Transformer multi-mode fault diagnosis method, system, equipment and medium

Abstract

The invention discloses a transformer multi-mode fault diagnosis method, system and device and a medium, and relates to the field of transformer fault diagnosis. The method comprises the following steps: acquiring multi-modal data of a transformer and preprocessing the multi-modal data to obtain preprocessed multi-modal data; based on a retrieval enhancement generation technology, constructing a transformer fault knowledge base according to a transformer fault diagnosis report and a maintenance guide rule type industry standard; inputting the preprocessed multi-modal data into a transformer fault diagnosis model for fault diagnosis to obtain a fault diagnosis result; converting the fault diagnosis result into a cue word vector according to a cue word template and a vector embedding model, matching the cue word vector with a vector index in a transformer fault knowledge base to obtain a transformer fault case with the highest similarity, and extracting similar case information; and generating and outputting a fault diagnosis report according to the fault diagnosis result and the similar case information. According to the method, the response speed and the resource utilization efficiency can be improved while the diagnosis precision is ensured.

Description

Technical Field

[0001] This application relates to the field of transformer fault diagnosis, and particularly to a multi-modal transformer fault diagnosis method, system, device, and medium. Background Art

[0002] With the rapid development of smart grids and the increasing complexity of power equipment, transformers, as core equipment in the power system, the monitoring and fault diagnosis of their operating states are particularly important. Traditional fault diagnosis methods mainly rely on single-modal data, such as electrical parameters, temperature, vibration, etc. These methods often show problems such as low diagnostic accuracy and slow response speed when facing complex and changeable fault situations. In addition, with the rapid development of artificial intelligence technology, especially the wide application of large models (such as GPT-4) in fields such as natural language processing and image recognition, new ideas and methods have been provided for transformer fault diagnosis.

[0003] However, although large models perform well in processing general domain data, there are still some limitations in their applications in specific fields (such as the power system). For example, when large models process power data with strong professionalism, the "hallucination" phenomenon may occur, that is, inaccurate or incorrect diagnostic results are generated without sufficient professional knowledge. In addition, large models have huge computational resource requirements and are difficult to meet the requirements of fault diagnosis tasks with high real-time requirements. Summary of the Invention

[0004] The purpose of this application is to provide a multi-modal transformer fault diagnosis method, system, device, and medium, which can improve the response speed and resource utilization efficiency while ensuring the diagnostic accuracy.

[0005] To achieve the above purpose, this application provides the following solutions:

[0006] In the first aspect, this application provides a multi-modal transformer fault diagnosis method, including:

[0007] Collect multi-modal data of the transformer and perform preprocessing to obtain preprocessed multi-modal data; the multi-modal data includes: image and video data, gas component data, and infrared thermal imaging data;

[0008] Based on the retrieval-augmented generation technology, construct a transformer fault knowledge base according to transformer fault diagnosis reports and industry standards such as maintenance guidelines; the transformer fault knowledge base includes: a number of vector indexes, and each vector index corresponds to a transformer fault case;

[0009] Input the preprocessed multimodal data into the transformer fault diagnosis model for fault diagnosis to obtain the fault diagnosis results. The transformer fault diagnosis model includes: a defect target detection model, an oil chromatography David triangle model, and an infrared thermal fault detection model. The fault diagnosis results include: defect detection results, status detection results, and overheating fault detection results.

[0010] According to the prompt template and the vector embedding model, convert the fault diagnosis results into prompt vectors, match the prompt vectors with the vector indexes in the transformer fault knowledge base to obtain the transformer fault case with the highest similarity, and extract the similar case information. The prompt template includes: fault type, fault characteristics, and detection methods. The similar case information includes: fault description, diagnosis process, and treatment measures.

[0011] Generate and output a fault diagnosis report based on the fault diagnosis results and the similar case information.

[0012] Optionally, collect and preprocess the multimodal data of the transformer to obtain the preprocessed multimodal data, including:

[0013] Collect the multimodal data of the transformer using a variety of sensors. The variety of sensors includes: cameras, oil chromatography analyzers, and infrared thermal imagers.

[0014] Clean, normalize, and extract features from the collected multimodal data to obtain the preprocessed multimodal data.

[0015] Optionally, based on the retrieval enhancement generation technology, construct a transformer fault knowledge base according to the transformer fault diagnosis report and industry standards such as maintenance guidelines, including:

[0016] Obtain the transformer fault diagnosis report and industry standards such as maintenance guidelines.

[0017] Slice the transformer fault diagnosis report and industry standards such as maintenance guidelines. Each slice represents a transformer fault case.

[0018] Use the vector embedding model to generate vector indexes for each slice to construct the transformer fault knowledge base.

[0019] Optionally, input the preprocessed multimodal data into the transformer fault diagnosis model for fault diagnosis to obtain the fault diagnosis results, including:

[0020] Input the image and video data in the preprocessed multimodal data into the defect target detection model to detect the defects in the appearance and internal structure of the transformer to obtain the defect detection results. The defect detection results include: whether there are crack defects, whether there are corrosion defects, whether there are deformation defects, and the defect locations.

[0021] Input the gas component data in the pre - processed multi - modal data into the oil chromatography David triangle model to detect the operating state of the transformer and obtain the state detection result; the state detection result includes: whether there is an overheating fault and whether there is a discharge fault;

[0022] Input the infrared thermal imaging data in the pre - processed multi - modal data into the infrared thermal fault detection model to detect the thermal abnormal area of the transformer and obtain the overheating fault detection result; the overheating fault detection result includes: whether there is a thermal abnormal area and the location of the thermal anomaly.

[0023] Optionally, convert the fault diagnosis result into a prompt vector according to the prompt template and the vector embedding model, match the prompt vector with the vector index in the transformer fault knowledge base to obtain the transformer fault case with the highest similarity, and extract the similar case information, including:

[0024] Convert the fault diagnosis result into a prompt according to the prompt template;

[0025] Use the vector embedding model to generate a prompt vector according to the prompt;

[0026] Match the prompt vector with the vector index in the transformer fault knowledge base to obtain the transformer fault case with the highest similarity, and extract the similar case information.

[0027] Optionally, generate and output a fault diagnosis report according to the fault diagnosis result and the similar case information, including:

[0028] Generate a fault diagnosis report according to the fault diagnosis result and the similar case information, and output the fault diagnosis report through the user interface or interface.

[0029] Optionally, the defect target detection model is constructed based on a deep learning algorithm; the oil chromatography David triangle model is constructed based on a machine learning algorithm; the infrared thermal fault detection model is constructed based on a deep learning algorithm.

[0030] In a second aspect, the present application provides a transformer multi - modal fault diagnosis system, including:

[0031] A multi - modal data acquisition and pre - processing module, configured to acquire the multi - modal data of the transformer and perform pre - processing to obtain the pre - processed multi - modal data; the multi - modal data includes: image and video data, gas component data, and infrared thermal imaging data;

[0032] The large model fault knowledge base construction module is used to construct a transformer fault knowledge base based on the retrieval augmented generation technology according to the transformer fault diagnosis report and industry standards such as maintenance guidelines; the transformer fault knowledge base includes: a number of vector indexes, and each vector index corresponds to a transformer fault case;

[0033] The dedicated small model fault diagnosis module is used to input the preprocessed multimodal data into the transformer fault diagnosis model for fault diagnosis to obtain a fault diagnosis result; the transformer fault diagnosis model includes: a defect target detection model, an oil chromatography David triangle model, and an infrared thermal fault detection model; the fault diagnosis result includes: a defect detection result, a status detection result, and an overheat fault detection result;

[0034] The large and small model fusion and decision-making module is used to convert the fault diagnosis result into a prompt word vector according to the prompt word template and the vector embedding model, match the prompt word vector with the vector indexes in the transformer fault knowledge base to obtain the transformer fault case with the highest similarity, and extract the similar case information; the prompt word template includes: fault type, fault feature, and detection method; the similar case information includes: fault description, diagnosis process, and treatment measures;

[0035] The fault diagnosis report generation and output module is used to generate and output a fault diagnosis report according to the fault diagnosis result and the similar case information.

[0036] In a third aspect, the present application provides a computer device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, and the processor executes the computer program to implement the above-mentioned transformer multimodal fault diagnosis method.

[0037] In a fourth aspect, the present application provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the above-mentioned transformer multimodal fault diagnosis method is implemented.

[0038] According to the specific embodiments provided by the present application, the following technical effects are disclosed in the present application:

[0039] The present application provides a transformer multimodal fault diagnosis method, system, device and medium. On the one hand, the large model technology, that is, the retrieval augmented generation technology, is used to process complex transformer fault diagnosis reports and industry standard texts such as maintenance guidelines, and a transformer fault knowledge base is extracted; on the other hand, a small model, that is, a transformer fault diagnosis model including a defect target detection model, an oil chromatography David triangle model, and an infrared thermal fault detection model, is used to quickly analyze the real-time collected multimodal data to achieve real-time fault early warning and diagnosis. Through the collaborative work of the large and small models, the present application can improve the response speed and resource utilization efficiency while ensuring the diagnosis accuracy. Description of the Drawings

[0040] To more clearly illustrate the technical solutions in the embodiments of the present application or the related art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0041] Figure 1 It is a schematic flowchart of the transformer multi-modal fault diagnosis method provided by the present application.

[0042] Figure 2 It is a module structure diagram of the transformer multi-modal fault diagnosis system provided by the present application. Detailed Embodiments

[0043] The following will clearly and completely describe the technical solutions in the embodiments of the present application in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, rather than all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present application.

[0044] The present application proposes a transformer multi-modal fault diagnosis method, system, device and medium, which combines the powerful generalization ability of the large model and the fast response characteristics of the small model, and realizes the accurate diagnosis of transformer faults through the fusion analysis of multi-modal data. Specifically, the present application uses the large model to process complex transformer fault diagnosis reports and industry standard texts such as maintenance guides, and extracts the transformer fault knowledge base; at the same time, the small model is used to quickly analyze multi-modal data such as electrical parameters and images collected in real time, and realize real-time fault warning and diagnosis. Through the collaborative work of the large and small models, the present application can improve the response speed and resource utilization efficiency of the system while ensuring the diagnosis accuracy.

[0045] To make the above objects, features and advantages of the present application more obvious and understandable, the following will further describe the present application in detail in conjunction with the drawings and specific embodiments.

[0046] In an exemplary embodiment, the present application provides a transformer multi-modal fault diagnosis method. In the embodiment of the present application, as Figure 1 shown, the method includes the following steps S1 to S5.

[0047] Step S1: Collect multi-modal data of the transformer and perform preprocessing to obtain the preprocessed multi-modal data. Among them, the multi-modal data includes: image and video data, gas composition data, and infrared thermal imaging data. In addition, it can also include electrical parameters (such as voltage, current, power factor), temperature data, vibration data, sound data, etc.

[0048] Specifically, in the data acquisition module, multi-modal data of the transformer is collected through a variety of sensors, and these data are transmitted to the data preprocessing module in real time through the data acquisition module. The variety of sensors includes: cameras, oil chromatographs, and infrared thermal imagers. In the data preprocessing module, operations such as cleaning, normalization, and feature extraction are performed on the collected multi-modal data to eliminate noise and outliers, ensuring the quality and consistency of the data. The preprocessed data will be divided into two parts: one part is used as the input for the large model, and the other part is used as the input for the small model.

[0049] Step S2: Based on the retrieval-augmented generation technology, construct a transformer fault knowledge base according to the transformer fault diagnosis reports and industry standards such as maintenance guidelines. Among them, the transformer fault knowledge base includes: a number of vector indexes, and each vector index corresponds to a transformer fault case.

[0050] Specifically, by combining the Retrieval-Augmented Generation (RAG) technology, a transformer fault knowledge base is constructed and applied to fault diagnosis. First, collect transformer fault diagnosis reports and industry standards such as maintenance guidelines from power companies, equipment manufacturers, and industry standards organizations. These data usually exist in text form, including PDF, Word documents, etc. Second, slice these documents, and each slice represents a transformer fault case. Finally, through a vector embedding model, generate vector indexes for each slice to construct a transformer fault knowledge base for subsequent retrieval and query.

[0051] Step S3: Input the preprocessed multi-modal data into the transformer fault diagnosis model for fault diagnosis to obtain the fault diagnosis result. Among them, the transformer fault diagnosis model includes: multiple dedicated small models such as a defect target detection model, an oil chromatography David triangle model, and an infrared thermal fault detection model. The fault diagnosis result includes: defect detection result, status detection result, and overheating fault detection result.

[0052] Specifically, input the preprocessed multi-modal data into multiple dedicated small models, and utilize the fast response characteristics of the small models to perform real-time fault diagnosis. The small models quickly analyze the input data through deep learning algorithms to identify possible fault patterns and abnormal situations.

[0053] The dedicated small models mainly include a defect target detection model, an oil chromatography David triangle model, and an infrared thermal fault detection model. Each small model focuses on solving a specific type of problem and provides high-precision diagnostic results.

[0054] The defect target detection model collects image and video data of the transformer through a high-resolution camera and sensors, and uses deep learning algorithms (such as YOLO, Faster R-CNN, etc.) to train the preprocessed data to identify and locate defects in the appearance and internal structure of the transformer, such as cracks, corrosion, deformation, etc., and provides high-precision defect detection results. The defect detection results include: whether there is a crack defect, whether there is a corrosion defect, whether there is a deformation defect, and the defect location.

[0055] The oil chromatography David triangle model collects gas component data in the transformer oil through an oil chromatography analyzer, uses machine learning algorithms (such as support vector machines, random forests, etc.) to train the preprocessed data, constructs a David triangle model, and judges the operating state of the transformer based on the gas components in the oil, and judges whether there are faults such as overheating and discharge inside the transformer, and provides high-precision state detection results. The state detection results include: whether there is an overheating fault and whether there is a discharge fault.

[0056] The infrared thermal fault detection model collects infrared thermal imaging data of the transformer through an infrared thermal imager, uses deep learning algorithms (such as CNN, U-Net, etc.) to train the preprocessed data, identifies and locates the thermal anomaly area of the transformer, and provides high-precision overheating fault detection results. The overheating fault detection results include: whether there is a thermal anomaly area and the location of the thermal anomaly.

[0057] Step S4: Convert the fault diagnosis result into a prompt word vector according to the prompt word template and the vector embedding model, match the prompt word vector with the vector index in the transformer fault knowledge base to obtain the transformer fault case with the highest similarity, and extract the similar case information. Among them, the prompt word template includes: fault type, fault feature, and detection method. The similar case information includes: fault description, diagnosis process, and treatment measures.

[0058] First, collect the output results of the three types of dedicated small models, including the diagnostic results of the defect target detection model, the oil chromatography David triangle model, and the infrared thermal fault detection model. These results usually exist in the form of structured data.

[0059] Secondly, construct a fusion prompt template to convert the output results of the small model into standardized prompts. The design of the prompt template should consider the following factors: Fault type: Clearly define the type of fault, such as mechanical fault, electrical fault, thermal fault, etc. Fault characteristics: Describe the specific characteristics of the fault, such as crack location, gas composition, thermal anomaly area, etc. Detection method: Record the detection methods used, such as image recognition, oil chromatography analysis, infrared thermal imaging, etc.

[0060] Finally, generate vector representations of the content in the prompt template through a vector embedding model. Match the generated vectors with the vector indices in the knowledge base, retrieve the case with the highest similarity, and extract the corresponding similar case information, including fault description, diagnosis process, treatment measures, etc.

[0061] Step S5: Generate and output a fault diagnosis report based on the fault diagnosis results and similar case information.

[0062] Specifically, generate a detailed fault diagnosis report based on the final fault diagnosis results, and output the report to the user or expert through the user interface or interface. The report content includes fault type, fault cause, fault location, recommended repair measures, etc.

[0063] In an exemplary embodiment, the present application provides a transformer multi-modal fault diagnosis system. In the embodiment of the present application, as Figure 2 shown, the system includes a multi-modal data acquisition and preprocessing module, a large model fault knowledge base construction module, a dedicated small model fault diagnosis module, a large and small model fusion and decision-making module, and a fault diagnosis report generation and output module. Among them, the multi-modal data acquisition and preprocessing module is connected to the large model fault knowledge base construction module and the dedicated small model fault diagnosis module, and is responsible for providing preprocessed multi-modal data. The large model fault knowledge base construction module is connected to the large and small model fusion and decision-making module, and is responsible for providing the retrieval and query functions of the fault knowledge base. The dedicated small model fault diagnosis module is connected to the large and small model fusion and decision-making module, and is responsible for providing high-precision fault diagnosis results. The large and small model fusion and decision-making module is connected to the fault diagnosis report generation and output module, and is responsible for generating and outputting the final fault diagnosis report. Among them, the large and small model fusion and decision-making module fuses the diagnosis results of three dedicated small models, namely, a defect target detection model, an oil chromatography David triangle model, and an infrared thermal fault detection model, through a customized prompt template.

[0064] Specifically, the multi-modal data acquisition and preprocessing module is used to collect and preprocess the multi-modal data of the transformer to obtain the preprocessed multi-modal data. The multi-modal data includes: image and video data, gas composition data, and infrared thermal imaging data. The large model fault knowledge base construction module is used to construct a transformer fault knowledge base based on the retrieval-augmented generation technology according to the transformer fault diagnosis report and industry standards such as maintenance guidelines. The transformer fault knowledge base includes: a number of vector indexes, and each vector index corresponds to a transformer fault case. The dedicated small model fault diagnosis module is used to input the preprocessed multi-modal data into the transformer fault diagnosis model for fault diagnosis to obtain the fault diagnosis result. The transformer fault diagnosis model includes: a defect target detection model, an oil chromatogram David triangle model, and an infrared thermal fault detection model. The fault diagnosis result includes: a defect detection result, a status detection result, and an overheating fault detection result. The large and small model fusion and decision-making module is used to convert the fault diagnosis result into a prompt word vector according to the prompt word template and the vector embedding model, match the prompt word vector with the vector indexes in the transformer fault knowledge base to obtain the transformer fault case with the highest similarity, and extract the similar case information. The prompt word template includes: fault type, fault characteristics, and detection methods. The similar case information includes: fault description, diagnosis process, and treatment measures. The fault diagnosis report generation and output module is used to generate and output a fault diagnosis report according to the fault diagnosis result and the similar case information.

[0065] The transformer multi-modal fault diagnosis method and system based on the fusion of large and small models provided by this application have the following beneficial effects compared with the existing single-modal transformer fault diagnosis methods and systems: Through the fusion of large and small models and the analysis of multi-modal data, it can more comprehensively and accurately identify the fault types and causes of the transformer, improving the diagnosis accuracy; The fast response characteristic of the small model enables the system to complete real-time fault diagnosis in a short time, meeting the real-time requirements of the power system for fault diagnosis.

[0066] In an exemplary embodiment, this application also provides a computer device, including a memory and a processor. A computer program is stored in the memory, and when the processor executes the computer program, the steps in the above method embodiments are implemented.

[0067] In an exemplary embodiment, this application also provides a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0068] In an exemplary embodiment, this application also provides a computer program product including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0069] In this application, all actions of obtaining signals, information, or data are carried out on the premise of complying with the corresponding data protection regulations and policies of the country where it is located and obtaining authorization from the owner of the corresponding device. It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use, and processing of relevant data need to comply with relevant laws and regulations.

[0070] Those of ordinary skill in the art can understand that all or part of the processes of implementing the methods in the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the above method embodiments. Among them, any reference to a memory, database, or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memories. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0071] The databases involved in the embodiments provided in this application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in this application can be general-purpose processors, central processors, graphics processors, digital signal processors, programmable logic devices, data processing logics based on quantum computing, etc., without limitation.

[0072] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered that the scope described in this specification is covered.

[0073] In this article, specific examples are used to elaborate on the principles and implementation manners of the present application. The description of the above embodiments is only used to help understand the method and its core idea of the present application; at the same time, for those of ordinary skill in the art, according to the idea of the present application, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation on the present application.

Claims

1. A transformer multi-mode fault diagnosis method, characterized in that: include: Collecting multimodal data of the transformer and preprocessing it to obtain preprocessed multimodal data; The multimodal data includes: image video data, gas composition data and infrared thermal imaging data; Based on the retrieval enhancement generation technology, a transformer fault knowledge base is constructed according to the transformer fault diagnosis report and maintenance guide industry standards; the transformer fault knowledge base includes: a number of vector indexes, each vector index corresponds to a transformer fault case; The preprocessed multimodal data is input into a transformer fault diagnosis model for fault diagnosis to obtain a fault diagnosis result; the transformer fault diagnosis model includes: a defect target detection model, an oil chromatography David triangle model and an infrared thermal fault detection model; the fault diagnosis result includes: a defect detection result, a state detection result and an overheating fault detection result; The fault diagnosis result is converted into a prompt word vector according to the prompt word template and the vector embedding model, and the prompt word vector is matched with the vector index in the transformer fault knowledge base to obtain the transformer fault case with the highest similarity, and extract similar case information; the prompt word template includes: fault type, fault characteristics and detection method; the similar case information includes: fault description, diagnosis process and treatment measures; Generate and output a fault diagnosis report based on the fault diagnosis results and similar case information.

2. The transformer multi-mode fault diagnosis method according to claim 1, characterized in that: Collect and preprocess the transformer's multimodal data to obtain preprocessed multimodal data, including: Using multiple sensors to collect multimodal data of the transformer; the multiple sensors include: a camera, an oil chromatograph and an infrared thermal imager; The collected multimodal data are cleaned, normalized and feature extracted to obtain preprocessed multimodal data.

3. The transformer multi-mode fault diagnosis method according to claim 1, characterized in that: Based on the retrieval enhancement generation technology, a transformer fault knowledge base is constructed according to the transformer fault diagnosis report and maintenance guide industry standards, including: Obtain industry standards for transformer fault diagnosis reports and maintenance guidelines; Slice the industry standards for transformer fault diagnosis reports and maintenance guidelines, with each slice representing a transformer fault case; A vector embedding model is used to generate a vector index for each slice and build a transformer fault knowledge base.

4. The transformer multi-mode fault diagnosis method according to claim 1, characterized in that: The preprocessed multimodal data is input into the transformer fault diagnosis model for fault diagnosis, and the fault diagnosis results are obtained, including: The image and video data in the preprocessed multimodal data are input into the defect target detection model to detect the defects of the appearance and internal structure of the transformer to obtain defect detection results; the defect detection results include: whether there is a crack defect, whether there is a corrosion defect, whether there is a deformation defect and the defect location; The gas composition data in the pre-processed multimodal data is input into the oil chromatogram David triangle model to detect the operating state of the transformer and obtain a state detection result; the state detection result includes: whether there is an overheating fault and whether there is a discharge fault; The infrared thermal imaging data in the preprocessed multimodal data is input into the infrared thermal fault detection model, and the thermal abnormality area of ​​the transformer is detected to obtain an overheating fault detection result; the overheating fault detection result includes: whether there is a thermal abnormality area and the thermal abnormality position.

5. The transformer multi-mode fault diagnosis method according to claim 1, characterized in that: The fault diagnosis results are converted into prompt word vectors according to the prompt word template and vector embedding model. The prompt word vectors are matched with the vector indexes in the transformer fault knowledge base to obtain the transformer fault cases with the highest similarity, and similar case information is extracted, including: Convert the fault diagnosis results into prompt words according to the prompt word template; Use the vector embedding model to generate prompt word vectors based on the prompt words; The prompt word vector is matched with the vector index in the transformer fault knowledge base to obtain the transformer fault case with the highest similarity and extract similar case information.

6. The transformer multi-mode fault diagnosis method according to claim 1, characterized in that: Generate and output a fault diagnosis report based on the fault diagnosis results and similar case information, including: A fault diagnosis report is generated based on the fault diagnosis results and similar case information, and the fault diagnosis report is output through a user interface or interface.

7. The transformer multi-mode fault diagnosis method according to claim 1, characterized in that: The defect target detection model is constructed based on a deep learning algorithm; the oil chromatography David triangle model is constructed based on a machine learning algorithm; and the infrared thermal fault detection model is constructed based on a deep learning algorithm.

8. A transformer multi-modal fault diagnosis system, characterized in that: include: A multimodal data acquisition and preprocessing module is used to acquire and preprocess the multimodal data of the transformer to obtain preprocessed multimodal data; The multimodal data includes: image video data, gas composition data and infrared thermal imaging data; A large model fault knowledge base construction module is used to construct a transformer fault knowledge base based on the retrieval enhancement generation technology and according to the transformer fault diagnosis report and maintenance guide industry standards; the transformer fault knowledge base includes: a number of vector indexes, each vector index corresponds to a transformer fault case; A dedicated small model fault diagnosis module is used to input the preprocessed multimodal data into the transformer fault diagnosis model for fault diagnosis to obtain fault diagnosis results; the transformer fault diagnosis model includes: defect target detection model, oil chromatography David triangle model and infrared thermal fault detection model; the fault diagnosis results include: defect detection results, state detection results and overheating fault detection results; The large and small model fusion and decision module is used to convert the fault diagnosis results into prompt word vectors according to the prompt word template and the vector embedding model, match the prompt word vector with the vector index in the transformer fault knowledge base, obtain the transformer fault case with the highest similarity, and extract similar case information; the prompt word template includes: fault type, fault characteristics and detection method; the similar case information includes: fault description, diagnosis process and treatment measures; The fault diagnosis report generation and output module is used to generate and output the fault diagnosis report based on the fault diagnosis results and similar case information.

9. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the transformer multi-modal fault diagnosis method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the transformer multi-modal fault diagnosis method according to any one of claims 1 to 7 is implemented.

Images