A method for intelligent generation of operation and maintenance scripts based on large models

Abstract

The present invention relates to a method for intelligently generating operation and maintenance scripts based on a large model, and relates to the field of network operation and maintenance technology. The method comprises the following steps: constructing an asset database and establishing an operation and maintenance network topology diagram; performing word segmentation, extraction, and matching based on operation and maintenance requirements actively input by manual operation and maintenance or recommended by automatic operation and maintenance analysis; constructing an indicator-parameter form; and selecting a script module based on the operating environment of the target asset, filling it in according to the indicator-parameter form, and deploying the script to the target asset. The method performs word segmentation, extraction, and matching based on the operation and maintenance requirements of a historical operation and maintenance library, and performs cluster analysis to construct a matching library with representative indicator names and representative parameter formats. This method reduces the need for refined segmentation of operation and maintenance requirements or batch operations, improves operation and maintenance efficiency, and reduces the difficulty and error rate of operation and maintenance. In addition, based on whether the same operation and maintenance requirements exist in the historical operation and maintenance library, the method generates precise operation and maintenance requirements, reference operation and maintenance requirements, and associated operation and maintenance requirements, thereby improving the accuracy and robustness of automatic operation and maintenance.

Description

Technical Field

[0001] The present invention relates to the field of network operation and maintenance technology, and in particular to a method for intelligently generating operation and maintenance scripts based on a large model. Background Art

[0002] With the upgrading of enterprise business models, the level of intelligence and informatization of enterprises has been continuously improved, and the scale of various business systems has been continuously expanding. The improvement of intelligence, informatization and scale has led to a rapid increase in the scale of equipment, which has brought great pressure and challenges to equipment operation and maintenance personnel.

[0003] To ensure the secure and stable operation of these hardware and software devices and business systems, operations and maintenance personnel are required to perform daily operations. Operation and maintenance scripts can effectively address a large amount of tedious and repetitive work. However, traditional methods for generating operation and maintenance scripts rely on the personnel's expertise and experience, making them inefficient and error-prone. Furthermore, because different hardware and software devices often have different data formats, batch operations on even the same type of device systems require different scripts, which is a significant workload. Therefore, we propose a method for intelligently generating operation and maintenance scripts based on large models. Summary of the Invention

[0004] The purpose of the present invention is to provide a method for intelligently generating operation and maintenance scripts based on a large model.

[0005] The technical problems solved by the present invention are:

[0006] (1) How to perform word segmentation, extraction and matching based on the operation and maintenance requirements of the historical operation and maintenance library, and perform cluster analysis to build a matching library with representative indicator names and representative parameter formats, so as to solve the problem in the existing technology that the operation and maintenance requirements description of the operation and maintenance personnel needs to be specific to a certain target asset, and avoid the problem of needing to make detailed divisions in the description;

[0007] (2) How to perform root cause analysis on the current alarm event based on the operation and maintenance big model, and determine whether the current alarm event is a previous alarm, a similar alarm, or an unfamiliar alarm event based on the historical operation and maintenance database, and then take different response measures to improve the accuracy and adaptability of automated operation and maintenance;

[0008] The present invention can be implemented through the following technical solution: A method for intelligently generating operation and maintenance scripts based on a large model, comprising the following steps:

[0009] Step 1: Build an asset database and establish an operation and maintenance network topology diagram;

[0010] Step 2: Perform word segmentation, extraction, and matching based on the operation and maintenance requirements actively input by manual operation and maintenance or recommended by automatic operation and maintenance analysis to obtain an indicator-parameter form;

[0011] Actively input operation and maintenance requirements are manually input by the operation and maintenance personnel. The corresponding matching relationship between clustered indicators and parameters is obtained according to the analysis, extraction and matching process, and an indicator-parameter form is generated.

[0012] Automatic operation and maintenance analysis recommends operation and maintenance requirements. Based on the intelligent alarm of the operation and maintenance large model, the alarm event archive is obtained. According to the alarm location, alarm event type and alarm cause in the alarm event archive, the historical operation and maintenance database is traversed and queried. Based on whether there are identical alarms and historical operation and maintenance requirements, precise operation and maintenance requirements, reference operation and maintenance requirements, and related operation and maintenance requirements are obtained. These requirements are analyzed, extracted and matched to generate indicator-parameter tables.

[0013] Step 3. Select the script module according to the target asset's operating environment and fill in the indicator-parameter form, then deploy the script to the target asset.

[0014] A further technical improvement of the present invention is that the specific process of performing word segmentation, extraction and matching based on operation and maintenance requirements includes:

[0015] Obtain historical operation and maintenance requirements, use automatic word segmentation and annotation tools to annotate the indicator names and corresponding parameters in the historical operation and maintenance requirements, and perform manual review and correction to obtain the preliminary requirement text;

[0016] Perform text cleaning on the preliminary requirements text;

[0017] Segment the initial demand text according to the annotations to construct a vocabulary in key-value format, and vectorize the words in the vocabulary to obtain high-dimensional space vectors;

[0018] A clustering algorithm is used to reduce the dimensionality of high-dimensional space vectors to obtain multiple clusters. The name of the most frequent representative indicator in each cluster is obtained and the representative parameter format of the corresponding parameter is recorded. Then, a representative indicator set and a representative parameter format set mapped to it are obtained, which together constitute a matching library.

[0019] When new operation and maintenance requirements arise, a vocabulary is obtained in the above manner, and the indicator names in the vocabulary are vectorized and their cosine similarity with the vectorized representative name indicators in the representative indicator set is calculated. The indicators with cosine similarity higher than the threshold are added to the corresponding clusters, and the representative indicator names of the clusters are used as key values. The format of the value corresponding to the key value is limited to the corresponding parameter format of the mapped representative parameter format set.

[0020] A further technical improvement of the present invention is that after a new operation and maintenance requirement is added to a corresponding cluster, the indicator name with the highest frequency in the corresponding cluster is re-counted and the matching library is updated.

[0021] A further technical improvement of the present invention is that the process of automatically generating and recommending three types of operation and maintenance requirements based on the alarm event archive includes:

[0022] (1) When there are identical alarms, the operation and maintenance requirement instructions for handling the corresponding historical alarms are directly extracted from the historical operation and maintenance database and defined as precise operation and maintenance requirements;

[0023] (2) When there are no identical alarms, filter the historical alarms with the same alarm event type and alarm cause and generate a candidate set. Extract the corresponding historical alarms from the candidate set according to the node type and node function where the alarm occurred. Extract the operation and maintenance requirements for handling the corresponding alarms from the historical operation and maintenance library and define them as reference operation and maintenance requirements.

[0024] (3) When an unfamiliar alarm occurs, an alarm correlation analysis is performed on each node on the business main chain where the node where the current fault occurs is located, and the corresponding operation and maintenance requirements of the alarms with strong correlation are defined as associated operation and maintenance requirements.

[0025] A further technical improvement of the present invention is that: the specific steps of performing alarm correlation analysis on each node on the service main chain where the node where the current alarm occurs is located and obtaining the associated operation and maintenance requirements include:

[0026] Determine the business main chain where the node generating the unfamiliar alarm is located;

[0027] Obtain alarm events of each node within a certain time range on the main chain of the business;

[0028] Align the alarm event occurrence time and the abnormal data occurrence time that generated the alarm on the timeline and unify the timestamp format, then arrange the alarm events in chronological order;

[0029] Calculate the temporal correlation, spatial correlation, and event type correlation between the unfamiliar alarm and another alarm event respectively and sum them up to obtain the total correlation;

[0030] The historical operation and maintenance requirements of the top M associated alarm events with the greatest correlation in the sorting are defined as the associated operation and maintenance requirements.

[0031] A further technical improvement of the present invention is that the total correlation R g The calculation formula is:

[0032] R g =P1*r time +P2*r distance +P3*r type ;

[0033] in, r type =r(A1,A2);

[0034] r time 、r distance 、r type They respectively represent the time correlation, distance correlation, and event type correlation between a certain alarm event and the current unknown alarm. P1, P2, and P3 are the weight distribution coefficients of the above three correlations, and P1+P2+P3=1. e is a natural constant in mathematics. λ and μ are the attenuation coefficients related to the time interval and the logical distance between nodes, respectively. t0 and t1 are the occurrence time of the unknown alarm and the occurrence time of the alarm of another related node, respectively. k represents the data flow impact coefficient. Q represents the average data flow on the link segment between the node representing the current unknown alarm and the related node on the service main chain.

[0035] r(A1, A2) represents the Pearson correlation coefficient between the two column vectors A1 and A2 of the type correlation matrix A, and the two column vectors are the type attribute vectors of the alarm event type.

[0036] A further technical improvement of the present invention is that: for reference operation and maintenance requirements, all reference operation and maintenance requirements are listed and organized and recommended to operation and maintenance personnel. If the operation and maintenance personnel do not perform a selection operation within the response time, the operation and maintenance requirements corresponding to the operating environment, equipment type and alarm level of the alarm node are directly selected to perform word segmentation extraction and matching, construct an indicator-parameter form, and fill in the operation and maintenance script for automatic operation and maintenance.

[0037] A further technical improvement of the present invention is that for related operation and maintenance needs, consumption is carried out one by one in the queue in the order of relevance, and recommendations are sent to the operation and maintenance personnel. If the operation and maintenance personnel do not perform any operation within the response time, word segmentation extraction and matching are directly performed, and an indicator-parameter form is constructed to fill in the operation and maintenance script.

[0038] A further technical improvement of the present invention is that for the indicator-parameter form generated using reference operation and maintenance requirements or associated operation and maintenance requirements, an optional operation rollback label needs to be set. When the corresponding generated operation and maintenance script cannot resolve the current alarm, the operation rollback is executed, and another operation and maintenance requirement corresponding to the operating environment, equipment type and alarm level of the alarm node is selected, or the operation and maintenance requirement in another correlation sequence queue is consumed to regenerate the script.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] 1. The present invention performs word segmentation, extraction and matching based on the operation and maintenance requirements of the historical operation and maintenance library, and performs cluster analysis to construct a matching library with representative indicator names and representative parameter formats. This allows operation and maintenance personnel to enter the operation and maintenance requirement description without considering issues such as the specific target asset operating environment, without the need to distinguish scripts, and without the need for fine-grained division in batch operations, thereby reducing workload. At the same time, the construction of the matching library is based on the indicator names, and errors in the operation and maintenance requirement indicators can be easily discovered in the matching library, thereby reducing the error rate.

[0041] 2. The present invention performs root cause analysis on the current alarm event based on the operation and maintenance big model, and determines whether the current alarm event is an alarm that has occurred before, an alarm of the same type, or an unfamiliar alarm event based on the historical operation and maintenance library, and then adopts different response measures to improve the accuracy and robustness of automated operation and maintenance; at the same time, the operation and maintenance requirements and related operation and maintenance requirements are referred to and sent to the operation and maintenance personnel for confirmation when executing the automated script generation, which improves accuracy; and automatically rolls back and re-edits the operation and maintenance script when the alarm cannot be eliminated, which can avoid new problems of the target asset in the case of invalid operation and maintenance, and the trial and error process itself also provides data reference for the operation and maintenance personnel to achieve fast and accurate operation and maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] To facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.

[0043] Figure 1 The figure is a schematic diagram of the execution flow of the method of the present invention.

[0044] Figure 2 The figure is a flow chart of the word segmentation, extraction and matching method of the present invention. DETAILED DESCRIPTION

[0045] In order to further illustrate the technical means and effects adopted by the present invention to achieve the predetermined purpose of the invention, the specific implementation methods, structures, features and effects of the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0046] See also Figure 1 As shown, a method for intelligently generating operation and maintenance scripts based on a large model includes:

[0047] Step 1: Build an asset database and establish an operation and maintenance network topology diagram

[0048] Integrate SNMP agent program in the operation and maintenance server to collect and organize information about device hardware, servers, network nodes and applications in the local area network to obtain an asset database;

[0049] The device or service information in the asset database includes IP address, operating environment, function type, etc.

[0050] Step 2: Analyze operation and maintenance requirements and generate an operation and maintenance task form

[0051] Operation and maintenance requirements come from input requirements from operation and maintenance personnel and automatic operation and maintenance requirements from operation and maintenance alarms;

[0052] Operation and maintenance personnel can actively input operation and maintenance requirements, such as: "Upgrade the version of xx application of xx device to version xxv at a set time"; through word segmentation extraction and matching, parameter extraction and indicator matching are performed to form a corresponding indicator-parameter form, which is defined as an operation and maintenance task form; the indicators here include operation indicators, object indicators, time indicators, etc.

[0053] Automatic O&M requirements based on O&M alarms means the O&M system automatically generates O&M requirements based on O&M alarms. This automatic generation of O&M requirements is based on analysis of static O&M rules and a dynamic historical O&M library, selecting the optimal O&M method and generating O&M requirements. The indicator-parameter form is then organized based on the O&M requirements.

[0054] Among them, Figure 2 As shown in FIG, the implementation process of the word segmentation extraction and matching method includes:

[0055] S11: Obtain the operation and maintenance requirements from historical operations and maintenance, select an automatic word segmentation and annotation tool with self-learning capabilities (such as THULAC, which uses RNN recurrent neural networks to achieve self-learning) to annotate the indicator names and corresponding parameters in the historical operations and maintenance. The annotated results are manually reviewed and corrected by operation and maintenance personnel to obtain the preliminary requirement text, which is the text with annotated labels.

[0056] Manual review and correction occurs during the training preparation phase of the word segmentation, extraction, and matching model. During the manual correction process, the original annotations of the erroneous samples, the corrected annotations, and the reasons for the corrections are recorded to form a correction log. This log is then imported into the automatic word segmentation and annotation tool for feedback and optimization.

[0057] S12: Clean the preliminary requirement text to remove irrelevant information in the text, such as stop words, punctuation marks, and special characters;

[0058] S13: Segment the preliminary demand text according to the annotations and construct a vocabulary. The vocabulary is presented in a key-value pairing format. Word2Vec is used to vectorize the words in the vocabulary to obtain a high-dimensional space vector of the corresponding words. The high-dimensional space vector includes text position attributes, format attributes, information attributes, etc.

[0059] S14: After performing dimensionality reduction processing using a density-based clustering algorithm such as DBSCAN, multiple highly correlated clusters are obtained. The corresponding indicator name with the highest frequency in the cluster is selected and defined as the representative indicator name. At the same time, the representative parameter format of the parameter corresponding to the representative indicator name is recorded.

[0060] Thus, a large number of representative indicator names representing different meanings and corresponding matching parameter formats are obtained, and a representative indicator set and a representative parameter format set mapped thereto are generated accordingly, which together constitute a matching library.

[0061] S15: After the operation and maintenance requirements are annotated and segmented to obtain a vocabulary, the indicator name vocabulary in the vocabulary is vectorized and the cosine similarity with the vectorized representative name indicator in the representative indicator set is calculated. A comparison threshold is set, and those above the comparison threshold are added to the cluster, so that the representative indicator name is used as the key value for the subsequent generation of the indicator-parameter form, and the format of the value value corresponding to the key value is limited to the corresponding parameter format of the mapped representative parameter format set; then the indicator name with the highest frequency in the corresponding cluster is re-counted to update the matching library; the representative indicator name and the corresponding parameter format are obtained according to the matching, and the indicator-parameter form is filled.

[0062] Specifically, the process of automatically generating and organizing operation and maintenance requirements into indicator-parameter forms includes:

[0063] S21: Analyze and process logs based on large models

[0064] The above-mentioned big model refers to the operation and maintenance big model, which refers to a large-scale operation and maintenance management model built based on big data and machine learning technologies. For example, IBM AIOps and Alibaba Cloud AIOps are both relatively mature intelligent operation and maintenance platforms with functions such as data collection, organization, intelligent alarms, root cause analysis, and automated response. They will not be described in detail here.

[0065] Use the big model to collect operation and maintenance logs and start intelligent alarms. In this application, alarm events with an alarm level not lower than WARN level are used as the objects of automatic operation and maintenance. The big model locates the alarm location based on the topological network and abnormal log information and performs root cause analysis.

[0066] In this way, the alarm location, alarm event type, occurrence time, short-term alarm frequency and alarm cause can be obtained, and the alarm event file can be integrated to obtain the alarm event file, which is generated in the form of key-value pairs.

[0067] S22: Based on the alarm occurrence location, alarm event type, and alarm occurrence cause in the alarm event file, the historical operation and maintenance database is searched to check whether there is the same historical alarm:

[0068] When identical historical alarms exist, the operation and maintenance requirement instructions for handling the corresponding historical alarms are directly extracted from the historical operation and maintenance database and defined as precise operation and maintenance requirements;

[0069] When there are no completely identical alarms, historical alarms with the same alarm event type and alarm cause are filtered and a candidate set is generated. Then, the type (hardware, service, middleware or application) and node function of the terminal node corresponding to the location where the alarm occurred are determined. Historical alarms with consistent node types and node functions are retrieved from the candidate set, and the operation and maintenance requirement instructions for handling the corresponding historical alarms are extracted and defined as reference operation and maintenance requirements.

[0070] S23: When an unfamiliar alarm occurs, that is, when no accurate operation and maintenance requirements or reference operation and maintenance requirements can be found in the historical operation and maintenance database, an alarm correlation analysis is performed on the alarms of each node on the main business chain where the node where the current alarm occurs is located. Specifically:

[0071] (1) Obtain all alarm information on the service main chain of the node within a set time range before the occurrence of the strange alarm. The service main chain refers to the communication link for receiving and generating data flow of the corresponding node;

[0072] (2) Align the alarm information with the time of the abnormal data event that triggered the alarm information on the time axis and unify the timestamp format. Then, sort the alarm information according to the alarm occurrence sequence after the alignment time axis;

[0073] (3) Count the occurrence ratio of each alarm event type and the distribution of each alarm event type in each associated alarm node, and determine the correlation between different time types of each associated alarm node and the current unfamiliar alarm. The calculation formula is:

[0074] R g =P1*r time +P2*r distance +P3*r type ;

[0075] Among them, R g Indicates the correlation between a certain alarm event and the current unknown alarm, r time 、r distance 、r type Respectively represent the time correlation, distance correlation, and event type correlation between a certain alarm event and the current unknown alarm. P1, P2, and P3 are the weight distribution coefficients of the above three correlations, and P1+P2+P3=1;

[0076] Time correlation is calculated based on the time interval between the current strange alarm and the alarm events of each associated node. The shorter the time interval, the greater the correlation. Wherein, e is a natural constant in mathematics, λ is the attenuation coefficient of the time interval, which ranges from 0.18 to 0.45 and can be adjusted according to the actual situation, t0 and t1 are the time when the strange alarm occurs and the time when the alarm of another related node occurs, respectively;

[0077] Distance correlation is calculated based on the logical distance between the node of the current strange alarm and the associated node on the business main chain. The closer the distance, the greater the correlation. The logical distance here represents the number of hops (number of nodes) that the data flow between the two nodes passes through. e is a natural mathematical constant, μ is the attenuation coefficient for the logical distance between nodes, ranging from 0.15 to 0.8 and adjustable based on actual conditions, k represents the data flow impact coefficient, and Q represents the average data flow on the link segment between the node currently receiving the strange alarm and the associated node on the service main chain.

[0078] For type correlation, a type correlation matrix A is defined. The type correlation matrix A contains the type attribute vector of the current strange alarm and the type attribute vector of another related alarm. The vector dimensions in the type attribute vector include level, layer (system layer, middle layer, service layer or application layer), device type (communication device, server or execution terminal), operating environment, alarm duration, etc. The Pearson correlation coefficient between two vectors in the type correlation matrix A is calculated, that is, r type =r(A1,A2), where A1 and A2 are two column vectors of the type correlation matrix respectively.

[0079] (4) Arrange the above-mentioned related alarm events in descending order of relevance, extract the operation and maintenance requirement instructions of the top M related alarm events with the largest relevance, and obtain the related operation and maintenance requirements. M is generally set to 5.

[0080] S24: For precise operation and maintenance requirements, directly perform the above word segmentation extraction and matching to build an indicator-parameter form;

[0081] For reference operation and maintenance requirements, all reference operation and maintenance requirements are listed and sorted and recommended to the operation and maintenance personnel for selection. If the operation and maintenance personnel do not respond within the set time limit, the operation and maintenance requirements corresponding to the operating environment, device type and alarm level of the alarm node are selected from the reference operation and maintenance requirements for word segmentation extraction and matching. An indicator-parameter form is constructed, and an optional operation fallback label is set;

[0082] For the first M sequential queues of associated operation and maintenance requirements, the operation and maintenance requirement with the highest ranking is selected each time and recommended to the operation and maintenance personnel for selection. If the operation and maintenance personnel do not respond after the set period, word segmentation and extraction matching are directly performed, an indicator-parameter form is constructed, and an optional operation fallback label is set. If the current alarm still exists after automated operation and maintenance is subsequently performed using the automatically generated script, the operation and maintenance requirement with the highest ranking in the sequential queue is marked as invalid consumption, and the second highest ranking operation and maintenance requirement is selected to re-execute the above operation.

[0083] Step 3. Select the script template and fill in the indicator-parameter form

[0084] Query the operating environment of the target asset for automated operation and maintenance in the indicator-parameter form from the asset database, select a script template that matches the operating environment, use placeholders for both key and value values ​​in the script template, replace the placeholders with the corresponding values ​​in the indicator-parameter form, and automatically generate an operation and maintenance script. Deploy the operation and maintenance script to the target asset.

[0085] The above description is merely a preferred embodiment of the present invention and does not constitute any form of limitation to the present invention. Although the present invention has been disclosed as a preferred embodiment as above, it is not intended to limit the present invention. Any person skilled in the art can make some changes or modifications to equivalent embodiments using the technical contents disclosed above without departing from the scope of the technical solution of the present invention. However, any simple modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A method for intelligently generating operation and maintenance scripts based on a large model, characterized by: The steps include: Step 1: Build an asset database and establish an operation and maintenance network topology diagram; Step 2: Perform word segmentation, extraction, and matching based on the operation and maintenance requirements recommended by the automatic operation and maintenance analysis to obtain an indicator-parameter form; Automatic operation and maintenance analysis recommends operation and maintenance requirements. Based on the intelligent alarm of the operation and maintenance large model, the alarm event archive is obtained. According to the alarm location, alarm event type and alarm cause in the alarm event archive, the historical operation and maintenance database is traversed and queried. Based on whether there are identical alarms and historical operation and maintenance requirements, precise operation and maintenance requirements, reference operation and maintenance requirements, and related operation and maintenance requirements are obtained. These requirements are analyzed, extracted and matched to generate indicator-parameter tables. Step 3: Select the script module based on the target asset's operating environment and fill in the indicator-parameter form, then deploy the script to the target asset; The process of automatically generating and recommending three types of operation and maintenance requirements based on alarm event archives includes: (1) When there are identical alarms, the operation and maintenance requirement instructions for handling the corresponding historical alarms are directly extracted from the historical operation and maintenance database and defined as precise operation and maintenance requirements; (2) When there are no identical alarms, filter the historical alarms with the same alarm event type and alarm cause and generate a candidate set. Extract the corresponding historical alarms from the candidate set according to the node type and node function where the alarm occurred. Extract the operation and maintenance requirements for handling the corresponding alarms from the historical operation and maintenance library, and define them as reference operation and maintenance requirements. (3) When an unfamiliar alarm occurs, perform alarm correlation analysis on each node on the business main chain where the node where the current fault occurs is located, and define the corresponding operation and maintenance requirements of the alarms with strong correlation as related operation and maintenance requirements; Perform alarm correlation analysis on each node on the main business chain where the node where the current alarm occurs is located. The specific steps to obtain the associated operation and maintenance requirements include: Determine the business main chain where the node generating the unfamiliar alarm is located; Obtain alarm events of each node within a certain time range on the main chain of the business; Align the alarm event occurrence time and the abnormal data occurrence time that generated the alarm on the timeline and unify the timestamp format, then arrange the alarm events in chronological order; Calculate the temporal correlation, spatial correlation, and event type correlation between the unfamiliar alarm and another alarm event respectively and sum them up to get the total correlation. The formula is: ； in, 、 、 ; Respectively represent the time correlation, distance correlation and event type correlation between a certain alarm event and the current unknown alarm. are the weight distribution coefficients of the above three correlations, and ; e is a natural constant in mathematics, 、 are the attenuation coefficients for time interval and node logical distance, respectively. The occurrence time of the strange alarm and the occurrence time of another related node alarm respectively, Indicates the data traffic impact coefficient, Indicates the average data traffic volume on the link segment between the node representing the current strange alarm and the associated node on the business main chain; Two column vectors representing the type correlation matrix A The Pearson correlation coefficient between them, the two column vectors are the type attribute vectors of the alarm event type; The historical operation and maintenance requirements of the top M associated alarm events with the greatest correlation in the sorting are defined as the associated operation and maintenance requirements.

2. The method for intelligently generating operation and maintenance scripts based on a large model according to claim 1, characterized in that: The specific process of word segmentation, extraction and matching based on operation and maintenance requirements includes: Obtain historical operation and maintenance requirements, use automatic word segmentation and annotation tools to annotate the indicator names and corresponding parameters in the historical operation and maintenance requirements, and perform manual review and correction to obtain the preliminary requirement text; Perform text cleaning on the preliminary requirements text; Segment the initial demand text according to the annotations to construct a vocabulary in key-value format, and vectorize the words in the vocabulary to obtain high-dimensional space vectors; A clustering algorithm is used to reduce the dimensionality of high-dimensional space vectors to obtain multiple clusters. The name of the most frequent representative indicator in each cluster is obtained and the representative parameter format of the corresponding parameter is recorded. Then, a representative indicator set and a representative parameter format set mapped to it are obtained, which together constitute a matching library. When new operation and maintenance requirements arise, a vocabulary is obtained in the above manner, and the indicator names in the vocabulary are vectorized and their cosine similarity with the vectorized representative name indicators in the representative indicator set is calculated. The indicators with cosine similarity higher than the threshold are added to the corresponding clusters, and the representative indicator names of the clusters are used as key values. The format of the value corresponding to the key value is limited to the corresponding parameter format of the mapped representative parameter format set.

3. The method for intelligently generating operation and maintenance scripts based on a large model according to claim 2, characterized in that: After the new operation and maintenance requirements are added to the corresponding cluster, the indicator names with the highest frequency in the corresponding cluster are re-counted and the matching library is updated.

4. The method for intelligently generating operation and maintenance scripts based on a large model according to claim 1, characterized in that: For reference operation and maintenance requirements, all reference operation and maintenance requirements are listed and organized and recommended to the operation and maintenance personnel. If the operation and maintenance personnel do not make a selection within the response time, the operation and maintenance requirements corresponding to the operating environment, equipment type and alarm level of the alarm node are directly selected to perform word segmentation extraction and matching, build an indicator-parameter form, and fill in the operation and maintenance script for automatic operation and maintenance.

5. The method for intelligently generating operation and maintenance scripts based on a large model according to claim 1, characterized in that: For related operation and maintenance requirements, they are consumed one by one in the queue in the order of relevance, and recommendations are sent to the operation and maintenance personnel. If the operation and maintenance personnel do not perform any operation within the response time, word segmentation extraction and matching will be directly performed, and an indicator-parameter form will be constructed to fill in the operation and maintenance script.

6. The method for intelligently generating operation and maintenance scripts based on a large model according to claim 4, characterized in that: For the indicator-parameter form generated using the reference operation and maintenance requirements or associated operation and maintenance requirements, an optional operation rollback label must be set. When the corresponding generated operation and maintenance script cannot resolve the current alarm, the operation rollback is performed, and another operation and maintenance requirement corresponding to the operating environment, equipment type and alarm level of the alarm node is selected, or the operation and maintenance requirement in another correlation sequence queue is consumed to regenerate the script.

Images