A RAG model-based equipment intelligent operation and maintenance system and method

Abstract

The present invention discloses an intelligent equipment operation and maintenance system and method based on a RAG model, and belongs to the technical field of information analysis and processing. The system includes a RAG model management module, a history record module, an intelligent analysis module and an equipment operation and maintenance platform; the RAG model management module is used to establish a RAG model and collect equipment operation and maintenance data; the RAG model is controlled to retrieve and push solutions in a vector database; the history record module is used to collect the retrieval data in the equipment operation and maintenance platform when the equipment is abnormal; the intelligent analysis module is used to analyze the retrieval data in the equipment operation and maintenance platform when the equipment is abnormal, and determine the query vector and corresponding weight value to be added for each alarm information; the equipment operation and maintenance platform is used to monitor the real-time operation data of the equipment, and when an abnormal situation occurs in the equipment, the corresponding alarm information is generated; the generated alarm information and the corresponding solution are digitally displayed.

Description

Technical Field

[0001] The present invention relates to the technical field of information analysis and processing, and specifically to an intelligent equipment operation and maintenance system and method based on a RAG model. Background Art

[0002] The RAG model (Retrieval-Augmented Generation) is a technical framework that combines retrieval and generation capabilities. It aims to enhance the output of the generation model by retrieving relevant information from an external knowledge base. The basic idea is to utilize the generation capabilities of large language models and obtain more accurate and relevant information through retrieval mechanisms. Although traditional large language models have been trained with a large amount of data, the content they generate is still not accurate enough and lacks the latest information. The RAG model introduces external databases to enable the model to refer to more information, thereby improving the quality and accuracy of the generated content.

[0003] The RAG model mainly consists of two core stages: data preparation and retrieval generation. In the data preparation stage, the system extracts and stores data from various data sources, formats the data into smaller data blocks, converts them into vector representations using an encoder, and stores them in a vector database. In the retrieval generation stage, the query text is converted into a query vector, and a similarity search is performed in the vector database to find the data blocks that are most similar to the query vector. These data blocks are input into the generation model together with the query text to generate the final response.

[0004] In the process of intelligent operation and maintenance management of equipment through the RAG model, when an abnormal situation occurs in the equipment, the system can directly push solutions to the relevant abnormal situation; however, for fixed types of equipment abnormal situations, the solutions pushed by the system often depend on the introduction of an external database, and when the external database remains unchanged, the pushed solutions are often the same. The system lacks adaptability and cannot adjust the pushed solutions to meet the needs of managers, thereby reducing the accuracy of retrieval and the efficiency of handling equipment abnormalities. Summary of the invention

[0005] The purpose of the present invention is to provide an equipment intelligent operation and maintenance system and method based on the RAG model to solve the problems raised in the above background technology.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: an intelligent operation and maintenance method for equipment based on a RAG model, the method comprising the following steps:

[0007] S1. Establish a RAG model, collect equipment operation and maintenance data, divide the collected equipment operation and maintenance data into data blocks, and convert the divided data blocks into numerical vectors and store them in a vector database;

[0008] S2. Collecting the search data in the equipment operation and maintenance platform when the equipment is abnormal; the search data includes the text content searched by the operation and maintenance management personnel within a fixed time window when the equipment is abnormal and the number of searches for different text contents;

[0009] S3. Analyze the retrieval data collected in the equipment operation and maintenance platform when the equipment is abnormal, determine the corresponding alarm information when the equipment is abnormal, and determine the query vector that needs to be added for each alarm information and the weight value of the added query vector according to the text content retrieved by the operation and maintenance management personnel within a fixed time window and the number of retrievals of different text contents;

[0010] S4. Monitor the real-time operation data of the equipment. When an equipment abnormality occurs, generate corresponding alarm information according to the equipment abnormality; determine the added query vector and the corresponding weight value according to the generated alarm information, and obtain the retrieved query vector; according to the retrieved query vector, the RAG model retrieves the pushed solution in the vector database, and sends the alarm information and the pushed solution to the equipment operation and maintenance platform;

[0011] S5. The equipment operation and maintenance platform digitally displays the generated alarm information and pushed solutions.

[0012] An intelligent equipment operation and maintenance system based on a RAG model, the system comprising a RAG model management module, a history recording module, an intelligent analysis module and an equipment operation and maintenance platform;

[0013] The RAG model management module is used to establish a RAG model, collect equipment operation and maintenance data, divide the collected equipment operation and maintenance data into data blocks, and convert the divided data blocks into numerical vectors and store them in a vector database; according to the generated alarm information, determine the added query vector and the corresponding weight value to obtain the retrieved query vector; according to the retrieved query vector, the RAG model retrieves the pushed solution in the vector database, and sends the alarm information and the pushed solution to the equipment operation and maintenance platform;

[0014] The history record module is used to collect the search data in the equipment operation and maintenance platform when the equipment is abnormal; the search data includes the text content searched by the operation and maintenance management personnel within a fixed time window when the equipment is abnormal and the number of searches for different text contents;

[0015] The intelligent analysis module is used to analyze the retrieval data collected in the equipment operation and maintenance platform when the equipment is abnormal, determine the corresponding alarm information when the equipment is abnormal, and determine the query vector that needs to be added to each alarm information and the weight value of the added query vector according to the text content retrieved by the operation and maintenance management personnel within a fixed time window and the number of retrievals of different text contents;

[0016] The equipment operation and maintenance platform is used to monitor the real-time operation data of the equipment. When an equipment abnormality occurs, corresponding alarm information is generated according to the equipment abnormality; the generated alarm information and the pushed solution are digitally displayed.

[0017] Compared with the prior art, the beneficial effects achieved by the present invention are: providing a RAG model equipment intelligent operation and maintenance system and method, by analyzing the retrieval data in different equipment operation and maintenance platforms when the equipment is abnormal, the query vector and the corresponding weight value to be added to each alarm information are continuously adjusted, so that the solution pushed by the system is more in line with the needs of the operation and maintenance management personnel, and the adaptability of the system is improved; by calculating the cosine similarity and analyzing the correlation between the query vector and each data block, the accuracy of the solution pushed by the system is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a schematic diagram of the steps of a device intelligent operation and maintenance method based on a RAG model of the present invention;

[0019] Figure 2 It is a structural schematic diagram of an equipment intelligent operation and maintenance system based on a RAG model of the present invention. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] See also Figure 1-Figure 2 , the present invention provides a technical solution:

[0022] See also Figure 1 In the first embodiment, a method for intelligent operation and maintenance of equipment based on a RAG model is provided. By building an intelligent dialogue equipment operation and maintenance platform, intelligent operation and maintenance of substations is created to improve the efficiency of handling abnormal equipment in substations. The method includes the following steps:

[0023] S1. Establish a RAG model, collect equipment operation and maintenance data, divide the collected equipment operation and maintenance data into data blocks, and convert the divided data blocks into numerical vectors and store them in the vector database.

[0024] Furthermore, the equipment operation and maintenance data contains different document information; the collected equipment operation and maintenance data are divided into data blocks, wherein each data block has a corresponding numerical vector; the pushed solution is queried through the query vector, the method is: calculate the cosine similarity between the query vector and the numerical vector corresponding to each data block, compare the calculated cosine similarity with a first threshold, when the calculated cosine similarity is greater than the first threshold, the data blocks are associated, when the calculated cosine similarity is less than the first threshold, the data blocks are not associated; when retrieving the pushed solution in the vector database through the RAG model, the document information in which the number of associated data blocks exceeds the second threshold is used as the pushed solution.

[0025] It should be noted that the query vector is used to determine the solution to be retrieved and pushed; the equipment operation and maintenance data includes but is not limited to equipment instructions, equipment fault solution manual, equipment repair work order disposal results, and equipment operation and maintenance management specifications; wherein each type of equipment operation and maintenance data contains different document information; the collected equipment operation and maintenance data is divided into data blocks, that is, the different document information in each type of equipment operation and maintenance data is divided into data blocks.

[0026] Further, the method for calculating the cosine similarity between the query vector and the corresponding numerical vector of each data block is as follows: Determine the query vector to be retrieved , and determine the numerical vector corresponding to each data block in the vector database ;according to and , calculate the cosine similarity between the query vector and the corresponding numerical vector of each data block, according to the calculation formula:

[0027] ;

[0028] in, Represents the first The numerical vector corresponding to the data blocks; ; Indicates the number of data blocks divided in the vector database; express and The cosine similarity of express and The dot product of Represents the query vector The model, , Represents the query vector The Quantity; Represents a numeric vector The model, , Represents a numeric vector The Quantity; ; Represents the number of components. The number of components in the numeric vector and the query vector is the same.

[0029] In this embodiment, the data block segmentation criteria include but are not limited to a fixed number of characters, sentence boundaries, paragraph boundaries and semantic units; the query vector is determined based on the alarm information and the retrieved text content; the first threshold and the second threshold are set by the operation and maintenance management personnel, the first threshold is used to determine the retrieved data block range, and the second threshold is used to determine the retrieved document information range.

[0030] In this implementation, the process of using document information with a number of associated data blocks exceeding a second threshold as a pushed solution is: determining document information with a number of associated data blocks exceeding a second threshold, using a pre-trained large language model to capture the complex patterns and structures of the language in these document information, and integrating them into a coherent context, and finally obtaining the generated text, which is the pushed solution.

[0031] S2. Collect the search data in the equipment operation and maintenance platform when the equipment is abnormal; the search data includes the text content searched by the operation and maintenance management personnel within a fixed time window when the equipment is abnormal and the number of searches for different text contents.

[0032] It should be noted that the fixed time window is directly determined by the operation and maintenance management personnel to ensure that the collected retrieval data is the retrieval of relevant alarm information by the operation and maintenance management personnel when the equipment is abnormal.

[0033] S3. Analyze the retrieval data collected in the equipment operation and maintenance platform when the equipment is abnormal, determine the corresponding alarm information when the equipment is abnormal, and determine the query vector that needs to be added for each alarm information and the weight value of the added query vector based on the text content retrieved by the operation and maintenance management personnel within a fixed time window and the number of retrievals of different text contents.

[0034] Specifically, the method steps are:

[0035] S31, analyzing the retrieval data collected in the equipment operation and maintenance platform when the equipment is abnormal, and determining the text content that the operation and maintenance management personnel retrieve within a fixed time window under each type of alarm information;

[0036] S32, summarizing the text contents retrieved by the operation and maintenance management personnel within a fixed time window for each type of alarm information, respectively determining different query vectors corresponding to the text contents, and determining the frequency of occurrence of each query vector according to the number of retrievals of different text contents;

[0037] S33: different query vectors corresponding to the text content under each alarm information type are used as query vectors to be added, and the frequency of occurrence of each query vector is used as the weight value of the query vector.

[0038] It should be noted that by analyzing the retrieval data in the equipment operation and maintenance platform when the equipment is abnormal, different query vectors corresponding to the text content are determined. The higher the frequency of the same query vector, the higher the attention of the operation and maintenance management personnel to related issues. At this time, the weight value of the query vector is increased. By continuously adjusting the query vector and the corresponding weight value to be added for each alarm information, the solution pushed by the system can better meet the needs of the operation and maintenance management personnel.

[0039] S4. Monitor the real-time operation data of the equipment. When an equipment abnormality occurs, generate corresponding alarm information according to the equipment abnormality; determine the added query vector and the corresponding weight value based on the generated alarm information to obtain the retrieved query vector; based on the retrieved query vector, the RAG model retrieves the pushed solution in the vector database, and sends the alarm information and the pushed solution to the equipment operation and maintenance platform.

[0040] Specifically, the method steps are:

[0041] S41. Monitor the real-time operation data of the equipment and generate corresponding alarm information when an abnormal situation occurs in the equipment;

[0042] S42: Determine the query vector corresponding to the alarm information according to the generated alarm information , and determine the different query vectors corresponding to the text content retrieved by the operation and maintenance manager within a fixed time window under this alarm information type and the frequency of each query vector ;

[0043] S43, according to and ,RAG model searches in the vector database, takes the frequency of each query vector as the weight value of the query vector, and determines the cosine similarity between the retrieved query vector and the corresponding numerical vector of each data block, according to the calculation formula:

[0044] ;

[0045] in, Represents the query vector for retrieval and The cosine similarity between ; It indicates the amount of data in the collected search data that the operation and maintenance management personnel did not perform text content search within a fixed time window; Indicates the total amount of retrieved data collected; express and The cosine similarity of express and The cosine similarity of ; Indicates the number of query vector types corresponding to the text content retrieved by the operation and maintenance management personnel in the generated alarm information within a fixed time window; Indicates the first column corresponding to the text content retrieved by the operation and maintenance manager within a fixed time window under the generated alarm information type The frequency of occurrence of the query vector; Indicates the first column corresponding to the text content retrieved by the operation and maintenance manager within a fixed time window under the generated alarm information type Query vector;

[0046] S44. According to the cosine similarity between the retrieved query vector and the numerical vector corresponding to each data block, the document information having the number of associated data blocks exceeding the second threshold is used as a pushed solution, and the generated alarm information and the corresponding solution are sent to the equipment operation and maintenance platform.

[0047] It should be noted that, by analyzing the collected search data, the more the query vector corresponding to the alarm information conforms to the pushed solution, the less text content search operations the operation and maintenance management personnel will perform. The total amount of data retrieved The larger the proportion; by calculating the cosine similarity between the query vector corresponding to the alarm information and the added query vector and each data block in the vector database, and assigning weights respectively, the correlation between the query vector and each data block is analyzed, and the document information with the number of associated data blocks exceeding the second threshold is used as the push solution, which improves the accuracy of the system push solution and indirectly improves the efficiency of handling equipment anomalies.

[0048] In this embodiment, the retrieved query vector includes the added query vector and the query vector corresponding to the alarm information; the operating data includes but is not limited to the equipment current, voltage, temperature, vibration amplitude and noise; the equipment abnormality indicates that there is an abnormality in the operating data during the operation of the equipment; by monitoring the real-time operating data of the equipment, when the parameters in the operating data are normal, the equipment is normal at this time; when the parameters in the operating data exceed the set safety threshold, the parameters are abnormal, that is, the equipment is abnormal; at this time, the corresponding alarm information is generated according to the abnormal parameters; for example, when the noise is abnormal, the alarm information is that there is mechanical damage to the equipment, and the query vector corresponding to the mechanical damage of the equipment is determined, and the different query vectors corresponding to the text content retrieved by the operation and maintenance management personnel within a fixed time window when the equipment is mechanically damaged and the frequency of occurrence of each query vector are determined, and the query vector added for the mechanical damage of the equipment and the corresponding weight value are determined. The RAG model searches in the vector database to determine the cosine similarity between the retrieved query vector and the corresponding numerical vector of each data block, and finally obtains the solution to the mechanical damage of the equipment as the relevant equipment fault solution manual and the equipment maintenance work order disposal result.

[0049] S5. The equipment operation and maintenance platform digitally displays the generated alarm information and pushed solutions.

[0050] In this implementation, the operation and maintenance managers can quickly take corresponding preventive measures through the alarm information displayed and solutions pushed by the equipment operation and maintenance platform, reduce the safety risks caused by equipment abnormalities, and improve the safety of the substation.

[0051] Furthermore, when the pushed solution is not sufficient to resolve the current equipment anomaly, the operation and maintenance manager can directly search on the equipment operation and maintenance platform and input the search text content; when the operation and maintenance manager searches through the equipment operation and maintenance platform, the text content is converted into a query vector based on the text content retrieved by the operation and maintenance manager, and the pushed solution is directly searched in the vector database based on the converted query vector.

[0052] See also Figure 2 , in the second embodiment: a device intelligent operation and maintenance system based on a RAG model is provided, the system includes a RAG model management module, a history record module, an intelligent analysis module and a device operation and maintenance platform;

[0053] The RAG model management module is used to establish a RAG model, collect equipment operation and maintenance data, divide the collected equipment operation and maintenance data into data blocks, and convert the divided data blocks into numerical vectors and store them in a vector database; according to the generated alarm information, determine the added query vector and the corresponding weight value to obtain the retrieved query vector; according to the retrieved query vector, the RAG model retrieves the pushed solution in the vector database, and sends the alarm information and the pushed solution to the equipment operation and maintenance platform;

[0054] The history record module is used to collect the search data in the equipment operation and maintenance platform when the equipment is abnormal; the search data includes the text content searched by the operation and maintenance management personnel within a fixed time window when the equipment is abnormal and the number of searches for different text contents;

[0055] The intelligent analysis module is used to analyze the retrieval data collected in the equipment operation and maintenance platform when the equipment is abnormal, determine the corresponding alarm information when the equipment is abnormal, and determine the query vector that needs to be added to each alarm information and the weight value of the added query vector according to the text content retrieved by the operation and maintenance management personnel within a fixed time window and the number of retrievals of different text contents;

[0056] The equipment operation and maintenance platform is used to monitor the real-time operation data of the equipment. When an equipment abnormality occurs, corresponding alarm information is generated according to the equipment abnormality; the generated alarm information and the pushed solution are digitally displayed.

[0057] In this embodiment:

[0058] The RAG model management module establishes the RAG model, collects equipment operation and maintenance data, divides the collected equipment operation and maintenance data into data blocks, and converts the divided data blocks into numerical vectors and stores them in the vector database;

[0059] The history record module collects the retrieval data in the equipment operation and maintenance platform when the equipment is abnormal, and sends the collected retrieval data to the intelligent analysis module;

[0060] The intelligent analysis module analyzes the retrieval data collected in the equipment operation and maintenance platform when the equipment is abnormal, determines the corresponding alarm information when the equipment is abnormal, and determines the query vector and weight value of each alarm information according to the text content retrieved by the operation and maintenance management personnel within a fixed time window and the number of retrievals of different text contents; sends the query vector and weight value of each alarm information to the RAG model management module;

[0061] The equipment operation and maintenance platform monitors the real-time operation data of the equipment. When an abnormal situation occurs, it generates corresponding alarm information according to the abnormal situation and sends the generated alarm information to the RAG model management module.

[0062] The RAG model management module determines the added query vector and the corresponding weight value according to the generated alarm information, and obtains the retrieved query vector; according to the retrieved query vector, the RAG model retrieves the pushed solution in the vector database, and sends the alarm information and the pushed solution to the equipment operation and maintenance platform;

[0063] The equipment operation and maintenance platform digitally displays the generated alarm information and pushed solutions.

[0064] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A device intelligent operation and maintenance method based on a RAG model, characterized in that: The method comprises the following steps: S1. Establish a RAG model, collect equipment operation and maintenance data, divide the collected equipment operation and maintenance data into data blocks, and convert the divided data blocks into numerical vectors and store them in a vector database; The equipment operation and maintenance data includes different document information; the collected equipment operation and maintenance data is divided into data blocks, wherein each data block has a corresponding numerical vector; the pushed solution is queried through the query vector, and the method is: the cosine similarity between the query vector and the numerical vector corresponding to each data block is calculated, and the calculated cosine similarity is compared with a first threshold value, when the calculated cosine similarity is greater than the first threshold value, the data blocks are associated, and when the calculated cosine similarity is less than the first threshold value, the data blocks are not associated; when the pushed solution is retrieved in the vector database through the RAG model, the document information with the number of associated data blocks exceeding the second threshold value is used as the pushed solution; the first threshold value is used to determine the range of the retrieved data blocks, and the second threshold value is used to determine the range of the retrieved document information; S2. Collecting the search data in the equipment operation and maintenance platform when the equipment is abnormal; the search data includes the text content searched by the operation and maintenance management personnel within a fixed time window when the equipment is abnormal and the number of searches for different text contents; S3. Analyze the retrieval data collected in the equipment operation and maintenance platform when the equipment is abnormal, determine the corresponding alarm information when the equipment is abnormal, and determine the query vector that needs to be added for each alarm information and the weight value of the added query vector according to the text content retrieved by the operation and maintenance management personnel within a fixed time window and the number of retrievals of different text contents; S4. Monitor the real-time operation data of the equipment. When an equipment abnormality occurs, generate corresponding alarm information according to the equipment abnormality; determine the added query vector and the corresponding weight value according to the generated alarm information, and obtain the retrieved query vector; according to the retrieved query vector, the RAG model retrieves the pushed solution in the vector database, and sends the alarm information and the pushed solution to the equipment operation and maintenance platform; Specifically, the method steps of step S4 are: S41. Monitor the real-time operation data of the equipment and generate corresponding alarm information when an abnormal situation occurs in the equipment; S42. Determine the query vector Q corresponding to the alarm information according to the generated alarm information. 0 , and determine the different query vectors corresponding to the text content retrieved by the operation and maintenance manager within a fixed time window under this alarm information type and the frequency of each query vector S43, according to and The RAG model searches the vector database, takes the frequency of each query vector as the weight value of the query vector, and determines the cosine similarity between the retrieved query vector and the corresponding numerical vector of each data block, according to the calculation formula: in, Represents the query vector and D i The cosine similarity between them; M represents the amount of data in the collected retrieval data that the operation and maintenance management personnel did not perform text content retrieval within a fixed time window; H represents the total amount of collected retrieval data; cosine(Q 0 ,D i ) indicates Q 0 With D i The cosine similarity of express With D i cosine similarity; g = 1, 2, ..., s; s represents the number of query vector types corresponding to the text content retrieved by the operation and maintenance management personnel in the generated alarm information within a fixed time window; Indicates the frequency of occurrence of the g-th query vector corresponding to the text content retrieved by the operation and maintenance manager within a fixed time window under the generated alarm information type; represents the g-th query vector corresponding to the text content retrieved by the operation and maintenance manager within a fixed time window under the generated alarm information type; D i Represents the numerical vector corresponding to the i-th data block in the vector database; S44, according to the cosine similarity between the retrieved query vector and the numerical vector corresponding to each data block, the document information having the number of associated data blocks exceeding the second threshold is used as a pushed solution, and the generated alarm information and the corresponding solution are sent to the device operation and maintenance platform; S5. The equipment operation and maintenance platform digitally displays the generated alarm information and pushed solutions.

2. According to claim 1, a method for intelligent operation and maintenance of equipment based on the RAG model is characterized in that: The method for calculating the cosine similarity between the query vector and the numerical vector corresponding to each data block is as follows: determine the query vector Q to be retrieved, and determine the numerical vectors D1, D2, ..., D corresponding to each data block in the vector database. n ; According to Q and D1, D2, ..., D n , calculate the cosine similarity between the query vector and the corresponding numerical vector of each data block, according to the calculation formula: Among them, D i represents the numerical vector corresponding to the i-th data block in the vector database; i = 1, 2, ..., n; n represents the number of data blocks divided in the vector database; cosine(Q,D i ) represents Q and D i Cosine similarity of Q·D i Indicates Q and D i The dot product of ||Q|| represents the modulus of the query vector Q. Q j represents the jth component in the query vector Q; ||D i || represents a numerical vector D i The model, Represents a numerical vector D i The j-th component in ; j = 1, 2, ..., m; m represents the number of components.

3. According to claim 2, a method for intelligent operation and maintenance of equipment based on the RAG model is characterized in that: The method steps of step S3 are: S31, analyzing the retrieval data collected in the equipment operation and maintenance platform when the equipment is abnormal, and determining the text content that the operation and maintenance management personnel retrieve within a fixed time window under each type of alarm information; S32, summarizing the text contents retrieved by the operation and maintenance management personnel within a fixed time window for each type of alarm information, respectively determining different query vectors corresponding to the text contents, and determining the frequency of occurrence of each query vector according to the number of retrievals of different text contents; S33: different query vectors corresponding to the text content under each alarm information type are used as query vectors to be added, and the frequency of occurrence of each query vector is used as the weight value of the query vector.

4. According to claim 3, a method for intelligent operation and maintenance of equipment based on the RAG model is characterized in that: When the operation and maintenance manager searches through the equipment operation and maintenance platform, the text content is converted into a query vector based on the text content retrieved by the operation and maintenance manager, and the pushed solution is directly retrieved in the vector database based on the converted query vector.

5. An intelligent operation and maintenance system for equipment based on a RAG model, applying an intelligent operation and maintenance method for equipment based on a RAG model as claimed in any one of claims 1 to 4, characterized in that: The system includes RAG model management module, history record module, intelligent analysis module and equipment operation and maintenance platform; The RAG model management module is used to establish a RAG model, collect equipment operation and maintenance data, divide the collected equipment operation and maintenance data into data blocks, and convert the divided data blocks into numerical vectors and store them in a vector database; according to the generated alarm information, determine the added query vector and the corresponding weight value to obtain the retrieved query vector; according to the retrieved query vector, the RAG model retrieves the pushed solution in the vector database, and sends the alarm information and the pushed solution to the equipment operation and maintenance platform; The history record module is used to collect the search data in the equipment operation and maintenance platform when the equipment is abnormal; the search data includes the text content searched by the operation and maintenance management personnel within a fixed time window when the equipment is abnormal and the number of searches for different text contents; The intelligent analysis module is used to analyze the retrieval data collected in the equipment operation and maintenance platform when the equipment is abnormal, determine the corresponding alarm information when the equipment is abnormal, and determine the query vector that needs to be added to each alarm information and the weight value of the added query vector according to the text content retrieved by the operation and maintenance management personnel within a fixed time window and the number of retrievals of different text contents; The equipment operation and maintenance platform is used to monitor the real-time operation data of the equipment. When an equipment abnormality occurs, corresponding alarm information is generated according to the equipment abnormality; the generated alarm information and the pushed solution are digitally displayed.

Images