A Pumped Storage Power Station Construction and Management System Based on a Large Language Model

Abstract

The present invention belongs to the technical field of energy storage power station construction management, and discloses a pumped storage power station construction management system based on a large language model. The present invention generates a large language model data database for pumped storage power stations through a model construction module, and trains, evaluates, and fine-tunes the large language model to generate a large language model for pumped storage power station construction management. The intelligent AI assistant module processes input questions and generates answers, drawings, and plans based on the database. The model correction module adjusts the model and corrects the generated plans. The resource library module stores the generated plans and iterates accurate versions. The authorization management module implements user access control, and the mobile terminal module is used to convert web terminal functions. The present invention can realize question and answer of professional knowledge of pumped storage power stations, reduce the management difficulty of managers, and generate plans and drawings according to user needs.

Description

Technical Field

[0001] This invention belongs to the field of pumped storage power station construction and management technology, specifically relating to a pumped storage power station construction and management system based on a large language model. Background Technology

[0002] Large language models are deep learning-based language processing models with powerful computing capabilities. They integrate multiple modal information and can process various types of data simultaneously, such as text, images, and audio, enabling cross-modal understanding and generation. They have wide applications in natural language processing, computer vision, speech recognition, and other fields, and are commonly used in tasks such as voice assistants, intelligent customer service, machine translation, and image description generation, providing people with more intelligent and convenient services and experiences.

[0003] In addition to the advantages mentioned above, large language models also have obvious drawbacks, mainly including the following two points: (1) they cannot provide relatively accurate answers for related professional fields; (2) they cannot generate drawings for the characteristics of pumped storage power stations; (3) the results are highly dependent on the training data. If the training data is insufficient, the generated results will be inaccurate. Currently, there are many large language models on the market. These large language models can answer users' routine questions, but their response results related to pumped storage power stations are almost meaningless. Summary of the Invention

[0004] This invention aims to address the shortcomings of existing technologies by providing a pumped storage power station construction management system based on a large language model. This system integrates technologies such as professional data integration, data preprocessing, intelligent AI, large language models, and natural language processing, enabling question-and-answer sessions on professional knowledge related to pumped storage power stations, reducing the management difficulty for managers, and generating plans and drawings based on user needs.

[0005] To achieve the above technical objectives, the present invention adopts the following technical solution:

[0006] A pumped storage power station construction and management system based on a large language model includes a model building module, an intelligent AI assistant module, a model correction module, a resource library module, and an authorization management module.

[0007] The model building module is used to collect data and generate a large language model database for pumped storage power stations. Based on the database, the Transformer neural network model architecture is selected, and Zero-shot Learning, Few-shot Learning and RLHF technologies are combined to generate a large language model for the construction and management of pumped storage power stations.

[0008] The intelligent AI assistant module is used to connect to the model building module, process the input questions to obtain the input sequence, extract the correlation between the input sequences using a self-attention mechanism, match the solutions that meet the requirements based on the correlation, and generate drawings and plans based on the large language model data database of pumped storage power stations.

[0009] The model correction module is used to correct and adjust the large language model for pumped storage power station construction and management generated by the model building module, and to conduct industry expert correction and adjustment of the solutions, drawings and schemes generated by the intelligent AI assistant module.

[0010] The resource library module is used to store the answers, drawings and solutions generated by the intelligent AI assistant module. Users interact with the intelligent AI assistant module based on the content stored in the resource library. The system iterates and generates an accurate version of the stored content according to the user's needs, and then stores the accurate version again.

[0011] The authorization management module is used to input user attributes into the system through dialogue between the user and the intelligent AI assistant module, and to control user access based on the user attributes.

[0012] Preferably, the specific content executed by the model building module includes:

[0013] A1. Basic Data Support: Collect basic data on pumped storage power stations to form a large language model data database for pumped storage power stations;

[0014] A2. Data Preprocessing: Data preprocessing is performed on the large language model data database of pumped storage power stations. Database deduplication and semantic similarity calculation methods are used to clean the text in the database, delete duplicate or irrelevant content, and format the preprocessed database.

[0015] A3. Model Generation: Select the Transformer neural network model architecture, and choose the number of parameters, hidden units and hyperparameters based on its ability to process sequential data and capture long-term dependencies. Combine Zero-shot Learning, Few-shot Learning and RLHF techniques to generate a large language model.

[0016] A4. Model Training: The large language model is trained using a formatted database. During training, the next word in a sentence is predicted based on the previous one or several words. Backpropagation and stochastic gradient descent optimization algorithms are used to adjust the parameters of the large language model to generate a large language model for the construction and management of pumped storage power stations.

[0017] A5. Evaluation and Adjustment: Evaluate the model performance using perplexity and downstream task performance metrics, and adjust the model parameters.

[0018] Preferably, the specific content executed by the intelligent AI assistant module includes:

[0019] B1. Using the intelligent AI assistant module to connect to the model building module, a large language model for the construction and management of pumped storage power stations is obtained. The problem is input into the intelligent AI assistant module. The intelligent AI assistant module obtains the input content of the problem, performs contextual understanding and extracts the input content, performs tagging and splitting of text content, and performs semantic and structural splitting of image and document content to obtain a tagging, semantic and structural input sequence.

[0020] B2. Execute the self-attention mechanism: Obtain the input sequence in execution content B1, perform a linear transformation, use the self-attention mechanism to obtain the attention weight distribution, and weight each element in the input sequence according to the distribution weight to extract the correlation between the input content and form each initial representation and the degree of representation correlation.

[0021] B3. Process the initial representation and representation correlation obtained in execution content B2, reduce the judgment of noun part of speech in the initial representation, improve the judgment of verb part of speech, infer the probability of the next word appearing, and obtain an input sequence with a representation correlation greater than the preset value. Then, through the similarity matching method, match the input sequence with the large language model data database of pumped storage power stations to generate a solution that meets the requirements.

[0022] B4. Input the input sequence obtained in execution content B1 into the large language model for the construction and management of pumped storage power stations, and generate drawings using image rendering methods based on the large language model data database of pumped storage power stations.

[0023] B5. Input the input sequence obtained from execution content B1 into the large language model for pumped storage power station construction and management, obtain the solution based on the large language model database for pumped storage power stations, and convert the solution into PDF or WORD format using the AIOffice method.

[0024] Preferably, the specific steps for executing content B4 are as follows:

[0025] B401. Perform a linear transformation on the labeled, semantic, and structured input sequence in execution content B1. Input the linearly transformed input sequence into the model. Use image rendering methods to obtain the first set of engineering drawings of the required quantity. Then, perform labeling and encoding processing on the first set of engineering drawings and compare and analyze it with the input sequence to obtain the first set of differences.

[0026] B402. The input sequence, the first set of engineering drawings after labeling and encoding, and the first set of differences are fused together. Based on the fused input sequence, the first set of engineering drawings after labeling and encoding, and the first set of differences, an image rendering method is used to generate a second set of engineering drawings. The second set of engineering drawings is then labeled and encoded and compared with the input sequence to obtain the second set of differences.

[0027] B403. The input sequence, the first set of engineering drawings after labeling and encoding, the second set of engineering drawings after labeling and encoding, and the second set of differences are fused together. Based on the fused input sequence, the first set of engineering drawings after labeling and encoding, the second set of engineering drawings after labeling and encoding, and the second set of differences, an image rendering method is used to generate a third set of engineering drawings.

[0028] B404. Merge the first set of engineering drawings, the second set of engineering drawings, and the third set of engineering drawings to obtain the total set of engineering drawings. Tag and encode the total set of engineering drawings, compare and analyze it with the input sequence, select several engineering drawings with differences less than the preset value, record them as the set of drawings that meet the requirements, and apply scale and coded content annotations to the drawings.

[0029] Preferably, the specific steps for executing content B5 are as follows:

[0030] B501. Perform a linear transformation on the tagged, semantic, and structured input sequence in execution content B1 to generate a scheme outline set. Perform tagged and semantic processing on each level of the scheme outline set. Merge the tagged and semantic processed scheme outline sets and input them into the model. Based on the input sequence and the tagged and semantic processed scheme outline set, match it with the large language model data database of pumped storage power stations to obtain the content corresponding to each level of the scheme outline.

[0031] B502. Tag, semanticize, and structure the content corresponding to each level of the matched scheme outline, generate drawings using image rendering methods, generate multiple attributes for the drawings by combining the input sequence content, and further match the drawing content to generate drawing number, drawing association description, and drawing remarks.

[0032] B503 uses the Blade template engine and WebKit / Blink rendering engine to arrange and combine the content of each level of the matched scheme outline and the generated drawings to obtain the scheme. The scheme is then converted into PDF or WORD format using AIOffice.

[0033] Preferably, the model correction module performs the following tasks: integrating professional data and expertise; adjusting the large language model for pumped storage power station construction and management by adjusting the large language model and prompt words; and using a restricted knowledge base to perform industry expert correction on the generated solutions, drawings, and plans. Specifically:

[0034] C1. Adjusting the large language model: Collect information on the pumped storage industry, generate a database for a specified domain, train the large language model for the construction and management of pumped storage power stations based on the database for the specified domain, adjust the model’s performance in NPL tasks in a specific domain, and achieve model self-correction.

[0035] C2. Adjust prompt words: Further realize model self-correction by designing template technology;

[0036] C3. Restricted Knowledge Base: This feature involves data analysis, cleaning, correction, and transformation of the database of large-scale language model data for pumped storage power stations. Invalid data is removed, and valid data is tagged. Combined with precise answers from industry experts, in-depth sentence correction and corresponding accuracy correction are performed to achieve industry expert correction of the generated solutions, drawings, and plans.

[0037] Preferably, the specific steps for executing content C2 are as follows:

[0038] C201, Design Template: Input a problem into a large language model for the construction and management of pumped storage power stations. Extract prompt words from the input problem. Assuming the extracted prompt word is x, use the design template technique to transform the prompt word x into X containing empty slots. Fill the empty slots of X multiple times with different fill values ​​to obtain a set X of X containing different fill values. 集 ;

[0039] C202, Answer Search: For set X 集 By searching the database of large language models for pumped storage power station construction and management, multiple answers were obtained.

[0040] C203, Answer Correction: For the multiple answers obtained in step C202, the optimal answer is found through deep traversal, realizing model self-correction.

[0041] Preferably, the specific content executed by the resource repository module includes:

[0042] D1. Set the question as the initial key point, retrieve the closest preset question and answer through the initial key point, use the retrieved closest preset question and answer as context, and concatenate it with the question to generate the final key point;

[0043] D2. Labeling the final key points to obtain the corresponding labels. Based on the final key points and their corresponding labels, asynchronous addressing is performed using a large language model combined with the rsync algorithm.

[0044] D3. Utilize Flink's asynchronous addressing IO technology in both unordered and ordered modes to store the content required in the context as the generated result in the resource library.

[0045] Preferably, the specific content executed by the authorization management module includes:

[0046] E1. Users communicate with the intelligent AI assistant module in natural language. The intelligent AI assistant module obtains text, images, and documents in the context, performs tagging and attribute processing, extracts user attribute parameters, and creates an underlying user table based on the user attribute parameters.

[0047] E2. The user communicates with the intelligent AI assistant module in natural language. The system compares the content obtained by the intelligent AI assistant module with the underlying user table to obtain the difference content, and compares whether the user is included in the underlying user table. If the user is included, the difference content is extended to the underlying user table to complete the entry and editing of the user table.

[0048] E2. By communicating with the intelligent AI assistant module in natural language, it can obtain text, images, documents, and other content from the context to complete user access control and authentication.

[0049] Preferably, the pumped storage power station construction management system based on a large language model also includes a mobile terminal module, and the model building module, intelligent AI assistant module, model correction module, resource library module and authorization management module are all web-based functions;

[0050] The mobile module is used to convert web-based functions into mobile-based functions. These mobile functions, combined with the intelligent AI assistant module, enable question answering and resource retrieval from the resource library module.

[0051] Preferably, the specific content executed by the mobile module includes:

[0052] By linking the web and mobile terminals through data communication technology, the functions and data of the web terminal are transformed, and the intelligent AI assistant module works with the functions of the mobile terminal to answer questions, generate solutions, and complete resource acquisition.

[0053] Compared with the prior art, the beneficial effects of the present invention are:

[0054] (1) This invention integrates technologies such as data integration, data preprocessing, intelligent AI, large language model and natural language processing to ensure the intelligent response capability of the intelligent AI assistant module;

[0055] (2) This invention uses the massive data of pumped storage power stations as basic data support, which can realize Q&A of professional knowledge of pumped storage power stations, suggestions for improvement of construction content during the construction period, etc., which greatly reduces the management difficulty of managers.

[0056] (3) The present invention enables administrators to input user information and roles into the system simply by talking to the intelligent AI assistant module, thereby providing authorization management services to users and applications, realizing the user access control mechanism, replacing the traditional web-based management process, ensuring that relevant personnel can quickly get started and use the system, and simplifying the development and maintenance of the system.

[0057] (4) This invention utilizes natural language processing (NPL) to implement a dialogue system, combining large language models and machine learning techniques to enable data to be transmitted quickly in both directions, greatly improving the real-time performance and convenience for users when receiving and processing messages or tasks.

[0058] (5) Based on the "Specifications for Compiling Feasibility Study Reports for Hydropower Projects" (NB / T 11013-2022) and combined with the actual use of the project, this invention establishes an intelligent management system for pumped storage power stations, providing the ability to handle professional issues for personnel at all stages and positions during the construction period, and laying the foundation for achieving the construction requirements of "digital power plants".

[0059] (6) This invention introduces online collaborative technology, which combines large language model and pumped storage power station large language model data database, and can supplement the professional issues of model response from multiple sources to enrich the model capabilities, and plays a significant role in improving the standardization and normalization of response capabilities.

[0060] (7) This invention solves the problems of difficulty in answering professional questions and backlog of repetitive work in the construction of pumped storage power stations. It realizes intelligent response and automated learning accumulation. By classifying, sorting and learning from various construction drawings of previous projects, users only need to input the relevant requirements for the construction of pumped storage power stations to generate plans and drawings. If the results are inaccurate, more parameters can be input. The model can be further adjusted on the current basis through the resource library module to obtain a drawing that meets the requirements. Technicians can easily improve the drawing and put it into use, reducing the huge amount of drawing work. The series of results can provide reference value for the digital and intelligent application of projects under construction or already built. Attached Figure Description

[0061] Figure 1 This is an architecture diagram of the pumped storage power station construction and management system based on a large language model according to the present invention.

[0062] Figure 2This is a technical roadmap for the pumped storage power station construction and management system based on a large language model, as described in an embodiment of the present invention. Detailed Implementation

[0063] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0064] Example 1

[0065] Combination Figure 1 and Figure 2 As shown, this embodiment of the invention provides a pumped storage power station construction management system based on a large language model, including a model building module, an intelligent AI assistant module, a model correction module, a resource library module, and an authorization management module;

[0066] The model building module is used to collect data and generate a large language model database for pumped storage power stations. Based on the database, the Transformer neural network model architecture is selected, and Zero-shot Learning, Few-shot Learning and RLHF technologies are combined to generate a large language model for the construction and management of pumped storage power stations.

[0067] Specifically, the model building module provides the basic support for the system. By uploading and inputting various professional data such as project information, geographic information, personnel scale, equipment data, drawing information, industry standards and specifications, documents, the role of pumped storage power stations, energy consumption, project information, and installed capacity, various professionals integrate a large amount of data collected from the Internet based on various pumped storage projects, including power station construction archives, technical requirements, personnel composition, construction progress, books, articles, websites, etc., to generate a large language model database for pumped storage power stations. The database encompasses various data information related to pumped storage power stations, and the large language model is trained based on the database. The model can automatically parse relevant data for machine learning. After learning, it can provide professional capability support in subsequent dialogues, forming a large language model for the construction and management of pumped storage power stations.

[0068] The intelligent AI assistant module is used to connect to the model building module, process the input questions to obtain the input sequence, extract the correlation between the input sequences using a self-attention mechanism, match the solutions that meet the requirements based on the correlation, and generate drawings and plans based on the large language model data database of pumped storage power stations.

[0069] Specifically, the intelligent AI assistant module can provide users with professional answers to their questions, handle user inquiries, and process sequential data according to context. It adopts a self-attention mechanism to establish the correlation between various elements in the input sequence. By encoding and decoding words or tags in the input sequence, it captures the semantic and grammatical relationships between words, thereby generating coherent and professional answers, including answering pumped storage questions, generating drawings, and generating solutions.

[0070] The model correction module is used to correct and adjust the large language model for pumped storage power station construction and management generated by the model building module, and to conduct industry expert correction and adjustment of the solutions, drawings and schemes generated by the intelligent AI assistant module.

[0071] Specifically, the model correction module can correct and adjust the model and correct the response results to the question. When a user asks a relevant question, the system's response may differ from the actual situation. Relevant professionals can correct the corresponding question and determine the standard of each answer. The system learns, summarizes and analyzes multiple answers. When encountering similar questions in the future, the system can provide users with relatively correct response content based on the correction results.

[0072] The resource library module is used to store the answers, drawings and solutions generated by the intelligent AI assistant module. Users interact with the intelligent AI assistant module based on the content stored in the resource library. The system iterates and generates an accurate version of the stored content according to the user's needs, and then stores the accurate version again.

[0073] Specifically, the resource library module can uniformly save the solutions, drawings, solutions, source code and other content generated by the system. Users can conduct in-depth dialogues with the intelligent AI assistant module based on the content stored in the resource library. The system iteratively generates more accurate and compliant versions according to the user's in-depth requirements. At the same time, the basis for the generation of each version is stored and labeled, which makes it convenient for users to uniformly manage the response results and review the process, improving the convenience of operation and the traceability of the response results.

[0074] The authorization management module is used to input user attributes into the system through dialogue between the user and the intelligent AI assistant module, and to control user access based on the user attributes;

[0075] Specifically, the authorization management module enables administrators to input user information and roles into the system simply by interacting with the intelligent AI assistant module, providing authorization management services to users and applications, implementing user access control mechanisms, replacing the traditional web-based personnel management process, and simplifying system development and maintenance;

[0076] Furthermore, the pumped storage power station construction management system based on a large language model also includes a mobile terminal module;

[0077] The model building module, intelligent AI assistant module, model correction module, resource library module, and authorization management module are all web-based functionalities.

[0078] The mobile module is used to convert web-based functions into mobile-based functions. The mobile functions, combined with the intelligent AI assistant module, complete the question answering and resource acquisition from the resource library module.

[0079] Specifically, mobile devices such as smartphones and tablets can be used to perform operations such as inquiry, answering, and solution design based on underlying model data, achieving the goal of multi-terminal integration, improving the efficiency and convenience of user use, and realizing a closed loop from basic to solution throughout the entire process and lifecycle;

[0080] The pumped storage power station construction management system based on a large language model can answer questions and generate solutions related to the construction of pumped storage power stations. For example, it can respond to the construction process of pumped storage power stations, generate construction plans for ventilation tunnels of pumped storage power stations, and answer questions about the construction precautions for flood discharge tunnels.

[0081] Example 2

[0082] like Figure 1 and Figure 2 As shown, a pumped storage power station construction and management system based on a large language model includes a model building module, an intelligent AI assistant module, a model correction module, a resource library module, and an authorization management module.

[0083] The model building module is used to collect data and generate a large language model database for pumped storage power stations. Based on the database, the Transformer neural network model architecture is selected, and Zero-shot Learning, Few-shot Learning and RLHF technologies are combined to generate a large language model for the construction and management of pumped storage power stations.

[0084] The intelligent AI assistant module is used to connect to the model building module, process the input questions to obtain the input sequence, extract the correlation between the input sequences using a self-attention mechanism, match the solutions that meet the requirements based on the correlation, and generate drawings and plans based on the large language model data database of pumped storage power stations.

[0085] The model correction module is used to correct and adjust the large language model for pumped storage power station construction and management generated by the model building module, and to conduct industry expert correction and adjustment of the solutions, drawings and schemes generated by the intelligent AI assistant module.

[0086] The resource library module is used to store the answers, drawings and solutions generated by the intelligent AI assistant module. Users interact with the intelligent AI assistant module based on the content stored in the resource library. The system iterates and generates an accurate version of the stored content according to the user's needs, and then stores the accurate version again.

[0087] The authorization management module is used to input user attributes into the system through dialogue between the user and the intelligent AI assistant module, and to control user access based on the user attributes;

[0088] Furthermore, the pumped storage power station construction management system based on a large language model also includes a mobile terminal module, while the model building module, intelligent AI assistant module, model correction module, resource library module, and authorization management module are all web-based functions.

[0089] The mobile module is used to convert web-based functions into mobile-based functions. The mobile functions, combined with the intelligent AI assistant module, complete the question answering and resource acquisition from the resource library module.

[0090] Based on this, in this embodiment, the functions performed by the model building module include:

[0091] Data preprocessing is performed on the database of large language models for pumped storage power stations. Based on the database, a large language model for the construction and management of pumped storage power stations is trained, and the parameters of the model are adjusted. The performance of the model is evaluated and the parameters are adjusted to generate a large language model for the construction and management of pumped storage power stations.

[0092] The specific execution steps are as follows:

[0093] A1. Basic Data Support: Collect basic data on pumped storage power stations to form a large language model data database for pumped storage power stations;

[0094] A2. Data Preprocessing: Data preprocessing is performed on the large language model data database of pumped storage power stations. Database deduplication and semantic similarity calculation methods are used to clean the text in the database, delete duplicate or irrelevant content, and format the preprocessed database.

[0095] A3. Model Generation: Select the Transformer neural network model architecture. Its ability to efficiently process sequential data and capture long-term dependencies determines the model size. Select the number of model parameters, hidden units, and hyperparameters, such as learning rate and batch size. Combine Zero-shot Learning, Few-shot Learning, and RLHF (Reinforcement Learning Based on Human Feedback) techniques to generate a large language model (Fuyao AI).

[0096] A4. Model Training: The large language model is trained using a formatted database. During training, the next word in a sentence is predicted based on the previous one or several words. Backpropagation and stochastic gradient descent optimization algorithms are used to adjust the parameters (weights and biases) of the large language model to generate a large language model for the construction and management of pumped storage power stations.

[0097] A5. Evaluation and Adjustment: Evaluate model performance using perplexity (a measure of the quality of model predictions) and downstream task performance metrics, adjust model parameters (specifically performance, accuracy, etc.), and implement adjustments to improve specific aspects of model performance.

[0098] Example 3

[0099] like Figure 1 and Figure 2 As shown, a pumped storage power station construction and management system based on a large language model includes a model building module, an intelligent AI assistant module, a model correction module, a resource library module, and an authorization management module.

[0100] The model building module is used to collect data and generate a large language model database for pumped storage power stations. Based on the database, the Transformer neural network model architecture is selected, and Zero-shot Learning, Few-shot Learning and RLHF technologies are combined to generate a large language model for the construction and management of pumped storage power stations.

[0101] The intelligent AI assistant module is used to connect to the model building module, process the input questions to obtain the input sequence, extract the correlation between the input sequences using a self-attention mechanism, match the solutions that meet the requirements based on the correlation, and generate drawings and plans based on the large language model data database of pumped storage power stations.

[0102] The model correction module is used to correct and adjust the large language model for pumped storage power station construction and management generated by the model building module, and to conduct industry expert correction and adjustment of the solutions, drawings and schemes generated by the intelligent AI assistant module.

[0103] The resource library module is used to store the answers, drawings and solutions generated by the intelligent AI assistant module. Users interact with the intelligent AI assistant module based on the content stored in the resource library. The system iterates and generates an accurate version of the stored content according to the user's needs, and then stores the accurate version again.

[0104] The authorization management module is used to input user attributes into the system through dialogue between the user and the intelligent AI assistant module, and to control user access based on the user attributes;

[0105] Furthermore, the pumped storage power station construction management system based on a large language model also includes a mobile terminal module, while the model building module, intelligent AI assistant module, model correction module, resource library module, and authorization management module are all web-based functions.

[0106] The mobile module is used to convert web-based functions into mobile-based functions. The mobile functions, combined with the intelligent AI assistant module, complete the question answering and resource acquisition from the resource library module.

[0107] Based on this, in this embodiment, the intelligent AI assistant module performs the following functions:

[0108] The intelligent AI assistant module is used to connect to the large language model for the construction and management of pumped storage power stations obtained in step A4, and to perform visual interaction to generate a user interface. The questions are input into the intelligent AI assistant module. After obtaining the input questions, the intelligent AI assistant module performs contextual understanding and performs content tagging, semantic and structural processing, executes a self-attention mechanism, and generates answers, drawings and solutions based on the large language model data database of pumped storage power stations.

[0109] The specific execution steps are as follows:

[0110] B1. Using the intelligent AI assistant module to connect to the model building module, a large language model for the construction and management of pumped storage power stations is obtained. The problem is input into the intelligent AI assistant module. The intelligent AI assistant module obtains the input content of the problem, performs contextual understanding and extracts the input content, performs tagging and splitting of text content, and performs semantic and structural splitting of image and document content to obtain a tagging, semantic and structural input sequence.

[0111] B2. Execute the self-attention mechanism: Obtain the input sequence in execution content B1, perform a linear transformation, use the self-attention mechanism to obtain the attention weight distribution, and weight each element in the input sequence according to the distribution weight to extract the correlation between the input content and form each initial representation and the degree of representation correlation.

[0112] B3. Generating a solution based on the large language model database of pumped storage power stations: Process the initial representations and their correlation in step B2, reduce the judgment of noun parts of speech in the initial representations, improve the judgment of verb parts of speech, infer the probability of the next word appearing, and obtain an input sequence with a correlation greater than the preset value. Then, match the input sequence with the large language model database of pumped storage power stations using a similarity matching method to generate a solution that meets the requirements.

[0113] B4. Input the input sequence obtained in execution content B1 into the large language model for the construction and management of pumped storage power stations, and generate drawings using image rendering methods based on the large language model data database of pumped storage power stations.

[0114] B5. Input the input sequence obtained in execution content B1 into the large language model for the construction and management of pumped storage power stations, obtain the solution based on the large language model data database of pumped storage power stations, and convert the solution into PDF or WORD format using the AIOffice method.

[0115] Specifically, the steps for executing content B4 are as follows:

[0116] B401. Perform a linear transformation on the labeled, semantic, and structured input sequence in execution content B1. Input the linearly transformed input sequence into the model and use object-order rendering to obtain the first set of engineering drawings of the required quantity. Then, perform labeling and encoding on the first set of engineering drawings and compare and analyze it with the input sequence. Specifically, depth-first traversal and breadth-first traversal techniques can be used for comparison and analysis to obtain the first set of differences.

[0117] B402. The input sequence, the first set of engineering drawings after labeling and encoding, and the first set of differences are fused together. Based on the fused input sequence, the first set of engineering drawings after labeling and encoding, and the first set of differences, an image rendering method is used to generate a second set of engineering drawings. The second set of engineering drawings is then labeled and encoded, and compared with the input sequence to obtain the second set of differences.

[0118] B403. The input sequence, the first set of engineering drawings after labeling and encoding, the second set of engineering drawings after labeling and encoding, and the second set of differences are fused together. Based on the fused input sequence, the first set of engineering drawings after labeling and encoding, the second set of engineering drawings after labeling and encoding, and the second set of differences, an image rendering method is used to generate a third set of engineering drawings.

[0119] B404. Merge the first set of engineering drawings, the second set of engineering drawings, and the third set of engineering drawings to obtain the total set of engineering drawings. Tag and encode the total set of engineering drawings, compare and analyze it with the input sequence, select several engineering drawings with differences less than the preset value, record them as the set of drawings that meet the requirements, and apply scale and coded content annotations to the drawings.

[0120] The specific steps for executing content B5 are as follows:

[0121] B501. Perform a linear transformation on the labeled, semantic, and structured input sequence in execution content B1 to generate a set of scheme outlines. Specifically, the set of scheme outlines can be generated using the mind chain technique. Each level of the scheme outline set is labeled and semantically processed. The labeled and semantically processed scheme outline sets are then merged and input into the model. Based on the input sequence and the labeled and semantically processed scheme outline sets, natural language processing methods are used to match the data with the large language model database of pumped storage power stations to obtain the content corresponding to each level of the scheme outline.

[0122] B502. Tag, semanticize, and structure the content corresponding to each level of the matched scheme outline, generate drawings using image rendering methods, and generate multiple attributes (such as width, height, length, etc.) for the drawings by combining the input sequence content. Specifically, using text segmentation and semantic technology, and combining the input sequence content, generate multiple attributes for the drawings, and further match the drawing content to generate drawing number, drawing association description, drawing remarks, etc.

[0123] B503 uses the Blade template engine and WebKit / Blink rendering engine to arrange and combine the content of each level of the matched scheme outline and the generated drawings to obtain the scheme. The scheme is then converted into PDF or WORD format using AIOffice.

[0124] Example 4

[0125] like Figure 1 and Figure 2 As shown, a pumped storage power station construction and management system based on a large language model includes a model building module, an intelligent AI assistant module, a model correction module, a resource library module, and an authorization management module.

[0126] The model building module is used to collect data and generate a large language model database for pumped storage power stations. Based on the database, the Transformer neural network model architecture is selected, and Zero-shot Learning, Few-shot Learning and RLHF technologies are combined to generate a large language model for the construction and management of pumped storage power stations.

[0127] The intelligent AI assistant module is used to connect to the model building module, process the input questions to obtain the input sequence, extract the correlation between the input sequences using a self-attention mechanism, match the solutions that meet the requirements based on the correlation, and generate drawings and plans based on the large language model data database of pumped storage power stations.

[0128] The model correction module is used to correct and adjust the large language model for pumped storage power station construction and management generated by the model building module, and to conduct industry expert correction and adjustment of the solutions, drawings and schemes generated by the intelligent AI assistant module.

[0129] The resource library module is used to store the answers, drawings and solutions generated by the intelligent AI assistant module. Users interact with the intelligent AI assistant module based on the content stored in the resource library. The system iterates and generates an accurate version of the stored content according to the user's needs, and then stores the accurate version again.

[0130] The authorization management module is used to input user attributes into the system through dialogue between the user and the intelligent AI assistant module, and to control user access based on the user attributes;

[0131] Furthermore, the pumped storage power station construction management system based on a large language model also includes a mobile terminal module, while the model building module, intelligent AI assistant module, model correction module, resource library module, and authorization management module are all web-based functions.

[0132] The mobile module is used to convert web-based functions into mobile-based functions. The mobile functions, combined with the intelligent AI assistant module, complete the question answering and resource acquisition from the resource library module.

[0133] Based on this, in this embodiment, the functions performed by the model correction module include:

[0134] By integrating professional data and expertise, the system adjusts the large language model for the construction and management of pumped storage power stations by adjusting the large language model and prompt words. It also uses a restricted knowledge base to adjust the generated answers, drawings and plans by industry experts, thereby further enhancing the system's professional capabilities for pumped storage power stations.

[0135] The specific execution steps are as follows:

[0136] C1. Adjusting the large language model: Collect information on the pumped storage industry, generate a database for a specified domain, train the large language model for the construction and management of pumped storage power stations based on the database for the specified domain, adjust the model’s performance in NPL tasks in a specific domain, and achieve model self-correction.

[0137] Specifically, domain-specific NPL tasks include sentiment analysis, entity recognition, text, image, speech, and video content classification, and dialogue generation.

[0138] C2. Adjust prompt words: Further realize model self-correction by designing template technology;

[0139] C3. Restricted Knowledge Base: This involves data analysis, cleaning, transformation, and correction of the large language model database for pumped storage power stations. Invalid data is removed, and valid data is tagged. Industry experts mark the accuracy of generated answers and assign weights to them, resulting in expert-provided precise answers. These expert-provided answers are then combined to perform deep sentence correction and corresponding accuracy correction, enabling expert-provided correction of generated solutions, drawings, and plans, thus achieving the return of clean data.

[0140] Furthermore, the specific steps for executing content C2 are as follows:

[0141] C201, Design Template: Input a problem into a large language model for the construction and management of pumped storage power stations. Extract prompt words from the input problem. Assuming the extracted prompt word is x, use the design template technique to transform the prompt word x into X containing empty slots. Fill the empty slots of X multiple times with different fill values ​​to obtain a set X of X containing different fill values. 集 ;

[0142] C202, Answer Search: For set X 集 By searching the database of large language models for pumped storage power station construction and management, multiple answers were obtained.

[0143] C203, Answer Correction: For the multiple answers obtained in step C202, the optimal answer is found through deep traversal, realizing model self-correction.

[0144] Example 5

[0145] like Figure 1 and Figure 2 As shown, a pumped storage power station construction and management system based on a large language model includes a model building module, an intelligent AI assistant module, a model correction module, a resource library module, and an authorization management module.

[0146] The model building module is used to collect data and generate a large language model database for pumped storage power stations. Based on the database, the Transformer neural network model architecture is selected, and Zero-shot Learning, Few-shot Learning and RLHF technologies are combined to generate a large language model for the construction and management of pumped storage power stations.

[0147] The intelligent AI assistant module is used to connect to the model building module, process the input questions to obtain the input sequence, extract the correlation between the input sequences using a self-attention mechanism, match the solutions that meet the requirements based on the correlation, and generate drawings and plans based on the large language model data database of pumped storage power stations.

[0148] The model correction module is used to correct and adjust the large language model for pumped storage power station construction and management generated by the model building module, and to conduct industry expert correction and adjustment of the solutions, drawings and schemes generated by the intelligent AI assistant module.

[0149] The resource library module is used to store the answers, drawings and solutions generated by the intelligent AI assistant module. Users interact with the intelligent AI assistant module based on the content stored in the resource library. The system iterates and generates an accurate version of the stored content according to the user's needs, and then stores the accurate version again.

[0150] The authorization management module is used to input user attributes into the system through dialogue between the user and the intelligent AI assistant module, and to control user access based on the user attributes;

[0151] Furthermore, the pumped storage power station construction management system based on a large language model also includes a mobile terminal module, while the model building module, intelligent AI assistant module, model correction module, resource library module, and authorization management module are all web-based functions.

[0152] The mobile module is used to convert web-based functions into mobile-based functions. The mobile functions, combined with the intelligent AI assistant module, complete the question answering and resource acquisition from the resource library module.

[0153] Based on this, in this embodiment, the resource library module performs the following: it connects to the large language model based on the intelligent AI assistant module, captures key points by obtaining content from the context, asynchronously addresses the database, assembles relevant content into various results and stores them in the resource library, and further reviews the response content, files, drawings, etc. in the resource library to put forward in-depth and detailed requirements for the system, resulting in more accurate results, as follows:

[0154] D1. Set the question as the initial key point, retrieve the closest preset question and answer through the initial key point, use the retrieved closest preset question and answer as context, and concatenate it with the question to generate the final key point;

[0155] D2. Labeling the final key points to obtain the corresponding labels. Based on the final key points and their corresponding labels, asynchronous addressing is performed using a large language model combined with the rsync algorithm.

[0156] D3. Utilizing Flink's asynchronous addressing I / O technology in both unordered and ordered modes, the content required in the context is stored as the generated result in the resource library, allowing users to preview and further refine the generated result.

[0157] Example 6

[0158] like Figure 1 and Figure 2 As shown, a pumped storage power station construction and management system based on a large language model includes a model building module, an intelligent AI assistant module, a model correction module, a resource library module, and an authorization management module.

[0159] The model building module is used to collect data and generate a large language model database for pumped storage power stations. Based on the database, the Transformer neural network model architecture is selected, and Zero-shot Learning, Few-shot Learning and RLHF technologies are combined to generate a large language model for the construction and management of pumped storage power stations.

[0160] The intelligent AI assistant module is used to connect to the model building module, process the input questions to obtain the input sequence, extract the correlation between the input sequences using a self-attention mechanism, match the solutions that meet the requirements based on the correlation, and generate drawings and plans based on the large language model data database of pumped storage power stations.

[0161] The model correction module is used to correct and adjust the large language model for pumped storage power station construction and management generated by the model building module, and to conduct industry expert correction and adjustment of the solutions, drawings and schemes generated by the intelligent AI assistant module.

[0162] The resource library module is used to store the answers, drawings and solutions generated by the intelligent AI assistant module. Users interact with the intelligent AI assistant module based on the content stored in the resource library. The system iterates and generates an accurate version of the stored content according to the user's needs, and then stores the accurate version again.

[0163] The authorization management module is used to input user attributes into the system through dialogue between the user and the intelligent AI assistant module, and to control user access based on the user attributes;

[0164] Furthermore, the pumped storage power station construction management system based on a large language model also includes a mobile terminal module, while the model building module, intelligent AI assistant module, model correction module, resource library module, and authorization management module are all web-based functions.

[0165] The mobile module is used to convert web-based functions into mobile-based functions. The mobile functions, combined with the intelligent AI assistant module, complete the question answering and resource acquisition from the resource library module.

[0166] Based on this, in this embodiment, the content executed by the authorization management module includes:

[0167] Through the system's human-computer dialogue, the intelligent AI assistant module replaces the traditional webpage to achieve user authorization and access control;

[0168] E1. Users communicate with the intelligent AI assistant module in natural language. The intelligent AI assistant module obtains text, images, and documents in the context, performs tagging and attribute processing, extracts user attribute parameters, and creates an underlying user table based on the user attribute parameters.

[0169] E2. The user communicates with the intelligent AI assistant module in natural language. The system compares the content obtained by the intelligent AI assistant module with the underlying user table to obtain the difference content, and compares whether the user is included in the underlying user table. If the user is included, the difference content is extended to the underlying user table to complete the entry and editing of the user table.

[0170] E2. By communicating with the intelligent AI assistant module in natural language, the system can obtain text, images, documents, and other content from the context to complete user access control and authentication.

[0171] Specifically, if the input is "Zhang San 12345678901", the large language model system automatically splits the content into "Zhang San" and "12345678901", and checks the user table to see if user Zhang San exists, obtaining a set of all users who are Zhang San; it then compares the attributes of all users in the set to determine if "12345678901" is a unique identifier among multiple attributes of a certain user Zhang San. If it is not determined to be a unique identifier, the intelligent AI assistant module obtains more information from the user and performs the addressing judgment again. If it is determined to be a unique identifier, the user behavior is analyzed (such as login IP, login location, login device, etc.).

[0172] If the user behavior analysis is normal, the user is allowed to log in and use the large language model system normally. If the user behavior analysis is abnormal, the core fields are determined from the user's multiple attributes (Zhang San's attributes contain fields that can prove his identity, such as ID card number or set verification code). These fields are obtained from the user through the intelligent AI assistant module. Only after successful verification can the user use the system normally to ensure user data security.

[0173] Furthermore, the specific content executed by the mobile module includes:

[0174] By linking the web and mobile terminals through data communication technology, the functions and data of the web terminal are transformed, and the intelligent AI assistant module works with the functions of the mobile terminal to answer questions, generate solutions, and complete resource acquisition.

[0175] The above description is merely an embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the scope of the present invention should be included within the protection scope of the present invention.

Claims

1. A pumped storage power station construction and management system based on a large language model, characterized in that, It includes a model building module, an intelligent AI assistant module, a model correction module, a resource library module, and an authorization management module; The model building module is used to collect data and generate a large language model database for pumped storage power stations. Based on the database, the Transformer neural network model architecture is selected, and Zero-shot Learning, Few-shot Learning and RLHF technologies are combined to generate a large language model for the construction and management of pumped storage power stations. The intelligent AI assistant module is used to connect to the model building module, process the input questions to obtain the input sequence, extract the correlation between the input sequences using a self-attention mechanism, match the solutions that meet the requirements based on the correlation, and generate drawings and plans based on the large language model data database of pumped storage power stations. The model correction module is used to correct and adjust the large language model for pumped storage power station construction and management generated by the model building module, and to conduct industry expert correction and adjustment of the solutions, drawings and schemes generated by the intelligent AI assistant module. The resource library module is used to store the answers, drawings and solutions generated by the intelligent AI assistant module. Users interact with the intelligent AI assistant module based on the content stored in the resource library. The system iterates and generates an accurate version of the stored content according to the user's needs, and then stores the accurate version again. The authorization management module is used to input user attributes into the system through dialogue between the user and the intelligent AI assistant module, and to control user access based on the user attributes; The specific content executed by the intelligent AI assistant module includes: B1. Using the intelligent AI assistant module to connect to the model building module, a large language model for the construction and management of pumped storage power stations is obtained. The problem is input into the intelligent AI assistant module. The intelligent AI assistant module obtains the input content of the problem, performs contextual understanding and extracts the input content, performs tagging and splitting of text content, and performs semantic and structural splitting of image and document content to obtain a tagging, semantic and structural input sequence. B2. Execute the self-attention mechanism: Obtain the input sequence in execution content B1, perform a linear transformation, use the self-attention mechanism to obtain the attention weight distribution, and weight each element in the input sequence according to the distribution weight to extract the correlation between the input content and form each initial representation and the degree of representation correlation. B3. Process the initial representation and representation correlation obtained in execution content B2, reduce the judgment of noun part of speech in the initial representation, improve the judgment of verb part of speech, infer the probability of the next word appearing, and obtain an input sequence with a representation correlation greater than the preset value. Then, through the similarity matching method, match the input sequence with the large language model data database of pumped storage power stations to generate a solution that meets the requirements. B4. Input the input sequence obtained in execution content B1 into the large language model for the construction and management of pumped storage power stations, and generate drawings using image rendering methods based on the large language model data database of pumped storage power stations. B5. Input the input sequence obtained in execution content B1 into the large language model for pumped storage power station construction and management, obtain the solution based on the large language model data database for pumped storage power stations, and convert the solution into PDF or WORD format using the AI ​​Office method.

2. The pumped storage power station construction management system based on a large language model according to claim 1, characterized in that, The specific tasks performed by the model building module include: A1. Basic Data Support: Collect basic data on pumped storage power stations to form a large language model data database for pumped storage power stations; A2. Data Preprocessing: Data preprocessing is performed on the large language model data database of pumped storage power stations. Database deduplication and semantic similarity calculation methods are used to clean the text in the database, delete duplicate or irrelevant content, and format the preprocessed database. A3. Model Generation: Select the Transformer neural network model architecture, and choose the number of parameters, hidden units and hyperparameters of the model based on its ability to process sequential data and capture long-term dependencies. Combine Zero-shot Learning, Few-shot Learning and RLHF techniques to generate a large language model. A4. Model Training: The large language model is trained using a formatted database. During training, the next word in a sentence is predicted based on the previous one or several words. Backpropagation and stochastic gradient descent optimization algorithms are used to adjust the parameters of the large language model to generate a large language model for the construction and management of pumped storage power stations. A5. Evaluation and Adjustment: Evaluate the model performance using perplexity and downstream task performance metrics, and adjust the model parameters.

3. The pumped storage power station construction management system based on a large language model according to claim 1, characterized in that, The specific steps for executing content B4 are as follows: B401. Perform a linear transformation on the labeled, semantic, and structured input sequence in execution content B1. Input the linearly transformed input sequence into the model. Use image rendering methods to obtain the first set of engineering drawings of the required quantity. Then, perform labeling and encoding processing on the first set of engineering drawings and compare and analyze it with the input sequence to obtain the first set of differences. B402. The input sequence, the first set of engineering drawings after labeling and encoding, and the first set of differences are fused together. Based on the fused input sequence, the first set of engineering drawings after labeling and encoding, and the first set of differences, an image rendering method is used to generate a second set of engineering drawings. The second set of engineering drawings is then labeled and encoded and compared with the input sequence to obtain the second set of differences. B403. The input sequence, the first set of engineering drawings after labeling and encoding, the second set of engineering drawings after labeling and encoding, and the second set of differences are fused together. Based on the fused input sequence, the first set of engineering drawings after labeling and encoding, the second set of engineering drawings after labeling and encoding, and the second set of differences, an image rendering method is used to generate a third set of engineering drawings. B404. Merge the first set of engineering drawings, the second set of engineering drawings, and the third set of engineering drawings to obtain the total set of engineering drawings. Tag and encode the total set of engineering drawings, compare and analyze it with the input sequence, select several engineering drawings with differences less than the preset value, record them as the set of drawings that meet the requirements, and apply scale and coded content annotations to the drawings.

4. The pumped storage power station construction management system based on a large language model according to claim 1, characterized in that, The specific steps for executing content B5 are as follows: B501. Perform a linear transformation on the tagged, semantic, and structured input sequence in execution content B1 to generate a scheme outline set. Perform tagged and semantic processing on each level of the scheme outline set. Merge the tagged and semantic processed scheme outline sets and input them into the model. Based on the input sequence and the tagged and semantic processed scheme outline set, match it with the large language model data database of pumped storage power stations to obtain the content corresponding to each level of the scheme outline. B502. Tag, semanticize, and structure the content corresponding to each level of the matched scheme outline, generate drawings using image rendering methods, generate multiple attributes for the drawings by combining the input sequence content, and further match the drawing content to generate drawing number, drawing association description, and drawing remarks. B503 uses the Blade template engine and WebKit / Blink rendering engine to arrange and combine the content of each level of the matched scheme outline and the generated drawings to obtain the scheme. The scheme is then converted into PDF or WORD format using AI Office methods.

5. The pumped storage power station construction management system based on a large language model according to claim 1, characterized in that, The model correction module performs the following tasks: integrating professional data and expertise; adjusting the overall language model for pumped storage power station construction and management by modifying the large language model and prompt words; and using a restricted knowledge base to perform industry expert correction on the generated solutions, drawings, and plans. Specifically: C1. Adjusting the large language model: Collect information on the pumped storage industry, generate a database for a specified domain, train the large language model for the construction and management of pumped storage power stations based on the database for the specified domain, adjust the model’s performance in NPL tasks in a specific domain, and achieve model self-correction. C2. Adjust prompt words: Further realize model self-correction by designing template technology; C3. Restricted Knowledge Base: This feature involves data analysis, cleaning, correction, and transformation of the database of large-scale language model data for pumped storage power stations. Invalid data is removed, and valid data is tagged. Combined with precise answers from industry experts, in-depth sentence correction and corresponding accuracy correction are performed to achieve industry expert correction of the generated solutions, drawings, and plans.

6. The pumped storage power station construction management system based on a large language model according to claim 5, characterized in that, The specific steps for executing content C2 are as follows: C201, Design Template: Input a problem into a large language model for the construction and management of pumped storage power stations. Extract prompt words from the input problem. Assuming the extracted prompt word is x, use template design technology to transform prompt word x into X containing empty slots. Fill the empty slots of X multiple times with different fill values ​​to obtain a set of X containing different fill values. ; C202, Answer Search: For Sets By searching the database of large language models for pumped storage power station construction and management, multiple answers were obtained. C203, Answer Correction: For the multiple answers obtained in step C202, the optimal answer is found through deep traversal, realizing model self-correction.

7. The pumped storage power station construction management system based on a large language model according to claim 1, characterized in that, The specific tasks executed by the resource repository module include: D1. Set the question as the initial key point, retrieve the closest preset question and answer through the initial key point, use the retrieved closest preset question and answer as context, and concatenate it with the question to generate the final key point; D2. Labeling the final key points to obtain the corresponding labels. Based on the final key points and their corresponding labels, asynchronous addressing is performed using a large language model combined with the rsync algorithm. D3. Utilize Flink's asynchronous addressing IO technology in both unordered and ordered modes to store the content required in the context as the generated result in the resource library.

8. The pumped storage power station construction management system based on a large language model according to claim 1, characterized in that, The specific tasks performed by the authorization management module include: E1. Users communicate with the intelligent AI assistant module in natural language. The intelligent AI assistant module obtains text, images, and documents in the context, performs tagging and attribute processing, extracts user attribute parameters, and creates an underlying user table based on the user attribute parameters. E2. The user communicates with the intelligent AI assistant module in natural language. The system compares the content obtained by the intelligent AI assistant module with the underlying user table to obtain the difference content, and compares whether the user is included in the underlying user table. If the user is included, the difference content is extended to the underlying user table to complete the entry and editing of the user table. E3. By communicating with the intelligent AI assistant module in natural language, it obtains text, images, and document content from the context to complete user access control and authentication.

9. The pumped storage power station construction management system based on a large language model according to claim 1, characterized in that, The pumped storage power station construction and management system based on a large language model also includes a mobile terminal module, while the model building module, intelligent AI assistant module, model correction module, resource library module, and authorization management module are all web-based functions. The mobile module is used to convert web-based functions into mobile-based functions. These mobile functions, combined with the intelligent AI assistant module, enable question answering and resource retrieval from the resource library module.

10. The pumped storage power station construction management system based on a large language model according to claim 9, characterized in that, The specific content executed by the mobile module includes: By linking the web and mobile terminals through data communication technology, the functions and data of the web terminal are transformed, and the intelligent AI assistant module works with the functions of the mobile terminal to answer questions, generate solutions, and complete resource acquisition.

Images