Hydroelectric generating set fault case quick retrieval method, equipment and medium

Abstract

The invention discloses a hydroelectric generating set fault case rapid retrieval method and device and a medium, and the retrieval method comprises the following steps: 1, obtaining multi-modal data, 2, extracting data features, 3, obtaining the measurement information, image information and text information of an actual current set, calculating the feature vector of the current set through a representation model, and calculating the fault case of the current set according to the feature vector. According to the retrieval method, a vector library and feature embedding are adopted to represent fault case information, a large amount of redundancy is avoided, the retrieval method is very compact, the retrieval speed is increased, the capacity of concurrent processing of mass data is achieved, and the retrieval efficiency is improved. The method has a wider application range, and solves the problems of low retrieval efficiency, high data redundancy, insufficient matching precision and weak concurrent processing capability in the prior art.

Description

Technical Field

[0001] The present invention relates to the technical field of fault case retrieval for hydroelectric generating units, and particularly to a method, device, and medium for quickly retrieving fault cases of hydroelectric generating units. Background Art

[0002] With the continuous construction of UHV projects, hydroelectric generating units are increasingly shifting from traditional power generation functions to peak shaving and frequency modulation functions. Their regulation frequency and depth have both increased significantly compared to the past. During operation, they are more affected by various factors such as erosion, wear, and fatigue. The performance of each component will gradually degrade with the increase in service time, leading to the occurrence of serious accidents.

[0003] Currently, the main reasons for faults in hydroelectric generating units mainly include the following aspects:

[0004] Mechanical faults: Axis bending deformation: Poor concentricity of the unit's main shaft, loose connection of the main shaft joint, improper adjustment of the bearing, and excessive clearance can cause vibration after startup; Impeller wear: The impeller of the water turbine may be worn after long-term erosion by water hammer force, affecting the hydrodynamic performance of the unit; Main shaft wear: The impeller and the main shaft adopt an interference fit. Long-term operation will cause metal fatigue, key wear, or key rolling phenomenon, resulting in radial movement between the inner hole of the impeller and the main shaft.

[0005] Electrical faults: Two-point grounding of the rotor circuit: It will cause abnormal increase in rotor current, which may damage the equipment in the excitation circuit and cause severe vibration of the unit; Overheating of the generator: Caused by the active load or stator current exceeding the rated range, malfunction of the temperature measurement device, problems in the cooling system, etc.; Fault of the excitation system: Such as loss of excitation fault, which may be caused by loose contact of the AC side switch contact of the excitation power supply, resulting in a large contact resistance and thus the loss of excitation fault.

[0006] Hydraulic balance problems: Vibration and noise of the unit: Caused by reasons such as excessive clearance of the rotating components of the unit, obvious swing, and unbalanced weight.

[0007] Other faults: Aging of winding insulation: The generator winding is prone to oxidation and aging during long-term operation in a water environment, resulting in short circuits; Cable damage and pole collapse: May be caused by aging or external damage, affecting power transmission; Embankment damage and sluice gate failure: May cause floods to pour into the power station, resulting in equipment short circuits or damage.

[0008] Predictive maintenance of hydropower stations through big data is the current mainstream development trend. The computer monitoring systems, online monitoring systems, etc. currently deployed in hydropower stations generate a large amount of heterogeneous data every day.

[0009] The hydroelectric generating unit is the core equipment of the hydropower generation system, and its operating status directly affects the power generation efficiency and safety. Due to the complex structure, variable operating environment, and diverse fault types of the hydroelectric generating unit, fault diagnosis and handling rely on a large number of historical cases. However, the traditional fault case retrieval method has the following problems:

[0010] Low retrieval efficiency: Traditional methods are based on keyword or simple rule matching, making it difficult to handle multi-modal data (such as text, measurement information, images, etc.), and the retrieval speed is slow.

[0011] High data redundancy: The historical case data is large in volume and redundant, resulting in high storage and retrieval costs.

[0012] Insufficient matching accuracy: Traditional methods are difficult to accurately extract fault features, resulting in poor relevance of retrieval results.

[0013] Weak concurrent processing ability: Unable to efficiently handle the concurrent retrieval requirements of massive data.

[0014] Therefore, there is an urgent need for a fast retrieval method, device, and medium for hydroelectric generating unit fault cases that are efficient, accurate, and have the ability to process massive data. Summary of the Invention

[0015] To overcome the above problems, the object of the present invention is to provide a fast retrieval method, device, and medium for hydroelectric generating unit fault cases. The retrieval method realizes the fast retrieval of hydroelectric generating unit fault cases, uses a vector library and feature embedding to represent fault case information, which is neither highly redundant nor very compact, speeds up the retrieval speed, has the ability to concurrently process massive data, has a wider application range, and solves the problems of low retrieval efficiency, high data redundancy, insufficient matching accuracy, and weak concurrent processing ability in the prior art.

[0016] The technical solution adopted by the present invention is as follows:

[0017] A fast retrieval method for hydroelectric generating unit fault cases includes the following steps:

[0018] Step 1: Acquisition of multi-modal data, which is obtained relying on the data interface or Kafka of the unified data platform of the hydroelectric generating unit to obtain measurement information, image information, and text information.

[0019] The measurement information is the information of various parts of the historical hydroelectric generating unit, the image information is the infrared thermal image information of key parts, and the text information is the inspection description information;

[0020] Step 2: Data feature extraction, using a large model based on the Transformer structure to extract features from the measurement information, image information, and text information in Step 1. All features are identified using a one-dimensional vector array, and the features are stored in the vector library. The specific data feature extraction method is:

[0021] The measured information extracts the trend change characteristics of the measurement with a window of 256 time points.

[0022] The image information is sliced into 5 layers of 32*32 slices to calculate features.

[0023] The text information extracts the summary description features of the text with a sentence as the window.

[0024] Step 3: Obtain the measured information, image information, and text information of the actual current unit, and calculate the feature vector of the current unit through the characterization model.

[0025] Step 4: Calculate the similarity probability between the feature vector of the current unit and the historical fault vector.

[0026] Step 5: Fault case result retrieval. According to the calculation result of the similarity probability in Step 4, compare the similarity probability value with the threshold to obtain the fault case result.

[0027] Step 6: Fault handling. According to the fault case result in Step 5, query the corresponding conclusion and fault handling decision suggestions.

[0028] As a further description of the present invention, the measured information of each part of the hydro-generator unit in Step 1 includes vibration, swing, temperature, oil pressure, oil level, current, and voltage information, and the key parts include the upper guide bearing position, the lower guide bearing position, and the water guide bearing position.

[0029] As a further description of the present invention, the calculation formula for the feature vector in Step 3 is:

[0030] ,

[0031] Where

[0032] is the vector of measurement data,

[0033] is the vector of image data,

[0034] is the vector of text data,

[0035] represents concatenating the measurement, image, and text data,

[0036] S is the number of vectors after characterizing this type of fault,

[0037] i is the number of faults,

[0038] ClassPrototype is the prototype result obtained by calculating the current fault.

[0039] As a further description of the present invention, the calculation formula for the similarity probability in step 4 is as follows:

[0040] ,

[0041] where

[0042] represents the Euclidean distance,

[0043] is the fault vector for comparative retrieval,

[0044] is the current vector,

[0045] k is the number of fault vectors to be retrieved,

[0046] p is the calculated similarity probability, and it is used to determine which fault the current vector is closest to through the similarity probability.

[0047] As a further description of the present invention, the measurement information in step 2 is extracted using the open-source Moriai model, the image information is extracted using the DINOV2 model, and the text information is extracted using the BGE model.

[0048] As a further description of the present invention, the vector library in step 2 uses the Milvus vector library.

[0049] As a further description of the present invention, the threshold in step 5 is 0.85.

[0050] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the fast retrieval method for hydro-generator unit fault cases described above.

[0051] A computer-readable storage medium stores a computer program. When the computer program is executed by a processor, it implements the fast retrieval method for hydro-generator unit fault cases described above.

[0052] Advantages of the present invention:

[0053] A fast retrieval method for hydropower unit fault cases. The retrieval method extracts spatial feature vectors from measurement information, image information, text information, etc. of the hydropower unit to obtain a feature sequence, extracts key feature vectors from the feature vectors, establishes a vector index library for the key feature vectors, extracts the key feature vectors of various features of the current unit, performs similarity retrieval in the distributed vector index library according to the key feature vectors, and outputs the retrieval results with a similarity greater than the system preset threshold. The retrieval method with this structure realizes the fast retrieval of hydropower unit fault cases, which is neither highly redundant nor very compact, speeds up the retrieval speed, has the ability to process a large amount of concurrent data, has a wider application range, and solves the problems of low retrieval efficiency, high data redundancy, insufficient matching accuracy, and weak concurrent processing ability in the prior art.

[0054] A fast retrieval method for hydropower unit fault cases. The measurement information in step 2 is extracted using the open-source Moriai model, the image information is extracted using the DINOV2 model, and the text information is extracted using the BGE model. For different information, a feature extraction model suitable for this information is used for feature extraction to ensure the accuracy of feature extraction and achieve precise and efficient retrieval. Brief Description of the Drawings

[0055] Figure 1 is a flowchart of a fast retrieval method for hydropower unit fault cases proposed by the present invention;

[0056] Figure 2 is a diagram of the interactive data analysis module of the hydropower unit for the fast retrieval method of hydropower unit fault cases proposed by the present invention. Detailed Embodiments

[0057] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following detailed description of the specific embodiments of the present invention is made in conjunction with the drawings in the specification. Obviously, the described embodiments are part of the embodiments of the present invention, not all of them. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0058] Many specific details are set forth in the following description to facilitate a thorough understanding of the present invention, but the present invention may be implemented in other ways different from those described herein. Those skilled in the art can make similar generalizations without departing from the spirit of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0059] Second, the "one embodiment" or "embodiment" referred to herein means a specific feature, structure, or characteristic that may be included in at least one implementation manner of the present invention. The appearances of "in one embodiment" in different places in this specification do not all refer to the same embodiment, nor are they separate or alternative embodiments that are mutually exclusive with other embodiments.

[0060] The present invention will be described in detail in conjunction with schematic diagrams. When describing the embodiments of the present invention in detail, for the convenience of explanation, the cross-sectional views showing the device structure will be enlarged locally in a non-general proportion, and the schematic diagrams are only examples and should not limit the scope of protection of the present invention here. In addition, in actual production, three-dimensional spatial dimensions including length, width, and depth should be included.

[0061] At the same time, in the description of the present invention, it should be noted that the orientation or positional relationship indicated by terms such as "upper, lower, inner, and outer" is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation to the present invention. In addition, the terms "first, second, or third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0062] Unless otherwise clearly defined and limited in the present invention, the terms "installed, connected, and coupled" should be understood in a broad sense. For example: it can be a fixed connection, a detachable connection, or an integral connection; it can also be a mechanical connection, an electrical connection, or a direct connection, and can also be indirectly connected through an intermediate medium, or can be the communication inside two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations.

[0063] As Figures 1-2 shown, it shows the specific implementation manner of the present invention:

[0064] A method for quickly retrieving fault cases of a hydropower unit, comprising the following steps:

[0065] Step 1: Multimodal data acquisition, which is acquired relying on the data interface of the unified data platform of the hydropower unit or Kafka, and the measurement information, image information, and text information are acquired.

[0066] The measurement information is the information of each part of the historical hydropower unit, the image information is the infrared thermal image information of the key parts, and the text information is the inspection description information;

[0067] Step 2: Data feature extraction. The measurement information, image information, and text information in Step 1 are subjected to feature extraction using a large model based on the Transformer structure. All features are identified using a one-dimensional vector array, and the features are stored in a vector library. The specific data feature extraction method is as follows:

[0068] For the measurement information, taking 256 time points as a window, extract the trend change features of the measurement.

[0069] For the image information, it is sliced into 5 layers of 32*32 slices, and features are calculated.

[0070] For the text information, taking one sentence as a window, extract the summary description features of the text.

[0071] Step 3: Obtain the measurement information, image information, and text information of the actual current unit, and calculate the feature vector of the current unit through a characterization model.

[0072] Step 4: Calculate the similarity probability between the feature vector of the current unit and the historical fault vector.

[0073] Step 5: Fault case result retrieval. According to the similarity probability calculation result in Step 4, compare this similarity probability value with a threshold to obtain the fault case result.

[0074] Step 6: Fault handling. According to the fault case result in Step 5, query the corresponding conclusion and fault handling decision suggestions.

[0075] In this embodiment, as Figure 1 shown, the retrieval method extracts spatial feature vectors from the measurement information, image information, text information, etc. of the hydropower unit to obtain a feature sequence, extracts key feature vectors from the feature vectors, establishes a vector index library for the key feature vectors, extracts the key feature vectors of various features of the current unit, performs similarity retrieval in the distributed vector index library according to the key feature vectors, and outputs the retrieval results with a similarity greater than the system preset threshold. Using the retrieval method with this structure to achieve fast retrieval of hydropower unit fault cases, it is neither highly redundant nor very compact, speeds up the retrieval speed, has the ability to handle a large amount of concurrent data, has a wider application range, and solves the problems of low retrieval efficiency, high data redundancy, insufficient matching accuracy, and weak concurrent processing ability in the prior art.

[0076] In this embodiment, as Figure 2 shown, the feature vectors generated from historical fault cases are stored in the vector library, the feature vectors of the current monitoring information are calculated, and through the retrieval of the similarity retrieval module, the fault finding of the current monitoring information is realized. The similarity retrieval module is the similarity probability calculation in Step 4 of this fast retrieval method.

[0077] In this embodiment, the main purpose of finding fault cases of the hydropower unit is to summarize experience and lessons, improve the equipment safety operation management ability, and strengthen the sense of responsibility. By collecting, sorting, and analyzing various fault cases that occur in the hydropower unit, the causes of faults and preventive measures can be summarized, helping the operation and maintenance personnel better understand the occurrence mechanism of faults, so as to avoid the occurrence of similar problems in future work; improving the equipment safety operation management ability by establishing a fault case library can help relevant personnel intuitively understand the phenomena, causes, and preventive measures of accidents and faults, enhance the prevention and handling ability of the same type of accidents and faults, and further consolidate the work of safe production; strengthening the sense of responsibility through the study and self-examination of fault cases can strengthen the sense of responsibility of the operation and inspection personnel for the safe and stable operation of the equipment, improve the ability to handle equipment safety accidents, and change from passive prevention to active control; the construction of the fault case library can enable the leadership organization to understand the safe operation status of the equipment in a timely and accurate manner, discover problems and make correct decisions, which is conducive to the development of supervision and management work; through the analysis and management of fault cases, problems in equipment design and maintenance can be found, the equipment structure can be optimized, the maintenance strategy can be improved, and the overall performance and reliability of the equipment can be enhanced. The ultimate goal of the rapid retrieval method for fault cases of the hydropower unit is to improve the safety and reliability of the equipment. Through this retrieval method, accurate fault finding can be quickly achieved, so that the hydropower unit can operate normally and ensure its operation stability.

[0078] Embodiment 2

[0079] Specifically, the measurement information of each part of the hydropower unit in step 1 includes vibration, swing, temperature, oil pressure, oil level, current, and voltage information, and the key parts include the upper guide bearing position, the lower guide bearing position, and the water guide bearing bush position.

[0080] In this embodiment, the specific types of information of the hydropower unit are described, so that the monitoring factors in the calculation process of the feature vector in this retrieval method are consistent, and accurate retrieval results can be obtained. At the same time, the key parts of the hydropower unit are described to achieve precise monitoring.

[0081] Embodiment 3

[0082] Specifically, the calculation formula of the feature vector in step 3 is:

[0083] ,

[0084] Among them,

[0085] is the vector of measurement data,

[0086] is the vector of image data,

[0087] is a vector of text data,

[0088] indicating the concatenation of measurement, image, and text data,

[0089] S is the number of vectors after characterizing this type of fault,

[0090] i is the number of faults,

[0091] ClassPrototype is the prototype result obtained from the current fault calculation.

[0092] In this embodiment, the feature vector of the current monitoring information is calculated using the above formula for the feature vector.

[0093] Specifically, the formula for calculating the similarity probability in step 4 is:

[0094] ,

[0095] where,

[0096] represents the Euclidean distance,

[0097] is the fault vector for comparative retrieval,

[0098] is the current vector,

[0099] k is the number of fault vectors to be retrieved,

[0100] p is the calculated similarity probability, and it is used to determine which fault the current vector is closest to through the similarity probability.

[0101] In this embodiment, the probability calculation between the current vector and the historical fault vectors is performed using the above formula for the similarity probability, and it is determined which fault the current vector is closest to.

[0102] Embodiment 4

[0103] Specifically, the measurement information in step 2 is extracted using the open-source Moriai model, the image information is extracted using the DINOV2 model, and the text information is extracted using the BGE model.

[0104] In this embodiment, for different information, a feature extraction model suitable for that information is used for feature extraction to ensure the accuracy of feature extraction and achieve precise and efficient retrieval.

[0105] In this embodiment, the measurement information slices the waveforms of the upper guide, lower guide, and water guide through the open-source Moriai model, sets the slice information according to the unit rotation frequency information by rounding the rotation frequency * 32, and brings it into the Moriai model for calculation to obtain the representation; the image information is calculated through the DINOV2 model, selects the Large version of the DINOV2 model, and calculates to obtain the infrared thermal image representation information of the upper guide, lower guide, and water guide bearing bushes; the text information is obtained by converting the training description information and is extracted through the BGE system, and the BGE-M3 model is selected to extract features.

[0106] Specifically, the vector library in step 2 uses the Milvus vector library.

[0107] In this embodiment, using the Milvus vector library can achieve reliable and effective storage, and it has the following advantages:

[0108] High performance: Milvus has excellent vector search performance and can achieve fast approximate nearest neighbor search (ANN) in large-scale vector datasets. It adopts an optimized index structure and search algorithm, and can complete query operations within milliseconds, maintaining high efficiency even in the face of massive data.

[0109] Scalability: Milvus supports horizontal expansion and can easily handle the growing data volume and query load. By adding more nodes, the cluster scale can be expanded to achieve distributed storage and processing of data, ensuring the stability and performance of the system in large-scale application scenarios.

[0110] Multimodal support: Milvus can not only process vector representations of multimedia data such as images, audio, and video, but also process vectors of other types of data such as text, and has wide applicability in multimodal applications, meeting the diverse needs of different fields.

[0111] Ease of use: Milvus provides a simple and easy-to-use API, supports multiple programming languages such as Python, Java, Go, etc., and is convenient for developers to integrate and use. At the same time, it also provides a visual management interface, which is convenient for users to perform operations such as data management, index creation, and query debugging.

[0112] Data security: Milvus supports security mechanisms such as data encryption and access control to ensure the security and privacy of data. It can encrypt the vector data stored in the database to prevent data leakage; at the same time, through access control policies, it restricts the access rights of different users to data.

[0113] Seamless Integration: Milvus is compatible with deep learning pipelines and supports integration with tools such as PyTorch, TensorFlow, and OpenAI models.

[0114] Open Source Community Support: Milvus is a completely open-source project, and users can freely host and customize the deployment. The support of the open-source community enables Milvus to be continuously iterated and optimized to meet the needs of different users.

[0115] Specifically, the threshold in step 5 is 0.85.

[0116] In this embodiment, the threshold is set to 0.85. When the calculated similarity probability is greater than 0.85, it indicates that the current monitoring information has a high similarity with the historical fault case. At this time, the fault handling method of the monitoring result can be implemented with reference to the conclusions and suggestions of the historical fault case. Setting a threshold of 0.85 can ensure that the retrieval results provided by this retrieval method are within the actual fault range. If the threshold is too small, the number of retrieval results is large and it is impossible to quickly obtain accurate results. If the threshold is too large, the retrieval result range is too small, and it may not be possible to directly obtain the final accurate result from all the retrieval results, affecting the efficiency of fault handling. Therefore, setting a threshold of 0.85 can achieve fast and efficient retrieval on the basis of ensuring the accuracy of the retrieval results.

[0117] Embodiment Five

[0118] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the fast retrieval method for water turbine generator set fault cases described in Embodiment 1.

[0119] Embodiment Six

[0120] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements the fast retrieval method for water turbine generator set fault cases described in Embodiment 1.

[0121] Those skilled in the art should understand that Embodiments Five and Six of the present invention can be provided as methods and computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0122] The present invention is described with reference to the flowcharts and / or block diagrams of methods, computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram, and the combination of flows and / or blocks in the flowchart and / or block diagram can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or multiple blocks.

[0123] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including instruction means that implement the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or multiple blocks.

[0124] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, so that the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or multiple blocks.

[0125] The preferred embodiments of the present invention have been described in detail above in conjunction with the accompanying drawings. However, the present invention is not limited to the above embodiments. Within the knowledge scope of those of ordinary skill in the art, various changes can be made without departing from the gist of the present invention.

[0126] Many other changes and modifications can be made without departing from the concept and scope of the present invention. It should be understood that the present invention is not limited to specific embodiments, and the scope of the present invention is defined by the appended claims.

Claims

1. A rapid retrieval method for fault cases of a hydropower unit, characterized in that, It includes the following steps: Step 1: Multimodal data acquisition, which is obtained relying on the data interface of the unified data platform of the hydropower unit or Kafka, and measurement information, image information, and text information are obtained. The measurement information is the information of each part of the historical hydropower unit, the image information is the infrared thermal image information of the key parts, and the text information is the inspection description information. Step 2: Data feature extraction. The measurement information, image information, and text information in Step 1 are subjected to feature extraction using a large model based on the Transformer structure. All features are identified using a one-dimensional vector array, and the features are stored in the vector library. The specific data feature extraction method is as follows: For the measurement information, with 256 time points as the window, the trend change features of the measurement are extracted. For the image information, it is sliced into 5 layers of 32*32 slices to calculate features. For the text information, with one sentence as the window, the summary description features of the text are extracted. Step 3: Obtain the measurement information, image information, and text information of the actual current unit, and calculate the feature vector of the current unit through the characterization model. Step 4: Calculate the similarity probability between the current unit feature vector and the historical fault vector. Step 5: Fault case result retrieval. According to the similarity probability calculation result in Step 4, compare the similarity probability value with the threshold to obtain the fault case result. Step 6: Fault handling. According to the fault case result in Step 5, query the corresponding conclusion and fault handling decision suggestions.

2. A rapid retrieval method for hydropower unit fault cases according to claim 1, characterized in that The measurement information of each part of the hydropower unit in Step 1 includes vibration, swing, temperature, oil pressure, oil level, current, and voltage information, and the key parts include the upper guide bearing position, lower guide bearing position, and water guide bearing bush position.

3. A rapid retrieval method for hydropower unit fault cases according to claim 1, characterized in that The calculation formula for the feature vector in Step 3 is: ， Where, is a vector of measurement data, is a vector of image data, is a vector of text data, Indicates the connection of measurement, image, and text data, S is the number of vectors after the characterization of this type of fault, i is the number of faults, ClassPrototype is the prototype result obtained by calculating the current fault.

4. A fast retrieval method for a hydropower unit fault case according to claim 1, characterized in that The calculation formula for the similarity probability in Step 4 is: ， Where, represents the Euclidean distance, The fault vector for comparative retrieval, is the current vector, k is the number of fault vectors to be retrieved, p is the calculated similarity probability, and it is used to determine which fault the current vector is closest to through the similarity probability.

5. A method for quickly retrieving fault cases of a hydropower unit according to claim 1, characterized in that, The measurement information in Step 2 is extracted using the open-source Moriai model, the image information is extracted using the DINOV2 model, and the text information is extracted using the BGE model.

6. A rapid retrieval method for hydropower unit fault cases according to claim 1, characterized in that, The vector library in Step 2 uses the Milvus vector library.

7. A rapid retrieval method for hydropower unit fault cases according to claim 1, characterized in that The threshold in Step 5 is 0.

85.

8. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method for quickly retrieving hydropower unit fault cases according to any one of claims 1 to 7.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the method for quickly retrieving hydropower unit fault cases according to any one of claims 1 to 7.

Images