Device fault diagnosis method and related device based on knowledge graph and large model

Abstract

The application relates to the technical field of power equipment fault diagnosis, and discloses a device fault diagnosis method based on a knowledge graph and a large model and related equipment. The method obtains power equipment to-be-diagnosed data and converts the to-be-diagnosed data into text data suitable for large model processing, performs abnormality analysis on the pretreated text data by using the large model, identifies abnormal states in the equipment operation process and forms a preliminary fault diagnosis result, generates a query request according to the fault diagnosis result, and performs retrieval by using a preset knowledge base containing a vector library and a knowledge graph library, obtains structured associated knowledge related to the equipment abnormal phenomenon, inputs the structured associated knowledge and the fault diagnosis result into the large model, so that the large model can comprehensively analyze and judge the preliminary diagnosis result in combination with the equipment fault associated relationship, the generated equipment fault diagnosis report can contain more sufficient fault associated basis, the accuracy of fault cause analysis and disposal suggestions is improved, and the explainability of the diagnosis result is enhanced.

Description

Technical Field

[0001] This invention relates to the field of power equipment fault diagnosis technology, and in particular to a fault diagnosis method and related equipment based on knowledge graphs and large models. Background Technology

[0002] With the increasing demand for intelligent operation and maintenance of power equipment, using artificial intelligence technology to analyze the operating status and diagnose faults of power equipment has become an important means to improve equipment operation and maintenance efficiency. In existing technologies, artificial intelligence models are typically used to analyze power equipment monitoring data to identify abnormal equipment conditions and generate fault diagnosis results.

[0003] However, existing fault diagnosis methods for power equipment based on large models usually analyze the input data directly and generate a diagnostic report. Since the large model itself lacks the ability to make inferences based on the professional knowledge of power equipment, it is difficult to fully combine the relationship between equipment defects, fault causes and disposal measures. This results in fault diagnosis reports with insufficient professional knowledge support, low diagnostic accuracy and poor interpretability. Summary of the Invention

[0004] In view of this, in order to solve the above-mentioned technical problems, the present invention provides a device fault diagnosis method and related equipment based on knowledge graphs and large models.

[0005] The first aspect of this invention provides a device fault diagnosis method based on knowledge graphs and large models, the method comprising:

[0006] Acquire the diagnostic data of the power equipment and convert the diagnostic data into text data;

[0007] The text data is preprocessed, and anomaly analysis is performed on the preprocessed text data using a preset large model to determine the fault diagnosis result corresponding to the power equipment.

[0008] A query request is generated based on the fault diagnosis result, and a search is performed in a preset knowledge base based on the query request to obtain structured association knowledge corresponding to the fault diagnosis result; wherein, the preset knowledge base includes a vector library and a knowledge graph library, and the knowledge graph library is used to store entities related to equipment faults and the association relationships between entities;

[0009] The structured association knowledge and the fault diagnosis results are input into the large model, which combines the equipment fault associations in the structured association knowledge to diagnose and judge the fault diagnosis results and generate an equipment fault diagnosis report.

[0010] In one example, the preprocessing of the text data and the anomaly analysis of the preprocessed text data using a preset large model to determine the fault diagnosis result corresponding to the power equipment include:

[0011] The text data is cleaned and information is extracted to extract state information that characterizes the changes in the operating state of the power equipment, and the preprocessed text data is formed based on the extracted state information.

[0012] The large model is used to perform state analysis on the preprocessed text data to identify anomalies in the text data.

[0013] Based on the preset prompts, the equipment defect phenomena corresponding to the abnormal points are summarized, and the fault diagnosis results are generated.

[0014] In one example, the step of retrieving information from a preset knowledge base according to the query request includes:

[0015] The query request is sent to the preset knowledge base via an interface call, and the search results returned by the preset knowledge base are obtained.

[0016] In one example, the step of generating a query request based on the fault diagnosis result and retrieving the structured related knowledge corresponding to the fault diagnosis result from a preset knowledge base based on the query request includes:

[0017] Semantic extraction is performed on the equipment defect phenomena in the fault diagnosis results, and a query request for retrieving equipment fault-related knowledge is generated based on the extracted defect phenomenon information.

[0018] The query request is vectorized to obtain a query vector;

[0019] Based on the semantic matching relationship between the query vector and the document content in the vector library, candidate entity information corresponding to the query request is determined;

[0020] Based on the candidate entity information, the corresponding source entity is determined in the knowledge graph database, and a topological relationship instance associated with the source entity is obtained;

[0021] Based on the matching relationship between the topological relationship instance and the query request, the corresponding target entity and the relationship between entities are determined, forming the structured association knowledge.

[0022] In one example, the step of inputting the structured association knowledge and the fault diagnosis results into the large model, so that the large model combines the equipment fault association relationships in the structured association knowledge to diagnose and judge the fault diagnosis results, includes:

[0023] Based on the fault diagnosis results, the structured association knowledge, and the preset prompt words, construct the diagnostic analysis input context corresponding to the large model;

[0024] Based on the diagnostic analysis input context, the large model uses the equipment fault association relationships in the structured association knowledge to perform cause analysis and handling association analysis on the abnormal phenomena in the fault diagnosis results, and generates the equipment fault diagnosis report.

[0025] In one example, based on the diagnostic analysis input context, the large model utilizes the equipment fault correlation relationships in the structured correlation knowledge to perform cause analysis and handling correlation analysis on the abnormal phenomena in the fault diagnosis results, generating the equipment fault diagnosis report, including:

[0026] Based on the large model, semantic association analysis is performed on the diagnostic analysis input context according to the preset prompt words, and the comprehensive diagnostic information is determined by combining the association between the equipment anomaly information and the equipment fault association.

[0027] The equipment fault diagnosis report is generated based on the comprehensive diagnostic information.

[0028] Secondly, the present invention also provides a device for diagnosing equipment faults based on knowledge graphs and large models, the device comprising:

[0029] The data conversion module is used to acquire the diagnostic data of the power equipment and convert the diagnostic data into text data.

[0030] The diagnostic module is used to preprocess the text data and perform anomaly analysis on the preprocessed text data using a preset large model to determine the fault diagnosis result corresponding to the power equipment.

[0031] The query module is used to generate a query request based on the fault diagnosis result, and to search in a preset knowledge base according to the query request to obtain structured association knowledge corresponding to the fault diagnosis result; wherein, the preset knowledge base includes a vector library and a knowledge graph library, and the knowledge graph library is used to store entities related to equipment faults and the association relationships between entities;

[0032] The reporting module is used to input the structured association knowledge and the fault diagnosis results into the large model, so that the large model combines the equipment fault association relationships in the structured association knowledge to diagnose and judge the fault diagnosis results and generate an equipment fault diagnosis report.

[0033] Thirdly, the present invention also provides an electronic device, the electronic device including a memory and a processor, the memory storing a computer program, the computer program being executed by the processor causing the processor to perform the steps of the device fault diagnosis method based on knowledge graphs and large models as described in the first aspect.

[0034] Fourthly, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed, implements the steps of the device fault diagnosis method based on knowledge graphs and large models as described in the first aspect.

[0035] Fifthly, the present invention also provides a computer program product comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions, wherein, when the program instructions are executed by a computer, the computer performs the steps of the device fault diagnosis method based on knowledge graphs and large models as described in the first aspect.

[0036] As can be seen from the above technical solutions, this invention acquires the data to be diagnosed from power equipment and converts it into text data suitable for processing by a large model, enabling equipment operating status information to participate in the fault analysis process in a unified semantic form. Furthermore, the large model is used to perform anomaly analysis on the preprocessed text data, identifying abnormal states during equipment operation and forming preliminary fault diagnosis results, achieving intelligent understanding of complex equipment state changes. Based on this, a query request is generated according to the fault diagnosis results, and a pre-set knowledge base including a vector library and a knowledge graph library is used for retrieval to obtain structured association knowledge related to equipment anomalies. This allows the correlation between equipment defect phenomena, possible fault causes, and corresponding processing information to be introduced into the subsequent diagnostic process. Furthermore, the structured association knowledge and fault diagnosis results are input into the large model together, enabling the large model to comprehensively evaluate the preliminary diagnosis results by combining equipment fault correlations, avoiding the problem of insufficient diagnostic basis caused by relying solely on the large model's own existing knowledge. By using the above methods, while improving the efficiency of equipment fault analysis by leveraging large models, the introduction of professional domain-related knowledge provided by knowledge graphs enables the final generated equipment fault diagnosis report to contain more comprehensive fault correlation evidence, improve the accuracy of fault cause analysis and handling suggestions, enhance the interpretability of diagnostic results, and thus improve the reliability of fault diagnosis in the intelligent operation and maintenance of power equipment. Attached Figure Description

[0037] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0038] Figure 1 This is an application environment diagram of a device fault diagnosis method based on knowledge graphs and large models provided in an embodiment of the present invention;

[0039] Figure 2 A flowchart of a device fault diagnosis method based on knowledge graphs and large models provided in this embodiment of the invention;

[0040] Figure 3 A schematic diagram of the structure of a device fault diagnosis device based on knowledge graphs and large models provided in an embodiment of the present invention;

[0041] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0042] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0043] The device fault diagnosis method based on knowledge graphs and large models provided in this application can be applied to, for example... Figure 1In the application environment shown, terminal 101 communicates with server 102 via a network. A data storage system can store the data that server 102 needs to process. The data storage system can be integrated onto server 102 or placed on a cloud or other network server. Terminal 101 or server 102 executes a device fault diagnosis method based on knowledge graphs and large models. This method includes: acquiring the data to be diagnosed from power equipment and converting it into text data; preprocessing the text data and using a preset large model to perform anomaly analysis on the preprocessed text data to determine the fault diagnosis result corresponding to the power equipment; generating a query request based on the fault diagnosis result and retrieving the structured association knowledge corresponding to the fault diagnosis result from a preset knowledge base; wherein the preset knowledge base includes a vector library and a knowledge graph library, the knowledge graph library being used to store entities related to device faults and the relationships between entities; inputting the structured association knowledge and the fault diagnosis result into the large model, enabling the large model to combine the device fault associations in the structured association knowledge to diagnose and judge the fault diagnosis result and generate a device fault diagnosis report.

[0044] The electrical equipment can include transformers, circuit breakers, GIS equipment, and other power transmission and transformation equipment. The data to be diagnosed can include equipment operating status data, online monitoring data, test and inspection data, inspection record data, and historical defect record data.

[0045] Terminal 101 can be, but is not limited to, various personal computers, laptops, smartphones, and tablets.

[0046] Server 102 can be a standalone physical server, a server cluster or distributed system consisting of multiple physical servers, or a cloud server that provides cloud computing services.

[0047] like Figure 2 As shown, this application provides a device fault diagnosis method based on knowledge graphs and large models, which can be applied to... Figure 1 Taking terminal 101 or server 102 as an example, the explanation includes the following steps S1 to S4. Wherein:

[0048] Step S1: Obtain the diagnostic data of the power equipment and convert the diagnostic data into text data.

[0049] In this embodiment, the data to be diagnosed may include monitoring data files generated during the operation of power equipment. For example, in a transformer fault diagnosis scenario, the data to be diagnosed may include: oil chromatography detection data, temperature change data, load operation data, and historical maintenance records.

[0050] Since large language models primarily process information in the form of natural language, the raw device data needs to be textualized first.

[0051] In the specific implementation process, since the operating data of power equipment usually exists in structured or semi-structured form, such as Excel spreadsheets, database records, monitoring system output files, etc., and large models usually cannot directly understand the above data formats, the field information in the equipment monitoring file can be read by document parsing, and the equipment number, detection time, parameter change trend and abnormal description information can be converted into text form.

[0052] Therefore, this step requires using a rule-based conversion algorithm to convert diagnostic data in different formats into text data described in natural language according to unified rules, while retaining key monitoring values, detection time, equipment parameters, and other core information from the original data to ensure that the large model can accurately identify the data content.

[0053] Step S2: Preprocess the text data and use a pre-set large model to perform anomaly analysis on the preprocessed text data to determine the fault diagnosis results corresponding to the power equipment.

[0054] Because equipment monitoring data may contain inconsistent data formats, invalid fields, and redundant information, preprocessing of the text data is necessary. Preprocessing includes data cleaning and status information extraction. Data cleaning involves removing invalid fields, duplicate descriptions, and information that does not reflect the equipment status. Status information extraction involves extracting information from the text that reflects changes in the equipment status, such as parameters exceeding thresholds, continuous parameter changes, or significant differences from historical status.

[0055] For example: The original text is: "The operating data of a certain device in the past three months is as follows..." After extraction, it becomes: "The detection parameters of a certain device continue to rise during operation, showing an abnormal trend."

[0056] The above processing can reduce the impact of irrelevant information on the large model analysis process and improve the accuracy of anomaly identification.

[0057] Subsequently, the large model is invoked to perform anomaly analysis on the preprocessed text data. In this embodiment, the large model is used to complete the semantic understanding of the equipment status, thereby identifying the abnormal points in the operating status of the power equipment through the large model. Combined with the knowledge reserves of the large model in the field of power equipment faults, the possible types of equipment defects are initially located.

[0058] Among them, the large model can be a commercial large model, and there is no restriction on the type of large model.

[0059] Step S3: Generate a query request based on the fault diagnosis results, and search the preset knowledge base according to the query request to obtain the structured association knowledge corresponding to the fault diagnosis results; wherein, the preset knowledge base includes a vector library and a knowledge graph library, and the knowledge graph library is used to store the entities related to the device fault and the association relationships between the entities.

[0060] Since the initial fault diagnosis results output by the large model mainly describe abnormal phenomena, while actual fault analysis requires further determination of possible fault causes, this embodiment generates a knowledge query request based on the fault diagnosis results. This query request is used to obtain related knowledge from the equipment fault knowledge base.

[0061] The preset knowledge base includes a vector library and a knowledge graph library. The vector library is used to realize semantic similarity retrieval. The equipment fault-related entities and the relationships between entities in the knowledge graph library are stored in the form of triples. The triples include the source entity, the target entity, and the relationship information used to represent the relationship between the source entity and the target entity.

[0062] By querying the preset knowledge base, the structured knowledge corresponding to the currently identified fault diagnosis result can be quickly located. This structured knowledge is a structured knowledge information used to assist in equipment fault diagnosis, based on entities such as equipment, defect phenomena, fault causes, and handling measures in the knowledge graph, as well as the relationships between entities.

[0063] Step S4: Input the structured association knowledge and fault diagnosis results into the large model, so that the large model can combine the equipment fault association relationship in the structured association knowledge to diagnose and judge the fault diagnosis results and generate an equipment fault diagnosis report.

[0064] In this process, structured association knowledge and fault diagnosis results are input into a large model as the input context. The large model then uses the equipment fault association relationships in the structured association knowledge to perform cause analysis and handling association analysis on the abnormal phenomena in the fault diagnosis results, and generates an equipment fault diagnosis report.

[0065] This equipment fault diagnosis report includes a description of the abnormal equipment phenomenon, an analysis of the fault cause, and corresponding handling information, used to characterize the fault diagnosis results of power equipment and the basis for subsequent handling. Specifically, the description of the abnormal equipment phenomenon describes the abnormal state or defect identified by the large model; the fault cause analysis analyzes the possible causes corresponding to the abnormal phenomenon based on the equipment fault relationships in the knowledge graph; and the handling information combines the related knowledge in the knowledge base to output corresponding handling measures or operation and maintenance suggestions.

[0066] It should be noted that this embodiment of the application acquires the data to be diagnosed from power equipment and converts it into text data suitable for processing by a large model, enabling equipment operating status information to participate in the fault analysis process in a unified semantic form. Furthermore, the large model is used to perform anomaly analysis on the preprocessed text data, identifying abnormal states during equipment operation and forming preliminary fault diagnosis results, achieving intelligent understanding of complex equipment state changes. Based on this, a query request is generated according to the fault diagnosis results, and a pre-set knowledge base including a vector library and a knowledge graph library is used for retrieval to obtain structured association knowledge related to equipment anomalies. This allows the correlation between equipment defect phenomena, possible fault causes, and corresponding processing information to be introduced into the subsequent diagnostic process. Furthermore, the structured association knowledge and the fault diagnosis results are jointly input into the large model, enabling the large model to comprehensively evaluate the preliminary diagnosis results by combining equipment fault correlations, avoiding the problem of insufficient diagnostic basis caused by relying solely on the large model's own existing knowledge. By using the above methods, while improving the efficiency of equipment fault analysis by leveraging large models, the introduction of professional domain-related knowledge provided by knowledge graphs enables the final generated equipment fault diagnosis report to contain more comprehensive fault correlation evidence, improve the accuracy of fault cause analysis and handling suggestions, enhance the interpretability of diagnostic results, and thus improve the reliability of fault diagnosis in the intelligent operation and maintenance of power equipment.

[0067] In some embodiments, text data is preprocessed, and a pre-defined large model is used to perform anomaly analysis on the preprocessed text data to determine the fault diagnosis results corresponding to the power equipment. This includes: cleaning and extracting information from the text data to extract state information that characterizes changes in the operating state of the power equipment, and forming preprocessed text data based on the extracted state information; performing state analysis on the preprocessed text data using a large model to identify anomalies in the text data; and summarizing the equipment defect phenomena corresponding to the anomalies based on pre-defined prompts to generate fault diagnosis results.

[0068] First, the text data is cleaned and information is extracted to extract state information that represents changes in the operating status of power equipment, and preprocessed text data is formed based on the extracted state information.

[0069] Text data can be text information converted from power equipment operation monitoring data, used to describe the state changes of the equipment during operation. Since equipment monitoring data usually contains various types of information such as basic equipment information, acquisition time, operating parameters, and detection records, some of which is not directly related to fault analysis, the text data needs to be preprocessed to reduce the impact of invalid information on the subsequent large-scale model analysis process.

[0070] For example, when diagnosing faults in converter transformers, the original monitoring data may contain oil chromatography data corresponding to multiple time points:

[0071] "On March 1, 2023, offline chromatography detected an acetylene content of 1.63 μL / L, and subsequent tests showed that the acetylene content gradually increased to 2.53 μL / L." After information extraction, the conclusion can be: "The acetylene content in the equipment insulating oil continued to increase."

[0072] This status information can more directly reflect abnormal changes during equipment operation, providing effective input for subsequent status analysis of large models.

[0073] In this embodiment, the large model is used to perform semantic understanding and state analysis on the preprocessed text data. By analyzing the device state description, parameter change trends and abnormal behavior information contained in the text, the abnormal points in the device operation state are identified.

[0074] Anomalies can be understood as abnormal changes in the equipment's operation that differ from its normal operating state. Examples include equipment detection parameters exceeding preset ranges of interest, a certain status parameter continuously increasing or decreasing, or significant differences between the equipment's operating state and its historical normal state.

[0075] Unlike traditional methods that rely on fixed thresholds, this embodiment utilizes a large model to analyze the contextual relationships within the text. It not only focuses on changes in a single parameter but also combines temporal changes and semantic relationships in the device status description to identify potential anomalies.

[0076] For example, by setting the prompt message as follows:

[0077]

[0078] Based on this, for the following text: "The acetylene content in the insulating oil of a converter transformer showed a slight increase, and subsequent shortening of the offline chromatographic analysis cycle revealed a continuous increase in the acetylene content," the large model can identify "continuous increase in acetylene content" as an abnormal state and use this information as the basis for subsequent defect induction.

[0079] Furthermore, in this embodiment, in order to ensure that the output of the large model meets the requirements of equipment fault diagnosis, prompt information can be preset to limit the way the large model analyzes and summarizes anomalies.

[0080] The prompts can be used to guide the large model: identify the corresponding defect based on the anomaly description; standardize the description of the anomaly; and output fault diagnosis results suitable for knowledge retrieval.

[0081] For example, the preset prompt message may include: "Based on the description of the anomaly you found, summarize the equipment defect phenomenon and describe the defect phenomenon in one sentence."

[0082] Based on this prompt, the large model will not directly output long analysis text, but will instead summarize the abnormal state into a standardized defect description.

[0083] For example: Inputting the abnormal information: "Acetylene content exceeds the warning value and shows a continuous increasing trend." After large-scale model induction, the result is: "Acetylene content abnormal growth defect".

[0084] The fault diagnosis results are used to generate subsequent query requests and further retrieve structured related knowledge from the knowledge base that relates to the defect phenomenon.

[0085] In some embodiments, retrieving data from a preset knowledge base based on a query request includes: sending a query request to the preset knowledge base via an API call and obtaining the retrieval results returned by the preset knowledge base.

[0086] In this embodiment, when searching in the preset knowledge base according to the query request, the data interaction between the query request and the preset knowledge base can be realized through interface call.

[0087] Specifically, after completing the device anomaly analysis and generating a query request, the query request is sent to the pre-configured knowledge base retrieval interface, which then calls the corresponding knowledge base retrieval service and obtains the retrieval results returned by the knowledge base.

[0088] The knowledge base retrieval interface is used to realize the information transfer between the large model diagnosis process and the external knowledge base, so that the equipment fault diagnosis process can call the knowledge retrieval function according to the preset workflow, without having to directly integrate the knowledge base into the large model.

[0089] In one implementation, the knowledge base retrieval process can be implemented via an HTTP RESTful API call. Specifically, the knowledge base retrieval function can be pre-encapsulated as an independent retrieval service and provided as a callable tool for the fault diagnosis process. After the large model completes anomaly analysis and generates a query request, the query request is sent as an API call parameter to the retrieval service, which then performs the knowledge base query and returns the corresponding search results.

[0090] In one specific implementation, a new "Retrieve Knowledge Base" workflow is added by calling GraphRAG via the HTTP RESTful API interface, published as a tool, and called by other workflows.

[0091] The "Start" node is the input to the knowledge base retrieval workflow, including two parameters: knowledge base ID and keywords. The knowledge base ID can be obtained from the knowledge base details page. The keywords are the query request generated above.

[0092] The "Separate Knowledge Base ID" node is used to convert the knowledgeId parameter into an array, separated by commas.

[0093] The "Generate Request Based on Keyword" node is used to encapsulate the keyword parameter into the HTTP Body of GraphRAG.

[0094] The "Iteration" node traverses the separated knowledge base IDs and makes an HTTP request for each knowledge base ID. The body of the HTTP request is the HTTP body generated by the "Generate Request Based on Keywords" node.

[0095] The “Format Text” node converts the JSON returned by GraphRAG into a text representation.

[0096] The "End" node is used to return the text description to the caller.

[0097] By using the aforementioned interface call method, this embodiment decouples the equipment fault diagnosis process from the knowledge base retrieval process, enabling the large model to dynamically call professional knowledge resources according to actual diagnostic needs. On the one hand, this improves the flexibility and scalability of the fault diagnosis process; on the other hand, by calling the knowledge base through an interface, the large model can avoid directly relying on its own training knowledge for judgment, allowing the diagnostic process to combine real-time retrieved equipment fault domain knowledge, thereby improving the accuracy and interpretability of subsequent diagnostic results.

[0098] In some embodiments, a query request is generated based on the fault diagnosis result, and a search is performed in a preset knowledge base based on the query request to obtain structured associated knowledge corresponding to the fault diagnosis result. This includes: semantically extracting the equipment defect phenomena in the fault diagnosis result, and generating a query request for retrieving equipment fault associated knowledge based on the extracted defect phenomenon information; vectorizing the query request to obtain a query vector; determining candidate entity information corresponding to the query request based on the semantic matching relationship between the query vector and the document content in the vector library; determining the corresponding source entity in the knowledge graph library based on the candidate entity information, and obtaining the topological relationship instance associated with the source entity; and determining the corresponding target entity and the relationship between entities based on the matching relationship between the topological relationship instance and the query request to form structured associated knowledge.

[0099] First, semantic extraction is performed on the equipment defect phenomena in the fault diagnosis results, and a query request is generated based on the extracted defect phenomenon information to retrieve equipment fault-related knowledge.

[0100] In this embodiment, the fault diagnosis result is generated by the aforementioned large-scale model anomaly analysis process, and it is mainly used to describe the abnormal state or defect phenomenon of the current power equipment. Since the fault diagnosis result is usually represented in natural language, and the knowledge in the knowledge base is organized in the form of entities and entity relationships, it is necessary to first extract key semantic information that can characterize the fault features of the equipment from the fault diagnosis result, and then generate a query request based on this information.

[0101] After receiving a query request, the knowledge base extracts the entity parts and retrieves similar entity names from the vector database. Based on these entity names, it expands the subgraph in the graph database to obtain the relationships between them.

[0102] In this embodiment, since the information stored in the equipment fault knowledge base may be described in different textual ways, if only keyword matching is used, knowledge that is semantically similar but textually different may not be retrieved.

[0103] Therefore, the query request is input into the vectorization model, which then transforms the query request into a query vector that can represent its semantic features.

[0104] In this embodiment, the vector library is used to store document information related to device failures and corresponding entity information.

[0105] Specifically, the query vector is matched with the semantic vectors corresponding to each document in the vector library, and candidate documents related to the current query request are determined based on the degree of semantic correlation between the two.

[0106] Among them, semantic matching relationship is used to represent the degree of relevance between the query request and the existing knowledge content in the knowledge base.

[0107] For example, similarity values ​​are calculated using the request query vector and the key fields of each document in the vector database to find the top few most similar documents. The entity names can be obtained from the key fields of these top few most similar documents as candidate entity information. Generally, based on cosine similarity, the top few semantically closest documents are retrieved from the vector database, and the entity names of these top few documents are extracted as candidate entity information.

[0108] A knowledge graph is a semantic network that represents real-world knowledge in a structured form. Its core consists of entities and relations, enabling the organization of scattered, heterogeneous information into a machine-understandable and reasonable graph structure. Entities are the basic nodes in a knowledge graph, representing objects or concepts with independent semantics within a power grid system. Entities are distinguishable and independently existing things, such as equipment or defects. Relationships are semantic edges connecting entities, used to describe the logic, causality, or correlation between them; for example, a certain model of equipment may experience a certain defect.

[0109] The entities in the knowledge graph can include equipment entities, equipment component entities, defect phenomenon entities, defect cause entities, and defect handling measure entities. These entities are connected by relationships to form a topological structure.

[0110] For example: The possible cause of water ingress defect in oil level gauge is uneven sealing surface. Here, "water ingress defect in oil level gauge" is the source entity; "uneven sealing surface" is the target entity; and "possible cause" is the relationship between the entities.

[0111] Therefore, after identifying the candidate entity, we can further query the relation edges connected to the entity in the knowledge graph to obtain the topological relation instances related to the current device defect.

[0112] In this embodiment, since the same entity in the knowledge graph may be associated with multiple target entities, it is necessary to further filter the associations related to the current fault diagnosis scenario by combining the query request.

[0113] For example: Source entity: "Abnormal increase in acetylene content", the possible causes include: internal discharge, insulation aging; the treatment measure is to carry out further testing.

[0114] Based on the current query request: "Possible reasons for abnormal increase in acetylene content", we prioritize filtering relational instances related to "possible reasons" to form structured relational knowledge such as: "Abnormal increase in acetylene content - possible reason - internal discharge".

[0115] Among them, structured association knowledge is a knowledge representation form composed of equipment fault-related entities and the relationships between entities, which is used to input into the large model for comprehensive diagnosis and judgment.

[0116] Through the above process, this embodiment first uses the defect phenomenon information in the fault diagnosis results to generate a query request, and then combines the semantic retrieval capability of the vector library and the relational reasoning capability of the knowledge graph library to realize the acquisition of related fault knowledge from abnormal phenomena.

[0117] Compared to methods that rely solely on keyword searches or depend on the knowledge of the large model itself for diagnosis, this embodiment can proactively retrieve relevant professional knowledge relationships based on the current abnormal state of the device. This enables the large model to perform diagnostic analysis based on more accurate and domain-relevant knowledge, thereby improving the reliability of fault cause analysis and handling suggestions.

[0118] For example, the main process for retrieving knowledge is as follows:

[0119] Semantic entity retrieval: When GraphRAG receives a user query request, it first uses a vectorization model to encode the query text into a query vector, and calculates the similarity (such as cosine similarity) with the key fields of each document in the vector database. It then retrieves the top few documents with the most similar semantics and extracts the candidate entity names contained in the key fields from them.

[0120] Expanding the subgraph: After obtaining the first few candidate entities, GraphRAG queries the knowledge graph database for all the edges associated with these entities, and combines the original query semantics to filter the edge types or attributes based on relevance, retaining only the relational edges that highly match the query intent; finally, the system not only returns the filtered edge list, but also the original document content associated with each edge.

[0121] In one specific embodiment, based on the actual business needs of power grid fault diagnosis, the ontology model of the knowledge graph is systematically defined starting from typical fault scenarios.

[0122] The entity types are defined in Table 1 below:

[0123] Table 1:

[0124]

[0125] The relation types are defined in Table 2 below:

[0126] Table 2:

[0127]

[0128] In some embodiments, structured association knowledge and fault diagnosis results are input into a large model, enabling the large model to combine the equipment fault association relationships in the structured association knowledge to diagnose and judge the fault diagnosis results. This includes: constructing a diagnostic analysis input context corresponding to the large model based on the fault diagnosis results, structured association knowledge, and preset prompt words; and using the large model to perform cause analysis and handling association analysis on the abnormal phenomena in the fault diagnosis results based on the equipment fault association relationships in the structured association knowledge, and generating an equipment fault diagnosis report.

[0129] The diagnostic analysis input context is used to provide the large model with complete fault diagnosis analysis conditions, enabling the large model to simultaneously acquire the current abnormal phenomenon of the equipment, the equipment fault knowledge related to the abnormal phenomenon, and the prompt information used to limit the diagnostic analysis method.

[0130] The preset prompts are used to guide the large model to perform analysis according to the equipment fault diagnosis task. The preset prompts can be set according to the equipment fault diagnosis requirements, for example:

[0131]

[0132] This method enables large models to complete subsequent analyses according to pre-set diagnostic logic.

[0133] Furthermore, after receiving the diagnostic analysis input context, the large model not only uses its own semantic understanding ability to analyze abnormal phenomena, but also combines the equipment fault correlations provided by the knowledge graph to perform correlation reasoning.

[0134] Specifically, based on the diagnostic analysis input context, the large model uses the equipment fault correlation relationship in the structured association knowledge to perform cause analysis and treatment correlation analysis on the abnormal phenomena in the fault diagnosis results, and generates an equipment fault diagnosis report. This includes: based on the large model, performing semantic correlation analysis on the diagnostic analysis input context according to preset prompt words, and determining comprehensive diagnostic information by combining the correlation between equipment abnormal information and equipment fault correlation relationship; and generating an equipment fault diagnosis report based on the comprehensive diagnostic information.

[0135] The large model first performs semantic understanding of the equipment anomaly information in the diagnostic analysis input context to determine the anomaly object that needs further analysis. Then, based on the equipment fault association relationship in the structured association knowledge, it analyzes the association between the anomaly object and related entities in the knowledge graph, thereby determining the comprehensive diagnostic information that matches the current anomaly state.

[0136] Among them, semantic association analysis is a large model that uses natural language understanding to match different information in the input context to determine the correspondence between equipment abnormal information and equipment failure, such as the association between defect phenomena and failure causes, the association between defect phenomena and equipment components, and the association between defect phenomena and handling measures.

[0137] In this embodiment, preset prompts can be set according to the equipment fault diagnosis requirements to guide the large model through the following process: identifying corresponding fault-related knowledge based on abnormal equipment phenomena; analyzing possible fault causes based on entity relationships in the knowledge graph; determining corresponding handling directions based on related knowledge; and outputting analysis results in the form of a preset diagnostic report.

[0138] The above prompts enable the large model to analyze the equipment fault diagnosis task and generate comprehensive diagnostic information. This comprehensive diagnostic information is used to characterize the final diagnostic result formed by the large model after combining equipment anomaly information and the correlation between equipment faults.

[0139] Comprehensive diagnostic information can include equipment anomaly confirmation information, fault correlation analysis information, and handling correlation information. Among them, equipment anomaly confirmation information describes the current abnormal state of the equipment; fault correlation analysis information describes the possible causes of the fault determined by the correlation relationship based on the knowledge graph; and handling correlation information describes the direction of further detection or handling corresponding to the current abnormal state.

[0140] Subsequently, an equipment fault diagnosis report is generated based on the comprehensive diagnostic information. The equipment fault diagnosis report is not directly generated by the large model based on the original equipment data, but is formed by introducing the equipment fault correlation provided by the knowledge graph on the basis of the existing preliminary fault diagnosis results, so that the large model can combine professional domain knowledge to complete the fault assessment.

[0141] Through the above implementation method, this embodiment uses preset prompt words to control the large model to perform targeted semantic association analysis on the diagnostic analysis input context, so that the large model can match the equipment abnormality information with the equipment fault association relationship in the knowledge graph, instead of just reasoning based on its own training knowledge.

[0142] Compared to directly generating fault diagnosis reports using large models, this embodiment introduces structured, correlated knowledge, giving the diagnostic process clear knowledge sources and correlational evidence, thus improving the reliability of fault cause analysis and handling recommendations. Simultaneously, by comprehensively understanding anomaly information and knowledge relationships through large models, the completeness and interpretability of equipment fault diagnosis reports can be improved, addressing the problem of insufficient professional knowledge correlation in existing equipment fault diagnosis processes.

[0143] Based on the same inventive concept, this application also provides a knowledge graph and large model-based device for implementing the above-mentioned knowledge graph and large model-based device fault diagnosis method.

[0144] The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more device fault diagnosis device embodiments based on knowledge graphs and large models provided below can be found in the limitations of the device fault diagnosis method based on knowledge graphs and large models above, and will not be repeated here.

[0145] like Figure 3 As shown in the figure, this application provides a device fault diagnosis apparatus based on knowledge graphs and large models, the apparatus comprising:

[0146] The data conversion module 100 is used to acquire the diagnostic data of the power equipment and convert the diagnostic data into text data.

[0147] The diagnostic module 200 is used to preprocess the text data and perform anomaly analysis on the preprocessed text data using a preset large model to determine the fault diagnosis results corresponding to the power equipment.

[0148] The query module 300 is used to generate a query request based on the fault diagnosis result, and to search in the preset knowledge base according to the query request to obtain the structured association knowledge corresponding to the fault diagnosis result; wherein, the preset knowledge base includes a vector library and a knowledge graph library, and the knowledge graph library is used to store the entities related to the device fault and the association relationships between the entities;

[0149] The report module 400 is used to input structured association knowledge and fault diagnosis results into the large model, so that the large model can combine the equipment fault association relationship in the structured association knowledge to diagnose and judge the fault diagnosis results and generate equipment fault diagnosis reports.

[0150] In some embodiments, the diagnostic module 200 is configured to:

[0151] The text data is cleaned and information is extracted to extract state information that represents changes in the operating status of power equipment, and preprocessed text data is generated based on the extracted state information.

[0152] A large model is used to perform state analysis on the preprocessed text data to identify outliers in the text data.

[0153] Based on preset prompts, the equipment defects corresponding to the anomalies are summarized, and fault diagnosis results are generated.

[0154] In some embodiments, the query module 300 is used for:

[0155] The query request is sent to the preset knowledge base via an API call, and the search results returned by the preset knowledge base are obtained.

[0156] In some embodiments, the query module 300 is used for:

[0157] Semantic extraction is performed on the equipment defect phenomena in the fault diagnosis results, and a query request for retrieving equipment fault-related knowledge is generated based on the extracted defect phenomenon information.

[0158] The query request is vectorized to obtain the query vector;

[0159] Based on the semantic matching relationship between the query vector and the document content in the vector library, candidate entity information corresponding to the query request is determined;

[0160] Based on the candidate entity information, determine the corresponding source entity in the knowledge graph database and obtain the topological relationship instance associated with the source entity;

[0161] Based on the matching relationship between topological relationship instances and query requests, the corresponding target entities and the relationships between entities are determined, forming structured relational knowledge.

[0162] In some embodiments, the reporting module 400 is used for:

[0163] Based on the fault diagnosis results, structured association knowledge, and preset prompts, construct the diagnostic analysis input context corresponding to the large model;

[0164] Based on the diagnostic analysis input context, the large model utilizes the equipment fault correlation relationships in the structured association knowledge to perform cause analysis and handling correlation analysis on the abnormal phenomena in the fault diagnosis results, and generates an equipment fault diagnosis report.

[0165] In some embodiments, the reporting module 400 is used for:

[0166] Based on the large model, semantic association analysis is performed on the input context of diagnostic analysis according to preset prompt words, and comprehensive diagnostic information is determined by combining the association between equipment anomaly information and equipment fault correlation.

[0167] A device fault diagnosis report is generated based on the comprehensive diagnostic information.

[0168] like Figure 4 As shown in the figure, this application provides an electronic device. The electronic device 10 includes a memory 20 and a processor 30. The memory 20 stores a computer program. When the computer program is executed by the processor 30, the processor 30 performs the steps of the device fault diagnosis method based on knowledge graph and large model as described in the above embodiment.

[0169] This application provides a computer-readable storage medium storing a computer program thereon. When the computer program is executed, it implements the steps of the device fault diagnosis method based on knowledge graphs and large models as described in the above embodiments.

[0170] This application provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions, wherein when the program instructions are executed by a computer, the computer performs the steps of the device fault diagnosis method based on knowledge graphs and large models as described in the above embodiments.

[0171] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, electronic devices, computer storage media, and computer program products described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0172] In the embodiments of this application, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design that is described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design. Specifically, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0173] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, apparatus, product, or device that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0174] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0175] In the several embodiments provided by this invention, it should be understood that the disclosed devices, electronic devices, computer storage media, computer program products, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0176] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0177] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0178] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for executing all or part of the steps of the methods described in the various embodiments of the present invention through a computer device (which may be a personal computer, a server, or a network device, etc.). The aforementioned storage medium includes: USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, optical disks, and other media capable of storing program code.

[0179] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for diagnosing equipment faults based on knowledge graphs and large models, characterized in that, The method includes: Acquire the diagnostic data of the power equipment and convert the diagnostic data into text data; The text data is preprocessed, and anomaly analysis is performed on the preprocessed text data using a preset large model to determine the fault diagnosis result corresponding to the power equipment. A query request is generated based on the fault diagnosis result, and a search is performed in a preset knowledge base based on the query request to obtain structured association knowledge corresponding to the fault diagnosis result; wherein, the preset knowledge base includes a vector library and a knowledge graph library, and the knowledge graph library is used to store entities related to equipment faults and the association relationships between entities; The structured association knowledge and the fault diagnosis results are input into the large model, which combines the equipment fault associations in the structured association knowledge to diagnose and judge the fault diagnosis results and generate an equipment fault diagnosis report.

2. The equipment fault diagnosis method based on knowledge graphs and large models according to claim 1, characterized in that, The process of preprocessing the text data and performing anomaly analysis on the preprocessed text data using a pre-defined large model to determine the fault diagnosis result corresponding to the power equipment includes: The text data is cleaned and information is extracted to extract state information that characterizes the changes in the operating state of the power equipment, and the preprocessed text data is formed based on the extracted state information. The large model is used to perform state analysis on the preprocessed text data to identify anomalies in the text data. Based on the preset prompts, the equipment defect phenomena corresponding to the abnormal points are summarized, and the fault diagnosis results are generated.

3. The equipment fault diagnosis method based on knowledge graphs and large models according to claim 1, characterized in that, The step of searching the preset knowledge base according to the query request includes: The query request is sent to the preset knowledge base via an interface call, and the search results returned by the preset knowledge base are obtained.

4. The equipment fault diagnosis method based on knowledge graphs and large models according to claim 1, characterized in that, The step of generating a query request based on the fault diagnosis result and retrieving structured related knowledge corresponding to the fault diagnosis result from a preset knowledge base based on the query request includes: Semantic extraction is performed on the equipment defect phenomena in the fault diagnosis results, and a query request for retrieving equipment fault-related knowledge is generated based on the extracted defect phenomenon information. The query request is vectorized to obtain a query vector; Based on the semantic matching relationship between the query vector and the document content in the vector library, candidate entity information corresponding to the query request is determined; Based on the candidate entity information, the corresponding source entity is determined in the knowledge graph database, and a topological relationship instance associated with the source entity is obtained; Based on the matching relationship between the topological relationship instance and the query request, the corresponding target entity and the relationship between entities are determined, forming the structured association knowledge.

5. The equipment fault diagnosis method based on knowledge graphs and large models according to claim 1, characterized in that, The step of inputting the structured association knowledge and the fault diagnosis results into the large model, so that the large model combines the equipment fault association relationships in the structured association knowledge to diagnose and judge the fault diagnosis results, includes: Based on the fault diagnosis results, the structured association knowledge, and the preset prompt words, construct the diagnostic analysis input context corresponding to the large model; Based on the diagnostic analysis input context, the large model uses the equipment fault association relationships in the structured association knowledge to perform cause analysis and handling association analysis on the abnormal phenomena in the fault diagnosis results, and generates the equipment fault diagnosis report.

6. The equipment fault diagnosis method based on knowledge graphs and large models according to claim 5, characterized in that, Based on the diagnostic analysis input context, the large model utilizes the equipment fault correlation relationships in the structured association knowledge to perform cause analysis and handling correlation analysis on the abnormal phenomena in the fault diagnosis results, generating the equipment fault diagnosis report, including: Based on the large model, semantic association analysis is performed on the diagnostic analysis input context according to the preset prompt words, and the comprehensive diagnostic information is determined by combining the association between the equipment anomaly information and the equipment fault association. The equipment fault diagnosis report is generated based on the comprehensive diagnostic information.

7. A device for diagnosing equipment faults based on knowledge graphs and large models, characterized in that, The device includes: The data conversion module is used to acquire the diagnostic data of the power equipment and convert the diagnostic data into text data. The diagnostic module is used to preprocess the text data and perform anomaly analysis on the preprocessed text data using a preset large model to determine the fault diagnosis result corresponding to the power equipment. The query module is used to generate a query request based on the fault diagnosis result, and to search in a preset knowledge base according to the query request to obtain structured association knowledge corresponding to the fault diagnosis result; wherein, the preset knowledge base includes a vector library and a knowledge graph library, and the knowledge graph library is used to store entities related to equipment faults and the association relationships between entities; The reporting module is used to input the structured association knowledge and the fault diagnosis results into the large model, so that the large model combines the equipment fault association relationships in the structured association knowledge to diagnose and judge the fault diagnosis results and generate an equipment fault diagnosis report.

8. An electronic device, characterized in that, The electronic device includes a memory and a processor. The memory stores a computer program, which, when executed by the processor, causes the processor to perform the steps of the device fault diagnosis method based on knowledge graphs and large models as described in any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed, it implements the steps of the device fault diagnosis method based on knowledge graphs and large models as described in any one of claims 1-6.

10. A computer program product, characterized in that, The computer program product includes a computer program stored on a non-transitory computer-readable storage medium, the computer program including program instructions, wherein when the program instructions are executed by a computer, the computer performs the steps of the device fault diagnosis method based on knowledge graphs and large models as described in any one of claims 1-6.

Images