maintenance service plans, thereby improving the efficiency and accuracy of operation and maintenance services.

[0007] In order to achieve the above-mentioned purpose, the user-side substation operation and maintenance service solution intelligent generation system of the present invention includes:

[0008] The domain knowledge vectorization processing module is used to vectorize professional documents related to substation operation and maintenance, extract key information, and establish a vector database for substation operation and maintenance;

[0009] The query preprocessing module is used to extract entities from the content input by the operation and maintenance personnel through the human-computer interface to obtain the keywords and basic elements required to generate the operation and maintenance plan;

[0010] A knowledge retrieval module, which is used to reorganize the query keyword entities into vectorized query questions and retrieve the operation and maintenance service solution knowledge points corresponding to the vectorized query questions from the vector database;

[0011] A solution generation module is used to generate prompt words and the keywords for inputting a large language model according to the retrieved knowledge points, and to generate an operation and maintenance service solution using the large language model;

[0012] The solution output module is used to output the generated operation and maintenance service solution to the operation and maintenance personnel in a structured format.

[0013] In the intelligent generation system of the user-side substation operation and maintenance service solution, the domain knowledge vectorization processing module specifically includes:

[0014] The file reading unit is used to read domain documents in various formats, including operation and maintenance management regulations and operation and maintenance procedures for various equipment. The document formats include HTML, docx, markdown, and PDF.

[0015] Format conversion unit, used to convert domain documents in various formats into txt text format;

[0016] Data cleaning unit, which is used to remove tags and specific elements in documents to ensure clean text and reduce text noise;

[0017] A text segmentation unit is used to segment the text by referring to each separator symbol and to segment the text into different semantic blocks;

[0018] The text vectorization unit is used to vectorize the divided text using one or more of the M3E, BGE, and BCE methods, and save it into the vector database.

[0019] In the user-side substation operation and maintenance service solution intelligent generation system, the query preprocessing module specifically includes:

[0020] An interactive interface unit is used to provide an interactive interface for operation and maintenance personnel to input the information necessary to form an operation and maintenance plan;

[0021] The keyword extraction unit is used to extract a keyword model according to the content input by the user through the interactive interface unit, the keyword model is represented by a json object, and keywords are generated based on the keyword model.

[0022] In the intelligent generation system of the user-side substation operation and maintenance service solution, the knowledge retrieval module specifically includes:

[0023] A query rewriting unit, for generating corresponding query content for each type of operation and maintenance task of each type of equipment according to the keywords provided by the query preprocessing module;

[0024] The query question vectorization unit is used to vectorize multi-channel query contents by using one or more of the M3E, BGE, and BCE methods, organize them into vectorized query questions, and retrieve the operation and maintenance service solution knowledge points corresponding to the vectorized query questions from the vector database.

[0025] In the intelligent generation system of the operation and maintenance service plan of the user-side substation, the plan generation module specifically includes:

[0026] A query result recall unit is used to recall several segmented operation and maintenance service solution knowledge points with high similarity for each query;

[0027] A scheme generation prompt word unit, used for forming a scheme generation prompt word according to the query result, keywords and a preset prompt word template;

[0028] The large language model generation unit is used to input prompt words into the large language model to generate an operation and maintenance service plan.

[0029] The present invention also provides a method for intelligently generating an operation and maintenance service plan for a user-side substation, which comprises the following steps:

[0030] (1) The domain knowledge vectorization processing module vectorizes professional documents related to substation operation and maintenance, extracts key information, and establishes a vector database for substation operation and maintenance;

[0031] (2) The query preprocessing module extracts entities from the content input by the operation and maintenance personnel through the human-computer interface to obtain the keywords and basic elements required to generate the operation and maintenance plan;

[0032] (3) The knowledge retrieval module reorganizes the queried keyword entities into vectorized query questions, and retrieves operation and maintenance service solution knowledge points corresponding to the vectorized query questions from the vector database;

[0033] (4) The solution generation module generates prompt words and the keywords for inputting the large language model based on the retrieved knowledge points, and generates an operation and maintenance service solution using the large language model;

[0034] (5) The solution output module outputs the generated operation and maintenance service solution to the operation and maintenance personnel in a structured format.

[0035] In the method for intelligently generating an operation and maintenance service plan for a user-side substation, the step (1) specifically includes the following steps:

[0036] (11) The file reading unit reads domain documents in various formats, including operation and maintenance management regulations and operation and maintenance procedures for various equipment. The document formats include HTML, docx, markdown, and PDF.

[0037] (12) The format conversion unit converts domain documents of various formats into txt text format;

[0038] (13) The data cleaning unit removes the tags and specific elements in the document to ensure that the text is clean and reduce the noise of the text;

[0039] (14) the text segmentation unit segments the text with reference to each separator symbol, and divides the text into different semantic blocks;

[0040] (15) The text vectorization unit uses one or more of the M3E, BGE, and BCE methods to vectorize the divided text and save it into the vector database.

[0041] In the method for intelligently generating an operation and maintenance service plan for a user-side substation, the step (2) specifically includes the following steps:

[0042] (21) The interactive interface unit provides an interactive interface for operation and maintenance personnel to input information necessary to form an operation and maintenance plan;

[0043] (22) The keyword extraction unit extracts a keyword model according to the user input content. The keyword model is represented by a json object, and keywords are generated based on the keyword model.

[0044] In the method for intelligently generating an operation and maintenance service plan for a user-side substation, the step (3) specifically includes the following steps:

[0045] (31) the query rewriting unit generates corresponding query content for each operation and maintenance task of each type of equipment according to the keywords provided by the query preprocessing module;

[0046] (32) The query question vectorization unit uses one or more of the M3E, BGE, and BCE methods to vectorize the multi-channel query content, organize it into a vectorized query question, and retrieve the operation and maintenance service solution knowledge points corresponding to the vectorized query question from the vector database.

[0047] In the method for intelligently generating an operation and maintenance service plan for a user-side substation, the step (4) specifically includes the following steps:

[0048] (41) The query result recall unit recalls several segmented operation and maintenance service solution knowledge points with high similarity for each query;

[0049] (42) a scheme generating prompt word unit generates a scheme generating prompt word according to the query result, the keyword and the preset prompt word template;

[0050] The prompt words are input into the large language model through the large language model generation unit to generate an operation and maintenance service plan.

[0051] The user-side substation operation and maintenance service plan intelligent generation system and method of the present invention firstly vectorizes professional documents in the field of substation operation and maintenance to generate a vector database; then obtains keywords based on an interactive human-computer interface; then reformulates vectorized query questions based on the keywords; retrieves operation and maintenance service plan knowledge points corresponding to the vectorized query questions in the vector database; and then generates and outputs the corresponding operation and maintenance service plan, thereby effectively improving the efficiency and accuracy of operation and maintenance services. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 A technical schematic diagram of the intelligent generation system of the user-side substation operation and maintenance service solution of the present invention;

[0053] Figure 2 It is a flow chart of the method for intelligently generating the operation and maintenance service scheme of the user-side substation of the present invention. DETAILED DESCRIPTION

[0054] In order to more clearly understand the technical content of the present invention, the following embodiments are given in detail.

[0055] See also Figure 1 As shown, it is a technical principle diagram of the intelligent generation system of the user-side substation operation and maintenance service solution of the present invention.

[0056] In one embodiment, the user-side substation operation and maintenance service solution intelligent generation system, such as Figure 1 As shown, including:

[0057] The domain knowledge vectorization processing module is used to vectorize professional documents related to substation operation and maintenance, extract key information, and establish a vector database for substation operation and maintenance;

[0058] The query preprocessing module is used to extract entities from the content input by the operation and maintenance personnel through the human-computer interface to obtain the keywords and basic elements required to generate the operation and maintenance plan;

[0059] A knowledge retrieval module, which is used to reorganize the query keyword entities into vectorized query questions and retrieve the operation and maintenance service solution knowledge points corresponding to the vectorized query questions from the vector database;

[0060] A solution generation module is used to generate prompt words and the keywords for inputting a large language model according to the retrieved knowledge points, and to generate an operation and maintenance service solution using the large language model;

[0061] The solution output module is used to output the generated operation and maintenance service solution to the operation and maintenance personnel in a structured format.

[0062] In a more preferred embodiment, the domain knowledge vectorization processing module specifically includes:

[0063] The file reading unit is used to read domain documents in various formats, including operation and maintenance management regulations and operation and maintenance procedures for various equipment. The document formats include HTML, docx, markdown, and PDF.

[0064] Format conversion unit, used to convert domain documents in various formats into txt text format;

[0065] Data cleaning unit, which is used to remove tags and specific elements in documents to ensure clean text and reduce text noise;

[0066] A text segmentation unit is used to segment the text by referring to each separator symbol and to segment the text into different semantic blocks;

[0067] The text vectorization unit is used to vectorize the divided text using one or more of the M3E, BGE, and BCE methods, and save it into the vector database.

[0068] In a further preferred embodiment, the query preprocessing module specifically includes:

[0069] An interactive interface unit is used to provide an interactive interface for operation and maintenance personnel to input the information necessary to form an operation and maintenance plan;

[0070] The keyword extraction unit is used to extract a keyword model according to the content input by the user through the interactive interface unit, the keyword model is represented by a json object, and keywords are generated based on the keyword model.

[0071] In a further preferred embodiment, the knowledge retrieval module specifically includes:

[0072] A query rewriting unit, for generating corresponding query content for each type of operation and maintenance task of each type of equipment according to the keywords provided by the query preprocessing module;

[0073] The query question vectorization unit is used to vectorize multi-channel query contents by using one or more of the M3E, BGE, and BCE methods, organize them into vectorized query questions, and retrieve the operation and maintenance service solution knowledge points corresponding to the vectorized query questions from the vector database.

[0074] In a more preferred embodiment, the solution generation module specifically includes:

[0075] A query result recall unit is used to recall several segmented operation and maintenance service solution knowledge points with high similarity for each query;

[0076] A scheme generation prompt word unit, used for forming a scheme generation prompt word according to the query result, keywords and a preset prompt word template;

[0077] The large language model generation unit is used to input prompt words into the large language model to generate an operation and maintenance service plan.

[0078] The present invention also provides a method for intelligently generating an operation and maintenance service plan for a user-side substation. Figure 2 As shown, the following steps are included:

[0079] (1) The domain knowledge vectorization processing module vectorizes professional documents related to substation operation and maintenance, extracts key information, and establishes a vector database for substation operation and maintenance;

[0080] (2) The query preprocessing module extracts entities from the content input by the operation and maintenance personnel through the human-computer interface to obtain the keywords and basic elements required to generate the operation and maintenance plan;

[0081] (3) The knowledge retrieval module reorganizes the queried keyword entities into vectorized query questions, and retrieves operation and maintenance service solution knowledge points corresponding to the vectorized query questions from the vector database;

[0082] (4) The solution generation module generates prompt words and the keywords for inputting the large language model based on the retrieved knowledge points, and generates an operation and maintenance service solution using the large language model;

[0083] (5) The solution output module outputs the generated operation and maintenance service solution to the operation and maintenance personnel in a structured format.

[0084] In a more preferred embodiment, the step (1) specifically comprises the following steps:

[0085] (11) The file reading unit reads domain documents in various formats, including operation and maintenance management regulations and operation and maintenance procedures for various equipment. The document formats include HTML, docx, markdown, and PDF.

[0086] (12) The format conversion unit converts domain documents of various formats into txt text format;

[0087] (13) The data cleaning unit removes the tags and specific elements in the document to ensure that the text is clean and reduce the noise of the text;

[0088] (14) the text segmentation unit segments the text with reference to each separator symbol, and divides the text into different semantic blocks;

[0089] (15) The text vectorization unit uses one or more of the M3E, BGE, and BCE methods to vectorize the divided text and save it into the vector database.

[0090] In a further preferred embodiment, the step (2) specifically comprises the following steps:

[0091] (21) The interactive interface unit provides an interactive interface for operation and maintenance personnel to input information necessary to form an operation and maintenance plan;

[0092] (22) The keyword extraction unit extracts a keyword model according to the user input content. The keyword model is represented by a json object, and keywords are generated based on the keyword model.

[0093] In a further preferred embodiment, the step (3) specifically comprises the following steps:

[0094] (31) the query rewriting unit generates corresponding query content for each operation and maintenance task of each type of equipment according to the keywords provided by the query preprocessing module;

[0095] (32) The query question vectorization unit uses one or more of the M3E, BGE, and BCE methods to vectorize the multi-channel query content, organize it into a vectorized query question, and retrieve the operation and maintenance service solution knowledge points corresponding to the vectorized query question from the vector database.

[0096] In a more preferred embodiment, the step (4) specifically comprises the following steps:

[0097] (41) The query result recall unit recalls several segmented operation and maintenance service solution knowledge points with high similarity for each query;

[0098] (42) a scheme generating prompt word unit generates a scheme generating prompt word according to the query result, the keyword and the preset prompt word template;

[0099] The prompt words are input into the large language model through the large language model generation unit to generate an operation and maintenance service plan.

[0100] In practical applications, the user-side substation operation and maintenance service plan intelligent generation system of the present invention realizes vectorized storage of professional documents in the field of substation operation and maintenance, obtains the query intention of operation and maintenance personnel based on interactive dialogue, retrieves relevant field knowledge, and generates and outputs corresponding operation and maintenance service plans.

[0101] System technical schematic diagram Figure 1As shown in the figure, it includes domain knowledge vectorization module, query preprocessing module, knowledge retrieval module, solution generation module, solution output module, etc. In addition, the solution includes data resources such as vector database, keyword model and document template. The following is a detailed introduction to each application module and data resource:

[0102] 1. Application Module

[0103] (1) Domain knowledge vectorization: Professional documents related to substation operation and maintenance are vectorized to extract key information and establish a vector database for substation operation and maintenance.

[0104] (2) Query preprocessing: The operation and maintenance personnel are allowed to input information through an interactive human-computer interface. The system then extracts entities from the input content of the operation and maintenance personnel to obtain the keywords and basic elements necessary for generating an operation and maintenance plan.

[0105] (3) Knowledge retrieval: Reorganize the query questions based on the query keyword entities, vectorize the query questions, and retrieve relevant operation and maintenance service solution knowledge points from the vector database.

[0106] (4) Solution generation: Combine the retrieved knowledge points to generate prompt words for large language model input, and use the large language model to generate detailed operation and maintenance service solutions.

[0107] (5) Solution output: The generated operation and maintenance service solution is output in a structured format to facilitate execution and recording by operation and maintenance personnel.

[0108] 2. Data resources

[0109] (1) Vector database: a special database used to store domain knowledge. The basic principle of a vector database is to convert text into vectors and then store the vectors in the database. When a user enters a question, the question is converted into a vector, and then the database is searched for the most similar vector and context, and finally the text is returned to the user.

[0110] (2) Keywords: used to store some key elements required for the operation and maintenance plan, such as company name, substation name, maintenance time, task type, equipment name and number, project manager, etc.

[0111] (3) Document template: a document template used to generate formatting schemes. The formats include docx, markdown, html, etc.

[0112] In each application module, the domain knowledge vectorization module vectorizes the professional documents related to the operation and maintenance of the substation, extracts key information, and establishes a vector database for the operation and maintenance of the substation. It specifically includes the following units:

[0113] (1) Reading a file

[0114] Read domain documents in various formats, including operation and maintenance management regulations, operation and maintenance procedures and rules for various equipment, etc.; document formats are not limited to HTML, docx, markdown, pdf, etc.

[0115] (2) Format Conversion

[0116] Convert various document formats to text format (txt).

[0117] (3) Data cleaning

[0118] If the document format is HTML or markdown, you need to delete tags or specific elements to ensure the "purity" of the text and reduce text noise.

[0119] (4) Text segmentation

[0120] Split text into different semantic blocks by period, semicolon, question mark, and paragraph.

[0121] (5) Text vectorization

[0122] Vectorize the divided text and save it in the vector library. The vectorization method can be M3E, BGE, BCE and other methods.

[0123] The query preprocessing module specifically includes the following units:

[0124] (1)Interactive interface

[0125] 1) User Interface (UI) Design: Design an intuitive user interface that allows operation and maintenance personnel to input the information necessary to generate an operation and maintenance plan.

[0126] 2) Input content on the interface: company name, substation name, operation and maintenance start and end time, task type (such as overall maintenance, partial maintenance, routine inspection and testing, non-stop maintenance), equipment type (transformer, busbar, circuit breaker, disconnector, capacitor, reactor, power cable, grounding device, mutual inductor, etc.), equipment name, equipment operation and maintenance work content (sub-category operation and maintenance content for each type of equipment), project leader, etc. One operation and maintenance task is only for one substation of one company, and can include multiple equipment in the substation, and each equipment can have multiple equipment operation and maintenance work contents.

[0127] (2) Keyword extraction

[0128] The keyword model is extracted based on the user input content. The keywords include company name, substation name, operation and maintenance start and end time, task type, equipment type, equipment name, equipment operation and maintenance content, project leader, etc.

[0129] The keyword model can be represented by a JSON object. The following is an example model of a JSON object that describes an operation and maintenance task data object:

[0130]

[0131]

[0132] The knowledge retrieval module specifically includes the following units:

[0133] (1) Query rewriting

[0134] According to the keywords provided by the query preprocessing module, the corresponding query content is generated for each operation and maintenance task of each type of equipment. The query content includes "equipment type + operation and maintenance content", such as "35kV oil circuit breaker electromagnetic operating mechanism maintenance", "35kV oil circuit breaker mechanism secondary circuit maintenance".

[0135] (2) Query question vectorization

[0136] The vectorization method for multi-channel query content is the same as that for professional data text. The multi-channel query question vectorization data is input into the vector database for query and the query results are returned.

[0137] The solution generation module specifically includes the following units:

[0138] (1) Recall of query results

[0139] For each query, recall the top K segments with high similarity.

[0140] (2) Prompt words for solution generation

[0141] Generate prompt words based on the query results, keywords and preset prompt word templates. A typical prompt word template is as follows:

[0142] #Role: Substation operation and maintenance plan writing expert

[0143] #Task Description:

[0144] -As an expert in substation operation and maintenance plan writing, generate operation and maintenance plans based on background knowledge and comply with answer requirements

[0145] #Background knowledge:

[0146] -Operation and maintenance tasks: {{maintenanceTask}}

[0147] -{{context}}

[0148] #Skill

[0149] ##Skill 1: Knowledge retrieval and application

[0150] -Use the information in {{context}} to find relevant information quickly and accurately.

[0151] - Extract and integrate key information from the knowledge base based on user questions.

[0152] - For complex questions, be able to combine content from multiple documents to provide comprehensive answers.

[0153] ##Skill 2: Information integration and presentation

[0154] -Integrate information from multiple sources into a coherent answer.

[0155] -Present information in a clear and concise manner to ensure users can understand it quickly.

[0156] - Use appropriate paragraphs and headings in your answer to make it structured and easy to read.

[0157] ##Skill 3: Professional Answers

[0158] - Provide answers based on expertise and experience.

[0159] - Ensure your answers are professional and accurate, and avoid providing misleading or incomplete information.

[0160] -Cite specific paragraphs or data in {{context}} to enhance the credibility of your answer.

[0161] #Answer request

[0162] -Only answer questions related to the content in {{context}}.

[0163] -The output answer includes the following chapters: 1. Preparation basis; 2. Work content; 3. Operation and maintenance tasks; 4. Organizational measures; 5. Safety measures; 6. Technical measures; 7. Progress guarantee measures

[0164] - Ensure all answers are based on the information in {{context}} and do not introduce external unverified content.

[0165] -The output content is in markdown format.

[0166] - Be professional and accurate and avoid providing misleading or incomplete information.

[0167] - Do not answer questions that are outside the scope of the knowledge base, if additional information is required, clearly inform the user and suggest they provide more information.

[0168] Among them, {{maintenanceTask}} is the embedded operation and maintenance task keyword data object model; {{context}} is the content returned by the vector database search.

[0169] (3) Generate prompt words through a large language model

[0170] Input the prompt words into the large language model to generate an operation and maintenance plan.

[0171] The solution output module is specifically used for:

[0172] The output of the large language model is generally a marked text solution, such as markdown. The solution output module forms a formatted document according to the preset document template. The document template defines the display mode of the document format, such as the title font, body font, alignment, etc.

[0173] The technical advantages of the present invention are:

[0174] (1) Intelligence: Utilize large language models and search-enhanced query technology to realize intelligent generation of operation and maintenance service solutions.

[0175] (2) Efficiency: Quickly respond to queries from operation and maintenance personnel to improve the efficiency of operation and maintenance services.

[0176] (3) Accuracy: Based on the professional knowledge base, ensure the accuracy of the generated operation and maintenance service plan.

[0177] (4) Scalability: The system can be updated and expanded based on new professional documents and domain knowledge.

[0178] The following is a specific example describing how the system responds to a query generated for a substation operation and maintenance service solution:

[0179] 1. Vectorization of domain knowledge: 1) Read in pdf documents such as the operation and maintenance management regulations, operation and maintenance procedures, and operation and maintenance details of various equipment in the user's substation distribution equipment; 2) Convert the pdf documents into txt text format; 3) Split the text into different semantic blocks by period, semicolon, question mark, and paragraph; 3) Use the M3E algorithm to vectorize the text semantic blocks and save them in the vector database chroma.

[0180] 2. The operation and maintenance personnel enter the content through the interactive interface:

[0181] User company: Company A

[0182] Substation: 110kV substation

[0183] Operation and maintenance start time: 2024-11-11 13:30:00

[0184] Operation and maintenance end time: 2024-11-11 16:30:00

[0185] Operation and maintenance equipment:

[0186] (1) 35kV oil circuit breaker, 3501 switch, operation and maintenance content: electromagnetic operating mechanism maintenance, mechanism secondary circuit maintenance;

[0187] (2) 35kV oil circuit breaker, 3502 switch, operation and maintenance content: electromagnetic operating mechanism maintenance, mechanism secondary circuit maintenance;

[0188] 2. Query preprocessing: The system constructs a JSON data object based on user input, extracts all keywords, and establishes an operation and maintenance task data object model.

[0189] 3. Knowledge retrieval: The system rewrites the query content according to the keyword model and generates the following multi-way query:

[0190] (1) Maintenance of electromagnetic operating mechanism of 35kV oil circuit breaker

[0191] (2) Maintenance of secondary circuit of 35kV oil circuit breaker mechanism

[0192] For each query, the cosine similarity retrieval strategy is used to retrieve the K records with the highest scores from the vector database; K is a preset value, which can be 10.

[0193] 4. Solution generation: The prompt words are dynamically generated by combining the multi-channel query results returned by the query with the keyword object model and the preset prompt word template. The prompt words are input into the large language model to generate the operation and maintenance plan. The large language model uses Alibaba's Qwen 2.5 model, and the returned solution format is markdown.

[0194] 5. Solution output: Based on the markdown format text, apply the preset document template definition to generate a formatted PDF document of the operation and maintenance solution.

[0195] The user-side substation operation and maintenance service plan intelligent generation system and method of the present invention firstly vectorizes professional documents in the field of substation operation and maintenance to generate a vector database; then obtains keywords based on an interactive human-computer interface; then reformulates vectorized query questions based on the keywords; retrieves operation and maintenance service plan knowledge points corresponding to the vectorized query questions in the vector database; and then generates and outputs the corresponding operation and maintenance service plan, thereby effectively improving the efficiency and accuracy of operation and maintenance services.

[0196] In this specification, the present invention has been described with reference to specific embodiments thereof. However, it is apparent that various modifications and variations may be made without departing from the spirit and scope of the present invention. Therefore, the specification and drawings should be regarded as illustrative rather than restrictive.

Claims

1. An intelligent generation system for operation and maintenance service solutions of user-side substations, characterized in that: include: The domain knowledge vectorization processing module is used to vectorize professional documents related to substation operation and maintenance, extract key information, and establish a vector database for substation operation and maintenance; The query preprocessing module is used to extract entities from the content input by the operation and maintenance personnel through the human-computer interface interaction, and obtain the keywords and basic elements required to generate the operation and maintenance plan; A knowledge retrieval module, which is used to reorganize the query keyword entities into vectorized query questions and retrieve the operation and maintenance service solution knowledge points corresponding to the vectorized query questions from the vector database; A solution generation module is used to generate prompt words and the keywords for inputting a large language model according to the retrieved knowledge points, and to generate an operation and maintenance service solution using the large language model; The solution output module is used to output the generated operation and maintenance service solution to the operation and maintenance personnel in a structured format.

2. The user-side substation operation and maintenance service solution intelligent generation system according to claim 1 is characterized in that: The domain knowledge vectorization processing module specifically includes: The file reading unit is used to read domain documents in various formats, including operation and maintenance management regulations and operation and maintenance procedures for various equipment. The document formats include HTML, docx, markdown, and PDF. Format conversion unit, used to convert domain documents in various formats into txt text format; Data cleaning unit, which is used to remove tags and specific elements in documents to ensure clean text and reduce text noise; A text segmentation unit is used to segment the text by referring to each separator symbol and to segment the text into different semantic blocks; The text vectorization unit is used to vectorize the divided text using one or more of the M3E, BGE, and BCE methods, and save it into the vector database.

3. The intelligent generation system for operation and maintenance service solutions of user-side substation according to claim 2 is characterized in that: The query preprocessing module specifically includes: An interactive interface unit is used to provide an interactive interface for operation and maintenance personnel to input the information necessary to form an operation and maintenance plan; The keyword extraction unit is used to extract a keyword model according to the content input by the user through the human-computer interface unit. The keyword model is represented by a json object, and keywords are generated based on the keyword model.

4. The user-side substation operation and maintenance service solution intelligent generation system according to claim 3 is characterized in that: The knowledge retrieval module specifically includes: A query rewriting unit, for generating corresponding query content for each type of operation and maintenance task of each type of equipment according to the keywords provided by the query preprocessing module; The query question vectorization unit is used to vectorize multi-channel query contents by using one or more of the M3E, BGE, and BCE methods, organize them into vectorized query questions, and retrieve the operation and maintenance service solution knowledge points corresponding to the vectorized query questions from the vector database.

5. The user-side substation operation and maintenance service solution intelligent generation system according to claim 4 is characterized in that: The solution generation module specifically includes: A query result recall unit is used to recall several segmented operation and maintenance service solution knowledge points with high similarity for each query; A scheme generation prompt word unit, used for forming a scheme generation prompt word according to the query result, keywords and a preset prompt word template; The large language model generation unit is used to input prompt words into the large language model to generate an operation and maintenance service plan.

6. A method for intelligently generating an operation and maintenance service plan for a user-side substation, characterized in that: The following steps are involved: (1) The domain knowledge vectorization processing module vectorizes professional documents related to substation operation and maintenance, extracts key information, and establishes a vector database for substation operation and maintenance; (2) The query preprocessing module extracts entities from the content input by the operation and maintenance personnel through the human-computer interface to obtain the keywords and basic elements required to generate the operation and maintenance plan; (3) The knowledge retrieval module reorganizes the queried keyword entities into vectorized query questions, and retrieves operation and maintenance service solution knowledge points corresponding to the vectorized query questions from the vector database; (4) The solution generation module generates prompt words and the keywords for inputting the large language model based on the retrieved knowledge points, and generates an operation and maintenance service solution using the large language model; (5) The solution output module outputs the generated operation and maintenance service solution to the operation and maintenance personnel in a structured format.

7. The method for intelligently generating an operation and maintenance service plan for a user-side substation according to claim 6 is characterized in that: The step (1) specifically comprises the following steps: (11) The file reading unit reads domain documents in various formats, including operation and maintenance management regulations and operation and maintenance procedures for various equipment. The document formats include HTML, docx, markdown, and PDF. (12) The format conversion unit converts domain documents of various formats into txt text format; (13) The data cleaning unit removes the tags and specific elements in the document to ensure that the text is clean and reduce the noise of the text; (14) the text segmentation unit segments the text with reference to each separator symbol, and divides the text into different semantic blocks; (15) The text vectorization unit uses one or more of the M3E, BGE, and BCE methods to vectorize the divided text and save it into the vector database.

8. The method for intelligently generating an operation and maintenance service plan for a user-side substation according to claim 7 is characterized in that: The step (2) specifically comprises the following steps: (21) The interactive interface unit provides an interactive interface for operation and maintenance personnel to input information necessary to form an operation and maintenance plan; (22) The keyword extraction unit extracts a keyword model according to the user input content. The keyword model is represented by a json object, and keywords are generated based on the keyword model.

9. The method for intelligently generating an operation and maintenance service plan for a user-side substation according to claim 8 is characterized in that: The step (3) specifically comprises the following steps: (31) the query rewriting unit generates corresponding query content for each operation and maintenance task of each type of equipment according to the keywords provided by the query preprocessing module; (32) The query question vectorization unit uses one or more of the M3E, BGE, and BCE methods to vectorize the multi-channel query content, organize it into a vectorized query question, and retrieve the operation and maintenance service solution knowledge points corresponding to the vectorized query question from the vector database.

10. The method for intelligently generating an operation and maintenance service plan for a user-side substation according to claim 9, characterized in that: The step (4) specifically comprises the following steps: (41) The query result recall unit recalls several segmented operation and maintenance service solution knowledge points with high similarity for each query; (42) a scheme generating prompt word unit generates a scheme generating prompt word according to the query result, the keyword and the preset prompt word template; The prompt words are input into the large language model through the large language model generation unit to generate an operation and maintenance service plan.

Images