Hydropower station equipment knowledge graph construction method based on large language model and related system

Abstract

The invention provides a hydropower station equipment knowledge graph construction method based on a large language model and a related system, and belongs to the field of hydroelectric generating set equipment monitoring. Specifically, data generated in the operation and maintenance process of the hydropower station are obtained, and the data are cleaned and standardized; based on a large language model, extracting entities and relationships of the hydropower station equipment from the standardized data; the extracted entities and relations are corrected, a knowledge graph of the hydropower station equipment is constructed, and the accuracy and robustness of the knowledge graph are improved through iterative optimization; new data generated in the operation and maintenance process of the hydropower station are monitored in real time, and the knowledge graph is automatically updated and maintained; and applying the constructed hydropower station equipment knowledge graph to operation and maintenance management of the hydropower station. The new data in hydropower station operation and maintenance can be monitored in real time, the knowledge graph is automatically updated, it is ensured that the knowledge graph is always kept consistent with the latest equipment state and operation and maintenance requirements, and manual operation of knowledge updating is greatly reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of hydropower unit equipment monitoring, and specifically relates to a method for constructing a knowledge graph of hydropower station equipment based on a large language model and a related system. Background Art

[0002] As an important energy infrastructure, the operating efficiency and safety of hydropower stations are directly related to the stability and reliability of power supply. In recent years, with the rapid development of information technology, advanced technologies such as big data and artificial intelligence have been widely used in the hydropower industry. In particular, the introduction of knowledge graph technology has provided new solutions for the intelligent management of hydropower stations. The operation and maintenance process of hydropower stations involves a large amount of equipment data, operation and maintenance records, fault cases and expert experience. By structuring massive data, knowledge graphs can effectively improve data retrieval, analysis and decision support capabilities, and have become an important tool for hydropower station operation and maintenance management.

[0003] In the prior art, the construction of knowledge graphs for key equipment in hydropower stations mainly relies on traditional data mining and expert system methods. These methods usually include steps such as data collection, preprocessing, entity recognition, and relationship extraction, and use artificial rules or machine learning algorithms to extract and organize knowledge. However, with the increasing types and number of hydropower station equipment and the explosive growth of equipment operation data, traditional methods are unable to cope with complex and dynamic data and are difficult to meet the needs of efficient and accurate knowledge graph construction. First, traditional knowledge graph construction methods mainly rely on experts to manually organize and define knowledge, which is not only time-consuming and labor-intensive, but also easily limited by the personal experience and cognition of experts, resulting in insufficient coverage and accuracy of knowledge graphs. For complex industrial systems such as hydropower stations, with a wide variety of equipment types, complex structures, and changing operating environments, the efficiency and effect of manually constructing knowledge graphs are difficult to meet actual needs. Secondly, existing knowledge graph technology has certain limitations in processing unstructured data. The operation and maintenance data of hydropower stations include a large number of text records, sensor data, pictures, and videos, which are diverse and interrelated, requiring powerful natural language processing and multimodal data processing capabilities. However, traditional knowledge graph technology is often unable to cope with such complex data, and it is difficult to achieve comprehensive extraction and in-depth understanding of information, resulting in the construction of knowledge graphs that are not comprehensive and accurate enough. Finally, existing technologies also have obvious deficiencies in knowledge updating and maintenance. The equipment and technology of hydropower stations are constantly updated, and new data and knowledge are constantly generated during the operation and maintenance process. If the knowledge graph cannot be updated and maintained in a timely manner, it will gradually lose its guiding significance for actual operation and maintenance. Traditional knowledge graph updates often require a lot of manual operations again, which is inefficient and error-prone, and difficult to adapt to dynamically changing operation and maintenance needs. Summary of the invention

[0004] The purpose of the present invention is to overcome the problem that existing hydropower station equipment will continuously generate new data and knowledge during the operation and maintenance process and cannot be updated and maintained in time, and to provide a method and related system for constructing a knowledge graph of hydropower station equipment based on a large language model.

[0005] In order to achieve the above object, the present invention adopts the following technical solution: In a first aspect, the present invention provides a method for constructing a knowledge graph of hydropower station equipment based on a large language model, comprising the following steps: Obtain data generated during the operation and maintenance of hydropower stations, and clean and standardize the data; Based on the large language model, the entities and relationships of hydropower station equipment are extracted from the standardized data; Correct the extracted entities and relationships, build a knowledge graph of hydropower station equipment, and improve the accuracy and robustness of the knowledge graph through iterative optimization; Real-time monitoring of new data generated during the operation and maintenance of hydropower stations, and automatic updating and maintenance of knowledge graphs; The constructed knowledge graph of hydropower station equipment is applied to the operation and maintenance management of the hydropower station.

[0006] The entities and relationships of hydropower station equipment are extracted from the standardized data through the few-sample prompt word technology.

[0007] Based on the large language model, the entities and relationships of the hydropower station equipment are extracted from the standardized data. The specific method is as follows: Write the output corresponding to the input model text in the form of a list of triples, which will be used as a few-shot example in the next prompt word. Using the few-sample prompt word technology, open information extraction is performed on the new device-related input text based on the few-sample examples to obtain the triple list output by the model.

[0008] The extracted entities and relationships are corrected to construct a knowledge graph of hydropower station equipment, and the accuracy and robustness of the knowledge graph are improved through iterative optimization. The specific method is as follows: Get the extracted entities and relationships; Through the self-reflection mechanism, the extracted entities and relationships are corrected, and the knowledge graph is constructed based on the corrected entities and relationships; Based on similarity relationship retrieval, the structure in the knowledge graph is optimized, similar relationships are merged, and the redundancy of the knowledge graph is reduced. If it is found that the relationship is not merged properly, the entities and relationships are recalibrated and iterative optimization is performed.

[0009] The new data generated during the operation and maintenance of the hydropower station is monitored in real time, and the knowledge graph is automatically updated and maintained. The specific method is as follows: Automatically monitor new operation and maintenance data and fault information of hydropower station equipment, and capture knowledge update needs in real time; Re-extract entities and relationships and build and optimize the knowledge graph based on the newly added data; Associate new data with historical data to ensure the consistency of the knowledge graph.

[0010] The application of the constructed hydropower station equipment knowledge graph to the operation and maintenance management of the hydropower station also includes: Supports operation and maintenance personnel to quickly retrieve device information through natural language queries; Combine the equipment operation history and fault data in the knowledge graph to analyze the cause of the fault and generate predictive maintenance suggestions; Based on the generated fault causes and maintenance suggestions, assist in optimizing the operation and scheduling of the hydropower station.

[0011] The large language model is a pre-trained deep learning model that can understand and generate natural language text and has the ability of few-sample learning and self-reflection.

[0012] In a second aspect, the present invention provides a hydropower station equipment knowledge graph construction system based on a large language model, comprising: Data acquisition and preprocessing module, used to acquire data generated during the operation and maintenance of the hydropower station, and to clean and standardize the data; The information extraction module is used to extract entities and relationships of hydropower station equipment from the standardized data based on the large language model; The knowledge graph construction module is used to correct the extracted entities and relationships, build the knowledge graph of hydropower station equipment, and improve the accuracy and robustness of the knowledge graph through iterative optimization; The knowledge graph automatic update module is used to monitor the new data generated during the operation and maintenance of the hydropower station in real time, and automatically update and maintain the knowledge graph; The application module is used to apply the constructed knowledge graph of hydropower station equipment to the operation and maintenance management of the hydropower station.

[0013] In a third aspect, the present invention provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of a method for constructing a knowledge graph of hydropower station equipment based on a large language model when executing the computer program.

[0014] In a fourth aspect, the present invention provides a storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps of a method for constructing a knowledge graph of hydropower station equipment based on a large language model are implemented.

[0015] Compared with the prior art, the present invention has the following beneficial effects: The present invention provides a method and a related system for constructing a knowledge graph of hydropower station equipment based on a large language model, comprising the following steps: obtaining data generated during the operation and maintenance of a hydropower station, and cleaning and standardizing the data; extracting entities and relationships of hydropower station equipment from the standardized data based on a large language model; correcting the extracted entities and relationships, constructing a knowledge graph of hydropower station equipment, and improving the accuracy and robustness of the knowledge graph through iterative optimization; real-time monitoring of new data generated during the operation and maintenance of the hydropower station, automatically updating and maintaining the knowledge graph; and applying the constructed knowledge graph of hydropower station equipment to the operation and maintenance management of the hydropower station. Through the powerful semantic understanding ability of the large language model, the equipment-related knowledge can be automatically and quickly extracted from the multi-source and multi-modal data of the hydropower station, significantly improving the efficiency and coverage of the knowledge graph construction, and reducing the dependence on manual participation. It can monitor the new data in the operation and maintenance of the hydropower station in real time, automatically integrate the new and old data, and automatically update the knowledge graph to ensure that it is always consistent with the latest equipment status and operation and maintenance requirements, greatly reducing the manual operation of knowledge update.

[0016] Furthermore, relationship definitions are automatically generated based on the large language model; relationship correction is performed using the self-reflection mechanism of the large language model; and structural optimization of the knowledge graph is achieved through similar relationship retrieval.

[0017] Furthermore, with the help of the natural language processing and multimodal data analysis capabilities of the large language model, the present invention can process data in various forms such as text, pictures, videos, etc., comprehensively extract key information in the operation and maintenance of hydropower station equipment, and ensure the accuracy and completeness of the knowledge graph.

[0018] Furthermore, the knowledge graph of key equipment in hydropower stations can support practical applications such as fault diagnosis, predictive maintenance and optimized scheduling, helping operation and maintenance personnel to quickly obtain equipment status, analyze fault causes, and provide intelligent decision-making suggestions to improve the operating efficiency and safety of hydropower stations. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a flow chart of the present invention; Figure 2 It is a system diagram of the present invention; Figure 3 This is a system diagram of Example 3 of the present invention. DETAILED DESCRIPTION

[0020] In order to further understand the content of the present invention, the present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the embodiments are only for explaining the present invention and are not intended to limit it.

[0021] Embodiment 1: like Figure 1 As shown in FIG. 1 , the method for constructing a knowledge graph of hydropower station equipment based on a large language model includes the following steps: S1: Obtain the data generated during the operation and maintenance of the hydropower station, and clean and standardize the data; S2: Based on the large language model, the entities and relations of hydropower station equipment are extracted from the standardized data; S3: Correct the extracted entities and relationships, build a knowledge graph of hydropower station equipment, and improve the accuracy and robustness of the knowledge graph through iterative optimization; S4: Real-time monitoring of new data generated during the operation and maintenance of hydropower stations, and automatic updating and maintenance of knowledge graphs; S5: Apply the constructed knowledge graph of hydropower station equipment to the operation and maintenance management of the hydropower station.

[0022] Specifically, in S1, the multi-source data generated during the operation of the hydropower station are collected and preprocessed, including structured data (such as sensor data, equipment parameters, etc.) and unstructured data (such as equipment maintenance documents, equipment failure reports, equipment operation and maintenance records, etc.). Data collection and preprocessing are the basic steps of the present invention, the purpose of which is to ensure the comprehensiveness, standardization and consistency of the source data, and to provide high-quality data support for the subsequent construction of the knowledge graph. This step includes data collection, data cleaning and standardization.

[0023] S101, data collection: A large amount of multi-source data is usually generated during the operation and maintenance management of a hydropower station, and these data include structured data and unstructured data. Structured data mainly comes from the sensor system, equipment monitoring system, etc. of the hydropower station, such as equipment operating parameters, temperature, pressure, vibration and other types of monitoring data. Unstructured data mainly includes operation and maintenance records, fault reports, expert documents, etc. In order to ensure the comprehensiveness of the data, the present invention adopts an automated data collection method, which can obtain complete equipment operation and maintenance data from different sources such as the equipment management system, monitoring system, database, etc. of the hydropower station; Through the interface, it connects with the existing equipment management system and monitoring system of the hydropower station to obtain the equipment operation data regularly or in real time, for example, by extracting equipment-related numerical data from the database through API or OBDC.

[0024] Unstructured data collection: Through text mining technology, we collect text records, fault reports, expert analysis documents, etc. generated during the daily operation and maintenance of the hydropower station. For electronic documents, we obtain their PDF files; for paper documents, we use OCR technology to process scanned documents and store them electronically.

[0025] S102, Data cleaning and standardization: The collected raw data usually has certain noise and inconsistency, and may contain missing values, inconsistent formats, redundant data, etc. Therefore, data cleaning and preprocessing are important steps to ensure data quality.

[0026] Missing value processing: interpolate missing values ​​in structured data, such as using mean interpolation, linear interpolation, etc.; check the content of documents obtained through OCR recognition in unstructured data: use a large language model to check the OCR recognition results and correct possible recognition errors.

[0027] Outlier detection and processing: Detect outliers in structured data through statistical methods or anomaly detection algorithms. For example, when parameters such as temperature and pressure of equipment exceed the reasonable range, the model can automatically determine and mark or correct them.

[0028] Format standardization: Perform unified format conversion on structured data from different sources, such as time format, unit conversion, etc., to ensure that the data has a consistent format in subsequent processing steps.

[0029] Specifically, in S2, open information extraction: Based on the large language model, using the technology of few-sample prompt words, the entities (including subjects and objects) and relationships of key equipment of the hydropower station are automatically extracted from the preprocessed data to build a preliminary knowledge graph. Among them, the main focus is on entity information such as equipment type, equipment components, equipment status, and fault type, and the main focus is on relationship information such as the association between equipment and the relationship between the cause of the fault and the equipment status.

[0030] S201, few-sample example writing: Based on a sentence of input text related to hydropower station equipment, power equipment experts write the output of their open information extraction in the format of a triple list of [[subject_1, relation_1, object_1], [subject_2, relation_2, object_2], …], and this input-output pair is used as a few-sample example in the next prompt word; 1) First, select representative text related to hydropower station equipment as input, for example: "The turbine water guide mechanism has abnormal vibration, and upon inspection, it was found that the guide vane seal ring was worn, resulting in an excessive gap"; 2) Power equipment experts manually annotate the entities and their relationships in the text based on the information in the text. The annotation method is to extract the main entities (i.e., subject and object) in the sentence and identify the relationship between the subject and the object. The manual annotation output format of each input text is in the form of a triple list, with the specific structure as follows: [[subject_1, relation_1, object_1], [subject_2, relation_2, object_2], …] For example, for the text "The turbine water guide mechanism has abnormal vibration, and upon inspection, it was found that the guide vane seal ring was worn, resulting in an excessive gap", the corresponding expert annotation is: [[The water guide mechanism of the turbine has abnormal vibration], [Wear of guide vane seal ring leads to abnormal vibration], [The guide vane seal ring is worn, resulting in excessive clearance]] 3) The above input text and its corresponding triple output are used as a few-shot example for the prompt words of the subsequent large language model.

[0031] S202, open information extraction based on a large language model: Using the few-sample prompt word technology, let the large language model extract open information from the new device-related input text based on the above example, and also output it in the format of a triple list of [[subject_1, relationship_1, object_1], [subject_2, relationship_2, object_2],…]. Guided by the prompt words, the large language model can learn and understand the potential relationships between entities (such as equipment, components, faults, etc.), and automatically extract entities and their relationships from new device-related input text. This process can efficiently realize open information extraction using a small number of samples without the need for a large amount of labeled data. The specific implementation method is as follows: 1) Construct a prompt word template, including the following elements: -Task description: This is a task to extract entity relationships from text related to hydropower station equipment. -Output format description: Indicates the output format of triple list - Example part: contains the few sample examples prepared in step S201 -Text to be processed: Place new text that needs to be extracted 2) Examples of prompt words are as follows: Task: Extract entity and relationship information from the following hydropower station equipment-related text and output it in the form of a triple list.

[0032] Output format: [[subject, relation, object], ...] Sample input: The turbine water guide mechanism vibrates abnormally. Inspection shows that the guide vane seal is worn, causing the gap to be too large. Sample output: [[water turbine water guide mechanism, abnormal vibration occurs], [Wear of guide vane seal ring leads to abnormal vibration], [The guide vane seal ring is worn, resulting in excessive clearance]] Please process the following text: {Text to be processed} 3) Input the prompt word into the large language model and obtain the triple list output by the model. For example, for the input text "The pump unit has a fault and the motor is found to be overheated", the model output is: [[Pump unit, failure], [Motor, overheating], [Overheating, resulting in pump unit failure]] 4) Collect and organize the triplets output by the model as the basis for building the knowledge graph.

[0033] Through this process, open information extraction based on a large language model can automatically extract key information such as equipment and its status, components, faults, etc. from the text, and convert this information into a structured knowledge graph form, providing preliminary entity and relationship data for subsequent knowledge graph construction.

[0034] In this process, the few-sample prompt word technology enables the large language model to accurately identify the relevant entities and relationships of the equipment when faced with new and unseen equipment-related texts. It has good generalization ability and adaptability, thereby efficiently supporting the construction of the knowledge graph of key equipment in hydropower stations.

[0035] Specifically, in S3, knowledge graph construction and optimization: organize and structure the extracted subjects, relationships, and objects, build a knowledge graph of key equipment in the hydropower station, and optimize the accuracy and robustness of the knowledge graph through iteration, including the following steps: S301, knowledge graph relationship definition: Based on the large language model and using the few-sample prompt word technology, output the definition of the relationship extracted in the previous step; 1) Select representative relations: From the triples extracted in step S2, select representative relation types, such as "occurrence", "cause", "result", etc. 2) Write a few sample examples: Power equipment experts write corresponding definition examples for the selected relationship type. For example, for the relationship "occurs", you can write the following example: - Input text: The water guide mechanism of the turbine vibrates abnormally. Upon inspection, it is found that the guide vane seal ring is worn, resulting in an excessive gap. - The extracted triplet: [water turbine water guide mechanism, occurrence, abnormal vibration] -Relationship definition: "Occurrence" means that the status or fault of a device or component begins to appear at a specific point in time.

[0036] 3) Construct a prompt word template: Construct a prompt word template that includes task description, output format description, example part, and relationship to be processed. The specific structure is as follows: -Task description: This is a task to define the relationship between entities related to hydropower station equipment -Output format description: Describes the format of the relationship definition to be output -Example section: contains the few sample examples prepared in the previous step -Relations to be processed: Place the input text, the extracted triples, and the relationships to be defined 4) Examples of prompt words: Task: Define the relationships between the following hydropower plant equipment related entities.

[0037] Output format: {relationship}:{relationship definition} Sample input: Based on the following, please write a definition of the relationship "appearance" - Input text: The water guide mechanism of the turbine vibrates abnormally. Upon inspection, it is found that the guide vane seal ring is worn, resulting in an excessive gap. - The extracted triplet: [water turbine water guide mechanism, occurrence, abnormal vibration] Sample output: Occurrence: Indicates that a device or component's status or fault begins to appear at a specific point in time Please write a definition for the relationship "{relationship to be defined}" based on the following: - Input text: {input text} -Extracted triples: {triples extracted in step S2} 5) Input prompt words and obtain output: Input the prompt words into the large language model and obtain the relationship definition output by the model. For example, for the relationship "leads to", the model output is: Cause: indicates that one event or state is the cause of another event or state S302, knowledge graph relationship correction: Use the self-reflection mechanism of the large language model to correct the extracted relationships to ensure the correctness of the relationships in the knowledge graph; 1) Select the triples to be corrected: From the triples extracted in step S2, select the relation triples that need to be corrected 2) Write a few-sample example: Power equipment experts write correct triple examples for the triples to be corrected, and provide the original text and the relationship definition obtained in the previous step as a reference for model judgment. For example, for the wrong triple "[water turbine water guide mechanism, leads to, abnormal vibration]", the correct triple should be "[guide vane seal wear, leads to, abnormal vibration]" 3) Construct a prompt word template: Construct a prompt word template that includes task description, output format description, example part, and triples to be processed. The specific structure is as follows: -Task description: This is a task to determine the correctness of the extraction of entity triples related to hydropower station equipment. -Output format description: Indicates that the corrected triples need to be output -Example section: contains the few sample examples prepared in the previous step -Triples to be processed: Place the triples that need to be corrected 4) Examples of prompt words: Task: Determine the correctness of the extraction of entity triples related to hydropower station equipment Output format: [[subject, relation, object], ..., [subject, relation, object]] Sample input: Based on the following content, output the corrected triples. If the input triples are correct, just output the input triples directly. Original text: The turbine water guide mechanism experienced abnormal vibration. Upon inspection, it was found that the guide vane seal ring was worn, resulting in an excessive gap. The triplet to be corrected: [water guide mechanism of turbine, causes, abnormal vibration] Definition of the relationship "cause": indicates that one event or state is the cause of another event or state Sample output: [Wear of guide vane seal ring causes abnormal vibration] Based on the following content, output the corrected triples. If the input triples are correct, just output the input triples directly. Original text: {original text} Triplet to be corrected: {Triplet extracted in step S2} Definition of the relation "{relation in triple}": {relation definition obtained in step S301} 5) Input prompt words and obtain output: Input the prompt words into the large language model and obtain the corrected triples output by the model. For example, for the input triple "[abnormal vibration, resulting in, gap is too large]", the model output is: [Wear of the guide vane seal ring causes the gap to be too large] S303, knowledge graph structure optimization: Based on similarity relationship retrieval, the structure of the knowledge graph is optimized using a large language model, similar relationships are merged, and the redundancy of the knowledge graph is reduced; 1) Similar relationship retrieval: Take the relationship definition of the triple extracted in step S2 as the query, and retrieve the relationship with similar definition in the existing knowledge graph structure. The similarity here is obtained by calculating the cosine similarity of the semantic vector of the definition description. 2) Construct a prompt word template: Construct a prompt word template that includes task description, output format description, information to be processed, and alternative relationships. The specific structure is as follows: -Task description: This task is to optimize the knowledge graph structure and merge similar relationships. -Output format description: Indicates that the optimized triple set needs to be output - Information to be processed: The original text, the set of extracted triples and the definition of the relationship between the triples are given -Alternative relations: List the similar relations retrieved in the previous step and their definitions 3) Examples of prompt words: Task: Given a text, extracted relation triples and relation definitions, select the most appropriate candidate relation to replace it. If none of the candidate relations can replace the original relation, output the original triples. Output format: [[subject, relation, object], ...] Pending information: Original text: {original text} Triplet to be processed: {Triplet extracted in step S2} Relationship definition: {relationship definition obtained in step S301} Alternative relations: {similar relations retrieved in the previous steps and their definitions} 4) Input prompt words and obtain output: Input prompt words into the large language model and obtain the optimized triplet set output by the model. For example, for the following information to be processed: Original text: The pump unit failed and inspection revealed that the motor was overheating. Triplet to be processed: [Overheating, causing pump unit failure] Relationship definition: Causes: indicates that one action, event, or state directly causes another specific result or change. Alternative relationship: Cause: Indicates that one event or state is the cause of another event or state The output of the large language model is: [overheating, leading to pump unit failure] 5) Evaluation and iteration: Evaluate the optimized triple set to ensure that the merged similarity relationships are reasonable and do not lose important information. If it is found that some relationships are not merged properly, readjust the retrieval method and prompt words and perform iterative optimization.

[0038] Through the above steps, the accuracy, robustness and availability of the knowledge graph are effectively improved, providing a solid data foundation for the status monitoring, fault diagnosis and predictive maintenance of key equipment in hydropower stations. In practical applications, through the small sample prompt word technology and self-reflection mechanism of the large language model, the construction and optimization of the knowledge graph can gradually improve its quality in continuous iteration, so that this method can still maintain high adaptability and scalability when facing complex and dynamic hydropower station equipment information. At the same time, by reducing redundancy and optimizing the structure, the query efficiency of the knowledge graph has also been significantly improved, providing efficient data support for subsequent intelligent applications.

[0039] Specifically, in S4, through the continuous learning ability of the large language model, the new data generated during the operation and maintenance of the hydropower station is monitored in real time, and the knowledge graph is automatically updated and maintained to ensure that it is consistent with the actual operation and maintenance needs.

[0040] S401, automatically monitors the newly added operation and maintenance data and fault information of hydropower station equipment, and captures knowledge update needs in real time; By connecting with the hydropower station monitoring system and equipment management system, the system can capture new data generated by hydropower station equipment in real time, especially equipment operation status data, sensor data, fault reports, etc. Once these data are generated, the system will automatically identify and mark them as potential data sources for knowledge graph updates; For unstructured data (such as operation and maintenance records, fault reports), the system regularly scans text data sources to discover new equipment failures or operation and maintenance information. These data usually contain equipment failure descriptions, processing procedures, etc. The system extracts key information through natural language processing technology.

[0041] S402: re-extract open information and build and optimize the knowledge graph based on the newly added data, dynamically update device entities and relationships, and ensure the timeliness and accuracy of the knowledge graph; When new data is captured, the system will re-execute the knowledge extraction step to identify the newly added equipment entities and relationships. For example, when a new fault report describes a new fault type of a certain equipment, the system will automatically identify the fault and associate it with the corresponding equipment node; Based on the newly added data, the system will automatically adjust the structure of the knowledge graph. For example, if the new data describes a new component or a new functional module of a device, the graph will automatically generate the corresponding node and associate it with the existing device entity. It can also optimize and merge newly added relationships in the knowledge graph to reduce the redundancy of the knowledge graph.

[0042] S403, associate the new data with the historical data to ensure that the knowledge graph has long-term knowledge accumulation and consistency.

[0043] Specifically, in S5, the constructed knowledge graph of key equipment of the hydropower station will be applied to the operation and maintenance management of the hydropower station, providing intelligent decision-making support for equipment fault diagnosis, predictive maintenance and optimized scheduling.

[0044] S501, based on the natural language understanding capability of the large language model, supports operation and maintenance personnel to quickly retrieve key information such as equipment status and fault records through natural language queries. Operation and maintenance personnel can input questions in natural language, such as "What is the latest status of a generator?" or "What are the common causes of failure?" The system will automatically parse the question and quickly retrieve relevant equipment status, fault records and other information through the knowledge graph to return accurate answers. Operation and maintenance personnel can query the historical fault records, maintenance frequency, operating life and other information of a certain equipment. The system will automatically generate a health status report of the equipment to help operation and maintenance personnel conduct a comprehensive assessment.

[0045] S502, combining the equipment operation history and fault data in the knowledge graph, performs fault cause analysis and generates predictive maintenance suggestions to help operation and maintenance personnel detect potential problems in advance and reduce equipment downtime.

[0046] S503, through the equipment operation status and historical operation and maintenance records provided by the knowledge graph, based on the generated fault causes and maintenance suggestions, assist in optimizing the operation and scheduling of the hydropower station, and improve the overall operation efficiency and reliability of the system. In the event of an emergency failure or other emergency, the system can automatically generate an emergency plan based on the historical emergency processing data in the knowledge graph. The system will recommend the best emergency processing plan based on the cause of the failure, equipment status, and the operation of surrounding equipment to help operation and maintenance personnel quickly respond to emergencies and reduce losses.

[0047] Embodiment 2: like Figure 2 As shown in the figure, the hydropower station equipment knowledge graph construction system based on the large language model includes: Data acquisition and preprocessing module, used to acquire data generated during the operation and maintenance of the hydropower station, and to clean and standardize the data; The information extraction module is used to extract entities and relationships of hydropower station equipment from the standardized data based on the large language model; The knowledge graph construction module is used to correct the extracted entities and relationships, build the knowledge graph of hydropower station equipment, and improve the accuracy and robustness of the knowledge graph through iterative optimization; The knowledge graph automatic update module is used to monitor the new data generated during the operation and maintenance of the hydropower station in real time, and automatically update and maintain the knowledge graph; The application module is used to apply the constructed knowledge graph of hydropower station equipment to the operation and maintenance management of the hydropower station.

[0048] Embodiment 3: like Figure 3 As shown, the present invention also provides an electronic device 100 for constructing a knowledge graph of hydropower station equipment based on a large language model; the electronic device 100 includes a memory 101, at least one processor 102, a computer program 103 stored in the memory 101 and executable on the at least one processor 102, and at least one communication bus 104.

[0049] The memory 101 can be used to store the computer program 103, and the processor 102 implements the steps of the method for constructing a knowledge graph of hydropower station equipment based on a large language model described in Example 1 by running or executing the computer program stored in the memory 101 and calling the data stored in the memory 101. The memory 101 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data (such as audio data) created according to the use of the electronic device 100, etc. In addition, the memory 101 can include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other non-volatile solid-state storage devices.

[0050] The at least one processor 102 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The processor 102 may be a microprocessor or any conventional processor, etc. The processor 102 is the control center of the electronic device 100, and uses various interfaces and lines to connect various parts of the entire electronic device 100.

[0051] The memory 101 in the electronic device 100 stores a plurality of instructions to implement a method for constructing a knowledge graph of hydropower station equipment based on a large language model, and the processor 102 can execute the plurality of instructions to implement: Obtain data generated during the operation and maintenance of hydropower stations, and clean and standardize the data; Based on the large language model, the entities and relationships of hydropower station equipment are extracted from the standardized data; Correct the extracted entities and relationships, build a knowledge graph of hydropower station equipment, and improve the accuracy and robustness of the knowledge graph through iterative optimization; Real-time monitoring of new data generated during the operation and maintenance of hydropower stations, and automatic updating and maintenance of knowledge graphs; The constructed knowledge graph of hydropower station equipment is applied to the operation and maintenance management of the hydropower station.

[0052] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for constructing a knowledge graph of hydropower station equipment based on a large language model, characterized in that: The steps include: Obtain data generated during the operation and maintenance of hydropower stations, and clean and standardize the data; Based on the large language model, the entities and relationships of hydropower station equipment are extracted from the standardized data; Correct the extracted entities and relationships, build a knowledge graph of hydropower station equipment, and improve the accuracy and robustness of the knowledge graph through iterative optimization; Real-time monitoring of new data generated during the operation and maintenance of hydropower stations, and automatic updating and maintenance of knowledge graphs; The constructed knowledge graph of hydropower station equipment is applied to the operation and maintenance management of the hydropower station.

2. The method for constructing a knowledge graph of hydropower station equipment based on a large language model according to claim 1 is characterized in that: The entities and relationships of hydropower station equipment are extracted from the standardized data through the few-sample prompt word technology.

3. The method for constructing a knowledge graph of hydropower station equipment based on a large language model according to claim 2 is characterized in that: Based on the large language model, the entities and relationships of the hydropower station equipment are extracted from the standardized data. The specific method is as follows: Write the output corresponding to the input model text in the form of a list of triples, which will be used as a few-shot example in the next prompt word. Using the few-sample prompt word technology, open information extraction is performed on the new device-related input text based on the few-sample examples to obtain the triple list output by the model.

4. The method for constructing a knowledge graph of hydropower station equipment based on a large language model according to claim 1 is characterized in that: The extracted entities and relationships are corrected to construct a knowledge graph of hydropower station equipment, and the accuracy and robustness of the knowledge graph are improved through iterative optimization. The specific method is as follows: Get the extracted entities and relationships; Through the self-reflection mechanism, the extracted entities and relationships are corrected, and the knowledge graph is constructed based on the corrected entities and relationships; Based on similarity relationship retrieval, the structure in the knowledge graph is optimized, similar relationships are merged, and the redundancy of the knowledge graph is reduced. If it is found that the relationship is not merged properly, the entities and relationships are recalibrated and iterative optimization is performed.

5. The method for constructing a knowledge graph of hydropower station equipment based on a large language model according to claim 1 is characterized in that: The new data generated during the operation and maintenance of the hydropower station is monitored in real time, and the knowledge graph is automatically updated and maintained. The specific method is as follows: Automatically monitor new operation and maintenance data and fault information of hydropower station equipment, and capture knowledge update needs in real time; Re-extract entities and relationships and build and optimize the knowledge graph based on the newly added data; Associate new data with historical data to ensure the consistency of the knowledge graph.

6. The method for constructing a knowledge graph of hydropower station equipment based on a large language model according to claim 5 is characterized in that: The application of the constructed hydropower station equipment knowledge graph to the operation and maintenance management of the hydropower station also includes: Supports operation and maintenance personnel to quickly retrieve device information through natural language queries; Combine the equipment operation history and fault data in the knowledge graph to analyze the cause of the fault and generate predictive maintenance suggestions; Based on the generated fault causes and maintenance suggestions, assist in optimizing the operation and scheduling of the hydropower station.

7. The method for constructing a knowledge graph of hydropower station equipment based on a large language model according to claims 1 to 6, characterized in that: The large language model is a pre-trained deep learning model that can understand and generate natural language text and has the ability of few-sample learning and self-reflection.

8. A knowledge graph construction system for hydropower station equipment based on a large language model, characterized by: include: Data acquisition and preprocessing module, used to acquire data generated during the operation and maintenance of the hydropower station, and to clean and standardize the data; The information extraction module is used to extract entities and relationships of hydropower station equipment from the standardized data based on the large language model; The knowledge graph construction module is used to correct the extracted entities and relationships, build the knowledge graph of hydropower station equipment, and improve the accuracy and robustness of the knowledge graph through iterative optimization; The knowledge graph automatic update module is used to monitor the new data generated during the operation and maintenance of the hydropower station in real time, and automatically update and maintain the knowledge graph; The application module is used to apply the constructed knowledge graph of hydropower station equipment to the operation and maintenance management of the hydropower station.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for constructing a knowledge graph of hydropower station equipment based on a large language model as described in any one of claims 1 to 7 are implemented.

10. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for constructing a knowledge graph of hydropower station equipment based on a large language model as described in any one of claims 1 to 7 are implemented.

Images