An automated operation and maintenance management system and method based on knowledge graph

Abstract

The present application relates to the field of operation and maintenance management technology, and in particular to an automated operation and maintenance management system and method based on a knowledge graph. The method includes: obtaining operation and maintenance data, and constructing a reserved node type general knowledge graph based on the operation and maintenance data; obtaining fault detection information, analyzing the fault detection information and the reserved node type general knowledge graph, and determining event-driven self-growing nodes; obtaining performance monitoring information, analyzing the performance monitoring information and the reserved node type general knowledge graph, and determining cycle-driven self-growing nodes; on this basis, constructing a system-adaptive knowledge graph based on the reserved node type general knowledge graph; obtaining continuous operation information, and determining a highly adaptive knowledge graph based on the continuous operation information; implementing automated operation and maintenance measures based on the highly adaptive knowledge graph, and outputting an operation and maintenance report. Through the update and iteration of the knowledge graph, the automated operation and maintenance work based on the knowledge graph remains stable and efficient in the changing system operation and maintenance requirements.

Description

Technical Field

[0001] The present application relates to the field of operation and maintenance management technology, and in particular to an automated operation and maintenance management system and method based on a knowledge graph. Background Art

[0002] Modern systems are becoming increasingly complex, involving multiple technology stacks and a large number of devices and services. In this context, traditional operation and maintenance management faces challenges such as low efficiency and long response time. Automated operation and maintenance management technology based on knowledge graphs can effectively reduce manual intervention in the operation and maintenance process, improve the degree of automation of operation and maintenance work, and thus ensure system stability.

[0003] However, the existing automated operation and maintenance management technology based on knowledge graphs relies on knowledge graphs built based on a large amount of existing operation and maintenance data. As the system scale and running time grow, the knowledge graph has problems such as data redundancy and lack of rapid response capabilities to new threats or vulnerabilities, which increases the system security risks and makes it difficult to maintain good system stability. Summary of the invention

[0004] The present application provides an automated operation and maintenance management system and method based on knowledge graph to solve the above-mentioned technical problems.

[0005] In a first aspect, the present application provides an automated operation and maintenance management method based on a knowledge graph, the method comprising:

[0006] Obtaining operation and maintenance data, and constructing a reserved node-type general knowledge graph based on the operation and maintenance data;

[0007] Acquire fault detection information, analyze the fault detection information and the reserved node type general knowledge graph, and determine an event-driven self-growing node;

[0008] Acquire performance monitoring information, analyze the performance monitoring information and the reserved node type general knowledge graph, and determine a cycle-driven self-growing node;

[0009] Based on the reserved node type general knowledge graph, a system adaptive knowledge graph is constructed according to the event-driven self-growing nodes and the cycle-driven self-growing nodes;

[0010] Acquire continuous operation information, and perform floating adjustment on the event-driven self-growing node and the cycle-driven self-growing node according to the continuous operation information to determine a highly adaptive knowledge graph;

[0011] According to the highly adaptable knowledge graph, automated operation and maintenance measures are implemented, and an operation and maintenance report is output.

[0012] Through this solution, a general knowledge graph with reserved nodes is constructed based on operation and maintenance data, providing a basic framework for subsequent changes and expansions to the knowledge graph. By analyzing fault detection information, event-driven self-growing nodes are constructed to improve the adaptability of automated operation and maintenance work to different system faults. By analyzing performance monitoring information, cycle-driven self-growing nodes are constructed, so that automated operation and maintenance work can accurately periodically tune the system according to the cycle-driven self-growing nodes. Based on event-driven self-growing nodes and cycle-driven self-growing nodes, a system-adaptive knowledge graph is constructed to automatically complete the update and iteration of the knowledge graph, so that the automated operation and maintenance work based on the knowledge graph remains stable and efficient in the changing system operation and maintenance requirements. According to the continuous operation information, the knowledge graph is further adjusted to obtain a highly adaptable knowledge graph, which improves the matching degree between the knowledge graph and the personalized operation and maintenance requirements of the system, thereby improving the operation and maintenance effect. Finally, based on the highly adaptable knowledge graph, automated operation and maintenance measures are determined and implemented to realize automated operation and maintenance work based on the knowledge graph.

[0013] Optionally, constructing a reserved node type general knowledge graph according to the operation and maintenance data includes:

[0014] Analyze the operation and maintenance data, perform entity recognition on the operation and maintenance data, and determine an operation and maintenance entity node set;

[0015] Based on the operation and maintenance entity node set, performing entity relationship analysis and extraction on the operation and maintenance data to determine an operation and maintenance entity relationship set;

[0016] According to the operation and maintenance entity node set, adding an event-driven branch node and a cycle-driven branch node to each operation and maintenance entity node to form a reserved node type entity node set;

[0017] The reserved node type general knowledge graph is constructed according to the reserved node type entity node set and the operation and maintenance entity relationship set.

[0018] Through this solution, entity recognition is performed on the operation and maintenance data to obtain an operation and maintenance entity node set, and on this basis, entity relationship extraction is performed on the operation and maintenance data to obtain an operation and maintenance entity relationship set, so as to achieve structured processing of the operation and maintenance data, and then by adding event-driven branch nodes and cycle-driven branch nodes to each operation and maintenance entity node, the reserved node-type general knowledge graph can easily add new nodes and relationships in the future without making major modifications to the existing knowledge graph structure. This flexibility helps to adapt to rapidly changing or uncertain operation and maintenance needs.

[0019] Optionally, the fault detection information includes a timestamp, a fault entity, an abnormal parameter, and a maintenance record, and the analyzing the fault detection information and the reserved node type general knowledge graph to determine the event-driven self-growing node includes:

[0020] According to the fault entity, analyzing the reserved node type general knowledge graph to determine the fault operation and maintenance entity node;

[0021] According to the fault entity and the abnormal parameter, respectively determine an entity feature vector and a parameter feature vector;

[0022] Constructing a fault feature vector according to the entity feature vector and the parameter feature vector;

[0023] Based on the fault feature vector, a fault similarity analysis is performed on the fault operation and maintenance entity node in the reserved node type general knowledge graph to determine whether there is a highly similar fault;

[0024] If the highly similar fault does not exist, the event-driven self-growing node is constructed based on the fault entity, according to the timestamp, the fault feature vector, and the maintenance record.

[0025] Through this solution, the entity feature vector and parameter feature vector are determined according to the fault entity and abnormal parameters corresponding to the fault, and on this basis, a fault feature vector for describing the fault characteristics is constructed. According to the fault feature vector, special faults occurring in the system are captured from the fault detection information, and according to the fault entity, timestamp, fault feature vector and maintenance record used to describe the special fault, the event-driven self-growing node is constructed, so that the subsequent knowledge graph can be flexibly expanded according to the special fault, thereby enhancing the adaptability of the knowledge graph in operation and maintenance work.

[0026] Optionally, the event-driven self-growing node is constructed based on the fault entity according to the timestamp, the fault feature vector, and the maintenance record, specifically as follows:

[0027] ;

[0028] in, is the fault entity, is the timestamp, is the entity feature vector, is the parameter eigenvector, For said maintenance records, is the node constructor, is the fault operation and maintenance entity node, The corresponding node under the fault operation and maintenance entity node Fault feature vectors, is the fault feature vector, is the preset fault similarity threshold.

[0029] Through this scheme, mathematical analysis methods are used to construct the construction conditions of event-driven self-growing nodes. On the basis of satisfying the construction conditions, event-driven self-growing nodes are constructed through node constructors according to fault entities, timestamps, fault feature vectors and maintenance records, so as to realize the scientific and automated construction of event-driven self-growing nodes, make event-driven self-growing nodes more accurate, and improve the node update speed of subsequent knowledge graphs.

[0030] Optionally, the performance monitoring information includes unit time period, load entity and load fluctuation information, and the analyzing the performance monitoring information and the reserved node type general knowledge graph to determine the period-driven self-growing node includes:

[0031] Determining a starting load, a fluctuation amplitude, and an average load according to the load fluctuation information;

[0032] According to the load entity, analyzing the reserved node type general knowledge graph to determine the load operation and maintenance entity node;

[0033] Determining resource scheduling information based on the unit time period, according to the starting load, the fluctuation amplitude and the average load;

[0034] The cycle-driven self-growing node is constructed according to the load operation and maintenance entity node and the resource scheduling information.

[0035] Through this scheme, according to the load fluctuation information, the starting load, fluctuation amplitude and average load that reflect the load fluctuation characteristics are extracted. On this basis, the resource scheduling information within the load fluctuation period is analyzed and obtained. At the same time, according to the load entity, the corresponding load operation and maintenance entity node in the reserved node type general knowledge graph is retrieved, and according to the load operation and maintenance entity node and resource scheduling information, a cycle-driven self-growing node is constructed, so that the node contains the characteristics of load fluctuations and the corresponding scheduling means, further improving the efficiency and effectiveness of subsequent automated operation and maintenance based on the knowledge graph.

[0036] Optionally, the determining the resource scheduling information based on the unit time period according to the starting load, the fluctuation amplitude and the average load includes:

[0037] According to the unit time period, the starting load, the fluctuation amplitude and the average load, the resource allocation value for the load operation and maintenance entity node at different time points is determined, specifically as follows:

[0038] ;

[0039] in, For time point The resource allocation value under is the preset adjustment factor, is the average load, is the starting load, is the unit time period, is the fluctuation range, Adjust thresholds for presets;

[0040] The resource allocation value for the load operation and maintenance entity node at different time points is used as the resource scheduling information.

[0041] Through this solution, mathematical analysis methods are used to mathematically model the system load fluctuation according to the unit time period, starting load, fluctuation amplitude and average load, and the resource scheduling intervention conditions are constrained to accurately quantify the resource scheduling value. By combining the fluctuation model with threshold judgment, maintenance resources are adjusted and allocated in a targeted manner, so that the cycle-driven self-growing nodes can meet the actual load requirements and ensure the stability of operation and maintenance.

[0042] Optionally, the system-adaptive knowledge graph is constructed based on the reserved node-type general knowledge graph and according to the event-driven self-growing nodes and the cycle-driven self-growing nodes, including:

[0043] According to the corresponding fault entity in the event-driven self-growing node, determine the corresponding operation and maintenance entity node in the reserved node type general knowledge graph;

[0044] According to the operation and maintenance entity node, determining the event-driven branch node and the period-driven branch node in the operation and maintenance entity node;

[0045] The system-adaptive knowledge graph is determined by connecting the event-driven self-growing node and the cycle-driven self-growing node to the event-driven branch node and the cycle-driven branch node respectively.

[0046] Through this solution, event-driven self-growing nodes and cycle-driven self-growing nodes are respectively connected to the event-driven branch nodes and cycle-driven branch nodes reserved under the corresponding operation and maintenance entity nodes in the reserved node-type general knowledge graph, so as to realize the persistent processing of event-driven self-growing nodes and cycle-driven self-growing nodes, obtain a system-adaptive knowledge graph, and complete the dynamic update of the knowledge graph.

[0047] Optionally, the continuous operation information includes a fault node adjustment frequency and a periodic node adjustment frequency, and the floating adjustment of the event-driven self-growing node and the periodic-driven self-growing node according to the continuous operation information to determine a highly adaptive knowledge graph includes:

[0048] According to the fault node adjustment frequency, the event-driven self-growing node in the system adaptive knowledge graph is controlled to float upward under the corresponding event-driven branch node, and the number of event branch layers after the floating of the event-driven self-growing node is determined;

[0049] According to the periodic node adjustment frequency, the periodic driven self-growing node in the system adaptive knowledge graph is controlled to float upward under the corresponding periodic driven branch node, and the number of periodic branch layers after the floating of the periodic driven self-growing node is determined;

[0050] Determine an event node conversion strategy and a period node conversion strategy respectively according to the number of event branch layers after floating and the number of period branch layers after floating;

[0051] According to the event node conversion strategy and the periodic node conversion strategy, the system adaptive knowledge graph is adjusted to determine the highly adaptive knowledge graph.

[0052] Through this solution, the frequency of faulty nodes is adjusted, and the properties of event-driven self-growing nodes and periodic event-driven self-growing nodes that are triggered too frequently are converted, and the knowledge graph that has completed the node property conversion is used as a highly adaptable knowledge graph, so that the highly adaptable knowledge graph can clearly show maintenance personnel the event-driven self-growing nodes or periodic event-driven self-growing nodes that need to be processed separately and in a targeted manner, further optimize the overall operation and maintenance effect of the system, reduce the continuous resource occupation of automated operation and maintenance work, and thus improve the response speed of the entire system.

[0053] Optionally, the determining the event node conversion strategy and the periodic node conversion strategy respectively according to the number of event branch layers after floating and the number of periodic branch layers after floating includes:

[0054] The post-floating event branch number is compared with a preset event layer number threshold, and if the post-floating event branch number is less than the preset event layer number threshold, the event node conversion strategy is determined to convert the corresponding event-driven self-growing node into a fault elimination node;

[0055] The post-floating cycle branch layer number is compared with a preset cycle layer number threshold. If the post-floating cycle branch layer number is less than the preset cycle layer number threshold, the cycle node conversion strategy is determined to convert the corresponding cycle-driven self-growth node into a normal optimization node.

[0056] Through this solution, according to the comparison results of the post-floating event layer number and the post-floating cycle layer number with their corresponding layer thresholds, the corresponding event-driven self-growing node is converted into a troubleshooting node, or the corresponding cycle-driven self-growing node is converted into a normal optimization node, so as to remind professional maintenance personnel to optimize the system bottom layer for the troubleshooting node and the normal optimization node, so as to release the system resources occupied by the automated operation and maintenance work and improve the overall stability of the system.

[0057] In a second aspect, the present application provides an automated operation and maintenance management system based on a knowledge graph, the system comprising:

[0058] A general construction module, used to obtain operation and maintenance data, and to construct a reserved node type general knowledge graph based on the operation and maintenance data;

[0059] An event analysis module, used to obtain fault detection information, analyze the fault detection information and the reserved node type general knowledge graph, and determine an event-driven self-growing node;

[0060] A cycle analysis module, used to obtain performance monitoring information, analyze the performance monitoring information and the reserved node type general knowledge graph, and determine the cycle-driven self-growing node;

[0061] An adaptive construction module, used to construct a system adaptive knowledge graph based on the reserved node type general knowledge graph according to the event-driven self-growing nodes and the cycle-driven self-growing nodes;

[0062] An adaptation adjustment module, used for obtaining continuous operation information, and performing floating adjustment on the event-driven self-growing node and the cycle-driven self-growing node according to the continuous operation information to determine a highly adaptive knowledge graph;

[0063] An implementation module is used to implement automated operation and maintenance measures based on the highly adaptable knowledge graph and output an operation and maintenance report. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0065] Figure 1 A schematic diagram of an application scenario provided for an embodiment of the present application;

[0066] Figure 2 A flowchart of an automated operation and maintenance management method based on a knowledge graph provided in one embodiment of the present application;

[0067] Figure 3 A schematic diagram of the structure of an automated operation and maintenance management system based on a knowledge graph provided in one embodiment of the present application. DETAILED DESCRIPTION

[0068] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0069] In addition, the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article, unless otherwise specified, generally means that the associated objects before and after are in an "or" relationship.

[0070] The embodiments of the present application are further described in detail below in conjunction with the drawings in the specification.

[0071] Existing knowledge graph-based automated operation and maintenance management technology relies on a knowledge graph built based on a large amount of existing operation and maintenance data. As the system scale and running time grow, the knowledge graph has problems such as data redundancy and lack of rapid response capabilities to new threats or vulnerabilities, resulting in increased system security risks and difficulty in maintaining good system stability.

[0072] Based on this, the present application provides an automated operation and maintenance management system and method based on knowledge graph. Automatically complete the update and iteration of the knowledge graph, so that the automated operation and maintenance work based on the knowledge graph remains stable and efficient in the changing system operation and maintenance requirements. According to the continuous operation information, the knowledge graph is further adjusted to obtain a highly adaptable knowledge graph, improve the matching degree between the knowledge graph and the personalized operation and maintenance requirements of the system, and then improve the operation and maintenance effect. Finally, according to the highly adaptable knowledge graph, the automated operation and maintenance measures are determined and implemented to realize the automated operation and maintenance work based on the knowledge graph.

[0073] Figure 1 A schematic diagram of an application scenario provided by this application. In the process of automated operation and maintenance, the method provided by this application is applied to dynamically update and iterate the knowledge graph to improve the automated operation and maintenance effect while reducing data redundancy.

[0074] Specifically, the method of the present application is applied to any server, which communicates with the application log monitoring system, obtains the fault detection information, performance monitoring information and continuous operation information provided by the application log monitoring system through the server, and obtains the operation and maintenance information provided by professional operation and maintenance personnel. According to the operation and maintenance information, a general knowledge graph with reserved nodes is constructed to provide a basic framework for subsequent changes and expansions to the knowledge graph. By analyzing the fault detection information, an event-driven self-growing node is constructed to improve the adaptability of automated operation and maintenance work to different system faults. By analyzing the performance monitoring information, a cycle-driven self-growing node is constructed so that automated operation and maintenance work can be based on According to the cycle-driven self-growing nodes, the system is precisely periodically tuned. According to the event-driven self-growing nodes and the cycle-driven self-growing nodes, a system-adaptive knowledge graph is constructed to automatically update and iterate the knowledge graph, so that the automated operation and maintenance work based on the knowledge graph remains stable and efficient in the changing system operation and maintenance requirements. According to the continuous operation information, the knowledge graph is further adjusted to obtain a highly adaptive knowledge graph, thereby improving the matching degree between the knowledge graph and the personalized operation and maintenance requirements of the system, thereby improving the operation and maintenance effect. Finally, according to the highly adaptive knowledge graph, the automated operation and maintenance measures are determined and implemented to realize the automated operation and maintenance work based on the knowledge graph.

[0075] For specific implementation methods, please refer to the following embodiments.

[0076] Figure 2 This is a flow chart of an automated operation and maintenance management method based on a knowledge graph provided in an embodiment of the present application. The method of this embodiment can be applied to the server in the above scenario. Figure 2 As shown, the method includes:

[0077] S201. Obtain operation and maintenance data, and construct a reserved node-type general knowledge graph based on the operation and maintenance data.

[0078] The operation and maintenance data may be existing fault maintenance data corresponding to the current target operation and maintenance system type, and the operation and maintenance data may be provided by professional maintenance personnel.

[0079] The reserved node type general knowledge graph can be a knowledge graph constructed based on operation and maintenance data that reserves some blank nodes for subsequent node expansion.

[0080] Specifically, since there are a large number of highly similar operation and maintenance tasks between systems of the same type, structured information is extracted from a large amount of unstructured data information in the existing operation and maintenance materials, and the corresponding knowledge graph is constructed through knowledge representation technology, such as RDF (Resource Description Framework). During the construction phase, directional blank nodes are reserved for the knowledge nodes representing the operation and maintenance entities in the knowledge graph, providing a basic framework for subsequent changes and expansions to the knowledge graph.

[0081] S202. Obtain fault detection information, analyze the fault detection information and the reserved node-type general knowledge graph, and determine the event-driven self-growing node.

[0082] The fault detection information may be fault information detected in the current system, and the fault detection information may be obtained through an application log monitoring system.

[0083] The event-driven self-growing node can be a node that needs to be added to the current knowledge graph and corresponds to fault information that exists in the system but is not included in the reserved node type general knowledge graph.

[0084] Specifically, since the operation and maintenance knowledge in the reserved node type general knowledge graph is extracted based on the existing operation and maintenance data of similar systems, there is no guarantee that the operation and maintenance information in the reserved node type general knowledge graph can completely solve the faults in the current system. In order to improve the adaptability of the knowledge graph to the current system, when a fault that is not included in the reserved node type general knowledge graph occurs in the system, the fault monitoring information is abstractly analyzed and processed through mathematical analysis methods, and an event-driven self-growing node is constructed based on the extracted fault operation and maintenance information. The event-driven self-growing node represents that the system knowledge graph is driven by the event of "new fault", and the constructed node is used to provide key information for subsequent automated processing of similar faults.

[0085] S203. Obtain performance monitoring information, analyze the performance monitoring information and the reserved node-type general knowledge graph, and determine the cycle-driven self-growing node.

[0086] The performance monitoring information may be load performance information detected in the current system, and the performance monitoring information may be obtained through an application log monitoring system.

[0087] The period-driven self-growing node may be a knowledge graph node that contains periodic adjustment information of the system for the purpose of improving the system load performance.

[0088] Specifically, in addition to resolving faults within the system and maintaining the normal operation of the system, the purpose of operation and maintenance work is to pay attention to the performance of the system during operation and to optimize the system performance to maintain the good operation state of the system. The performance of the system is usually based on the workload it needs to bear at different times and shows periodic changes. That is, as the system working time increases and the work content solidifies, the load in the system will fluctuate at a fixed time period. The corresponding automated operation and maintenance work should provide more operating resources for the system when the system is highly loaded, and release more operating resources when the system is under low load to ensure the normal and stable operation of the system. The performance monitoring information is analyzed by mathematical means, the periodic load changes in the system are captured and quantified, and the cycle-driven self-growing nodes are constructed based on the extracted performance adjustment information. In this way, during the automated operation and maintenance process, the system can be accurately periodically tuned according to the cycle-driven self-growing nodes, so as to achieve the purpose of one-time construction and repeated use, and significantly improve the operation and maintenance efficiency.

[0089] S204. Based on the reserved node type general knowledge graph, a system-adaptive knowledge graph is constructed according to the event-driven self-growing nodes and the cycle-driven self-growing nodes.

[0090] The system-adaptive knowledge graph may be a knowledge graph that contains personalized operation and maintenance information within the current system.

[0091] Specifically, after constructing event-driven self-growing nodes and cycle-driven self-growing nodes respectively in the aforementioned manner, the event-driven self-growing nodes and cycle-driven self-growing nodes are respectively added to the reserved node-type general knowledge graph to automatically complete the update and iteration of the knowledge graph according to the personalized operation and maintenance needs of the system, so as to ensure that the automated operation and maintenance work based on the knowledge graph remains stable and efficient under the changing system operation and maintenance needs.

[0092] S205. Obtain continuous operation information, and make floating adjustments to event-driven self-growing nodes and cycle-driven self-growing nodes according to the continuous operation information to determine a highly adaptive knowledge graph.

[0093] The continuous operation information may be fault information and performance information during the continuous operation of the current system, and the continuous operation information may be obtained through the application log monitoring system.

[0094] Floating adjustment can be a process of adjusting the importance of different nodes under the current system operation and maintenance work.

[0095] A highly adaptable knowledge graph can be a knowledge graph that is highly matched with the long-term operation and maintenance requirements of the current system.

[0096] Specifically, when the system is running for a long time, as the operation and maintenance frequency triggered by the event-driven self-growing nodes and the cycle-driven self-growing nodes changes, the degree of influence of the event-driven self-growing nodes and the cycle-driven self-growing nodes on the operation and maintenance of the entire system will also change accordingly. For example, if an event-driven self-growing node is frequently triggered by the automated operation and maintenance process, it means that the system failure corresponding to the event-driven self-growing node is no longer an "occasional failure" but a "highly reproducible failure". Although the high-frequency automated operation and maintenance work performed on the event-driven self-growing node can temporarily solve the failure, it cannot eradicate the failure. , the resources occupied by high-frequency operation and maintenance work are too high, which will directly affect the overall performance of the system. At this time, the importance of the event-driven self-growing node in the current system operation and maintenance work needs to be improved. This is reflected in the knowledge graph, that is, the level of the event-driven self-growing node needs to float upward to remind professional operation and maintenance personnel that the fault corresponding to the event-driven self-growing node requires human intervention, and seek a "once and for all" solution to release the resources occupied by high-frequency operation and maintenance work in the system. By floating adjustments to event-driven self-growing nodes and cycle-driven self-growing nodes, a highly adaptable knowledge graph is obtained, which can further improve the operation and maintenance effect of the system.

[0097] S206. Implement automated operation and maintenance measures based on the highly adaptable knowledge graph and output an operation and maintenance report.

[0098] Automated operation and maintenance measures can be operation and maintenance measures for dealing with system failures and optimizations retrieved from a highly adaptable knowledge graph.

[0099] The operation and maintenance report can be report information that describes the system performance after the automated operation and maintenance is completed.

[0100] Specifically, after obtaining a highly adaptable knowledge graph that is highly matched with the current system operation and maintenance system, through knowledge graph-specific retrieval technology, such as standardized query language or vector retrieval means, according to the operation and maintenance needs of the current system, the corresponding operation and maintenance measures are quickly retrieved from the highly adaptable knowledge graph, and the corresponding operation and maintenance measures are automatically executed through automated script tools to realize automated operation and maintenance of the system, and the corresponding operation and maintenance reports are provided to professional maintenance personnel.

[0101] Through this solution, a general knowledge graph with reserved nodes is constructed based on operation and maintenance data, providing a basic framework for subsequent changes and expansions to the knowledge graph. By analyzing fault detection information, event-driven self-growing nodes are constructed to improve the adaptability of automated operation and maintenance work to different system faults. By analyzing performance monitoring information, cycle-driven self-growing nodes are constructed, so that automated operation and maintenance work can accurately periodically tune the system according to the cycle-driven self-growing nodes. Based on event-driven self-growing nodes and cycle-driven self-growing nodes, a system-adaptive knowledge graph is constructed to automatically complete the update and iteration of the knowledge graph, so that the automated operation and maintenance work based on the knowledge graph remains stable and efficient in the changing system operation and maintenance requirements. According to the continuous operation information, the knowledge graph is further adjusted to obtain a highly adaptable knowledge graph, which improves the matching degree between the knowledge graph and the personalized operation and maintenance requirements of the system, thereby improving the operation and maintenance effect. Finally, based on the highly adaptable knowledge graph, automated operation and maintenance measures are determined and implemented to realize automated operation and maintenance work based on the knowledge graph.

[0102] In some embodiments, the operation and maintenance data is analyzed, entity recognition is performed on the operation and maintenance data, and an operation and maintenance entity node set is determined; based on the operation and maintenance entity node set, entity relationship analysis and extraction are performed on the operation and maintenance data to determine an operation and maintenance entity relationship set; according to the operation and maintenance entity node set, event-driven branch nodes and cycle-driven branch nodes are added to each operation and maintenance entity node to form a reserved node type entity node set; based on the reserved node type entity node set and the operation and maintenance entity relationship set, a reserved node type general knowledge graph is constructed.

[0103] The operation and maintenance entity node set may be a set of knowledge graph nodes that map entities that require operation and maintenance within the system.

[0104] The operation and maintenance entity relationship set may be a set of operation and maintenance relationships between different operation and maintenance entities.

[0105] The event-driven branch node may be a knowledge graph node reserved for each operation and maintenance entity for expanding the event-driven self-growing node.

[0106] The cycle-driven branch node may be a knowledge graph node reserved by each operation and maintenance entity for expanding the cycle-driven self-growing node.

[0107] The reserved node type entity node set may be a set of operation and maintenance entity nodes that complete node reservation.

[0108] Specifically, in the process of constructing the reserved node type general knowledge graph, the entity recognition algorithm in the NPL (Natural Language Processing) library is used to extract the information of the system operation and maintenance entities in the operation and maintenance data to obtain the operation and maintenance entity node set. On this basis, the operation and maintenance relationship between the operation and maintenance entities is extracted through the relationship extraction algorithm in the NPL library to construct an operation and maintenance entity relationship set, and then two blank nodes are added to each operation and maintenance entity node. The two blank nodes are event-driven branch nodes and cycle-driven branch nodes respectively. The blank nodes can be used as placeholders in the current knowledge graph, allowing new nodes and relationships to be easily added in the future without major modifications to the existing knowledge graph structure. This flexibility helps to adapt to rapidly changing or uncertain operation and maintenance needs. According to the reserved node type entity node set and the operation and maintenance entity relationship set, a knowledge representation framework, such as RDF, is used to structure the feature vectors corresponding to the nodes and relationships in the reserved node type entity node set and the operation and maintenance entity relationship set to construct a reserved node type general knowledge graph.

[0109] Through this solution, entity recognition is performed on the operation and maintenance data to obtain an operation and maintenance entity node set, and on this basis, entity relationship extraction is performed on the operation and maintenance data to obtain an operation and maintenance entity relationship set, so as to achieve structured processing of the operation and maintenance data, and then by adding event-driven branch nodes and cycle-driven branch nodes to each operation and maintenance entity node, the reserved node-type general knowledge graph can easily add new nodes and relationships in the future without making major modifications to the existing knowledge graph structure. This flexibility helps to adapt to rapidly changing or uncertain operation and maintenance needs.

[0110] In some embodiments, according to the fault entity, the reserved node type general knowledge graph is analyzed to determine the fault operation and maintenance entity node; according to the fault entity and the abnormal parameters, the entity feature vector and the parameter feature vector are respectively determined; according to the entity feature vector and the parameter feature vector, a fault feature vector is constructed; based on the fault feature vector, a fault similarity analysis is performed on the fault operation and maintenance entity nodes in the reserved node type general knowledge graph to determine whether there are highly similar faults; if there are no highly similar faults, based on the fault entity, an event-driven self-growing node is constructed according to the timestamp, fault feature vector, and maintenance record.

[0111] Fault detection information includes timestamp, fault entity, abnormal parameters, and maintenance records.

[0112] The timestamp may be the time when the current fault occurs in the system.

[0113] The fault entity may be an entity that needs maintenance corresponding to the current fault. The entity is not limited to a specific physical device, but may also be a functional module in the system.

[0114] The abnormal parameter may be a system abnormal parameter caused by the current fault.

[0115] Maintenance records can be records of current fault operation and maintenance measures.

[0116] The fault operation and maintenance entity node may be the operation and maintenance entity node corresponding to the fault entity in the reserved node type general knowledge graph.

[0117] The entity feature vector may be feature vector information corresponding to the fault entity.

[0118] The parameter feature vector may be feature vector information corresponding to the abnormal parameter.

[0119] The fault feature vector may be vector information used to describe comprehensive fault features.

[0120] Specifically, the operation and maintenance entity node corresponding to the fault entity in the reserved node type general knowledge graph is retrieved through the knowledge graph retrieval technology, and is used as the fault operation and maintenance entity node. The fault entity and the abnormal parameters are mapped to the low-dimensional vector space through the word embedding model, such as Word2Vec (word tovector, word vector model), so as to obtain the entity feature vector and the parameter feature vector. The entity feature vector and the parameter feature vector are vector-concatenated to obtain the fault feature vector used to describe the overall characteristics of the fault. The fault feature vector is respectively compared with all the fault feature vectors under the corresponding fault operation and maintenance entity node in the reserved node type general knowledge graph. The similarity between the fault feature vectors is measured through a similarity evaluation algorithm, such as a cosine similarity algorithm, to determine whether there are highly similar faults in the knowledge graph constructed according to the existing operation and maintenance data. If there are no highly similar faults, it means that the current fault is not included in the current knowledge graph. At this time, an event-driven self-growing node for expanding the current reserved node type general knowledge graph is constructed based on the timestamp, fault feature vector and maintenance record used to describe the current fault.

[0121] Through this solution, the entity feature vector and parameter feature vector are determined according to the fault entity and abnormal parameters corresponding to the fault, and on this basis, a fault feature vector for describing the fault characteristics is constructed. According to the fault feature vector, special faults occurring in the system are captured from the fault detection information, and according to the fault entity, timestamp, fault feature vector and maintenance record used to describe the special fault, the event-driven self-growing node is constructed, so that the subsequent knowledge graph can be flexibly expanded according to the special fault, thereby enhancing the adaptability of the knowledge graph in operation and maintenance work.

[0122] In some embodiments, based on the fault entity, according to the timestamp, fault feature vector, and maintenance record, an event-driven self-growing node is constructed, specifically as follows:

[0123] (1)

[0124] in, is the fault entity, is the timestamp, is the entity feature vector, is the parameter eigenvector, To maintain records, is the node constructor, For fault maintenance entity nodes, The corresponding node under the fault operation and maintenance entity node Fault feature vectors, is the fault feature vector, is the preset fault similarity threshold.

[0125] The node constructor may be a function used to integrate various types of information to form a corresponding node.

[0126] The preset fault similarity threshold may be a similarity threshold capable of judging that two faults are highly similar.

[0127] Specifically, through formula (1) , restricting the node construction conditions. When the fault similarity (here the Euclidean distance is used to measure the similarity) between the fault feature vectors corresponding to all fault operation and maintenance entity nodes and the fault feature vector corresponding to the current fault entity is less than the preset fault similarity threshold, it means that the fault corresponding to the current fault entity is not included in the current knowledge graph. At this time, the node constructor in formula (1) is called to construct the corresponding event-driven self-growing node according to the fault entity, timestamp, fault feature vector and maintenance record. The node constructor here can use node construction algorithms such as support vector machine or principal component analysis.

[0128] Through this scheme, mathematical analysis methods are used to construct the construction conditions of event-driven self-growing nodes. On the basis of satisfying the construction conditions, event-driven self-growing nodes are constructed through node constructors according to fault entities, timestamps, fault feature vectors and maintenance records, so as to realize the scientific and automated construction of event-driven self-growing nodes, make event-driven self-growing nodes more accurate, and improve the node update speed of subsequent knowledge graphs.

[0129] In some embodiments, the starting load, fluctuation amplitude and average load are determined based on the load fluctuation information; based on the load entity, the reserved node type general knowledge graph is analyzed to determine the load operation and maintenance entity node; based on the unit time period, the resource scheduling information is determined based on the starting load, fluctuation amplitude and average load; based on the load operation and maintenance entity node and the resource scheduling information, a cycle-driven self-growing node is constructed.

[0130] The performance monitoring information includes unit time period, load entity and load fluctuation information.

[0131] The unit time period may be a unit time period of load fluctuation in the system.

[0132] The load entity may be an operation and maintenance entity corresponding to the load fluctuation in the system.

[0133] The load fluctuation information may be information reflecting load fluctuation characteristics, such as fluctuation amplitude, average load, and the like.

[0134] The starting load may be an initial load value when a periodic load fluctuation occurs in the system.

[0135] The fluctuation amplitude may be the load variation amplitude when the load fluctuates periodically.

[0136] The average load may be an average value of the load during periodic load fluctuations.

[0137] The load operation and maintenance entity node may be the operation and maintenance entity node corresponding to the load entity in the knowledge graph.

[0138] The resource scheduling information may be a system resource scheduling strategy for periodic load fluctuations with the purpose of improving system load performance.

[0139] Specifically, when periodic fluctuations are detected in the system, the collected load fluctuation information is analyzed to extract the starting load, fluctuation amplitude and average load that reflect the load fluctuation characteristics. The starting load can reflect the load performance of the system when the fluctuation occurs, and the fluctuation amplitude can reflect the degree of load fluctuation. The larger the fluctuation, the more resources need to be balanced within the fluctuation period. The average load can reflect the occupation of system resources within the fluctuation period. According to the load entity, the reserved node type general knowledge graph is searched to determine the corresponding load operation and maintenance entity node, and then through mathematical analysis, according to the unit time period, starting load, fluctuation amplitude and average load, the resource scheduling information within the load fluctuation period is obtained. Then, according to the load operation and maintenance entity node and the resource scheduling information, a node construction algorithm such as support vector machine or principal component analysis is used to construct a cycle-driven self-growing node.

[0140] Through this scheme, according to the load fluctuation information, the starting load, fluctuation amplitude and average load that reflect the load fluctuation characteristics are extracted. On this basis, the resource scheduling information within the load fluctuation period is analyzed and obtained. At the same time, according to the load entity, the corresponding load operation and maintenance entity node in the reserved node type general knowledge graph is retrieved, and according to the load operation and maintenance entity node and resource scheduling information, a cycle-driven self-growing node is constructed, so that the node contains the characteristics of load fluctuations and the corresponding scheduling means, further improving the efficiency and effectiveness of subsequent automated operation and maintenance based on the knowledge graph.

[0141] In some embodiments, the resource allocation value for the load operation and maintenance entity node at different time points is determined according to the unit time period, the starting load, the fluctuation amplitude and the average load, which is specifically the following formula (3):

[0142] (3)

[0143] in, For time point The resource allocation value under is the preset adjustment factor, is the average load, is the starting load, is the unit time period, is the fluctuation range, The preset adjustment threshold is used; the resource allocation value of the load operation and maintenance entity node at different time points is used as the resource scheduling information.

[0144] The resource allocation value may be a specific resource value that needs to be allocated to system resources at different time points in order to balance load fluctuations within the system.

[0145] The preset adjustment threshold may be a load fluctuation threshold indicating that automated operation and maintenance intervention is required for resource scheduling.

[0146] Specifically, it can be seen from the description of the above embodiment that the load fluctuation in the system changes periodically. Therefore, through the formula (3) The system's periodic load fluctuations are mathematically modeled, where the sine function can accurately express the periodic characteristics of the load fluctuations. The overall description of the load fluctuation state in the system around the starting load within a unit time period is The conditions for resource scheduling intervention are as follows: when a short-term, small-amplitude load fluctuation occurs in the system, the resource allocation value is 0, indicating that no resource scheduling will be performed. By combining the above load fluctuation model with threshold judgment, the resource allocation value can be accurately quantified.

[0147] Through this solution, mathematical analysis methods are used to mathematically model the system load fluctuation according to the unit time period, starting load, fluctuation amplitude and average load, and the resource scheduling intervention conditions are constrained to accurately quantify the resource scheduling value. By combining the fluctuation model with threshold judgment, maintenance resources are adjusted and allocated in a targeted manner, so that the cycle-driven self-growing nodes can meet the actual load requirements and ensure the stability of operation and maintenance.

[0148] In some embodiments, according to the corresponding fault entity in the event-driven self-growing node, the corresponding operation and maintenance entity node in the reserved node type general knowledge graph is determined; according to the operation and maintenance entity node, the event-driven branch node and the cycle-driven branch node in the operation and maintenance entity node are determined; and according to connecting the event-driven self-growing node and the cycle-driven self-growing node to the event-driven branch node and the cycle-driven branch node respectively, the system-adaptive knowledge graph is determined.

[0149] Specifically, after the event-driven self-growing nodes and the cycle-driven self-growing nodes are constructed, in order to avoid repeated construction of the same nodes, the event-driven self-growing nodes and the cycle-driven self-growing nodes need to be persisted to achieve the purpose of one-time construction and repeated use. The event-driven self-growing nodes and the cycle-driven self-growing nodes are respectively connected to the event-driven branch nodes and cycle-driven branch nodes reserved under the corresponding operation and maintenance entity nodes in the reserved node-type general knowledge graph to achieve persistent processing of the event-driven self-growing nodes and the cycle-driven self-growing nodes.

[0150] In some other implementations, during the construction of a system-adaptive knowledge graph, the operation and maintenance trigger frequency of faults and performance tuning under each operation and maintenance entity in the knowledge graph is collected through an application monitoring system, and the corresponding fault and performance tuning nodes with too low triggering frequency are removed from the current knowledge graph and aggregated into another blank knowledge graph. The traversal priority of the nodes in the current knowledge graph is set to be higher than the traversal priority of the pruned nodes, so as to improve the knowledge graph retrieval speed.

[0151] Through this solution, event-driven self-growing nodes and cycle-driven self-growing nodes are respectively connected to the event-driven branch nodes and cycle-driven branch nodes reserved under the corresponding operation and maintenance entity nodes in the reserved node-type general knowledge graph, so as to realize the persistent processing of event-driven self-growing nodes and cycle-driven self-growing nodes, obtain a system-adaptive knowledge graph, and complete the dynamic update of the knowledge graph.

[0152] In some embodiments, according to the fault node adjustment frequency, the event-driven self-growing nodes in the control system adaptive knowledge graph float upward under the corresponding event-driven branch nodes, and the number of event branch layers after floating of the event-driven self-growing nodes is determined; according to the periodic node adjustment frequency, the cycle-driven self-growing nodes in the control system adaptive knowledge graph float upward under the corresponding cycle-driven branch nodes, and the number of periodic branch layers after floating of the periodic driven self-growing nodes is determined; according to the number of event branch layers after floating and the number of periodic branch layers after floating, the event node conversion strategy and the periodic node conversion strategy are respectively determined; according to the event node conversion strategy and the periodic node conversion strategy, the system adaptive knowledge graph is adjusted to determine a highly adaptive knowledge graph.

[0153] The continuous operation information includes the fault node adjustment frequency and the periodic node adjustment frequency.

[0154] The fault node adjustment frequency may be the frequency at which the event-driven self-growing node is triggered by automated operation and maintenance work.

[0155] The periodic node adjustment frequency may be the frequency at which the periodic driven self-growing node is triggered by automated operation and maintenance work.

[0156] The post-floating event branch layer number may be the layer number of the event-driven branch node under the event-driven branch node after the event-driven self-growing node is float-adjusted.

[0157] The post-floating period branch layer number may be the layer number of the period-driven branch node under the period-driven branch node after the period-driven self-growth node is float-adjusted.

[0158] The event node conversion strategy may be a strategy for performing node property conversion on an event-driven self-growing node.

[0159] The periodic node conversion strategy may be a strategy for performing node property conversion on period-driven self-growing nodes.

[0160] Specifically, when the frequency of event-driven self-growing nodes or cycle-driven self-growing nodes being triggered by automated operation and maintenance work is too high, the nature of the event-driven self-growing nodes or cycle-driven self-growing nodes in the knowledge graph will change. Specifically, when the triggering frequency of an event-driven self-growing node is too high, its nature will change from "occasional" to "highly reproducible". At this time, the fault corresponding to the event-driven self-growing node is no longer an "occasional fault", but tends to be a "bottom-level normal anomaly", and its importance in the entire knowledge graph will be significantly improved. When the triggering frequency of a cycle-driven self-growing node is too high, its nature will change from "periodic" to "normal", and the tuning process corresponding to the cycle-driven self-growing node will not be It is then a "periodic tuning" and tends to a "normal optimization". Its importance in the entire knowledge graph will also be significantly improved. In the knowledge graph, the level of the event-driven self-growing node or the periodic event-driven self-growing node needs to float upward. When the level of the event-driven self-growing node or the periodic event-driven self-growing node reaches the height that requires targeted processing, the properties of the event-driven self-growing node or the periodic event-driven self-growing node need to be converted to corresponding properties. The knowledge graph that has completed the node property conversion will be used as a highly adaptable knowledge graph, so that the highly adaptable knowledge graph can clearly show the event-driven self-growing nodes or periodic event-driven self-growing nodes that need to be processed separately to the maintenance personnel.

[0161] Through this solution, the frequency of faulty nodes is adjusted, and the properties of event-driven self-growing nodes and periodic event-driven self-growing nodes that are triggered too frequently are converted, and the knowledge graph that has completed the node property conversion is used as a highly adaptable knowledge graph, so that the highly adaptable knowledge graph can clearly show maintenance personnel the event-driven self-growing nodes or periodic event-driven self-growing nodes that need to be processed separately and in a targeted manner, further optimize the overall operation and maintenance effect of the system, reduce the continuous resource occupation of automated operation and maintenance work, and thus improve the response speed of the entire system.

[0162] In some embodiments, the number of event branch layers after floating is compared with a preset event layer threshold. If the number of event branch layers after floating is less than the preset event layer threshold, the event node conversion strategy is determined to convert the corresponding event-driven self-growing node into a troubleshooting node; the number of cycle branch layers after floating is compared with a preset cycle layer threshold. If the number of cycle branch layers after floating is less than the preset cycle layer threshold, the cycle node conversion strategy is determined to convert the corresponding cycle-driven self-growing node into a normal optimization node.

[0163] The preset event layer threshold may be the layer threshold that needs to be reached when an event-driven self-growing node undergoes a property transformation. The preset event layer threshold may be set according to the distribution of knowledge graph layers.

[0164] The troubleshooting node can be a node that requires maintenance personnel to make targeted modifications to the underlying system to avoid highly reproducible corresponding faults.

[0165] The preset cycle layer number threshold may be the layer number threshold that needs to be reached when a cycle-driven self-growing node undergoes property transformation. The preset cycle layer number threshold may be set according to the distribution of the knowledge graph layers.

[0166] The normal optimization node may be a node that requires maintenance personnel to make targeted modifications to the underlying system to convert the existing periodic tuning method into the system normal optimization.

[0167] Specifically, the number of event branch layers after floating is compared with the preset event layer threshold. If the number of event branch layers after floating is less than the preset event layer threshold, it means that the "occasional failure" corresponding to the event-driven self-growing node has been converted into a "highly reproducible" failure. The failure corresponding to the event-driven self-growing node is no longer suitable for troubleshooting using automated operation and maintenance work. Maintenance personnel are required to intervene to modify the underlying operating logic of the system to eliminate the "reproducibility" of the failure and release the system resources that are continuously occupied by the automated operation and maintenance work. Professional maintenance personnel are reminded by converting the corresponding event-driven self-growing node into a troubleshooting node.

[0168] At the same time, the number of branch layers after floating is compared with the preset cycle layer threshold. If the number of branch layers after floating is less than the preset cycle layer threshold, it means that the "periodic" tuning corresponding to the cycle-driven self-growth node has been converted to "normal tuning". The tuning work corresponding to the cycle-driven self-growth node is no longer suitable for frequent periodic performance using automated operation and maintenance work. Maintenance personnel are required to intervene to perform normalized tuning of the system's underlying operating parameters to release the system resources that are continuously occupied by the tuning work for automated operation and maintenance work. Professional maintenance personnel are reminded by converting the corresponding cycle-driven self-growth node into a normal optimization node.

[0169] Through this solution, according to the comparison results of the post-floating event layer number and the post-floating cycle layer number with their corresponding layer thresholds, the corresponding event-driven self-growing node is converted into a troubleshooting node, or the corresponding cycle-driven self-growing node is converted into a normal optimization node, so as to remind professional maintenance personnel to optimize the system bottom layer for the troubleshooting node and the normal optimization node, so as to release the system resources occupied by the automated operation and maintenance work and improve the overall stability of the system.

[0170] Figure 3 A schematic diagram of the structure of an automated operation and maintenance management system based on a knowledge graph provided in one embodiment of the present application, such as Figure 3 As shown, an automated operation and maintenance management system 300 based on a knowledge graph in this embodiment includes: a general construction module 301, an event analysis module 302, a cycle analysis module 303, an adaptive construction module 304, an adaptive adjustment module 305 and an implementation module 306.

[0171] A general construction module 301 is used to obtain operation and maintenance data, and to construct a reserved node type general knowledge graph based on the operation and maintenance data;

[0172] An event analysis module 302 is used to obtain fault detection information, analyze the fault detection information and the reserved node type general knowledge graph, and determine an event-driven self-growing node;

[0173] The cycle analysis module 303 is used to obtain performance monitoring information, analyze the performance monitoring information and the reserved node type general knowledge graph, and determine the cycle-driven self-growing node;

[0174] An adaptive construction module 304 is used to construct a system adaptive knowledge graph based on the reserved node type general knowledge graph according to the event-driven self-growing nodes and the cycle-driven self-growing nodes;

[0175] The adaptation adjustment module 305 is used to obtain continuous operation information, and according to the continuous operation information, perform floating adjustment on the event-driven self-growing node and the cycle-driven self-growing node to determine a highly adaptive knowledge graph;

[0176] The implementation module 306 is used to implement automated operation and maintenance measures according to the highly adaptable knowledge graph and output an operation and maintenance report.

[0177] Optionally, the general building module 301 is specifically used for:

[0178] Analyze the operation and maintenance data, perform entity recognition on the operation and maintenance data, and determine an operation and maintenance entity node set;

[0179] Based on the operation and maintenance entity node set, performing entity relationship analysis and extraction on the operation and maintenance data to determine an operation and maintenance entity relationship set;

[0180] According to the operation and maintenance entity node set, adding an event-driven branch node and a cycle-driven branch node to each operation and maintenance entity node to form a reserved node type entity node set;

[0181] The reserved node type general knowledge graph is constructed according to the reserved node type entity node set and the operation and maintenance entity relationship set.

[0182] Optionally, the event analysis module 302 is specifically used to:

[0183] According to the fault entity, analyzing the reserved node type general knowledge graph to determine the fault operation and maintenance entity node;

[0184] According to the fault entity and the abnormal parameter, respectively determine an entity feature vector and a parameter feature vector;

[0185] Constructing a fault feature vector according to the entity feature vector and the parameter feature vector;

[0186] Based on the fault feature vector, a fault similarity analysis is performed on the fault operation and maintenance entity node in the reserved node type general knowledge graph to determine whether there is a highly similar fault;

[0187] If the highly similar fault does not exist, the event-driven self-growing node is constructed based on the fault entity, according to the timestamp, the fault feature vector, and the maintenance record.

[0188] Optionally, when the event analysis module 302 constructs the event-driven self-growing node based on the fault entity, according to the timestamp, the fault feature vector, and the maintenance record, the specific formula is as follows:

[0189] ;

[0190] in, is the fault entity, is the timestamp, is the entity feature vector, is the parameter eigenvector, For said maintenance records, is the node constructor, is the fault operation and maintenance entity node, The corresponding node under the fault operation and maintenance entity node Fault feature vectors, is the fault feature vector, is the preset fault similarity threshold.

[0191] Optionally, the period analysis module 303 is specifically used for:

[0192] Determining a starting load, a fluctuation amplitude, and an average load according to the load fluctuation information;

[0193] According to the load entity, analyzing the reserved node type general knowledge graph to determine the load operation and maintenance entity node;

[0194] Determining resource scheduling information based on the unit time period, according to the starting load, the fluctuation amplitude and the average load;

[0195] The cycle-driven self-growing node is constructed according to the load operation and maintenance entity node and the resource scheduling information.

[0196] Optionally, when the period analysis module 303 determines the resource scheduling information based on the unit time period, according to the starting load, the fluctuation amplitude and the average load, it includes:

[0197] According to the unit time period, the starting load, the fluctuation amplitude and the average load, the resource allocation value for the load operation and maintenance entity node at different time points is determined, specifically as follows:

[0198] ;

[0199] in, For time point The resource allocation value under is the preset adjustment factor, is the average load, is the starting load, is the unit time period, is the fluctuation range, Adjust thresholds for presets;

[0200] The resource allocation value for the load operation and maintenance entity node at different time points is used as the resource scheduling information.

[0201] Optionally, the adaptation construction module 304 is specifically used to:

[0202] According to the corresponding fault entity in the event-driven self-growing node, determine the corresponding operation and maintenance entity node in the reserved node type general knowledge graph;

[0203] According to the operation and maintenance entity node, determining the event-driven branch node and the period-driven branch node in the operation and maintenance entity node;

[0204] The system-adaptive knowledge graph is determined by connecting the event-driven self-growing node and the cycle-driven self-growing node to the event-driven branch node and the cycle-driven branch node respectively.

[0205] Optionally, the adaptation and adjustment module 305 is specifically used to:

[0206] According to the fault node adjustment frequency, the event-driven self-growing node in the system adaptive knowledge graph is controlled to float upward under the corresponding event-driven branch node, and the number of event branch layers after the floating of the event-driven self-growing node is determined;

[0207] According to the periodic node adjustment frequency, the periodic driven self-growing node in the system adaptive knowledge graph is controlled to float upward under the corresponding periodic driven branch node, and the number of periodic branch layers after the floating of the periodic driven self-growing node is determined;

[0208] Determine an event node conversion strategy and a period node conversion strategy respectively according to the number of event branch layers after floating and the number of period branch layers after floating;

[0209] According to the event node conversion strategy and the periodic node conversion strategy, the system adaptive knowledge graph is adjusted to determine the highly adaptive knowledge graph.

[0210] Optionally, when the adaptive adjustment module 305 determines the event node conversion strategy and the period node conversion strategy respectively according to the number of event branch layers after floating and the number of period branch layers after floating, the adaptive adjustment module 305 is specifically used to:

[0211] The post-floating event branch number is compared with a preset event layer number threshold, and if the post-floating event branch number is less than the preset event layer number threshold, the event node conversion strategy is determined to convert the corresponding event-driven self-growing node into a fault elimination node;

[0212] The post-floating cycle branch layer number is compared with a preset cycle layer number threshold. If the post-floating cycle branch layer number is less than the preset cycle layer number threshold, the cycle node conversion strategy is determined to convert the corresponding cycle-driven self-growth node into a normal optimization node.

[0213] The system of this embodiment can be used to execute the method of any of the above embodiments. The implementation principles and technical effects are similar and will not be described in detail here.

Claims

1. An automated operation and maintenance management method based on knowledge graph, characterized in that: include: Obtaining operation and maintenance data, and constructing a reserved node-type general knowledge graph based on the operation and maintenance data; Acquire fault detection information, analyze the fault detection information and the reserved node type general knowledge graph, and determine an event-driven self-growing node; Acquire performance monitoring information, analyze the performance monitoring information and the reserved node type general knowledge graph, and determine the cycle-driven self-growing node; Based on the reserved node type general knowledge graph, a system adaptive knowledge graph is constructed according to the event-driven self-growing nodes and the cycle-driven self-growing nodes; Acquire continuous operation information, and perform floating adjustment on the event-driven self-growing node and the cycle-driven self-growing node according to the continuous operation information to determine a highly adaptive knowledge graph; The floating adjustment represents the process of adjusting the importance of different nodes under the current system operation and maintenance work; Implement automated operation and maintenance measures based on the highly adaptable knowledge graph and output an operation and maintenance report; The fault detection information includes a timestamp, a fault entity, an abnormal parameter, and a maintenance record, and the analyzing the fault detection information and the reserved node type general knowledge graph to determine the event-driven self-growing node includes: According to the fault entity, analyzing the reserved node type general knowledge graph to determine the fault operation and maintenance entity node; According to the fault entity and the abnormal parameter, respectively determine an entity feature vector and a parameter feature vector; Constructing a fault feature vector according to the entity feature vector and the parameter feature vector; Based on the fault feature vector, a fault similarity analysis is performed on the fault operation and maintenance entity node in the reserved node type general knowledge graph to determine whether there is a highly similar fault; If the highly similar fault does not exist, constructing the event-driven self-growing node based on the fault entity, according to the timestamp, the fault feature vector, and the maintenance record; The performance monitoring information includes unit time period, load entity and load fluctuation information, and the analyzing the performance monitoring information and the reserved node type general knowledge graph to determine the period-driven self-growing node includes: Determining a starting load, a fluctuation amplitude, and an average load according to the load fluctuation information; According to the load entity, analyzing the reserved node type general knowledge graph to determine the load operation and maintenance entity node; Determining resource scheduling information based on the unit time period, according to the starting load, the fluctuation amplitude and the average load; The cycle-driven self-growing node is constructed according to the load operation and maintenance entity node and the resource scheduling information.

2. The method according to claim 1, characterized in that The step of constructing a reserved node type general knowledge graph based on the operation and maintenance data includes: Analyze the operation and maintenance data, perform entity recognition on the operation and maintenance data, and determine an operation and maintenance entity node set; Based on the operation and maintenance entity node set, performing entity relationship analysis and extraction on the operation and maintenance data to determine an operation and maintenance entity relationship set; According to the operation and maintenance entity node set, adding an event-driven branch node and a cycle-driven branch node to each operation and maintenance entity node to form a reserved node type entity node set; The reserved node type general knowledge graph is constructed according to the reserved node type entity node set and the operation and maintenance entity relationship set.

3. The method according to claim 1, characterized in that The event-driven self-growing node is constructed based on the fault entity according to the timestamp, the fault feature vector, and the maintenance record, specifically as follows: ; in, is the fault entity, is the timestamp, is the entity feature vector, is the parameter eigenvector, For said maintenance records, is the node constructor, is the fault operation and maintenance entity node, is the i-th fault feature vector corresponding to the fault operation and maintenance entity node, is the fault feature vector, is the preset fault similarity threshold.

4. The method according to claim 1, characterized in that The determining of resource scheduling information based on the unit time period and according to the starting load, the fluctuation amplitude and the average load includes: According to the unit time period, the starting load, the fluctuation amplitude and the average load, the resource allocation value for the load operation and maintenance entity node at different time points is determined, specifically as follows: ； in, is the resource allocation value at time point t, is the preset adjustment factor, is the average load, A is the starting load, P is the unit time period, is the fluctuation range, Adjust thresholds for presets; The resource allocation value for the load operation and maintenance entity node at different time points is used as the resource scheduling information.

5. The method according to claim 2, characterized in that: The method of constructing a system-adaptive knowledge graph based on the reserved node-type general knowledge graph and according to the event-driven self-growing nodes and the cycle-driven self-growing nodes includes: According to the corresponding fault entity in the event-driven self-growing node, determine the corresponding operation and maintenance entity node in the reserved node type general knowledge graph; According to the operation and maintenance entity node, determining the event-driven branch node and the period-driven branch node in the operation and maintenance entity node; The system-adaptive knowledge graph is determined by connecting the event-driven self-growing node and the cycle-driven self-growing node to the event-driven branch node and the cycle-driven branch node respectively.

6. The method according to claim 5, characterized in that The continuous operation information includes a fault node adjustment frequency and a periodic node adjustment frequency. The floating adjustment of the event-driven self-growing node and the periodic-driven self-growing node according to the continuous operation information to determine a highly adaptive knowledge graph includes: According to the fault node adjustment frequency, the event-driven self-growing node in the system adaptive knowledge graph is controlled to float and adjust under the corresponding event-driven branch node, and the number of event branch layers after the floating of the event-driven self-growing node is determined; According to the periodic node adjustment frequency, the periodic driven self-growing node in the system adaptive knowledge graph is controlled to float and adjust under the corresponding periodic driven branch node, and the number of periodic branch layers after the floating of the periodic driven self-growing node is determined; Determine an event node conversion strategy and a period node conversion strategy respectively according to the number of event branch layers after floating and the number of period branch layers after floating; According to the event node conversion strategy and the periodic node conversion strategy, the system adaptive knowledge graph is adjusted to determine the highly adaptive knowledge graph.

7. The method according to claim 6, characterized in that The determining, according to the number of event branch layers after floating and the number of period branch layers after floating, respectively, an event node conversion strategy and a period node conversion strategy comprises: The post-floating event branch number is compared with a preset event layer number threshold, and if the post-floating event branch number is less than the preset event layer number threshold, the event node conversion strategy is determined to convert the corresponding event-driven self-growing node into a fault elimination node; The post-floating cycle branch layer number is compared with a preset cycle layer number threshold. If the post-floating cycle branch layer number is less than the preset cycle layer number threshold, the cycle node conversion strategy is determined to convert the corresponding cycle-driven self-growth node into a normal optimization node.

8. An automated operation and maintenance management system based on knowledge graph, characterized in that: The method as claimed in any one of claims 1 to 7 comprises: A general construction module, used to obtain operation and maintenance data, and to construct a reserved node type general knowledge graph based on the operation and maintenance data; An event analysis module, used to obtain fault detection information, analyze the fault detection information and the reserved node type general knowledge graph, and determine an event-driven self-growing node; A cycle analysis module, used to obtain performance monitoring information, analyze the performance monitoring information and the reserved node type general knowledge graph, and determine the cycle-driven self-growing node; An adaptive construction module, used to construct a system adaptive knowledge graph based on the reserved node type general knowledge graph according to the event-driven self-growing nodes and the cycle-driven self-growing nodes; An adaptation adjustment module, used for obtaining continuous operation information, and performing floating adjustment on the event-driven self-growing node and the cycle-driven self-growing node according to the continuous operation information to determine a highly adaptive knowledge graph; An implementation module is used to implement automated operation and maintenance measures based on the highly adaptable knowledge graph and output an operation and maintenance report.

Images