LLM-based intelligent operation and maintenance processing method and device, medium and equipment

Abstract

The invention discloses an LLM-based intelligent operation and maintenance processing method and apparatus, a medium and a device. The method comprises the steps of constructing an entity database based on order sending data uploaded by a user; performing entity and entity relationship extraction on the order sending data uploaded by the user through an LLM model to obtain first entity relationship data; entity deduplication is carried out through an LLM model, and deduplication entity information is obtained; integrating and purifying the first entity relational data to obtain updated second entity relational data; and performing identification processing on an operation and maintenance scheduling demand instruction input by a user through the LLM model, and generating an operation and maintenance scheduling query result after extraction and simplification. According to the method, the entity database is constructed for the order sending data uploaded by the user through the LLM model, the success rate of effective entity and relation extraction and the recognition rate of equivalent entities and relations are greatly improved by utilizing the inductive reasoning capability of the LLM model, and more accurate entity recognition and entity relation recognition can be obtained.

Description

Technical Field

[0001] The present invention relates to the field of power operation and maintenance technology, and in particular to an LLM-based intelligent operation and maintenance processing method, device, medium and equipment. Background Art

[0002] The dispatching problem in the field of power operation and maintenance is traditionally solved manually. In the power system, several substations, transmission stations, and distribution stations are managed by the same operation and maintenance team. The dispatching team leader needs to dispatch the entire team's work every day according to the annual and monthly plans specified by the superior, combined with each person's place of residence, professional ability, interpersonal relationship, recent work intensity, and weather forecast information.

[0003] LLM technology is a large language model technology that can generate natural language text or understand the meaning of language text. However, applying it to the operation and maintenance of power systems is a problem that has not been effectively implemented at present.

[0004] The existing RAG technology stands for Retrieval Enhanced Generation, which is a technology that can improve the credibility of LLM generation and give LLM memory ability. GraphRAG technology essentially uses LLM to build a knowledge graph first, so as to avoid too many irrelevant texts from being retrieved during the retrieval stage. However, GraphRAG cannot accurately identify whether two words represent an entity or two relationships are equivalent relationships during the entity and relationship extraction process. The retrieval process of RAG is fuzzy retrieval. In the power operation and maintenance scenario, highly similar entities are very common. For example, except for the different numbers, the other information (including equipment that needs to be inspected and maintenance personnel) of 1# substation and 2# substation are basically the same. All RAG technologies based on fuzzy search cannot ensure that the information between two entities will not be confused during the retrieval stage.

[0005] Therefore, how to realize in the power system, according to the actual needs, to generate various query results through the LLM model, so as to be suitable for scenarios such as power operation and maintenance dispatching, power knowledge question and answer intelligent assistant, operation and maintenance technical training AI, and especially how to achieve fast and accurate operation and maintenance dispatching, is a technical problem that needs to be solved urgently. Summary of the invention

[0006] In view of the above problems, the present invention is proposed to provide an LLM-based intelligent operation and maintenance processing method, device, medium and equipment that overcome the above problems or at least partially solve the above problems.

[0007] Other features and advantages of the present invention will become apparent from the following detailed description, or may be learned in part by practice of the present invention.

[0008] According to a first aspect of an embodiment of the present invention, there is provided an intelligent operation and maintenance processing method based on LLM, the intelligent operation and maintenance processing method based on LLM comprising:

[0009] Build an entity database based on the dispatch data uploaded by users;

[0010] Extracting entities and entity relationships from the dispatch data uploaded by the user through the LLM model to obtain first entity relationship data, and storing the extracted first entity relationship data in the form of triples;

[0011] Based on the first entity relationship data and the entity database, perform entity deduplication through the LLM model to obtain deduplicated entity information;

[0012] Integrate and purify the first entity relationship data according to the deduplicated entity information to obtain updated second entity relationship data;

[0013] The operation and maintenance scheduling demand instructions input by the user are identified and processed through the LLM model, the entities to be processed in the operation and maintenance scheduling demand instructions are obtained, the entity relationships corresponding to all the entities to be processed are retrieved according to the second entity relationship data, and the operation and maintenance scheduling query results are generated after refinement and simplification.

[0014] In some embodiments of the present invention, constructing an entity database based on the dispatch data uploaded by the user includes:

[0015] Confirm the category of the dispatch data uploaded by the user, which includes a structured Excel document and an unstructured WORD document;

[0016] When the category of the dispatch data is a structured Excel document, constructing an entity database based on the structured Excel document;

[0017] When the category of the dispatch data is an unstructured WORD document, the entity information in the unstructured WORD document is extracted through the LLM model to construct an entity database.

[0018] In some embodiments of the present invention, extracting entities and entity relationships from the dispatch data uploaded by the user through the LLM model to obtain first entity relationship data, and storing the extracted first entity relationship data in the form of triples includes:

[0019] Defining extraction rules for extracting entities and entity relationships in the LLM model, wherein the extraction rules include types of extracted entities and types of relationships between entities;

[0020] Extracting entities and entity relationships in the dispatch data uploaded by the user based on the extraction rules through the LLM model;

[0021] The extracted entities and entity relationships are combined and stored in the form of triples to obtain first entity relationship data, and the extraction sources corresponding to the elements in the triples of the first entity relationship data are recorded.

[0022] In some embodiments of the present invention, performing entity deduplication through an LLM model based on the first entity relationship data and the entity database to obtain deduplicated entity information includes:

[0023] The first entity relationship data is traversed by using the LLM model to confirm whether there is overlap between entities in each triple, and the overlapping entities are removed and merged to obtain a first screening result;

[0024] Compare all entities in the first screening result with the entity database to confirm the corresponding relationship. If there is an entity in the first screening result that has no corresponding relationship with the entity database, delete the entity. If there is an entity in the first screening result that has a corresponding relationship with the entity database, rename the entity according to the name of the entity in the entity database.

[0025] After comparing all entities in the first screening result with the entity database in turn, they are deduplicated and merged to obtain the deduplicated entity information.

[0026] In some embodiments of the present invention, integrating and purifying the first entity relationship data according to the deduplicated entity information to obtain updated second entity relationship data includes:

[0027] According to the deduplicated entity information, confirm whether there are multiple entity relationships between any two identical entities in the first entity relationship data;

[0028] For two identical entities with multiple entity relationships, the multiple entity relationships are integrated and purified through the LLM model to obtain a corresponding entity summary relationship;

[0029] After integrating and purifying all entity relationships between two identical entities in the first entity relationship data, updated second entity relationship data is obtained.

[0030] In some embodiments of the present invention, after obtaining the updated second entity relationship data, the method further includes: purifying information of all entities in the second entity relationship data to obtain entity reference information, where the entity reference information is basic information data of the corresponding entity.

[0031] In some embodiments of the present invention, the identification and processing of the operation and maintenance scheduling demand instruction input by the user through the LLM model, obtaining the entity to be processed in the operation and maintenance scheduling demand instruction, retrieving the entity relationships corresponding to all the entities to be processed according to the second entity relationship data, and generating the operation and maintenance scheduling query result after refining and simplification includes:

[0032] The operation and maintenance scheduling demand instruction input by the user is identified and processed by the LLM model to obtain the entity to be processed in the operation and maintenance scheduling demand instruction, where the number of the entity to be processed is at least one;

[0033] According to the name of the entity to be processed, searching in the second entity relationship data for all entity relationships corresponding to the entity to be processed with the same name;

[0034] According to the entity relationship corresponding to the entity to be processed, the entity reference information of the entity to be processed is summarized to extract the prompt reference information;

[0035] After the prompt reference information is simplified according to a preset length threshold range, the operation and maintenance scheduling query result is generated.

[0036] According to a second aspect of an embodiment of the present invention, there is provided an intelligent operation and maintenance processing device based on LLM, the intelligent operation and maintenance processing device based on LLM comprising:

[0037] A database construction module is used to construct an entity database based on the dispatch data uploaded by users;

[0038] An entity extraction module, used for extracting entities and entity relationships from the dispatch data uploaded by the user through an LLM model to obtain first entity relationship data, and storing the extracted first entity relationship data in the form of triples;

[0039] An entity deduplication module, configured to perform entity deduplication through an LLM model based on the first entity relationship data and the entity database to obtain deduplicated entity information;

[0040] An entity relationship purification module, configured to integrate and purify the first entity relationship data according to the deduplicated entity information to obtain updated second entity relationship data;

[0041] The data query module is used to identify and process the operation and maintenance scheduling demand instructions input by the user through the LLM model, obtain the entities to be processed in the operation and maintenance scheduling demand instructions, retrieve the entity relationships corresponding to all the entities to be processed according to the second entity relationship data, and generate the operation and maintenance scheduling query results after refining and simplification.

[0042] According to a third aspect of an embodiment of the present invention, a computer-readable storage medium is provided, in which computer program instructions are stored. The computer program instructions are loaded and executed by a processor to implement the operations performed by any of the methods described above.

[0043] According to a fourth aspect of an embodiment of the present invention, there is provided an electronic device, comprising a processor and a memory, wherein the memory stores computer program instructions that can be executed by the processor, and when the processor executes the computer program instructions, it implements instructions of any of the methods described above.

[0044] The technical solution provided in the embodiments of the present invention has at least the following technical effects or advantages:

[0045] The embodiments of the present invention provide an intelligent operation and maintenance processing method, device, medium and equipment based on LLM. The method constructs an entity database for the dispatch data uploaded by the user through the LLM model, and utilizes the inductive reasoning ability of the LLM model to greatly improve the success rate of effective entity and relationship extraction and the recognition rate of equivalent entities and relationships, so as to obtain more accurate entity recognition and entity relationship recognition. Compared with the existing method of constructing knowledge graphs, the method is more flexible, highly applicable and more accurate.

[0046] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented according to the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0048] Figure 1 A schematic diagram of a flow chart of an intelligent operation and maintenance processing method based on LLM provided in an embodiment of the present invention;

[0049] Figure 2 A reference schematic diagram for generating operation and maintenance scheduling query results for the present invention;

[0050] Figure 3 A schematic diagram of the principle of an LLM-based intelligent operation and maintenance processing device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0051] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings.

[0052] Various structural schematic diagrams according to embodiments of the present disclosure are shown in the accompanying drawings. These figures are not drawn to scale, and some details are magnified and some details may be omitted for the purpose of clear expression. The shapes of various regions and layers shown in the figures and the relative sizes and positional relationships therebetween are only exemplary, and may deviate in practice due to manufacturing tolerances or technical limitations, and those skilled in the art may further design regions / layers with different shapes, sizes, and relative positions according to actual needs.

[0053] In the context of the present disclosure, when a layer / element is referred to as being "on" another layer / element, the layer / element may be directly on the other layer / element, or there may be an intervening layer / element therebetween. Additionally, if a layer / element is "on" another layer / element in one orientation, the layer / element may be "below" the other layer / element when the orientation is reversed. In the context of the present disclosure, similar or identical components may be denoted by the same or similar reference numerals.

[0054] In order to better understand the above-mentioned technical scheme, the above-mentioned technical scheme will be described in detail below in combination with specific implementation methods. It should be understood that the embodiments of the present disclosure and the specific features in the embodiments are detailed descriptions of the technical scheme of the present invention, rather than limitations on the technical scheme of the present invention. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0055] Figure 1 FIG. 1 is a flow chart of an intelligent operation and maintenance processing method based on LLM provided in an embodiment of the present invention. Figure 1 As shown, the LLM-based intelligent operation and maintenance processing method includes the following steps:

[0056] S1. Build an entity database based on the dispatch data uploaded by users;

[0057] In the embodiment of the present invention, the dispatch data is the operation and maintenance dispatching data for the substation, transmission station, distribution station and other scenes in the power system, as shown in the following Table 1:

[0058] Table 1 Dispatch data

[0059]

[0060] Among them, the embodiment of the present invention builds an entity database based on the dispatch data uploaded by the user, including: confirming the category of the dispatch data uploaded by the user, the category of the dispatch data including structured Excel documents and unstructured WORD documents; when the category of the dispatch data is a structured Excel document, building an entity database based on the structured Excel document; when the category of the dispatch data is an unstructured WORD document, extracting entity information in the unstructured WORD document through the LLM model to build an entity database.

[0061] Of course, it should be noted that the dispatch data may also be an electronic file in other formats. For example, the dispatch data may also be an unstructured Excel document. In this case, the entity information in the unstructured Excel document may also be extracted through the LLM model to construct an entity database.

[0062] The information contained in the dispatch data includes but is not limited to scenes, operators, power equipment, and power protection equipment and places that may be affected by the shutdown of the power station, or the commuting time between two scenes, what equipment is available in each scene, what skills the operators have, etc.

[0063] Optionally, the entities and entity relationships in the dispatch data may be defined by an LLM model, for example. The prompt words for extracting entity information by using the LLM model in an embodiment of the present invention may be, for example: "Read the following text: \n{document paragraph}\n\nWhich entities does this text describe? Note that an entity may be a power station, an operator, power equipment, or power protection equipment and places that may be affected by a power station shutdown. Please organize the results into a JSON list format."

[0064] S2. Extracting entities and entity relationships from the dispatch data uploaded by the user through the LLM model to obtain first entity relationship data, and storing the extracted first entity relationship data in the form of triples;

[0065] After constructing the entity database, the embodiment of the present invention extracts entities and entity relationships from the dispatch data uploaded by the user through the LLM model to obtain first entity relationship data, and stores the extracted first entity relationship data in the form of triples, including: defining extraction rules for extracting entities and entity relationships by the LLM model, the extraction rules including the types of extracted entities and the types of relationships between entities; extracting entities and entity relationships from the dispatch data uploaded by the user based on the extraction rules through the LLM model; combining the extracted entities and entity relationships and storing them in the form of triples to obtain first entity relationship data, and recording the extraction sources corresponding to the elements in the triples of the first entity relationship data.

[0066] In an embodiment of the present invention, entities and entity relationships are extracted from the dispatch data uploaded by the user through an LLM model, and prompt words are, for example: "Read the following text: \n{document paragraph}\n\nWhat entity relationships does this text describe? Note that an entity can be XX, XX, and XX. Please organize the results into a triple format of (entity 1, relationship, entity 2)".

[0067] The format of the triplet may be (entity 1, relationship, entity 2), for example, (Zhang San, good at, substation inspection), (Zhang San, colleague, Li Si), etc.

[0068] When the embodiment of the present invention extracts entities and entity relationships from dispatch data through the LLM model, it can preliminarily deduplicate the same expression of the same entity in the dispatch data, thereby improving recognition efficiency compared to knowledge graph technology.

[0069] S3. Based on the first entity relationship data and the entity database, perform entity deduplication through an LLM model to obtain deduplicated entity information;

[0070] After obtaining the first entity relationship data, the embodiment of the present invention performs entity deduplication through the LLM model based on the first entity relationship data and the entity database to obtain deduplicated entity information, including: traversing the first entity relationship data through the LLM model to confirm whether there is overlap between entities in each triple, deduplicating and merging the overlapping entities to obtain a first screening result; comparing all entities in the first screening result with the entity database respectively to confirm the corresponding relationship, if there is an entity in the first screening result that has no corresponding relationship with the entity database, the entity is deleted, if there is an entity in the first screening result that has a corresponding relationship with the entity database, the entity is renamed according to the name of the entity in the entity database; after comparing all entities in the first screening result with the entity database in turn, deduplicating and merging them respectively to obtain the deduplicated entity information.

[0071] The embodiment of the present invention uses the LLM model to deduplicate, aggregate, and match the first entity relationship data and the same entity in the entity database, confirm whether there are duplicate entities in the first entity relationship data, and whether the entities in the first entity relationship data exist in the entity database. If not, it is considered that the entity is irrelevant to the entity database, and the entity is discarded and deleted. If there is an entity corresponding to the entity database in the first screening result, it is renamed according to the name of the entity in the entity database to achieve a unique name for the same entity.

[0072] Among them, the matching of entities in the first screening result with entities in the entity database can be calculated and obtained through an existing relevance algorithm. For example, the relevance algorithm can be based on a lightweight text classification model such as BERT, or use another large language model in the form of rankRAG, which can be selected according to actual needs.

[0073] The prompt words for removing duplicates and merging the same entity in the first entity relationship data and the entity database through the LLM model are, for example: "Based on the context, determine whether {entity 1} and {entity 2} are the same entity. Reference materials such as: {...}".

[0074] S4. Integrate and purify the first entity relationship data according to the deduplicated entity information to obtain updated second entity relationship data;

[0075] After deduplicating entities to obtain deduplicated entity information, an embodiment of the present invention integrates and purifies the first entity relationship data based on the deduplicated entity information to obtain updated second entity relationship data, including: confirming whether there are multiple entity relationships between any two identical entities in the first entity relationship data based on the deduplicated entity information; for two identical entities with multiple entity relationships, integrating and purifying the multiple entity relationships through the LLM model to obtain a corresponding entity summary relationship; integrating and purifying the entity relationships between all two identical entities in the first entity relationship data to obtain updated second entity relationship data.

[0076] In the embodiment of the present invention, the prompt word for integrating and purifying the first entity relationship data through the LLM model is, for example: "Organize the following triples into one triple: \n{triple list}".

[0077] After obtaining the updated second entity relationship data, an embodiment of the present invention also includes: purifying information on all entities in the second entity relationship data to obtain entity reference information, wherein the entity reference information is basic information material of the corresponding entity and is used to describe the entity; and in order to keep the data concise, it is necessary to refine the entity reference information. In an embodiment of the present invention, the text description of the entity reference information does not exceed 100 tokens.

[0078] In the embodiment of the present invention, the LLM model is used to purify the information of all entities in the second entity relationship data, and the corresponding prompt words are, for example: "Briefly summarize and generalize the following text, and be careful not to lose key information: \n{text}".

[0079] In other embodiments of the present invention, the present invention can also construct a knowledge graph, import the information in the entity database into an open source knowledge graph tool, and display the entity data to the user in a visual form, so that the user can edit it according to actual needs to obtain entity data that meets actual needs.

[0080] After obtaining the second entity relationship data, the embodiment of the present invention may also enter the second entity relationship data into an entity database for data update to obtain a knowledge base used for subsequent search queries.

[0081] S5. Identify and process the operation and maintenance scheduling demand instructions input by the user through the LLM model, obtain the entities to be processed in the operation and maintenance scheduling demand instructions, retrieve the entity relationships corresponding to all the entities to be processed according to the second entity relationship data, and generate the operation and maintenance scheduling query results after refining and simplification.

[0082] The embodiment of the present invention uses an LLM model to identify and process the operation and maintenance scheduling demand instructions input by the user, obtains the entities to be processed in the operation and maintenance scheduling demand instructions, retrieves the entity relationships corresponding to all the entities to be processed according to the second entity relationship data, and generates the operation and maintenance scheduling query results after refining and simplification, including: identifying and processing the operation and maintenance scheduling demand instructions input by the user through the LLM model, obtaining the entities to be processed in the operation and maintenance scheduling demand instructions, and the number of the entities to be processed is at least one; according to the names of the entities to be processed, retrieving the entity relationships corresponding to all the entities to be processed with the same name in the second entity relationship data; according to the entity relationships corresponding to the entities to be processed, summarizing the entity reference information of the entities to be processed to extract prompt reference information; after simplifying the prompt reference information according to a preset length threshold range, the operation and maintenance scheduling query results are generated.

[0083] Specifically, see Figure 2 As shown, after obtaining the operation and maintenance scheduling demand instructions input by the user, the LLM model performs identification processing to obtain the entities to be processed in the operation and maintenance scheduling demand instructions, and the number of the entities to be processed is at least one; the entity relationships corresponding to all the entities to be processed with the same name are retrieved in the second entity relationship data (i.e., knowledge base retrieval), and the retrieval method is precise retrieval, that is, only when the names of the entities are exactly the same, the corresponding retrieval results are fed back; after retrieving the entity relationships corresponding to all the entities to be processed, the entity reference information of the entities to be processed is summarized to extract prompt reference information, and the prompt reference information can be added after the original prompt word as reference material.

[0084] Optionally, since the total prompt words plus the prompt reference information may result in data that is too long, an embodiment of the present invention further summarizes and refines the prompt reference information; the refining method is to let the LLM model summarize the text. If the summarized text is too long, the text summarized in the previous step is used as input to summarize again until a text that meets the length requirement is generated; the prompt words for the LLM model to summarize and refine the text are, for example: "Based on the following questions, make a brief summary of the information related to the following documents and questions."

[0085] After the physical database (or knowledge base) is constructed, the LLM-based intelligent operation and maintenance processing method described in the embodiment of the present invention can query all personnel, equipment, plans and power stations in the database, including their brief descriptions, for the operation and maintenance scheduling tasks; generate a preliminary scheduling list based on the retrieval results, and the data retrieved in the previous step will be deleted; generate a preliminary scheduling list through a large model, and identify various power equipment mentioned in the scheduling list; query the characteristics, precautions and other information of these equipment in the knowledge base (long document, corresponding to the database basis part); based on the supplementary precautions and the previously generated preliminary scheduling list, supplement detailed inspection instructions and precautions. At the same time, if it involves inspecting multiple devices at the same site, the order of inspecting these devices can be further planned, and finally generate an operation and maintenance scheduling query result for users to provide reference for intelligent operation and maintenance processing.

[0086] Compared with the prior art, the LLM-based intelligent operation and maintenance processing method described in the embodiment of the present invention has the following advantages:

[0087] 1. Use the LLM model to build an entity database for the dispatch data uploaded by users. Utilize the inductive reasoning ability of the LLM model to greatly improve the success rate of effective entity and relationship extraction and the recognition rate of equivalent entities and relationships, and obtain more accurate entity recognition and entity relationship recognition. Compared with the existing way of building knowledge graphs, it is more flexible, highly applicable, and more accurate.

[0088] 2. Confirm the corresponding relationship of entities through precise retrieval, which improves the retrieval efficiency and accuracy compared with the fuzzy search method used in the existing technology such as RAG solution;

[0089] 3. For operation and maintenance scheduling scenarios, the number of entities is relatively limited. Generally, they are all about dispatching personnel, scenarios, etc. Other irrelevant entities are ignored noise. Therefore, the LLM model uses the "task scenario-oriented" feature to further improve the retrieval efficiency and accuracy.

[0090] Based on the above embodiment, as the above Figure 1 The present invention provides an embodiment of an intelligent operation and maintenance processing device based on LLM, and the embodiment of the device is Figure 1 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices, see Figure 3 As shown, the LLM-based intelligent operation and maintenance processing device includes:

[0091] A database construction module 100 is used to construct an entity database based on the dispatch data uploaded by the user;

[0092] The entity extraction module 200 is used to extract entities and entity relationships from the dispatch data uploaded by the user through the LLM model to obtain first entity relationship data, and store the extracted first entity relationship data in the form of triples;

[0093] An entity deduplication module 300 is used to perform entity deduplication through an LLM model based on the first entity relationship data and the entity database to obtain deduplicated entity information;

[0094] An entity relationship purification module 400, configured to integrate and purify the first entity relationship data according to the deduplicated entity information to obtain updated second entity relationship data;

[0095] The data query module 500 is used to identify and process the operation and maintenance scheduling demand instructions input by the user through the LLM model, obtain the entities to be processed in the operation and maintenance scheduling demand instructions, retrieve the entity relationships corresponding to all the entities to be processed according to the second entity relationship data, and generate the operation and maintenance scheduling query results after refining and simplification.

[0096] The LLM-based intelligent operation and maintenance processing device described in the embodiment of the present invention can execute the LLM-based intelligent operation and maintenance processing method provided in the above embodiment. The LLM-based intelligent operation and maintenance processing device has the corresponding functional steps and beneficial effects of the LLM-based intelligent operation and maintenance processing method described in the above embodiment. Please refer to the above-mentioned embodiment of the LLM-based intelligent operation and maintenance processing method for details. The embodiment of the present invention will not be repeated here.

[0097] The embodiment of the present invention also provides an electronic device, which may include a processor and a memory, wherein the processor and the memory may be connected via a bus or other means. The processor may be a central processing unit (CPU). The processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components and other chips, or a combination of the above-mentioned various types of chips. The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs, non-transient computer executable programs and modules, such as program instructions / modules corresponding to the intelligent operation and maintenance processing method based on LLM in the embodiment of the present invention. The processor executes various functional applications and data processing of the processor by running the non-transient software programs, instructions and modules stored in the memory, that is, realizing the intelligent operation and maintenance processing method based on LLM in the above method embodiment.

[0098] The memory may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required by at least one function; the data storage area may store data created by the processor, etc. In addition, the memory may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. The one or more modules are stored in the memory, and when executed by the processor, the intelligent operation and maintenance processing method based on LLM in the above-mentioned method embodiment is executed. The specific details of the above-mentioned electronic device can be understood by corresponding to the corresponding descriptions and effects in the above-mentioned method embodiment, and will not be repeated here. Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and the program, when executed, may include the processes of the embodiments of the above-mentioned methods. Among them, the storage medium can be a read-only memory (ROM), a random access memory (RAM), a flash memory (Flash Memory), a hard disk drive (HDD) or a solid-state drive (SSD), etc.; the storage medium can also include a combination of the above types of memory.

[0099] In the description provided herein, a large number of specific details are described. However, it is understood that embodiments of the present invention can be practiced without these specific details. In some instances, well-known methods, structures and techniques are not shown in detail so as not to obscure the understanding of this description.

[0100] Similarly, it should be understood that in order to streamline the present disclosure and aid in understanding one or more of the various inventive aspects, in the above description of exemplary embodiments of the present invention, various features of the present invention are sometimes grouped together into a single embodiment, figure, or description thereof. However, this disclosed method should not be interpreted as reflecting the intention that the claimed invention requires more features than those expressly recited in each claim. Rather, as reflected in the claims, inventive aspects lie in less than all of the features of the individual embodiments previously disclosed. Therefore, the claims that follow the detailed description are hereby expressly incorporated into the detailed description, with each claim itself serving as a separate embodiment of the present invention.

[0101] It should be noted that the above-mentioned embodiments illustrate rather than limit the invention and that those skilled in the art will be able to design alternative embodiments without departing from the scope of the appended claims.

Claims

1. An intelligent operation and maintenance processing method based on LLM, characterized in that: The LLM-based intelligent operation and maintenance processing method includes: Build an entity database based on the dispatch data uploaded by users; Extracting entities and entity relationships from the dispatch data uploaded by the user through the LLM model to obtain first entity relationship data, and storing the extracted first entity relationship data in the form of triples; Based on the first entity relationship data and the entity database, perform entity deduplication through the LLM model to obtain deduplicated entity information; Integrate and purify the first entity relationship data according to the deduplicated entity information to obtain updated second entity relationship data; The operation and maintenance scheduling demand instructions input by the user are identified and processed through the LLM model, the entities to be processed in the operation and maintenance scheduling demand instructions are obtained, the entity relationships corresponding to all the entities to be processed are retrieved according to the second entity relationship data, and the operation and maintenance scheduling query results are generated after refinement and simplification.

2. The LLM-based intelligent operation and maintenance processing method according to claim 1, characterized in that: The step of constructing an entity database based on the dispatch data uploaded by the user includes: Confirm the category of the dispatch data uploaded by the user, which includes a structured Excel document and an unstructured WORD document; When the category of the dispatch data is a structured Excel document, constructing an entity database based on the structured Excel document; When the category of the dispatch data is an unstructured WORD document, the entity information in the unstructured WORD document is extracted through the LLM model to construct an entity database.

3. The LLM-based intelligent operation and maintenance processing method according to claim 1, characterized in that: The extracting entities and entity relationships from the dispatch data uploaded by the user through the LLM model to obtain first entity relationship data, and storing the extracted first entity relationship data in the form of triples includes: Defining extraction rules for extracting entities and entity relationships in the LLM model, wherein the extraction rules include types of extracted entities and types of relationships between entities; Extracting entities and entity relationships in the dispatch data uploaded by the user based on the extraction rules through the LLM model; The extracted entities and entity relationships are combined and stored in the form of triples to obtain first entity relationship data, and the extraction sources corresponding to the elements in the triples of the first entity relationship data are recorded.

4. The LLM-based intelligent operation and maintenance processing method according to claim 1, characterized in that: The performing entity deduplication through the LLM model based on the first entity relationship data and the entity database to obtain deduplicated entity information includes: The first entity relationship data is traversed by using the LLM model to confirm whether there is overlap between entities in each triple, and the overlapping entities are removed and merged to obtain a first screening result; Compare all entities in the first screening result with the entity database to confirm the corresponding relationship. If there is an entity in the first screening result that has no corresponding relationship with the entity database, delete the entity. If there is an entity in the first screening result that has a corresponding relationship with the entity database, rename the entity according to the name of the entity in the entity database. After comparing all entities in the first screening result with the entity database in turn, they are deduplicated and merged to obtain the deduplicated entity information.

5. The LLM-based intelligent operation and maintenance processing method according to claim 1, characterized in that: According to the deduplicated entity information, the first entity relationship data is integrated and purified to obtain updated second entity relationship data including: According to the deduplicated entity information, confirm whether there are multiple entity relationships between any two identical entities in the first entity relationship data; For two identical entities with multiple entity relationships, the multiple entity relationships are integrated and purified through the LLM model to obtain a corresponding entity summary relationship; After integrating and purifying all entity relationships between two identical entities in the first entity relationship data, updated second entity relationship data is obtained.

6. The LLM-based intelligent operation and maintenance processing method according to claim 1, characterized in that: After obtaining the updated second entity relationship data, the method further includes: purifying information of all entities in the second entity relationship data to obtain entity reference information, where the entity reference information is basic information material of the corresponding entity.

7. The LLM-based intelligent operation and maintenance processing method according to claim 6, characterized in that: The LLM model is used to identify and process the operation and maintenance scheduling demand instruction input by the user, obtain the entity to be processed in the operation and maintenance scheduling demand instruction, retrieve the entity relationship corresponding to all the entities to be processed according to the second entity relationship data, and generate the operation and maintenance scheduling query result after refining and simplifying, including: The operation and maintenance scheduling demand instruction input by the user is identified and processed by the LLM model to obtain the entity to be processed in the operation and maintenance scheduling demand instruction, where the number of the entity to be processed is at least one; According to the name of the entity to be processed, searching in the second entity relationship data for all entity relationships corresponding to the entity to be processed with the same name; According to the entity relationship corresponding to the entity to be processed, the entity reference information of the entity to be processed is summarized to extract the prompt reference information; After the prompt reference information is simplified according to a preset length threshold range, the operation and maintenance scheduling query result is generated.

8. An intelligent operation and maintenance processing device based on LLM, applied to the method according to any one of claims 1 to 7, characterized in that: The LLM-based intelligent operation and maintenance processing device includes: A database construction module is used to construct an entity database based on the dispatch data uploaded by users; An entity extraction module, used for extracting entities and entity relationships from the dispatch data uploaded by the user through an LLM model to obtain first entity relationship data, and storing the extracted first entity relationship data in the form of triples; An entity deduplication module, configured to perform entity deduplication through an LLM model based on the first entity relationship data and the entity database to obtain deduplicated entity information; An entity relationship purification module, configured to integrate and purify the first entity relationship data according to the deduplicated entity information to obtain updated second entity relationship data; The data query module is used to identify and process the operation and maintenance scheduling demand instructions input by the user through the LLM model, obtain the entities to be processed in the operation and maintenance scheduling demand instructions, retrieve the entity relationships corresponding to all the entities to be processed according to the second entity relationship data, and generate the operation and maintenance scheduling query results after refining and simplification.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer program instructions, which are loaded and executed by a processor to implement the operations performed by the method according to any one of claims 1 to 7.

10. An electronic device comprising a processor and a memory, characterized in that: The memory stores computer program instructions that can be executed by the processor, and when the processor executes the computer program instructions, the processor implements instructions of any one of the methods described in claims 1-7.

Images