Nuclear power operation data mining method and system based on large model

Abstract

The invention belongs to the technical field of nuclear power operation data management, and discloses a nuclear power operation data mining method and system based on a large model, and the method comprises the steps of data collection, preprocessing, conversion from a natural language to a database language (such as NL2SQL), database query, feature extraction, result analysis and visualization and the like. A large model is adopted to process nuclear power operation data, equipment monitoring data and the like, and prediction of potential faults, energy efficiency analysis and optimization decision support are achieved; the system comprises a data layer, a model layer, an analysis layer and an application layer. According to the method, the natural language problem can be efficiently processed, the problem proposed by the user can be accurately understood, and the matched data can be quickly found and organized into the natural language to be output.

Description

Technical Field

[0001] The present application belongs to the technical field of nuclear power operation data management, and in particular, relates to a nuclear power operation data mining method and system based on a large model. Background Art

[0002] With the rapid development of my country's nuclear energy industry, the amount of operating data of nuclear power units has shown explosive growth. These data contain rich information, which is of great significance for improving the safety, economy and reliability of nuclear power unit operation. However, traditional data analysis methods are inefficient in processing complex data and cannot effectively extract valuable information. How to mine valuable information from massive operating data has become a major challenge facing the current nuclear power industry.

[0003] At present, the nuclear power operation data analysis method mainly relies on manual experience and small model analysis, which has the following limitations:

[0004] (1) Manual analysis is inefficient, time-consuming and labor-intensive, and it is difficult to cope with massive amounts of data. During the operation of nuclear power plants, a large amount of monitoring data is generated, such as reactor power, coolant temperature and other parameters. These data are not only huge in quantity, but also require real-time analysis to monitor the status of the nuclear power plant. Manual analysis is not only time-consuming and labor-intensive, but also prone to errors. Especially when the amount of data surges, the speed of manual processing is far behind the speed of data generation, resulting in the possibility that key information may be omitted or delayed in processing. In addition, manual analysis will also have certain subjective factors, and there may be differences in analysis between different technicians, affecting accuracy.

[0005] (2) Traditional methods mainly rely on structured data (e.g., tabular data in a database) and are limited by the size and complexity of the data set, making it difficult to handle more complex or real-time changing data scenarios. However, unstructured data (e.g., log files, images, and videos) also contain important information, and traditional methods are difficult to effectively process these unstructured data.

[0006] (3) Small models (e.g., linear regression, decision trees) have limited analytical accuracy and are unable to capture the deep patterns in the data. Small models often cannot accurately capture the deep patterns and complex patterns in the data, resulting in inaccurate analysis results. In addition, small models are easily disturbed when processing noisy data.

[0007] (4) Existing methods have certain limitations in data preprocessing, feature extraction, and model building, which lead to inaccurate analysis results. Data preprocessing is an important step in data analysis, but existing technologies still have shortcomings in data cleaning, missing value processing, and outlier detection.

[0008] It can be seen that the existing nuclear power operation data analysis methods face many limitations, which not only limit the efficiency and accuracy of data analysis, but may also affect the safe operation and efficiency optimization of nuclear power plants.

[0009] As an advanced artificial intelligence technology, large language model (LLM) has made significant breakthroughs in many aspects in recent years and has attracted widespread attention both domestically and internationally. It can provide technical support for nuclear power operation data mining. First, by integrating large-scale parallel processing mechanisms, large models can quickly process and analyze massive data, effectively support cross-domain data integration, and build a comprehensive and interconnected data view. This data integration helps to identify cross-domain data patterns and provide a more comprehensive basis for decision-making. For example, large models can analyze equipment performance data, predict potential failures and provide early warnings, thereby significantly reducing the unexpected downtime of nuclear power plants. In addition, large models show extremely high flexibility and efficiency in processing complex operational data, such as calculating customer-level and regional-level indicators, providing in-depth operational insights, and optimizing resource allocation and operational strategies. Large models can also provide deeper data insights for nuclear power operation management platforms. These models have the ability to process and analyze various types of data (such as sensor data, operation logs, environmental data, etc.) and extract valuable information from them, which is crucial for understanding complex operating modes and identifying potential risks. At the same time, by optimizing data quality and integrity, the big model ensures the accuracy and reliability of the data, providing solid support for the safe and efficient operation of nuclear power plants. The application of these technologies not only improves the operation and maintenance efficiency of nuclear power plants, but also enhances the ability to cope with future uncertainties, which is the key to the nuclear power industry moving towards a higher technical level. Summary of the invention

[0010] The purpose of this application is to overcome the defects of the prior art and to provide a nuclear power operation data mining method and system based on a large model to achieve prediction of potential nuclear power failures, energy efficiency analysis and optimization decision support.

[0011] In order to achieve the above objectives, this application provides the following technical solutions:

[0012] In a first aspect, the present application provides a nuclear power operation data mining method based on a large model, comprising:

[0013] Step 1: Collect nuclear power operation data and establish a data aggregation center;

[0014] Step 2: Pre-process the data in the data aggregation center, store them in different databases according to data types, establish a nuclear power operation database, and obtain a sample data set;

[0015] Step 3: Use the sample data set and prompt words to pre-train and fine-tune the general large model base to obtain the nuclear power inspection large model;

[0016] Step 4: Obtain the problem to be solved. The nuclear power inspection model converts natural language into low-dimensional vectors, understands the problem and generates a database language to solve the problem.

[0017] Step 5: Call the corresponding API to execute the database language query and return the required data to the nuclear power inspection model;

[0018] Step 6: The nuclear power inspection model optimizes the query results, generates explanations in natural language to facilitate user understanding, and returns them to the user;

[0019] Step 7: Customize the visualization scheme to obtain the desired result presentation format.

[0020] In step 1, various data forms are obtained from the monitoring system of the nuclear power plant, historical operation records, equipment operating condition hidden danger data, etc.

[0021] In step 2, data preprocessing includes cleaning, denoising and normalizing the collected nuclear power operation data.

[0022] In step 2, building a nuclear power operation database includes:

[0023] Provide critical data aggregation and integration support, covering all operations from data storage to data management;

[0024] The data is stored in different databases according to the data type, and the relational database MySQL is used to store basic information.

[0025] In step 3, the universal large model base is fine-tuned using a chain of thought approach.

[0026] In step 5, different database languages ​​call different API interfaces to link to the corresponding database, execute database statements to query the required data, and return it to the nuclear power inspection model.

[0027] In step 6, the nuclear power inspection big model will perform secondary optimization on the query results and generate a paragraph of explanation of the query results in natural language. The big model will be used to generate a paragraph to describe the numerical and text results in the query form to facilitate user understanding.

[0028] In a second aspect, the present application provides a nuclear power operation data mining system based on a large model, comprising:

[0029] The data layer is used to aggregate and integrate data, covering all operations from data storage to data management;

[0030] The model layer is used to process and execute all AI-related tasks;

[0031] The analysis layer is used to apply large models for in-depth data mining and analysis, automatically extracting valuable business insights from data through natural language processing and intelligent reasoning;

[0032] The application layer is used to convert analysis results into specific business decisions and operational recommendations.

[0033] In a third aspect, the present application provides an electronic device including a memory and a processor, wherein computer-readable instructions are stored in the memory, and the above-mentioned method is implemented when the processor executes the computer-readable instructions.

[0034] In a fourth aspect, the present application provides a computer-readable storage medium, in which computer-readable instructions are stored, and the above-mentioned method is implemented when the computer-readable instructions are executed.

[0035] Compared with the prior art, the nuclear power operation data mining method and system based on large models provided by the present application have the following beneficial effects:

[0036] The method includes data collection, preprocessing, translation of natural language into database language (such as NL2SQL), database query, feature extraction, result analysis and visualization, etc. It uses large models to process nuclear power operation data, equipment monitoring data, etc. to achieve prediction of potential faults, energy efficiency analysis and optimization decision support.

[0037] This application can efficiently process natural language questions, accurately understand the questions raised by users, and quickly find matching data and organize it into natural language output.

[0038] Furthermore, the present application has data visualization capabilities, using large models to quickly analyze query results, calling visualization tools to generate charts, and speeding up the data visualization process.

[0039] Furthermore, the application has a simple and easy-to-use interface design, focuses on user experience, and continuously optimizes system functions and performance based on user usage habits and needs.

[0040] Furthermore, the present application has high performance and can process large amounts of data in a short period of time.

[0041] Furthermore, the present application has good scalability and can be upgraded and expanded in the future as demand grows.

[0042] Furthermore, during data processing and transmission, this application ensures the security of user data to prevent leakage and tampering; for sensitive information, appropriate encryption and desensitization measures are taken to protect user privacy. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solution of the present application, the following is a brief introduction to the drawings required for the technical description.

[0044] Figure 1 A flowchart of a nuclear power operation data mining method based on a large model provided for this application;

[0045] Figure 2 Schematic diagram of the nuclear power intelligent inspection assistant architecture provided for this application. DETAILED DESCRIPTION

[0046] The following is further explained in detail through specific implementation methods.

[0047] like Figure 1 and Figure 2 As shown, the present application provides a nuclear power operation data mining method based on a large model, which uses intelligent means to efficiently and accurately mine, analyze and utilize the massive data generated during the nuclear power operation process, including:

[0048] Step 1: Collect nuclear power operation data (data aggregation) and establish a data aggregation center;

[0049] Step 2: Pre-process the data in the data aggregation center, store them in different databases according to data types, establish a nuclear power operation database, and obtain a sample data set;

[0050] Step 3: Use the sample data set and prompt words to pre-train and fine-tune the general large model base to obtain the nuclear power inspection large model;

[0051] Step 4: Obtain the problem to be solved. The nuclear power inspection model converts natural language into low-dimensional vectors, understands the problem and generates a database language to solve the problem.

[0052] Step 5: Call the corresponding API to execute the database language query and return the required data to the nuclear power inspection model;

[0053] Step 6: The nuclear power inspection model optimizes the query results, generates explanations in natural language to facilitate user understanding, and returns them to the user;

[0054] Step 7: Users can customize the visualization scheme to obtain the desired result presentation format.

[0055] In step 1, collecting nuclear power operation data includes: collecting data related to nuclear power operation from various monitoring systems, sensor networks, log recording systems, etc. of nuclear power plants, including cooling system parameters, equipment maintenance records, fault alarm information, etc.

[0056] In step 1, various data forms are obtained from the monitoring system, historical operation records, equipment operating condition hidden danger data, etc. of the nuclear power plant, including production process data, equipment status data, personnel data, nuclear energy consumption data, inventory data, etc. Exemplarily, this application focuses on nuclear power operation data forms, including structured and unstructured data.

[0057] In step 1, a data aggregation center is established to receive data from different sources and ensure data integrity.

[0058] In step 2, data preprocessing includes: cleaning, denoising and normalizing the collected nuclear power operation data to eliminate outliers, fill missing values, unify data formats, and improve data quality. Data exists in different formats, including WORD\PDF\TXT\JSON\EXCEL, etc., and files are used for data extraction. Different processing tools are provided according to the format, such as OCR to extract text from PDF and WORD, and EXCEL is converted to CSV format files through a front-end program. Different word segmentation tools are selected according to the language of the document for combination, and participate in recognition processes such as part-of-speech tagging and named entity recognition.

[0059] In step 2, classified storage is performed, and the preprocessed data is stored in different databases according to the data type to form a nuclear power operation database system.

[0060] In step 2, building a nuclear power operation database includes: providing key data collection and integration support, covering all operations from data storage to data management, including multiple data storage methods such as time series database, vector database, relational database, document-based database, etc. Data is stored in different databases according to data type. For example, nuclear power equipment operating conditions are stored in InfluxDB2.0, graphic data are stored in Faiss, and relational database MySQL is used to store basic information such as entities, attributes, and relationships; time series data is usually stored in high-performance time series databases such as InfluxDB to ensure fast retrieval and efficient analysis of real-time data; data such as reactor core outlet temperature and system overall pressure are converted into high-dimensional vectors and stored in vector databases (Milvus or Faiss), etc. At the same time, the constructed database can provide fast similarity search and complex data analysis functions, making the retrieval and analysis of real-time data more efficient and helping to quickly respond to user needs.

[0061] In step 2, representative samples are extracted from the database to construct a sample data set for model training to ensure that the data set can fully reflect the actual situation of nuclear power operation.

[0062] Furthermore, this application uses feature extraction and deep learning and other technologies to extract high-dimensional features of data, providing a basis for subsequent analysis, responsible for language understanding, complex data analysis and reasoning tasks. It is carried out after data collection, cleaning and preprocessing, converting raw data into a form that can be used to train models or further analyze, with the purpose of:

[0063] 1) Remove redundant, irrelevant or noisy features to simplify the data and focus on effective information;

[0064] 2) Convert complex data, especially structured data, into an easily understandable format;

[0065] 3) Extract meaningful features from raw data to help subsequent analysts and models discover hidden information in the data.

[0066] After feature extraction, the data enters the exploratory data analysis stage, and through visualization, statistical analysis and other methods, common sense is used to understand the distribution, relationships and potential patterns of the data, and insights are given to assist decision-making when users need it (natural language commands are sufficient).

[0067] In step 3, a general large model with strong language understanding and generation capabilities is selected or developed as the base, such as the GPT series, BERT series, etc. Using the constructed sample data set and specific prompt words for the nuclear power field (for example, professional terms, common query patterns), the general large model is pre-trained and fine-tuned to enable it to deeply understand the complex issues related to nuclear power operation and generate accurate answers, thereby building a nuclear power inspection large model.

[0068] In one embodiment, the Chain of Thought (CoT) is used to fine-tune the general large model base ChatGLM (Zhipu Qingyan). The specific code used is as follows:

[0069] SystemPrompt="""

[0070] 'You are an experienced SQL expert who is proficient in database management.\n'

[0071] 'Please write efficient and optimized SQL queries according to the following requirements:\n'

[0072] '{{input}}\n'

[0073] 'Note: If there are conditions, be sure to use the `WHERE` clause to perform conditional queries instead of querying the entire table, and do not add additional query conditions. For example:\n'

[0074] 'SELECT*FROM table_name WHERE condition="?";\n'

[0075] 'Note: Return only the SQL query statement, do not provide any explanation or comments.'

[0076] 'Please ensure that your query syntax is correct and follows best practices.',

[0077] ”””

[0078] Thus, a large model of nuclear power inspection is obtained.

[0079] In step 4, the user raises questions or demands about nuclear power operation in natural language, and the nuclear power inspection model converts these natural language inputs into low-dimensional vector representations for semantic analysis.

[0080] In step 4, based on the understanding of the problem, the nuclear power inspection model automatically generates the corresponding database query language to accurately locate and extract the required data.

[0081] In step 4, the inspection model understands and translates the problem and generates the database language required to solve the problem.

[0082] In step 5, the generated database query language is executed by calling the API interface provided by the database to obtain and return the required data to the nuclear power inspection model. The nuclear power inspection model further processes and optimizes the returned data, and then generates detailed explanations and reports in an easy-to-understand natural language form.

[0083] In step 5, different database languages ​​call different API interfaces to link to the corresponding database, execute database statements to query the required data, and return it to the nuclear power inspection model. The large model understands and translates natural language questions to generate the required database language (such as NL2SQL), which will call the corresponding API, such as generating SQL query statements to call and operate data stored in the MySQL database, InfluxQL to query data in the InfluxDB database, and MQL to call data stored in the MongoDB database.

[0084] In step 6, the nuclear power inspection model performs secondary optimization on the query results and generates a description of the query results in natural language to facilitate users to further understand the retrieved results.

[0085] In step 7, the user can select or design a specific visualization scheme according to his or her own needs in order to more intuitively display the analysis results of the nuclear power operation data. Then, based on the user's selection, the optimized data and analysis results will be displayed to the user in the selected visualization form to assist the user in making decisions or taking corresponding measures.

[0086] In step 7, in order to better display data trends and assist in data analysis, the Nuclear Power Intelligent Inspection Assistant provides a variety of custom visualization solutions, and users can freely choose the drawing format, including line charts, bar charts, pie charts, funnel charts, etc.

[0087] In addition, based on the above-mentioned nuclear power operation data mining method based on a large model, the present application also provides a nuclear power operation data mining system based on a large model, including:

[0088] Data layer: The data layer provides key data collection and integration support for the entire platform, covering all operations from data storage to data management. The data layer includes a variety of data storage methods such as relational databases, time series databases, document-based databases, vector databases, and graph databases, etc., to ensure effective integration and efficient access of data, while supporting the intelligent functions of the upper application layer.

[0089] Model layer: The model layer is the core of the platform and is responsible for processing and executing all AI-related tasks. The model layer integrates a variety of large models, including public cloud models, private cloud large models, and fine-tuned large models, responsible for language understanding, complex data analysis and reasoning tasks, and provides high-quality AI services for specific fields such as nuclear power plant operation and maintenance.

[0090] Analysis layer: Apply large models for in-depth data mining and analysis, and automatically extract valuable business insights from data through natural language processing and intelligent reasoning. These analysis results help managers understand the operating status and potential risks of nuclear power plants and optimize resource allocation and operation strategies.

[0091] Application layer: Transforms analysis results into specific business decisions and operational recommendations. Through visualization tools and user-friendly interfaces, complex data analysis results are easy to understand and operate. In addition, the application layer is responsible for the real-time application of large models to ensure efficient and safe operation of nuclear power plants.

[0092] like Figure 2As shown in the figure, a variety of storage engines such as time series database, graph database, vector database, etc. are used to normalize and semantically associate heterogeneous inspection data, build a nuclear power database covering all elements such as equipment, components, personnel, processes, and faults, and a data mining system - nuclear power inspection intelligent assistant, which is applied to nuclear power industry inspection scenarios. When the inspection personnel need to count the work of the approval personnel, they only need to ask the nuclear power intelligent inspection assistant in natural language "how many times each approval personnel appeared", and the big model will analyze the semantics of the problem in real time. Based on various databases, the results of the retrieval are first displayed intuitively, and then the visualization method of "bar chart, pie chart" selected by the user is drawn. When the inspection personnel have doubts about the equipment status, they only need to ask the nuclear power intelligent inspection assistant in natural language "the temperature change of a certain electrical equipment, and draw a line chart", and the nuclear power intelligent inspection assistant can give a clear picture of the equipment temperature change over time.

[0093] In the nuclear power industry inspection scenario based on big model technology, the inspection personnel are equipped with intelligent devices (intelligent handheld terminals or smart helmets) and use multimodal sensors on nuclear power industrial equipment to collect massive data such as vibration, temperature, and images of equipment operation in real time, and upload them to the nuclear power data center. The data center uses multiple storage engines such as time series database, graph database, and vector database to normalize and semantically associate heterogeneous inspection data, and build a nuclear power knowledge graph covering all elements such as equipment, components, personnel, processes, and faults.

[0094] Exemplarily, the nuclear power operation data mining method based on the large model provided in this application includes the following steps:

[0095] S10, database construction;

[0096] S11, data aggregation: various data forms are obtained from the monitoring system of the nuclear power plant, historical operation records, equipment operating condition hidden danger data, etc., including production process data, equipment status data, personnel data, nuclear energy consumption data, inventory data, etc. The present invention focuses on nuclear power operation data forms, including structured and unstructured data;

[0097] S12. Data preprocessing: Clean, denoise and normalize the collected nuclear power operation data to improve data quality; data exists in different formats, including WORD\PDF\TXT\JSON\EXCEL, etc., and files are used for data extraction. Different processing tools are provided according to the format, such as OCR to extract text from PDF and WORD, and EXCEL is converted into CSV format files through a front-end program, etc. Different word segmentation tools are selected according to the language of the document for combination, and participate in recognition processes such as part-of-speech tagging and named entity recognition;

[0098] S13. Build a database: Provide key data collection and integration support, covering all operations from data storage to data management, including a variety of data storage methods such as time series databases, vector databases, relational databases, document-structured databases, etc. Store the data in step 12 in different databases according to data type. For example, the operating conditions of nuclear power equipment are stored in InfluxDB2.0, graphic data are stored in Faiss, and the relational database MySQL is used to store basic information such as entities, attributes, and relationships; time series data is usually stored in high-performance time series databases, such as InfluxDB, to ensure fast retrieval and efficient analysis of real-time data; data such as reactor core outlet temperature and system overall pressure are converted into high-dimensional vectors and stored in vector databases (Milvus or Faiss), etc. At the same time, the constructed database can provide fast similarity search and complex data analysis functions, making the retrieval and analysis of real-time data more efficient, and helping to quickly respond to user needs;

[0099] S20, nuclear power inspection large model: pre-training and fine-tuning the general large model base using the database obtained in step S10;

[0100] S21. Fine-tuning of the large model: In the present invention, the general large model base ChatGLM (Smart Spectrum Clear Words) is fine-tuned mainly by using the Chain of Thought (CoT) method, thereby obtaining a large model for nuclear power inspection;

[0101] S30, nuclear power intelligent inspection assistant: composed of the nuclear power operation database constructed in step S10 and the nuclear power inspection large model trained in step S20;

[0102] S31. The user asks questions in natural language, and the nuclear power intelligent inspection assistant converts the natural language extracted keywords into low-dimensional vectors and enters the nuclear power inspection model;

[0103] S32, the inspection model understands and translates the problem, and generates the database language required to solve the problem;

[0104] S33. Different database languages ​​call different API interfaces to link to corresponding databases, and execute database statements to query required data, and return them to the nuclear power inspection model;

[0105] S34. The nuclear power inspection model performs secondary optimization on the query results and generates a description of the query results in natural language to facilitate users to further understand the retrieved results;

[0106] S35. Data visualization: In order to better display data trends and assist in data analysis, the Nuclear Power Intelligent Inspection Assistant provides a variety of custom visualization solutions. Users can freely choose the drawing format, including line charts, bar charts, pie charts, funnel charts, etc.

[0107] In addition, the present application also provides an electronic device, including a memory and a processor, wherein the memory stores computer-readable instructions, and when the processor executes the computer-readable instructions, the above-mentioned large-model-based nuclear power operation data mining method is implemented.

[0108] In addition, the present application also provides a computer-readable storage medium, on which computer-readable instructions are stored. When the computer-readable instructions are executed, the above-mentioned large-model-based nuclear power operation data mining method is implemented.

[0109] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device.

[0110] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed in the present application should be covered within the protection scope of the present application.

Claims

1. A nuclear power operation data mining method based on a large model, characterized in that: include: Step 1: Collect nuclear power operation data and establish a data aggregation center; Step 2: Preprocess the data in the data aggregation center, store them in different databases according to data types, establish a nuclear power operation database, and obtain a sample data set; Step 3: Use the sample data set and prompt words to pre-train and fine-tune the general large model base to obtain the nuclear power inspection large model; Step 4: Obtain the problem to be solved. The nuclear power inspection model converts natural language into low-dimensional vectors, understands the problem and generates a database language to solve the problem. Step 5: Call the corresponding API to execute the database language query and return the required data to the nuclear power inspection model; Step 6: The nuclear power inspection model optimizes the query results, generates explanations in natural language to facilitate user understanding, and returns them to the user; Step 7: Customize the visualization scheme to obtain the desired result presentation format.

2. The method for nuclear power operation data mining based on a large model according to claim 1 is characterized in that: In step 1, various data forms are obtained from the monitoring system of the nuclear power plant, historical operation records, equipment operating condition hidden danger data, etc.

3. The method for nuclear power operation data mining based on a large model according to claim 1, characterized in that: In step 2, data preprocessing includes cleaning, denoising and normalizing the collected nuclear power operation data.

4. The method for nuclear power operation data mining based on a large model according to claim 1, characterized in that: In step 2, building a nuclear power operation database includes: Provide critical data aggregation and integration support, covering all operations from data storage to data management; The data is stored in different databases according to the data type, and the relational database MySQL is used to store basic information.

5. The method for nuclear power operation data mining based on a large model according to claim 1, characterized in that: In step 3, the universal large model base is fine-tuned using a chain of thought approach.

6. The method for nuclear power operation data mining based on a large model according to claim 1, characterized in that: In step 5, different database languages ​​call different API interfaces to link to the corresponding database, execute database statements to query the required data, and return it to the nuclear power inspection model.

7. The method for nuclear power operation data mining based on a large model according to claim 1, characterized in that: In step 6, the nuclear power inspection model performs secondary optimization on the query results and generates a description of the query results in natural language.

8. A nuclear power operation data mining system based on a large model, characterized in that: include: The data layer is used to aggregate and integrate data, covering all operations from data storage to data management; The model layer is used to process and execute all AI-related tasks; The analysis layer is used to apply large models for in-depth data mining and analysis, automatically extracting valuable business insights from data through natural language processing and intelligent reasoning; The application layer is used to convert analysis results into specific business decisions and operational recommendations.

9. An electronic device, characterized in that: The method comprises a memory and a processor, wherein the memory stores computer-readable instructions, and the processor implements the method according to any one of claims 1 to 7 when executing the computer-readable instructions.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-readable instructions, and when the computer-readable instructions are executed, the method according to any one of claims 1 to 7 is implemented.

Images