Nuclear power plant risk identification and management and control measure generation method

Abstract

The invention particularly relates to a nuclear power plant risk identification and management and control measure generation method. The method comprises the following steps: receiving nuclear power plant risk description data input by a user; preprocessing nuclear power plant risk description data input by a user, and inputting the preprocessed data into the LangChain framework integrated with the large language model; and the LangChain framework retrieves related data from a nuclear power plant risk management and control knowledge base constructed offline according to the nuclear power plant risk description data and performs reasoning to generate corresponding management and control measures. The invention further relates to a nuclear power plant risk identification and management and control measure generation device, computer equipment and a storage medium. According to the invention, nuclear power plant risk identification and management and control measure generation are automatically realized.

Description

Technical Field

[0001] The present invention relates to the technical field of nuclear power plant risk management and control, and in particular to a method, device, computer equipment and storage medium for generating risk identification and management and control measures for a nuclear power plant. Background Art

[0002] With the rapid development of the nuclear power industry, the safety and operating efficiency of nuclear power plants have become the focus of the industry. Nuclear power plant operations are highly complex, and effective risk identification and control measures are the key to ensuring safe operation. Traditional nuclear power plant risk identification and control measures generation methods usually rely on human experience and manual analysis, which has the following problems:

[0003] 1. Manual recognition may be limited by subjective factors and individual abilities, resulting in inconsistent and incomplete recognition results;

[0004] 2. Manual analysis is time-consuming and labor-intensive, and cannot respond promptly to challenges of changing risks and increasing complexity;

[0005] 3. When faced with large amounts of complex data and changing operating environments, traditional methods often find it difficult to achieve efficient generation and updating of management and control measures, thus affecting the safety and efficiency of nuclear power plant operations. Summary of the invention

[0006] In view of the fact that the existing nuclear power plant risk identification and control measure generation method is overly dependent on manual labor, there are problems such as incomplete risk identification, untimely risk response, low decision-making accuracy and efficiency. The present invention provides a nuclear power plant risk identification and control measure generation system, which utilizes the powerful natural language processing capability and modular design construction of the large language model Lang Chain framework to automatically realize the nuclear power plant risk identification and control measure generation, reduce the interference and errors of human factors, ensure the comprehensive identification of nuclear power plant risks, improve the nuclear power plant risk response speed and the accuracy and efficiency of decision-making; provide users with valuable reference when filling in nuclear power plant risks and control measures, effectively help users avoid potential nuclear power plant risks, and ensure the safe and stable operation of nuclear power plants; ultimately greatly improve the efficiency and accuracy of nuclear power plant risk management, and provide a strong guarantee for the safe operation of the nuclear power industry.

[0007] In order to achieve the above object, the present invention provides a method for generating risk identification and control measures for a nuclear power plant, comprising the following steps:

[0008] Receive the nuclear power plant risk description data input by the user; pre-process the nuclear power plant risk description data input by the user and input it into the LangChain framework integrating the large language model;

[0009] Based on the risk description data of nuclear power plants, the LangChain framework retrieves relevant data from the offline nuclear power plant risk management knowledge base and infers it to generate corresponding control measures.

[0010] As one possible approach, the risk description data of nuclear power plants include events or situations in nuclear power plant operations;

[0011] Preprocessing the nuclear power plant risk description data input by the user includes the following steps:

[0012] Clean the nuclear power plant risk description data to remove noise data that is irrelevant to the nuclear power plant risk description;

[0013] Perform word segmentation on the cleaned nuclear power plant risk description data, segment the sentences in the nuclear power plant risk description data into words, and perform part-of-speech tagging on the words;

[0014] Perform grammatical analysis on the statements and resolve their grammatical structure; perform vectorized processing on the statements.

[0015] As one of the feasible ways, building a risk management knowledge base for nuclear power plants offline includes the following steps:

[0016] Collect historical risk management and control data of nuclear power plants; determine the text format required by the risk management and control knowledge base of nuclear power plants; convert the historical risk management and control data of nuclear power plants into the text format required by the risk management and control knowledge base of nuclear power plants to form text;

[0017] The text is segmented and chunked to obtain chunks; the chunks are converted into vector data using a semantic vector model; the vector data is stored in a vector database to form a nuclear power plant risk management knowledge base;

[0018] The historical risk management and control data of nuclear power plants include event descriptions of historical risks of nuclear power plants, experience feedback and operational recommendations.

[0019] As one of the possible ways to achieve this, the BCEmbedding semantic vector model is used to convert the chunk into vector data.

[0020] As one of the implementable ways, the vector databases include Faiss, Annoy and Elasticsearch, with Faiss being preferred.

[0021] As one of the feasible ways, the LangChain framework integrates a large language model, including the following steps:

[0022] Install the Lang Chain framework;

[0023] Configure a large language model;

[0024] Create a prompt word template for risk identification in nuclear power plants;

[0025] The prompt word template and large language model for nuclear power plant risk identification are combined into the Chain module of the LangChain framework, and the Chain module is encapsulated through the Agents module of the LangChain framework; user input data is passed in by calling the interface provided by the Chain module.

[0026] As one of the feasible ways, the large language model is ChatGLM; LangChain framework retrieves relevant data from the offline nuclear power plant risk management knowledge base and makes inference analysis based on the risk description data of the nuclear power plant, and generates corresponding control measures, including the following steps:

[0027] The LangChain framework performs similarity matching between the nuclear power plant risk description vector data and the vector data in the nuclear power plant risk management knowledge base, and selects the topk vector data that are most similar to the nuclear power plant risk description vector data;

[0028] The LangChain framework adds the topk vector data selected as context and nuclear power plant risk description data to the prompt word template for nuclear power plant risk identification, generates prompt words for nuclear power plant risk identification and submits them to the large language model;

[0029] The large language model infers the risk identification prompt words of the nuclear power plant, generates corresponding control measures and returns them to the user through the LangChain framework.

[0030] As one of the possible ways, the LangChain framework matches similarity by calculating cosine similarity.

[0031] As one of the feasible ways, the method for generating risk identification and control measures for nuclear power plants also includes: fine-tuning the large language model integrated in the LangChain framework;

[0032] Fine-tuning the large language model integrated into the LangChain framework includes the following steps:

[0033] Convert the data format in the nuclear power plant risk management knowledge base into a format suitable for large language model training as a fine-tuning dataset;

[0034] Train large language models by fine-tuning datasets.

[0035] As one of the feasible ways, the method for generating risk identification and control measures for nuclear power plants also includes: collecting user feedback data, and optimizing the large language model according to the user feedback data;

[0036] Collect user feedback data and optimize the large language model based on the user feedback data, including the following steps:

[0037] Collect user feedback data; user feedback data includes user satisfaction and accuracy with nuclear power plant risk identification results and generated control measures;

[0038] Analyze user feedback data to identify deficiencies or areas for improvement of large language models;

[0039] Optimize the large language model based on the identified deficiencies or improvements of the large language model, including adjusting model parameters, updating training data, and optimizing prompt word templates;

[0040] The optimized large language model is reintegrated into the LangChain framework, and user feedback data continues to be collected to form an iterative cycle.

[0041] In order to achieve the above-mentioned object, the present invention also provides a nuclear power plant risk identification and control measure generation device, comprising:

[0042] The input and preprocessing module is used to receive the nuclear power plant risk description data input by the user; preprocess the nuclear power plant risk description data input by the user and then input it into the risk identification and control measures generation module;

[0043] The knowledge base construction module is used to construct the risk management knowledge base of nuclear power plants offline;

[0044] The risk identification and control measures generation module is equipped with a LangChain framework that integrates a large language model. It is used to retrieve relevant data from the nuclear power plant risk control knowledge base and infer and generate corresponding control measures based on the nuclear power plant risk description data;

[0045] Model fine-tuning module, used to fine-tune the large language model integrated in the LangChain framework;

[0046] The data management and model optimization module is used to store, update and maintain the risk management knowledge base of nuclear power plants; collect user feedback data and optimize the large language model integrated in the LangChain framework based on the user feedback data.

[0047] In order to achieve the above-mentioned purpose, the present invention also provides a computer device, including a memory and a processor, wherein the memory stores computer-readable instructions, and when the processor executes the computer-readable instructions, the steps of the above-mentioned nuclear power plant risk identification and control measure generation method are implemented.

[0048] In order to solve the above technical problems, the present invention provides a computer-readable storage medium, on which computer-readable instructions are stored. When the computer-readable instructions are executed, the steps of the above-mentioned nuclear power plant risk identification and control measure generation method are implemented.

[0049] Beneficial technical effects of the present invention:

[0050] The method, device, computer equipment and storage medium for generating risk identification and control measures for nuclear power plants of the present invention use the natural language processing and large language model capabilities of the LangChain framework and a deep learning algorithm to learn massive data, thereby realizing intelligent analysis of risk description data input by users, quickly identifying possible risks, and significantly improving the accuracy and efficiency of risk identification; at the same time, targeted control measures are generated in real time, greatly shortening the risk response time and ensuring the safety and stability of nuclear power plant operation; this intelligent risk management method not only reduces the manual burden, but also avoids the interference of human factors and improves the overall management efficiency; it supports intelligent learning and adaptive optimization, and can continuously optimize its identification ability and suggestion effect with the accumulation of time and data, ensuring that nuclear power plants maintain a high level of safety and operating efficiency in complex and changing environments, directly reducing the incidence of risk events in the operation of nuclear power plants, and reducing potential safety hazards, thereby significantly improving the safe operation level of nuclear power plants. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 A flowchart of an embodiment of a method for generating risk identification and control measures for a nuclear power plant according to the present invention;

[0052] Figure 2 It is a schematic structural view of an embodiment of a nuclear power plant risk identification and control measure generation device of the present invention.

[0053] Figure 3 The figure is a schematic structural view of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0054] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by technicians in the technical field to which this application belongs; the terms used in the specification of the application are only for the purpose of describing specific embodiments and are not intended to limit this application; the terms "including" and "having" and any variations thereof in the specification and claims of this application and the above-mentioned figure descriptions are intended to cover non-exclusive inclusions.

[0055] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0056] The technical solution of the present invention is clearly and completely described below in conjunction with the accompanying drawings and specific embodiments.

[0057] refer to Figure 1 , shows a flow chart of an embodiment of a method for generating risk identification and control measures for a nuclear power plant, the method for generating risk identification and control measures for a nuclear power plant comprising the following steps:

[0058] Receive the nuclear power plant risk description data input by the user; pre-process the nuclear power plant risk description data input by the user and input it into the LangChain framework integrating the large language model;

[0059] Based on the risk description data of nuclear power plants, the LangChain framework retrieves relevant data from the offline nuclear power plant risk management knowledge base and infers it to generate corresponding control measures.

[0060] In this embodiment, as one of the achievable methods, the nuclear power plant risk description data includes events or situations in the operation of the nuclear power plant, such as equipment failure or abnormal operation.

[0061] Preprocessing the nuclear power plant risk description data input by the user to ensure that the nuclear power plant risk description data input by the user can be effectively processed later. In this embodiment, as one of the achievable methods, preprocessing the nuclear power plant risk description data input by the user includes the following steps:

[0062] Clean the nuclear power plant risk description data and remove the noise data irrelevant to the nuclear power plant risk description to ensure the neatness of the nuclear power plant risk description data; the noise data irrelevant to the nuclear power plant risk description includes characters, symbols and spaces irrelevant to the nuclear power plant risk description;

[0063] Perform word segmentation on the cleaned nuclear power plant risk description data, and segment the continuous sentences in the nuclear power plant risk description data into words to facilitate subsequent analysis;

[0064] Tag the words with parts of speech to reveal their functions in sentences; parts of speech include nouns and verbs;

[0065] Perform grammatical analysis on the statements, resolve the grammatical structure of the statements, and further understand the relationships in the statements; perform vectorized processing on the statements.

[0066] By preprocessing the nuclear power plant risk description data input by the user, the unstructured and complex data form can be converted into a structured and easy-to-process data form, providing strong support for the subsequent risk identification and generation of control measures.

[0067] The construction of a nuclear power plant risk management knowledge base is a systematic process, involving multiple links such as text format specification, text processing and extraction. In this embodiment, as one of the feasible methods, the offline construction of a nuclear power plant risk management knowledge base includes the following steps:

[0068] Collect historical risk management and control data of nuclear power plants; determine the text format required by the risk management and control knowledge base of nuclear power plants; convert the historical risk management and control data of nuclear power plants into the text format required by the risk management and control knowledge base of nuclear power plants to form text;

[0069] The text is segmented and chunked to obtain chunks; the chunks are converted into vector data using a semantic vector model;

[0070] The vector data is stored in the vector database to form a risk management and control knowledge base for nuclear power plants.

[0071] For large texts, direct chunk processing may become infeasible due to input limitations such as the token limit of the model. In this case, the large text needs to be split into smaller parts. These parts should maintain semantic integrity as much as possible, such as by paragraph or chapter. The purpose of segmentation is to ensure that the size of each piece of text is suitable for model processing while minimizing the loss of contextual information.

[0072] Chunking text means dividing a large text into smaller, more manageable and processable text blocks. Dividing the entire text into multiple text blocks allows each text block to be independently processed such as text cleaning, word segmentation and vectorization.

[0073] After the text is cut into blocks, each block needs to be converted into a numerical vector. The semantic vector model can be used to achieve the vector representation of the text block. These vectors capture the deep semantic features of the text, making it possible to compare, search and analyze texts.

[0074] The vector data is stored in the vector database for subsequent retrieval and analysis. During the offline process of the nuclear power plant risk management knowledge base, attention should be paid to the accuracy and completeness of the data to ensure that the nuclear power plant risk management knowledge base constructed offline can meet the risk management needs of the nuclear power plant.

[0075] In this embodiment, as one of the feasible ways, the historical risk management and control data of the nuclear power plant includes event descriptions, experience feedback and operational suggestions of the historical risks of the nuclear power plant.

[0076] The historical risk management and control data of nuclear power plants is converted into the text format required by the nuclear power plant risk management and control knowledge base to facilitate unified management and processing. In this embodiment, as one of the achievable methods, the text format required by the nuclear power plant risk management and control knowledge base includes Markdown, PDF, WORD and TXT.

[0077] In this embodiment, as one of the achievable methods, the BCEmbedding semantic vector model is used to convert the chunk into vector data.

[0078] The BCEmbedding semantic vector model has powerful bilingual and cross-language semantic representation capabilities in Chinese and English, and is mainly composed of two parts: EmbeddingModel and RerankerModel; EmbeddingModel is responsible for generating semantic vectors, which are mainly used for semantic search and question-answering tasks; RerankerModel is responsible for optimizing semantic search results and accurate sorting. The EmbeddingModel adopts a dual encoder structure, which enables it to achieve efficient semantic retrieval in the first stage, which is especially important for processing large amounts of data and quickly responding to user query requests. The RerankerModel adopts a cross encoder structure, which can perform more in-depth semantic analysis and precise sequence arrangement in the second stage. This feature makes the BCEmbedding semantic vector model more efficient and accurate when processing complex semantic understanding tasks.

[0079] In this embodiment, as one of the implementable ways, the vector databases include Faiss, Annoy and Elasticsearch, which optimize the storage and similarity search operations of vector data.

[0080] In this embodiment, as one of the preferred methods, the vector database is Faiss. The index mechanism provided by Faiss is used to efficiently store vector data in Faiss for subsequent fast similarity search or retrieval. Faiss achieves efficient storage and query of large-scale vector data by optimizing the index structure and vector quantization technology, thereby supporting the application of text vectors in the fields of semantic search, recommendation systems, etc.

[0081] In this embodiment, as one of the feasible ways, the LangChain framework integrates a large language model, including the following steps:

[0082] Install the Lang Chain framework;

[0083] Configure a large language model;

[0084] Create a prompt word template for risk identification in nuclear power plants;

[0085] The prompt word template and large language model for nuclear power plant risk identification are combined into the Chain module of the LangChain framework, and the Chain module is encapsulated through the Agents module of the LangChain framework; user input data is passed in by calling the interface provided by the Chain module.

[0086] Through the above steps, developers can efficiently integrate large language models into their own applications and implement complex language processing functions.

[0087] The LangChain framework is a framework built around a large language model, which aims to simplify the development difficulty of large language model applications. It provides a set of tools, components, and interfaces to support the integration of large language models with other data sources, tools, etc., so as to build complex end-to-end applications. The LangChain framework itself does not develop large language models, but links different large language models by implementing common interfaces.

[0088] The LangChain framework is an open source Python library designed to simplify the integration and application development of large language models. By installing the LangChain framework through pip, you can easily obtain the tools and components required for large language models.

[0089] LangChain supports the integration of large language models, such as ChatGLM and Llama. During the integration process, you need to select a suitable large language model and configure the corresponding API key or authentication information so that the LangChain framework can successfully call the large language model for reasoning.

[0090] The prompt template is an important concept in the LangChain framework. It defines how to send requests to the large language model and helps convert user input into a format suitable for processing by the large language model. Developers can customize the prompt word template or select the predefined prompt word template provided by the LangChain framework according to specific needs. The present invention designs a prompt word template for nuclear power plant risk identification, which can guide the large language model to understand the risk description of the nuclear power plant and generate corresponding risk identification results.

[0091] Using the Chain module of the LangChain framework, components such as the prompt word template, large language model, and output parser are combined into an orderly execution process. The Chain module supports modular design, allowing developers to flexibly combine different components according to actual needs to achieve complex language processing tasks. The Agents module of the LangChain framework further encapsulates the Chain module so that it can execute action decisions based on user input and feedback. Finally, by calling the interface provided by the Chain module and passing in user input data, the LangChain framework will automatically execute the entire processing flow and return the response result of the large language model. Developers can further process or display the response results as needed.

[0092] In this embodiment, as one of the implementable methods, the large language model is ChatGLM;

[0093] Based on the risk description data of nuclear power plants, the LangChain framework retrieves relevant data from the offline nuclear power plant risk management knowledge base and performs reasoning and analysis to generate corresponding management and control measures, including the following steps:

[0094] The LangChain framework performs similarity matching between the nuclear power plant risk description vector data and the vector data in the nuclear power plant risk management knowledge base, and selects the topk vector data that are most similar to the nuclear power plant risk description vector data;

[0095] The LangChain framework adds the topk vector data selected as context and nuclear power plant risk description data to the prompt word template for nuclear power plant risk identification, generates prompt words for nuclear power plant risk identification and submits them to the large language model;

[0096] The large language model infers the risk identification prompt words of the nuclear power plant, generates corresponding control measures and returns them to the user through the LangChain framework.

[0097] ChatGLM uses the Transformer model as its core architecture. The Transformer model is a neural network structure based on the self-attention mechanism. It achieves efficient encoding and decoding of input sequences through multi-layer self-attention mechanisms and attention weights. Compared with traditional recurrent neural networks and convolutional neural networks, the Transformer model has better performance in processing long sequences and complex contexts.

[0098] In this embodiment, as one of the feasible ways, the LangChain framework matches the similarity by calculating the cosine similarity. A is the risk description vector data of the nuclear power plant, and B is a vector data of a vector database. The similarity between the two is:

[0099] The fine-tuning process of a large language model usually involves adaptive adjustment of the pre-trained model for a specific task. In order to achieve fine-tuning in combination with the nuclear power plant risk management knowledge base, in this embodiment, as one of the achievable methods, the method for generating risk identification and control measures for a nuclear power plant also includes: fine-tuning the large language model integrated with the LangChain framework;

[0100] Fine-tuning the large language model integrated into the LangChain framework includes the following steps:

[0101] Convert the data format in the nuclear power plant risk management knowledge base into a format suitable for large language model training as a fine-tuning dataset;

[0102] The large language model is trained by fine-tuning the dataset so that it can understand and apply this knowledge.

[0103] During the fine-tuning process, the model's parameters and structure can be optimized to improve the model's ability to identify, analyze and respond to nuclear power plant risk measures. Therefore, the fine-tuned large language model can better serve the risk management needs of nuclear power plants.

[0104] In this embodiment, as one of the achievable ways, the method for generating risk identification and control measures for a nuclear power plant further includes: collecting user feedback data, and optimizing a large language model according to the user feedback data;

[0105] Collect user feedback data and optimize the large language model based on the user feedback data, including the following steps:

[0106] Collect user feedback data such as user satisfaction and accuracy of nuclear power plant risk identification results and generated control measures;

[0107] Analyze user feedback data to identify deficiencies or areas for improvement of large language models;

[0108] Optimize the large language model based on the identified deficiencies or improvements of the large language model, including adjusting model parameters, updating training data, and optimizing prompt word templates to improve the accuracy and efficiency of the large language model;

[0109] The optimized large language model is reintegrated into the LangChain framework, and user feedback data continues to be collected to form an iterative cycle.

[0110] Through continuous user feedback data collection and large language model optimization, the performance of the large language model in risk identification and control measure generation in nuclear power plants is continuously improved.

[0111] refer to Figure 2As an implementation of the above method, the present invention provides an embodiment of a nuclear power plant risk identification and control measure generation device. The device embodiment corresponds to the above embodiment of the nuclear power plant risk identification and control measure generation method. The device can be specifically applied to various electronic devices.

[0112] The nuclear power plant risk identification and control measures generation device described in this embodiment includes:

[0113] The input and preprocessing module is used to receive the nuclear power plant risk description data input by the user; preprocess the nuclear power plant risk description data input by the user and then input it into the risk identification and control measures generation module;

[0114] The knowledge base construction module is used to construct the risk management knowledge base of nuclear power plants offline;

[0115] The risk identification and control measures generation module is equipped with a LangChain framework that integrates a large language model. It is used to retrieve relevant data from the nuclear power plant risk control knowledge base and infer and generate corresponding control measures based on the nuclear power plant risk description data;

[0116] Model fine-tuning module, used to fine-tune the large language model integrated in the LangChain framework;

[0117] The data management and model optimization module is used to store, update and maintain the risk management knowledge base of nuclear power plants; collect user feedback data and optimize the large language model integrated in the LangChain framework based on the user feedback data.

[0118] As an implementation of the above method, the present invention provides an embodiment of a computer device, and the embodiment of the computer device corresponds to the embodiment of the above method for generating risk identification and control measures for nuclear power plants.

[0119] The computer device described in this embodiment includes a memory, a processor and a network interface that are interconnected and communicated through a system bus. It should be noted that the figure only shows a computer device with a memory, a processor and a network interface, but it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to a microprocessor, an application-specific integrated circuit, a programmable gate array, a digital processor, an embedded device, etc.

[0120] The computer device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The computer device may interact with a user through a keyboard, a mouse, a remote controller, a touch pad, or a voice control device.

[0121] The memory includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory, random access memory, static random access memory, read-only memory, electrically erasable programmable read-only memory, programmable read-only memory, magnetic memory, disk, optical disk, etc. In some embodiments, the memory may be an internal storage unit of the computer device, such as a hard disk or memory of the computer device. In other embodiments, the memory may also be an external storage device of the computer device, such as a plug-in hard disk, smart memory card, secure digital card, flash memory card, etc. equipped on the computer device. Of course, the memory may also include both the internal storage unit of the computer device and its external storage device. In this embodiment, the memory is generally used to store the operating system and various application software installed on the computer device, such as the computer-readable instructions of the above-mentioned nuclear power plant risk identification and control measures generation method. In addition, the memory may also be used to temporarily store various types of data that have been output or are to be output.

[0122] In some embodiments, the processor may be a central processing unit, a controller, a microcontroller, a microprocessor, or other data processing chip. The processor is generally used to control the overall operation of the computer device. In this embodiment, the processor is used to run computer-readable instructions stored in the memory or process data, such as running computer-readable instructions of the above-mentioned nuclear power plant risk identification and control measures generation method.

[0123] The network interface may include a wireless network interface or a wired network interface, which is generally used to establish a communication connection between the computer device and other electronic devices.

[0124] As an implementation of the above method, the present invention provides an embodiment of a computer-readable storage medium, and the embodiment of the computer-readable storage medium corresponds to the embodiment of the above-mentioned nuclear power plant risk identification and control measure generation method.

[0125] The computer-readable storage medium described in this embodiment stores computer-readable instructions, and the computer-readable instructions can be executed by at least one processor to enable the at least one processor to perform the steps of the above-mentioned nuclear power plant risk identification and control measures generation method.

[0126] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A method for generating risk identification and control measures for a nuclear power plant, characterized in that: The steps include: Receive the nuclear power plant risk description data input by the user; pre-process the nuclear power plant risk description data input by the user and input it into the LangChain framework integrating the large language model; Based on the risk description data of nuclear power plants, the LangChain framework retrieves relevant data from the offline nuclear power plant risk management knowledge base and infers it to generate corresponding control measures.

2. The method for generating risk identification and control measures for a nuclear power plant according to claim 1, characterized in that: The nuclear power plant risk description data includes events or situations in the operation of the nuclear power plant; preprocessing the nuclear power plant risk description data input by the user includes the following steps: Clean the nuclear power plant risk description data to remove noise data that is irrelevant to the nuclear power plant risk description; Perform word segmentation on the cleaned nuclear power plant risk description data, segment the sentences in the nuclear power plant risk description data into words, and perform part-of-speech tagging on the words; Perform grammatical analysis on the statements and resolve their grammatical structure; perform vectorized processing on the statements.

3. The method for generating risk identification and control measures for a nuclear power plant according to claim 1, characterized in that: Offline construction of a nuclear power plant risk management knowledge base includes the following steps: Collect historical risk management and control data of nuclear power plants; determine the text format required by the risk management and control knowledge base of nuclear power plants; convert the historical risk management and control data of nuclear power plants into the text format required by the risk management and control knowledge base of nuclear power plants to form text; The text is segmented and chunked to obtain chunks; the chunks are converted into vector data using a semantic vector model; the vector data is stored in a vector database to form a nuclear power plant risk management knowledge base; The historical risk management and control data of nuclear power plants include event descriptions of historical risks of nuclear power plants, experience feedback and operational recommendations.

4. The method for generating risk identification and control measures for a nuclear power plant according to claim 1, characterized in that: The LangChain framework integrates a large language model, including the following steps: Install the Lang Chain framework; Configure a large language model; Create a prompt word template for risk identification in nuclear power plants; The prompt word template and large language model for nuclear power plant risk identification are combined into the Chain module of the LangChain framework, and the Chain module is encapsulated through the Agents module of the LangChain framework; user input data is passed in by calling the interface provided by the Chain module.

5. The method for generating risk identification and control measures for a nuclear power plant according to claim 4, characterized in that: The large language model is ChatGLM. Based on the risk description data of nuclear power plants, the LangChain framework retrieves relevant data from the offline nuclear power plant risk management knowledge base, performs reasoning and analysis, and generates corresponding management and control measures, including the following steps: The LangChain framework performs similarity matching between the nuclear power plant risk description vector data and the vector data in the nuclear power plant risk management knowledge base, and selects the top k vector data that are most similar to the nuclear power plant risk description vector data; The LangChain framework adds the filtered topk vector data as context and nuclear power plant risk description data to the prompt word template for nuclear power plant risk identification, generates prompt words for nuclear power plant risk identification and submits them to the large language model; The large language model infers the risk identification prompt words of the nuclear power plant, generates corresponding control measures and returns them to the user through the LangChain framework.

6. The method for generating risk identification and control measures for a nuclear power plant according to claim 1, characterized in that: Also includes: Fine-tune the large language model integrated by the LangChain framework; Fine-tune the large language model integrated by the LangChain framework, including the following steps: Convert the data format in the nuclear power plant risk management knowledge base into a format suitable for large language model training as a fine-tuning dataset; Train large language models by fine-tuning datasets.

7. The method for generating risk identification and control measures for a nuclear power plant according to claim 1, characterized in that: Also includes: Collecting user feedback data and optimizing the large language model based on the user feedback data; Collecting user feedback data and optimizing the large language model based on the user feedback data include the following steps: Collect user feedback data; user feedback data includes user satisfaction and accuracy with nuclear power plant risk identification results and generated control measures; Analyze user feedback data to identify deficiencies or areas for improvement of large language models; Optimize the large language model based on the identified deficiencies or improvements of the large language model, including adjusting model parameters, updating training data, and optimizing prompt word templates; The optimized large language model is reintegrated into the LangChain framework, and user feedback data continues to be collected to form an iterative cycle.

8. A device for generating risk identification and control measures for a nuclear power plant, characterized in that: include: An input and preprocessing module, used for receiving nuclear power plant risk description data input by a user; Pre-process the nuclear power plant risk description data input by the user and input it into the risk identification and control measures generation module; The knowledge base construction module is used to construct the risk management knowledge base of nuclear power plants offline; The risk identification and control measures generation module is equipped with a LangChain framework that integrates a large language model. It is used to retrieve relevant data from the nuclear power plant risk control knowledge base and infer and generate corresponding control measures based on the nuclear power plant risk description data; Model fine-tuning module, used to fine-tune the large language model integrated in the LangChain framework; Data management and model optimization module, used for storage, updating and maintenance of nuclear power plant risk management knowledge base; Collect user feedback data and optimize the large language model integrated in the LangChain framework based on the user feedback data.

9. A computer device comprising a memory and a processor, wherein the memory stores computer-readable instructions, characterized in that: When the processor executes the computer-readable instructions, the steps of the method for generating risk identification and control measures for a nuclear power plant as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having computer-readable instructions stored thereon, characterized in that: When the computer-readable instructions are executed, the steps of the method for generating risk identification and control measures for a nuclear power plant as described in any one of claims 1 to 7 are implemented.

Images