Large model knowledge base construction method and system for electric power communication management, operation and maintenance

Abstract

The invention belongs to the technical field of electric power communication, and provides an electric power communication management operation and maintenance-oriented large model knowledge base construction method and system, and the method comprises the steps: collecting structured and unstructured data in electric power communication management operation and maintenance, and carrying out the cleaning and standardization processing of the data through a distributed calculation framework; based on the processed data, performing entity extraction, relationship recognition and knowledge mapping by utilizing a domain fine-tuning large language model, and outputting an entity relationship triple and a semantic embedding vector; respectively storing the entity relationship triad and the semantic embedding vector into a graph database and a vector database to form a mixed storage architecture; based on a hybrid storage architecture, decision support is realized in combination with a hybrid retrieval strategy and a generative large language model. According to the invention, semantic understanding of data, dynamic knowledge fusion and intelligent decision support are realized.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of electric power communication, and particularly relates to a large model knowledge base construction method and system for electric power communication management and operation. BACKGROUND

[0002] The electric power communication system has the characteristics of complex equipment, various protocols and strong professional scenes. In the operation and maintenance process of the electric power backbone communication network, a large amount of data exists in the form of unstructured text (such as operation and maintenance logs and fault work orders). The traditional method relies on manual annotation and rule matching, which is low in efficiency and difficult to cover complex semantic scenes. The existing knowledge base is based on keyword retrieval, lacks deep understanding of natural language, and leads to low knowledge utilization rate and weak decision support capability. SUMMARY

[0003] The purpose of the present application is to provide a large model knowledge base construction method and system for electric power communication management and operation, in order to solve the problems of low knowledge utilization rate and weak decision support capability.

[0004] To achieve the above purpose, the present application adopts the following technical scheme: In a first aspect, the present application provides a large model knowledge base construction method for electric power communication management and operation, comprising: Collecting structured and unstructured data in electric power communication management and operation, and cleaning and standardizing the data through a distributed computing framework; Based on the processed data, using a domain fine-tuning large language model to perform entity extraction, relationship recognition and knowledge mapping, and outputting entity relationship triplets and semantic embedding vectors; Storing the entity relationship triplets and semantic embedding vectors into a graph database and a vector database respectively to form a hybrid storage architecture; Based on the hybrid storage architecture, combining a hybrid retrieval strategy and a generative large language model to realize decision support.

[0005] Further, the collecting of structured and unstructured data in electric power communication management and operation comprises: The structured data includes device status and network topology, and the unstructured data includes operation and maintenance work order text and maintenance records. After the structured and unstructured data is collected through TMS, optical transmission system professional network management and offline forms, a distributed computing framework is used for cleaning and standardization.

[0006] Further, the data cleaning is to filter outliers by threshold; and the data normalization is to eliminate dimension differences by using minimum-maximum scaling, as shown below:

[0007] wherein, and min and max values of the feature; The unstructured text preprocessing converts the operation and maintenance work order text into a JSON format, extracts the time, device ID, and key fields of the description content.

[0008] Further, based on the processed data, a domain fine-tuned large language model is used for entity extraction, relationship identification, and knowledge mapping, to output entity relationship triples and semantic embedding vectors, including: The domain fine-tuned large model LLM is used for semantic analysis of unstructured text to extract entities and relationships. The entity recognition uses a bidirectional Transformer model to label device names, fault types, and other entities, and the loss function is cross-entropy:

[0009] wherein, is the true label, is the model prediction probability; The relationship extraction calculates the correlation between entities through attention weight, and outputs entity-relation-entity triples. The semantic vector generation uses the token hidden state of the LLM as the text embedding vector.

[0010] Further, the entity relationship triples and semantic embedding vectors are stored in a graph database and a vector database respectively to form a hybrid storage architecture, including: Knowledge graph construction: triples are stored in a graph database. Semantic embedding vector storage: text embedding vectors are stored in a vector database, and cosine similarity is used for similarity calculation:

[0011] wherein, v 1 is the text embedding vector; v 2 is the query vector; Based on a distributed data stream engine, incremental updates are realized, and the update rule is:

[0012] wherein, is the knowledge base at time t, is the new data.

[0013] Further, based on the hybrid storage architecture, a hybrid retrieval strategy and generative large language model are used to realize decision support, including: The hybrid retrieval strategy includes keyword retrieval, semantic retrieval, and graph reasoning: Keyword retrieval uses Elasticsearch to match structured fields of devices. Semantic retrieval is based on vector similarity to find relevant cases; Graph reasoning calculates the comprehensive risk probability of fault propagation path:

[0014] Wherein, is the failure probability of the i-th edge in the path; Generative large language model is used for generative question answering, input user query "how to solve the optical cable interruption?", LLM generates answers based on retrieval results, and the probability formula is:

[0015] Wherein, Q is the user query, and S is the retrieval result.

[0016] Further, the decision support includes fault root cause analysis, treatment suggestion generation and risk warning, wherein the risk warning is triggered by threshold alarm rule: Trigger threshold alarm:

[0017] Alert is the alarm information; is the probability of failure.

[0018] In a second aspect, the present application provides a large model knowledge base construction system for power communication management and operation, comprising: A data acquisition module is used to acquire structured and unstructured data in power communication management and operation, and to clean and standardize the data through a distributed computing framework; A vector output module is used to perform entity extraction, relationship identification and knowledge mapping based on the processed data using a domain fine-tuning large language model, and to output entity relationship triplets and semantic embedding vectors; A storage module is used to store the entity relationship triplets and semantic embedding vectors into a graph database and a vector database respectively, forming a hybrid storage architecture; A decision output module is used to realize decision support based on the hybrid storage architecture, combined with hybrid retrieval strategy and generative large language model.

[0019] In a third aspect, the present application provides a computer device, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to realize the steps of the large model knowledge base construction method for power communication management and operation.

[0020] In a fourth aspect, the present application provides a computer readable storage medium, which stores a computer program, and the computer program, when executed by a processor, implements the steps of the method for constructing a large model knowledge base for power communication management operation and maintenance.

[0021] Compared with the prior art, the present application has the following technical effects: On the one hand, the present application breaks the information island by unifying the knowledge standard, realizes the efficient association and semantic analysis of cross-system data, and on the other hand, supports the large language model to complete intelligent question answering, fault tracing, strategy deduction and other tasks, significantly improves the knowledge reuse efficiency of operation and maintenance personnel, reduces the risk of human misjudgment, at the same time provides an interpretable decision basis for network dynamic optimization and resource intelligent scheduling, accelerates the paradigm upgrade of power communication from experience-driven to data intelligent-driven, and lays a knowledge cornerstone for the digital transformation of the new power system.

[0022] In view of the problem that a large amount of heterogeneous data (such as device logs, network topology, operation and maintenance work orders, and fault reports described in natural language) in the power backbone communication network is difficult to be efficiently integrated and intelligently utilized, the present application proposes a power backbone communication knowledge base construction method, which realizes semantic understanding of data, dynamic knowledge fusion and intelligent decision support. BRIEF DESCRIPTION OF DRAWINGS

[0023] Figure 1 The flowchart of the present application.

[0024] Figure 2 The multi-modal data acquisition and preprocessing schematic diagram of the present application.

[0025] Figure 3 The semantic analysis schematic diagram of the present application.

[0026] Figure 4 The dynamic knowledge fusion and storage schematic diagram of the present application.

[0027] Figure 5 The mixed retrieval schematic diagram of the present application. DETAILED DESCRIPTION

[0028] The present application is further described below in conjunction with the accompanying drawings: NOMENCLATURE: Power backbone communication network: The power backbone communication network is an important infrastructure supporting the stable operation of the power grid, and provides communication channels for production control type businesses such as relay protection, stability control, dispatching automation and integrated management type businesses. The power backbone communication network is mainly composed of SDH, OTN and other optical transmission equipment and communication optical cables, and is usually constructed synchronously with the primary line.

[0029] Large Language Models (LLMs): LLMs are a significant breakthrough in the field of natural language processing in recent years. The core technology is a deep learning model based on the Transformer architecture. The Transformer model uses a self-attention mechanism to achieve parallel processing of sequence data, making it excel at handling long-range dependencies. The self-attention mechanism can capture the relationships between different words in a sentence, helping the model understand complex language structures. LLMs can be further improved through scaling laws, which involve increasing the number of model parameters, the size of the training data, and computing resources to significantly improve model performance.

[0030] Knowledge Graph: A knowledge base that represents entities and their relationships in a graph structure, supporting semantic association reasoning.

[0031] Vector Database: Stores data through embedding vectors, supporting high-dimensional semantic similarity retrieval.

[0032] TMS: Transmit Management System, a communication digital management and operation software system promoted by State Grid Corporation, with functions such as data collection, data storage, and operation management.

[0033] Example 1, please refer to Figure 1 The present application provides a large model knowledge base construction method for power communication management and operation, including: Collecting structured and unstructured data in power communication management and operation, and cleaning and standardizing the data through a distributed computing framework; Based on the processed data, use the field fine-tuning large language model for entity extraction, relationship identification and knowledge mapping, output entity relationship triple and semantic embedding vector; Store the entity relationship triple and semantic embedding vector into the graph database and vector database respectively, forming a hybrid storage architecture; Based on the hybrid storage architecture, combined with hybrid retrieval strategy and generative large language model to realize decision support.

[0034] Fine-tune the model based on power communication corpus to enhance the accuracy of professional term understanding and relationship extraction; Collaborate between graph database and vector database, taking into account structured association and semantic retrieval; Real-time data-driven knowledge base dynamic evolution, combined with rule engine and generative model to improve decision reliability.

[0035] Embodiment 2, the application provides a large model knowledge base construction method for power communication management operation and maintenance, comprising: 1. Multi-modal data acquisition: integrate structured data (device parameters, performance indicators) and unstructured data (operation and maintenance text, voice recording).

[0036] 2. Semantic analysis driven by large language model: use LLM to perform entity extraction, relationship identification and knowledge graph mapping on natural language description of fault reports.

[0037] 3. Dynamic knowledge base construction: combine graph database and vector database to store structured knowledge graph and semantic vector, support complex query and reasoning.

[0038] 4. Intelligent question answering and decision engine: based on the generation ability of LLM and knowledge base retrieval, realize fault root cause analysis, treatment suggestion generation and risk warning.

[0039] 1. Multi-modal data acquisition: integrate structured data (device parameters, performance indicators) and unstructured data (operation and maintenance text, voice recording).

[0040] This module is responsible for integrating multi-source heterogeneous data in power backbone communication network, including structured data (device status, network topology, etc.) and unstructured data (operation and maintenance order text, maintenance record, etc.). After data is collected through TMS, optical transmission system professional network management and offline forms, distributed computing framework (such as Apache Spark) can be used for cleaning and standardization.

[0041] Data cleaning: filter outliers by threshold (such as marking as abnormal when optical power fluctuation exceeds ±3dB).

[0042] Data normalization: use min-max scaling to eliminate dimension difference, as follows: (1) where, and are the minimum and maximum values of the feature.

[0043] Unstructured text preprocessing: convert operation and maintenance order text to JSON format, extract key fields (time, device ID, description content).

[0044] 2. Semantic analysis driven by large language model: use LLM to perform entity extraction, relationship identification and knowledge graph mapping on natural language description of fault reports.

[0045] This module uses domain fine-tuned LLM (such as BERT variant trained based on power communication corpus) to perform semantic analysis on unstructured text, extract entities and relationships.

[0046] Entity Recognition: Use a bidirectional Transformer model to label entities such as device names and fault types. The loss function is cross-entropy: (2) where, is the true label, is the model prediction probability.

[0047] Relation Extraction: Calculate the correlation between entities using attention weights and output triples (e.g., "Cable A - Fault Causes - Site B Communication Interruption").

[0048] Semantic Vector Generation: Use the [CLS] token hidden state of the LLM as the text embedding vector.

[0049] 3. Dynamic Knowledge Base Construction: Combine graph databases and vector databases to store structured knowledge graphs and semantic vectors, supporting complex queries and reasoning.

[0050] This module will integrate the parsed knowledge into a hybrid storage architecture, supporting real-time updates and multi-modal queries.

[0051] Knowledge Graph Construction: Store triples in a graph database (Neo4j). Cypher query example: MATCH(a:Device)-[r:CAUSES]->(b:Fault) RETURN a, r, b Vector Storage: Store text embedding vectors in a vector database (e.g., Milvus). Similarity calculation uses cosine similarity: (3) Streamed Updates: Implement incremental updates based on a distributed data stream engine. Update rules: (4) where, is the knowledge base at time t, is the new data.

[0052] 4. Intelligent Question Answering and Decision Engine: Based on the generation capabilities of LLM and knowledge base retrieval, implement fault root cause analysis, treatment suggestion generation, and risk warning.

[0053] This module combines retrieval and generation techniques to achieve intelligent decision support. Mixed retrieval strategies include keyword retrieval, semantic retrieval, and graph reasoning.

[0054] Keyword Retrieval: Use Elasticsearch to match structured fields such as device IDs; Semantic Retrieval: Find related cases based on vector similarity; Graph reasoning: compute the comprehensive risk probability of fault propagation paths: (5) where, is the failure probability of the i-th edge in the path.

[0055] Generative question answering: input user query "How to solve the optical cable interruption?", LLM generates answers based on retrieval results, the probability formula is: (6) where Q is the user query and S is the retrieval result.

[0056] Risk warning: trigger threshold alarm (such as optical power falling by more than 5% for 3 consecutive times): (6) As a key infrastructure supporting the new power system, power communication network faces challenges such as massive terminal access, high-frequency information interaction, and terminal control extension. The traditional management and control mode has bottlenecks in data cognition, state monitoring, and intelligent decision-making. Large language models, with their excellent natural language processing and data analysis capabilities, can deeply mine the potential laws of power communication data, effectively improve system reliability, operation efficiency, and decision-making intelligence, and provide innovative technical paths for the digital transformation of the power industry.

[0057] The deep application of large language models in the field of power communication requires the construction of a professional power communication knowledge base as the underlying support. The power communication system covers a large amount of device parameters, network architecture, protocol standards, and operation and maintenance experience, and its complexity and professionalism require the knowledge base to have a structured and dynamic knowledge system to accurately cover device management, fault diagnosis, and business arrangement. By integrating device technical documents, historical operation and maintenance records, expert experience, and industry standards, the knowledge base not only provides high-quality training corpus for large language models, enhancing their understanding of power communication professional terms, logical relationships, and business rules, but also builds a full-link intelligent foundation covering "data-knowledge-decision". Its importance lies in: on the one hand, breaking down information silos by unifying knowledge standards, achieving efficient association and semantic analysis of cross-system data; on the other hand, supporting large language models to complete intelligent question answering, fault tracing, and strategy deduction, significantly improving the knowledge reuse efficiency of operation and maintenance personnel, reducing the risk of human error, and providing interpretable decision-making basis for network dynamic optimization and resource intelligent scheduling, accelerating the paradigm upgrade of power communication from experience-driven to data-intelligent, and laying a solid foundation for the digital transformation of the new power system.

[0058] Aiming at the problem that mass heterogeneous data (such as device logs, network topology, operation and maintenance work order, natural language description of fault report) in power backbone communication network is difficult to be efficiently integrated and intelligently utilized, the application provides a power backbone communication knowledge base construction method, realizing semantic understanding of data, dynamic knowledge fusion and intelligent decision support.

[0059] In another embodiment of the application, a large model knowledge base construction system for power communication management and operation is provided, which can be used to implement the large model knowledge base construction method for power communication management and operation described above. Specifically, the system comprises: A data acquisition module is configured to acquire structured and unstructured data in power communication management and operation, and clean and standardize the data through a distributed computing framework. A vector output module is configured to perform entity extraction, relationship identification and knowledge mapping based on the processed data using a field fine-tuning large language model, and output entity relationship triplets and semantic embedding vectors. A storage module is configured to store the entity relationship triplets and semantic embedding vectors into a graph database and a vector database respectively, forming a hybrid storage architecture. A decision output module is configured to realize decision support based on the hybrid storage architecture, combined with a hybrid retrieval strategy and a generative large language model.

[0060] The division of modules in the embodiments of the application is illustrative, and is only a logical functional division. In actual implementation, another division mode can be used. In addition, each functional module in each embodiment of the application can be integrated in one processor, or can be physically separated, or two or more modules can be integrated in one module. The integrated module can be realized in the form of hardware or in the form of a software functional module.

[0061] In another embodiment of the present application, a computer device is provided, which comprises a processor and a memory for storing a computer program comprising program instructions, and the processor is configured to execute the program instructions stored in the computer storage medium. The processor can be a central processing unit (CPU), and can 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 gate or transistor logic devices, discrete hardware components, etc., which are the computing core and control core of the terminal, and are suitable for implementing one or more instructions, and are specifically suitable for loading and executing one or more instructions in the computer storage medium to implement a corresponding method process or a corresponding function; the processor in the embodiments of the present application can be used for the operation of the large model knowledge base construction method for power communication management operation.

[0062] In another embodiment of the present application, the present application further provides a storage medium, specifically a computer readable storage medium (Memory), which is a memory device in a computer device, and is used for storing programs and data. It can be understood that the computer readable storage medium herein can include an internal storage medium in the computer device, and of course can also include an expansion storage medium supported by the computer device. The computer readable storage medium provides a storage space, and the storage space stores an operating system of the terminal. Moreover, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space, and these instructions can be one or more computer programs (including program codes). It should be noted that the computer readable storage medium herein can be a high-speed RAM memory, or a non-volatile memory such as at least one disk memory. One or more instructions stored in the computer readable storage medium can be loaded and executed by the processor to implement the corresponding steps of the large model knowledge base construction method for power communication management operation in the above embodiments.

[0063] Those skilled in the art will appreciate that embodiments of the application can be devised for a method, a system, or a computer program product. Accordingly, the present application can be embodied in the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application can take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage devices, etc.) embodying computer readable program code.

[0064] The present application is described in reference to the flowchart and / or block diagrams of the method, apparatus (system) and computer program product according to embodiments of the application. It will be understood that each block of the flowchart and / or block diagrams, and combinations of blocks in the flowchart and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks.

[0065] These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the function specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks.

[0066] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks.

[0067] Finally, it should be noted that the above-mentioned embodiments are merely intended for describing the technical solutions of the present application, but not for limiting it. Although the present application is described in detail with reference to the above embodiments, those skilled in the field should understand that the specific embodiments of the present application can be modified or replaced equivalently without departing from the spirit and scope of the present application, and any modification or equivalent replacement without departing from the spirit and scope of the present application should be covered in the protection scope of the claims of the present application.

Claims

1. A method for constructing a large-scale model knowledge base for power communication management and maintenance, characterized in that: include: Collect structured and unstructured data in power communication management and maintenance, and clean and standardize the data through a distributed computing framework; Based on the processed data, a domain-fine-tuned large language model is used to perform entity extraction, relation recognition and knowledge mapping, and output entity relation triples and semantic embedding vectors. Entity relation triples and semantic embedding vectors are stored in graph databases and vector databases respectively, forming a hybrid storage architecture; Decision support is achieved by combining a hybrid storage architecture with a hybrid retrieval strategy and a generative large language model.

2. The method for constructing a large-scale model knowledge base for power communication management and maintenance according to claim 1, characterized in that, The collected structured and unstructured data in power communication management and maintenance includes: Structured data includes equipment status and network topology, while unstructured data includes maintenance work order texts and repair records. After being collected through TMS, professional network management of optical transmission systems, and offline forms, structured and unstructured data are cleaned and standardized using a distributed computing framework.

3. The method for constructing a large-scale model knowledge base for power communication management and maintenance according to claim 2, characterized in that, Data cleaning involves filtering outliers using thresholds; data normalization uses min-max scaling to eliminate dimensional differences, as shown below: in, and These are the minimum and maximum values ​​of the feature; Unstructured text preprocessing converts maintenance work order text into JSON format and extracts key fields such as time, device ID, and description.

4. The method for constructing a large-scale model knowledge base for power communication management and maintenance according to claim 1, characterized in that, Based on the processed data, a domain-fine-tuned large language model is used for entity extraction, relation recognition, and knowledge mapping, outputting entity relation triples and semantic embedding vectors, including: We use a domain-fine-tuned large model LLM to perform semantic parsing on unstructured text and extract entities and relationships. Entity recognition uses a bidirectional Transformer model to label entities such as device names and fault types, with cross-entropy as the loss function. in, For real labels, Predict probabilities for the model; Relation extraction calculates the degree of association between entities using attention weights and outputs entity-relation-entity triples. Semantic vector generation uses the LLM's labeled hidden state as the text embedding vector.

5. The method for constructing a large-scale model knowledge base for power communication management and maintenance according to claim 1, characterized in that, The method of storing entity relation triples and semantic embedding vectors in a graph database and a vector database respectively, forming a hybrid storage architecture, includes: Knowledge graph construction: Triples are stored in a graph database; Semantic embedding vector storage: Text embedding vectors are stored in a vector database, and cosine similarity is used for similarity calculation. in, v 1 represents the text embedding vector; v 2 represents the query vector; Incremental updates are implemented based on a distributed data stream engine, with the following update rules: in, For the knowledge base at time t, This is newly added data.

6. The method for constructing a large-scale model knowledge base for power communication management and maintenance according to claim 1, characterized in that, The decision support based on a hybrid storage architecture, combined with a hybrid retrieval strategy and a generative large language model, includes: Hybrid retrieval strategies include keyword retrieval, semantic retrieval, and graph reasoning: Keyword retrieval uses Elasticsearch to match device structured fields; Semantic retrieval uses vector similarity to find relevant cases; Graph inference calculates the overall risk probability of the fault propagation path: in, Let be the failure probability of the i-th edge in the path; Generative Large Language Model (LLM) is used for generative question answering. When a user queries "How to resolve fiber optic cable interruptions?", the LLM generates an answer based on the search results, with the probability formula as follows: Where Q represents the user query and S represents the search results.

7. The method for constructing a large-scale model knowledge base for power communication management and maintenance according to claim 6, characterized in that, Decision support includes root cause analysis, response recommendations, and risk warnings, with risk warnings triggered by threshold alarm rules. Trigger threshold alarm: Alert indicates a warning message; This represents the probability of a failure occurring.

8. A large-scale model knowledge base construction system for power communication management and maintenance, characterized in that: include: The data acquisition module is used to collect structured and unstructured data in power communication management and maintenance, and to clean and standardize the data through a distributed computing framework. The vector output module is used to perform entity extraction, relation recognition and knowledge mapping based on the processed data using a domain-fine-tuned large language model, and output entity relation triples and semantic embedding vectors. The storage module is used to store entity relation triples and semantic embedding vectors to the graph database and vector database respectively, forming a hybrid storage architecture; The decision output module is used to provide decision support based on a hybrid storage architecture, combined with hybrid retrieval strategies and generative large language models.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method for constructing a large model knowledge base for power communication management and maintenance as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the method for constructing a large model knowledge base for power communication management and maintenance as described in any one of claims 1 to 7.

Images