Intelligent operation and maintenance decision system and method

Abstract

The application provides a kind of intelligent operation and maintenance decision system and method, comprising: data acquisition and processing module, for based on distributed acquisition mechanism from multiple source equipment terminal real-time acquisition operation and maintenance data, and to the operation and maintenance data collected are cleaned and standardized processing, obtain multi-source time series data;Data fusion module is used for the fusion of multi-source time series data, and based on fusion result carries out fault prediction, root cause positioning and situation analysis;Decision management module is used for based on analysis result from operation and maintenance knowledge graph matches historical case and generates decision suggestion, and based on visual interface decision suggestion is issued to each level personnel terminal and is displayed;Operation and maintenance optimization module is used for based on the result of issue according to visual interface receives operation and maintenance personnel submitted operation and maintenance disposal information, and based on operation and maintenance disposal information, operation and maintenance decision mechanism is optimized reversely. The accuracy and reliability of operation and maintenance decision are improved.

Description

Technical Field

[0001] This invention relates to the field of information technology operation and maintenance technology, and in particular to an intelligent operation and maintenance decision-making system and method. Background Technology

[0002] Currently, with the rapid development of digital infrastructure, the scale and complexity of government cloud platforms, cross-departmental collaborative systems, and critical information infrastructure are increasing day by day; In the current operation and maintenance management process, the technical architecture of systems built at different times is different, resulting in inconsistent data standards and incompatible interfaces, forming data silos, which makes it difficult to achieve cross-system global monitoring and correlation analysis; Current operations and maintenance work still heavily relies on human experience, with most alarms requiring manual handling. There is a lack of proactive risk prediction and intervention capabilities, making it difficult to meet the business requirements of high reliability and rapid response. In addition, existing intelligent operations and maintenance tools are insufficient in terms of data fusion capabilities, model interpretability, and compliance, and cannot meet the strict requirements for security and control in government and other fields. Therefore, in order to overcome the above-mentioned defects, the present invention provides an intelligent operation and maintenance decision-making system and method. Summary of the Invention

[0003] This invention provides an intelligent operation and maintenance decision-making system and method. It collects and cleans operation and maintenance data from various devices in real time, transforms it into standardized time series data for fusion analysis, thereby achieving early fault prediction, rapid root cause location, and accurate overall situation assessment. Secondly, it generates targeted suggestions based on historical experience matched with the analysis results and distributes them promptly through a visual interface, improving the speed of operation and maintenance response and the accuracy of decision-making. Finally, it receives feedback from operation and maintenance personnel on their actual actions, continuously optimizing the decision-making logic in reverse, thereby continuously enhancing operation and maintenance efficiency and the system's adaptability, and improving the accuracy and reliability of operation and maintenance decisions.

[0004] This invention provides an intelligent operation and maintenance decision-making system, comprising: The data acquisition and processing module is used to collect operation and maintenance data from multiple source device terminals in real time based on a distributed acquisition mechanism, and to clean and standardize the collected operation and maintenance data to obtain multi-source time series data. The data fusion module is used to fuse multi-source time series data and perform fault prediction, root cause localization, and situation analysis based on the fusion results. The decision management module is used to match historical cases from the operation and maintenance knowledge graph based on the analysis results and generate decision suggestions, and then distribute the decision suggestions to the terminals of personnel at all levels for display through a visual interface. The operation and maintenance optimization module is used to receive operation and maintenance handling information submitted by operation and maintenance personnel through a visual interface based on the distributed results, and to perform reverse optimization of the operation and maintenance decision-making mechanism based on the operation and maintenance handling information.

[0005] Preferably, an intelligent operation and maintenance decision-making system includes a data acquisition and processing module, comprising: The interface adaptation unit is used to determine the multi-source device terminals to be collected based on the operation and maintenance decision requirements, allocate communication interfaces to the multi-source device terminals to be collected respectively, and adapt the corresponding communication interfaces according to the transmission protocol of the multi-source device terminals based on the multi-protocol adapter. The data acquisition unit is used for: Access to multi-source device terminals is performed based on protocol adaptation results and communication interfaces, and the operation and maintenance database of multi-source device terminals is logged in based on the access results. The system collects operation and maintenance data from the operation and maintenance database sequentially based on time series data. Simultaneously, it generates timestamp labels based on the time series data and returns the collected operation and maintenance data after marking it with the timestamp labels.

[0006] Preferably, an intelligent operation and maintenance decision-making system includes a data acquisition and processing module, comprising: The data receiving unit is used to collect operation and maintenance data from multi-source device terminals in real time based on a distributed acquisition mechanism, and to receive the collected operation and maintenance data stream from the multi-source device terminals. It also determines the statistical characteristics of the numerical indicator sequence in the operation and maintenance data stream based on a sliding time window, and constructs a dynamic normal data pattern. Data processing unit, used for: The values ​​of the current operation and maintenance data points are compared with the normal data patterns. When the deviation exceeds the adaptive threshold set according to the historical distribution, the current operation and maintenance data points are identified as candidate anomalies. The candidate anomalies are then analyzed based on an unsupervised clustering algorithm to distinguish between occasional noise and correlated anomaly patterns. Candidate outliers that belong to sporadic noise are marked as invalid data, and consecutive candidate outliers that represent correlated outlier patterns are marked as outlier data. Candidate outliers marked as invalid data are deleted, and consecutive candidate outliers marked as anomalous data are smoothly interpolated and replaced with normal data within the preceding and following sliding time windows to obtain an effective operation and maintenance data stream. Based on a unified time benchmark, effective operation and maintenance data streams are time-series aligned and resampled to obtain time series data. Numerical indicators in the time series data are normalized, and textual information is segmented and vectorized to generate standardized multi-source time series data.

[0007] Preferably, an intelligent operation and maintenance decision-making system includes a data processing unit, comprising: The information splitting subunit is used to extract textual information from time series data and to segment the textual information into continuous word units based on a preset word segmentation dictionary and character sequence rules. Vectorized representation defines the unit, used for: Based on the segmented set of word units, the frequency of each word unit in the text information is determined, and the corresponding statistical features are determined based on the frequency of occurrence. Based on statistical features, each word unit is mapped to a numerical vector in a high-dimensional space, and the vectorized representation of each word unit in the text information is obtained based on the mapping result.

[0008] Preferably, an intelligent operation and maintenance decision-making system includes a data fusion module, comprising: The data acquisition unit is used to receive multi-source time series data and extract features from the multi-source time series data to obtain time series feature vectors of different data source characteristics in the multi-source device terminal. The data fusion unit is used for: A neural network model based on the attention mechanism performs dynamic weighting and correlation analysis on temporal feature vectors, and completes the fusion of temporal feature vectors at the feature layer to obtain fused data; The fused data is input into an ensemble learning framework, which includes multiple base models pre-trained based on historical operation and maintenance data. The fused data is analyzed independently based on the base models, and preliminary prediction results of each base model for the fused data are obtained based on the results of the independent analysis. At the same time, the performance characteristics of each base model in historical scenarios are obtained, and weights are dynamically assigned to each base model based on the performance characteristics. Based on the results of dynamic weight allocation, the preliminary prediction results of each base model on the fused data are weighted and decided to obtain the unified judgment characteristics of the fused data. The data analysis unit is used for fault prediction, root cause localization, and situation analysis based on unified judgment characteristics.

[0009] Preferably, an intelligent operation and maintenance decision-making system includes a data analysis unit, comprising: The fault prediction subunit is used to input the time series data in the unified judgment features into the pre-trained prediction model for analysis, obtain the predicted values ​​of key performance indicators at multiple future time points, compare the predicted values ​​with the actual observed values ​​collected in real time, and generate and output a fault warning signal when the actual observed values ​​continuously exceed the preset deviation threshold of the predicted values. Root cause localization subunit, used for: When a fault warning signal is present, the current abnormal indicators and timestamps are extracted from the unified judgment features, and a causal reasoning graph is constructed based on the association rules between historical alarms and performance indicator data, with system components as nodes and causal relationships between indicators as directed edges. Map the current abnormal indicators to the corresponding nodes in the causal inference graph, and perform backpropagation traversal based on the directed edges of the causal inference graph to determine the abnormal contribution of each upstream node. The upstream node with the highest contribution to the anomaly is identified as the root cause node of this anomaly. The situation analysis subunit is used for: Based on the root cause node, multiple dimensions of indicators such as resource utilization, service response time and error rate are extracted from the unified judgment features, and the multidimensional indicators are clustered based on the unsupervised clustering algorithm to obtain data groups of different situation categories. Based on the correspondence between the target values ​​of each dimension indicator in each situation category data group and the preset reference table, the health level and load level of each situation category data group are obtained, and the corresponding situation analysis results are obtained based on the health level and load level.

[0010] Preferably, an intelligent operation and maintenance decision-making system includes a decision management module, comprising: The feature information determination unit is used to extract feature information of the current fault scenario based on the results of fault prediction, root cause localization and situation analysis; The retrieval unit is used to search the operation and maintenance knowledge graph based on feature information to identify historical fault cases similar to the current fault. The target emergency response plan retrieval unit is used to obtain the case number of historical failure cases and retrieve them from the operation and maintenance knowledge graph based on the case number to determine the target emergency response plan associated with the historical failure case number. The decision recommendation determination unit is used to generate decision recommendations based on historical failure cases and target emergency plans. The decision recommendations include handling suggestions, operating procedures and precautions. The structured work order conversion unit is used to convert decision suggestions into structured work orders based on a preset IT service management system. The structured work order includes a fault description, recommended handling solution, responsible personnel and handling time limit. The distribution unit is used to obtain user roles, configure permissions based on user roles, distribute the corresponding structured orders to the corresponding level of visualization interface based on the permission configuration, and provide feedback and correction operations on decision suggestions based on the visualization interface.

[0011] Preferably, an intelligent operation and maintenance decision-making system includes a feature information determination unit, comprising: The target event graph construction subunit is used to construct a target event graph containing devices, services, applications, and topology relationships, using the root cause association nodes determined by the root cause localization results as the central nodes. The mapping subunit is used to map the fault prediction probability to the corresponding nodes and edges of the target event graph, thereby generating a target influence network containing the fault prediction probability. The macro-environment feature vector generation subunit is used to extract raw indicator data representing resource load, business traffic and security status based on the situation analysis results, and process the raw indicator data to generate macro-environment feature vectors. The encoding subunit is used to encode the target influence network into a graph structure, generate graph embedding features, and encode the graph embedding features and macro-environment feature vectors according to the feature encoding layer to generate a scene feature embedding representation with a unified dimension. The calibration subunit is used to obtain the feature recognition format of the operation and maintenance knowledge graph, calibrate the scene feature embedding representation according to the feature recognition format, and generate feature information of the current fault scene based on the calibration results.

[0012] Preferably, an intelligent operation and maintenance decision-making system includes an operation and maintenance optimization module, comprising: The information receiving unit is used to receive operation and maintenance personnel's operation and maintenance handling information in response to the issued decision suggestions based on a visual interface. The operation and maintenance handling information includes confirmation of the suggestions, parameter adjustments, or supplements to the handling plan. Information parsing unit, used for: The operation and maintenance information is analyzed to obtain the final handling steps after confirmation or correction by the operation and maintenance personnel. The final handling steps are then compared with the original decision recommendations to obtain feedback logs of the differences between the final handling steps and the handling steps in the original decision recommendations. Simultaneously, the final handling steps are executed, and the recovery data of the system under maintenance is obtained in real time under the final handling steps; The optimization unit is used to determine the operational effectiveness of the decision path based on feedback logs and recovery data of the system to be maintained under the final disposal steps, and to update the disposal suggestions and associated weights of the corresponding cases in the operational knowledge graph based on the operational effectiveness, thereby completing the reverse optimization of the operational decision mechanism.

[0013] This invention provides an intelligent operation and maintenance decision-making method, comprising: Step 1: Collect operation and maintenance data in real time from multiple source device terminals based on a distributed acquisition mechanism, and clean and standardize the collected operation and maintenance data to obtain multi-source time series data; Step 2: Fusion of multi-source time series data, and fault prediction, root cause localization and situation analysis based on the fusion results; Step 3: Based on the analysis results, match historical cases from the operation and maintenance knowledge graph and generate decision suggestions, and distribute the decision suggestions to the terminals of personnel at all levels for display through a visual interface; Step 4: Based on the distributed results, receive the operation and maintenance information submitted by the operation and maintenance personnel through the visualization interface, and optimize the operation and maintenance decision-making mechanism in reverse based on the operation and maintenance information.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: By collecting and cleaning maintenance data from various devices in real time and transforming it into standardized time series for fusion analysis, we can achieve early prediction of faults, rapid location of root causes, and accurate assessment of the overall situation. Secondly, based on the analysis results, we can match historical experience to generate targeted suggestions and distribute them in a timely manner through a visual interface, thereby improving the speed of maintenance response and the accuracy of decision-making. Finally, we can receive actual handling feedback from maintenance personnel and continuously optimize the decision-making logic in reverse, thereby continuously enhancing maintenance efficiency and the system's adaptability, and improving the accuracy and reliability of maintenance decisions.

[0015] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in this application.

[0016] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0017] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a structural diagram of an intelligent operation and maintenance decision-making system according to an embodiment of the present invention; Figure 2 This is a structural diagram of a data fusion module in an intelligent operation and maintenance decision-making system according to an embodiment of the present invention; Figure 3 This is a flowchart of an intelligent operation and maintenance decision-making method in an embodiment of the present invention. Detailed Implementation

[0018] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0019] Example 1: This example provides an intelligent operation and maintenance decision-making system, such as... Figure 1 As shown, it includes: The data acquisition and processing module is used to collect operation and maintenance data from multiple source device terminals in real time based on a distributed acquisition mechanism, and to clean and standardize the collected operation and maintenance data to obtain multi-source time series data. The data fusion module is used to fuse multi-source time series data and perform fault prediction, root cause localization, and situation analysis based on the fusion results. The decision management module is used to match historical cases from the operation and maintenance knowledge graph based on the analysis results and generate decision suggestions, and then distribute the decision suggestions to the terminals of personnel at all levels for display through a visual interface. The operation and maintenance optimization module is used to receive operation and maintenance handling information submitted by operation and maintenance personnel through a visual interface based on the distributed results, and to perform reverse optimization of the operation and maintenance decision-making mechanism based on the operation and maintenance handling information.

[0020] In this embodiment, the distributed acquisition mechanism refers to a system method that collects data simultaneously from multiple device endpoints to ensure the real-time nature and broad coverage of the data.

[0021] In this embodiment, multi-source device terminals refer to various hardware or software device endpoints that generate operation and maintenance data, such as servers, network devices, and applications.

[0022] In this embodiment, operation and maintenance data refers to various types of information generated during the operation and maintenance process, including logs, performance indicators, alarm signals, etc.

[0023] In this embodiment, multi-source time series data refers to data sequences from different sources arranged in chronological order, and which have been cleaned and standardized.

[0024] In this embodiment, the operations and maintenance knowledge graph refers to a structured knowledge base that contains entities, relationships, and historical event cases in the operations and maintenance field, used to support intelligent decision-making.

[0025] In this embodiment, historical cases refer to past operational failures or events and their corresponding handling solutions and results.

[0026] In this embodiment, the decision recommendation refers to the operation and maintenance guidance or solution generated based on data fusion analysis and knowledge graph matching.

[0027] In this embodiment, the operation and maintenance handling information refers to the feedback information submitted by the operation and maintenance personnel after implementing the decision suggestions, including the handling steps, effect evaluation, etc.

[0028] In this embodiment, the operation and maintenance decision-making mechanism refers to the rules, models, or algorithm system used in the system to generate decision suggestions, which can be iteratively optimized based on feedback.

[0029] The beneficial effects of the above technical solution are as follows: By collecting and cleaning operation and maintenance data from various devices in real time and converting it into standardized time series for fusion analysis, it is possible to achieve early prediction of faults, rapid location of root causes, and accurate assessment of the overall situation. Secondly, based on the analysis results, targeted suggestions are generated by matching historical experience and promptly distributed through a visual interface, which improves the speed of operation and maintenance response and the accuracy of decision-making. Finally, by receiving actual handling feedback from operation and maintenance personnel, the decision-making logic is continuously optimized in reverse, thereby continuously enhancing operation and maintenance efficiency and the system's adaptability, and improving the accuracy and reliability of operation and maintenance decisions.

[0030] Example 2: Based on Example 1, this example provides an intelligent operation and maintenance decision-making system, including a data acquisition and processing module, comprising: The interface adaptation unit is used to determine the multi-source device terminals to be collected based on the operation and maintenance decision requirements, allocate communication interfaces to the multi-source device terminals to be collected respectively, and adapt the corresponding communication interfaces according to the transmission protocol of the multi-source device terminals based on the multi-protocol adapter. The data acquisition unit is used for: Access to multi-source device terminals is performed based on protocol adaptation results and communication interfaces, and the operation and maintenance database of multi-source device terminals is logged in based on the access results. The system collects operation and maintenance data from the operation and maintenance database sequentially based on time series data. Simultaneously, it generates timestamp labels based on the time series data and returns the collected operation and maintenance data after marking it with the timestamp labels.

[0031] In this embodiment, the communication interface refers to the physical or logical connection point assigned to the device terminal for establishing a data communication link.

[0032] In this embodiment, a multi-protocol adapter refers to a tool or component that can identify and convert different transmission protocols to achieve compatible communication with various device terminals.

[0033] In this embodiment, the operation and maintenance database refers to a structured data storage system stored in the device terminal that contains operation and maintenance related data (such as logs, performance indicators, etc.).

[0034] In this embodiment, time series refers to a sequence of data points arranged in chronological order, used to guide the timing logic of data acquisition.

[0035] In this embodiment, the timestamp label refers to the identification information generated based on the collection time, which is used to mark the specific time point of each piece of operation and maintenance data.

[0036] The beneficial effects of the above technical solution are: by intelligently adapting to multiple communication protocols and interfaces, it ensures efficient and accurate collection of operation and maintenance data from various device terminals. At the same time, it collects data in an orderly manner according to time sequence and automatically adds time stamps, so that the data has clear time sequence and traceability, providing a reliable foundation for subsequent processing and analysis, and improving the automation and standardization level of data collection.

[0037] Example 3: Based on Example 1, this example provides an intelligent operation and maintenance decision-making system, including a data acquisition and processing module, comprising: The data receiving unit is used to collect operation and maintenance data from multi-source device terminals in real time based on a distributed acquisition mechanism, and to receive the collected operation and maintenance data stream from the multi-source device terminals. It also determines the statistical characteristics of the numerical indicator sequence in the operation and maintenance data stream based on a sliding time window, and constructs a dynamic normal data pattern. Data processing unit, used for: The values ​​of the current operation and maintenance data points are compared with the normal data patterns. When the deviation exceeds the adaptive threshold set according to the historical distribution, the current operation and maintenance data points are identified as candidate anomalies. The candidate anomalies are then analyzed based on an unsupervised clustering algorithm to distinguish between occasional noise and correlated anomaly patterns. Candidate outliers that belong to sporadic noise are marked as invalid data, and consecutive candidate outliers that represent correlated outlier patterns are marked as outlier data. Candidate outliers marked as invalid data are deleted, and consecutive candidate outliers marked as anomalous data are smoothly interpolated and replaced with normal data within the preceding and following sliding time windows to obtain an effective operation and maintenance data stream. Based on a unified time benchmark, effective operation and maintenance data streams are time-series aligned and resampled to obtain time series data. Numerical indicators in the time series data are normalized, and textual information is segmented and vectorized to generate standardized multi-source time series data.

[0038] In this embodiment, the sliding time window refers to a dynamic, continuously moving time interval used to calculate and analyze the statistical characteristics of the data stream within that time period in real time.

[0039] In this embodiment, the normal data pattern refers to a numerical feature model that characterizes the normal operating state of the system, calculated based on historical data within a sliding window.

[0040] In this embodiment, candidate outliers refer to data points to be reviewed whose numerical characteristics deviate from the "normal data pattern" by more than a certain threshold.

[0041] In this embodiment, unsupervised clustering algorithm refers to a machine learning algorithm that does not rely on pre-labeling and can automatically discover the inherent grouping structure of data. Here, it is used to classify outliers.

[0042] In this embodiment, the correlation anomaly pattern refers to the set of anomalies identified by an unsupervised clustering algorithm that are similar and continuous in time or features.

[0043] In this embodiment, sporadic noise refers to random anomalous data points that are determined to be isolated and have no clear pattern or correlation.

[0044] In this embodiment, the adaptive threshold refers to a boundary value that is dynamically calculated and adjusted based on the distribution characteristics of historical data, and is used to determine whether the data is abnormal.

[0045] In this embodiment, the effective operation and maintenance data stream refers to a continuous data stream of reliable quality obtained after cleaning processes such as anomaly removal and interpolation repair.

[0046] In this embodiment, a unified time base refers to a common time standard that is coordinated and set for all data sources, used for cross-source data alignment.

[0047] In this embodiment, time alignment refers to the operation of arranging and synchronizing all data according to a "unified time base".

[0048] In this embodiment, resampling refers to the process of converting non-uniform time series data into data with fixed time intervals.

[0049] In this embodiment, normalization refers to a data preprocessing method that scales numerical indicators to a uniform standard range (such as [0,1]).

[0050] In this embodiment, vectorization refers to the process of converting textual information into a numerical vector form that can be processed by a computer.

[0051] In this embodiment, standardized multi-source time series data refers to a time series data set that has been cleaned, aligned, resampled, and numerically transformed, and has a unified format that can be used for direct analysis.

[0052] The beneficial effects of the above technical solution are as follows: by dynamically learning the normal patterns of data and intelligently identifying outliers, and by effectively distinguishing random noise from potential associated faults through cluster analysis, the data is cleaned accurately. Secondly, the abnormal data is smoothed and repaired, and the time stamp and format are unified, transforming the messy raw data into high-quality, standardized time-series data, which greatly improves the reliability of subsequent analysis and prediction.

[0053] Example 4: Based on Example 3, this example provides an intelligent operation and maintenance decision-making system, including a data processing unit, comprising: The information splitting subunit is used to extract textual information from time series data and to segment the textual information into continuous word units based on a preset word segmentation dictionary and character sequence rules. Vectorized representation defines the unit, used for: Based on the segmented set of word units, the frequency of each word unit in the text information is determined, and the corresponding statistical features are determined based on the frequency of occurrence. Based on statistical features, each word unit is mapped to a numerical vector in a high-dimensional space, and the vectorized representation of each word unit in the text information is obtained based on the mapping result.

[0054] In this embodiment, semantically similar word units in high-dimensional space are close in distance in vector space.

[0055] In this embodiment, the word segmentation dictionary refers to a predefined set of professional terms and common terms in the field of operations and maintenance, which is used to guide the accurate segmentation of text.

[0056] In this embodiment, the character sequence rule refers to the rule defined based on character combination rules (such as punctuation and spaces) and used to assist in delineating word boundaries.

[0057] In this embodiment, a word unit refers to the smallest linguistic unit with independent semantics that is segmented from continuous text based on a "word segmentation dictionary" and "character sequence rules".

[0058] In this embodiment, high-dimensional space refers to an abstract mathematical space with a dimension much higher than that of the everyday three-dimensional space. Here, it is used to represent the semantic and statistical features of words through multi-dimensional numerical vectors.

[0059] The beneficial effects of the above technical solution are: by accurately decomposing unstructured text information and transforming it into standardized numerical vectors, it provides a unified, computable, high-quality data input for subsequent intelligent analysis and model processing, thereby improving the usability of text-based operation and maintenance information.

[0060] Example 5: Based on Example 1, this example provides an intelligent operation and maintenance decision-making system, such as... Figure 2 As shown, the data fusion module includes: The data acquisition unit is used to receive multi-source time series data and extract features from the multi-source time series data to obtain time series feature vectors of different data source characteristics in the multi-source device terminal. The data fusion unit is used for: A neural network model based on the attention mechanism performs dynamic weighting and correlation analysis on temporal feature vectors, and completes the fusion of temporal feature vectors at the feature layer to obtain fused data; The fused data is input into an ensemble learning framework, which includes multiple base models pre-trained based on historical operation and maintenance data. The fused data is analyzed independently based on the base models, and preliminary prediction results of each base model for the fused data are obtained based on the results of the independent analysis. At the same time, the performance characteristics of each base model in historical scenarios are obtained, and weights are dynamically assigned to each base model based on the performance characteristics. Based on the results of dynamic weight allocation, the preliminary prediction results of each base model on the fused data are weighted and decided to obtain the unified judgment characteristics of the fused data. The data analysis unit is used for fault prediction, root cause localization, and situation analysis based on unified judgment characteristics.

[0061] In this embodiment, the time-series feature vector refers to the numerical feature representation extracted from multi-source time-series data that can characterize the data's change over time.

[0062] In this embodiment, the neural network model of the attention mechanism refers to a neural network structure that uses attention computing units, which can automatically assign different importance weights to the input feature vector in order to focus on key information.

[0063] In this embodiment, fused data refers to comprehensive data formed at the feature level after dynamic weighting and correlation analysis of multi-source time-series feature vectors through an attention mechanism.

[0064] In this embodiment, the ensemble learning framework refers to a collaborative machine learning system framework that integrates multiple pre-trained independent models (base models) and combines their outputs through a certain strategy to improve overall performance.

[0065] In this embodiment, the base model refers to a single machine learning model pre-trained based on historical operation and maintenance data and included in the ensemble learning framework. Specifically, it can be a random forest, gradient boosting tree, or other similar models.

[0066] In this embodiment, the preliminary prediction result refers to the initial judgment or output about the fault or situation generated by each base model after independently analyzing the input fused data.

[0067] In this embodiment, performance characteristics refer to the performance indicators and behavioral patterns exhibited by each base model in historical operation and maintenance scenarios, such as accuracy, recall, or the ability to identify specific fault types.

[0068] In this embodiment, dynamic weight allocation refers to the process of calculating and assigning different importance coefficients to each base model in real time during the decision-making process based on the performance characteristics of each base model.

[0069] In this embodiment, the unified judgment feature refers to the final comprehensive analysis feature obtained by weighting and fusing the preliminary prediction results of each base model according to the dynamically allocated weights.

[0070] The beneficial effects of the above technical solution are as follows: by extracting the temporal features of multi-source data and using the attention mechanism for dynamic weighted fusion, the correlation between features is enhanced. The integrated learning framework integrates the independent analysis of multiple base models, dynamically adjusts the weights according to historical performance for weighted decision-making, and generates unified judgment features, thereby significantly improving the accuracy of fault prediction, the precision of root cause location, and the comprehensiveness of situation analysis, making operation and maintenance decisions more intelligent and reliable.

[0071] Example 6: Based on Example 5, this example provides an intelligent operation and maintenance decision-making system, including a data analysis unit, comprising: The fault prediction subunit is used to input the time series data in the unified judgment features into the pre-trained prediction model for analysis, obtain the predicted values ​​of key performance indicators at multiple future time points, compare the predicted values ​​with the actual observed values ​​collected in real time, and generate and output a fault warning signal when the actual observed values ​​continuously exceed the preset deviation threshold of the predicted values. Root cause localization subunit, used for: When a fault warning signal is present, the current abnormal indicators and timestamps are extracted from the unified judgment features, and a causal reasoning graph is constructed based on the association rules between historical alarms and performance indicator data, with system components as nodes and causal relationships between indicators as directed edges. Map the current abnormal indicators to the corresponding nodes in the causal inference graph, and perform backpropagation traversal based on the directed edges of the causal inference graph to determine the abnormal contribution of each upstream node. The upstream node with the highest contribution to the anomaly is identified as the root cause node of this anomaly. The situation analysis subunit is used for: Based on the root cause node, multiple dimensions of indicators such as resource utilization, service response time and error rate are extracted from the unified judgment features, and the multidimensional indicators are clustered based on the unsupervised clustering algorithm to obtain data groups of different situation categories. Based on the correspondence between the target values ​​of each dimension indicator in each situation category data group and the preset reference table, the health level and load level of each situation category data group are obtained, and the corresponding situation analysis results are obtained based on the health level and load level.

[0072] In this embodiment, the predicted value of key performance indicators refers to the estimated value given by the prediction model for the core performance indicators of the system (such as CPU utilization, memory usage, etc.) at a specific point in the future.

[0073] In this embodiment, the preset deviation threshold refers to a pre-set error range or limit used to determine whether the actual observed value deviates significantly from the predicted value.

[0074] In this embodiment, the causal reasoning graph refers to a system model represented graphically, where nodes represent system components and directed edges represent causal relationships between indicators or components.

[0075] In this embodiment, the abnormal contribution refers to the quantified value of the degree of influence of an upstream node on the current downstream abnormal node in the causal inference graph.

[0076] In this embodiment, the root cause node refers to the system component node that is determined to have the highest anomaly contribution in the causal reasoning graph and is the starting point of the current anomaly chain.

[0077] In this embodiment, multidimensional metrics refer to multiple parameters that measure the system status from different perspectives, such as resource utilization, service response time, and error rate.

[0078] In this embodiment, the situation category data group refers to the data group with similar characteristics formed after analyzing multidimensional indicators through an unsupervised clustering algorithm.

[0079] In this embodiment, the preset reference table refers to a predefined table that describes the correspondence between different indicator value ranges and specific states (such as healthy and sub-healthy).

[0080] In this embodiment, the health level refers to the level (such as excellent, good, medium, poor) determined after evaluating the stability and reliability of the system or component based on a "preset reference table".

[0081] In this embodiment, the load level refers to the level (such as light load, normal, heavy load) determined after assessing the current working pressure or resource consumption of the system or component based on a "preset reference table".

[0082] The beneficial effects of the above technical solution are: by predictive models, potential faults can be detected and warned in advance; when anomalies occur, the root cause of the problem can be accurately located through causal reasoning; at the same time, the overall operating status of the system can be evaluated by comprehensively assessing multi-dimensional indicators, clarifying its health status and load pressure, thereby achieving foresight, rapid location and overall control of operation and maintenance risks, and significantly improving the initiative and decision-making efficiency of operation and maintenance.

[0083] Example 7: Based on Example 1, this example provides an intelligent operation and maintenance decision-making system, including a decision management module, comprising: The feature information determination unit is used to extract feature information of the current fault scenario based on the results of fault prediction, root cause localization and situation analysis; The retrieval unit is used to search the operation and maintenance knowledge graph based on feature information to identify historical fault cases similar to the current fault. The target emergency response plan retrieval unit is used to obtain the case number of historical failure cases, and retrieve them from the operation and maintenance knowledge graph based on the case number to determine the target emergency response plan associated with the historical failure case number. The decision recommendation determination unit is used to generate decision recommendations based on historical failure cases and target emergency plans. The decision recommendations include handling suggestions, operating procedures and precautions. The structured work order conversion unit is used to convert decision suggestions into structured work orders based on a preset IT service management system. The structured work order includes a fault description, recommended handling solution, responsible personnel, and handling time limit. The distribution unit is used to obtain user roles, configure permissions based on user roles, distribute the corresponding structured orders to the corresponding level of visualization interface based on the permission configuration, and provide feedback and correction operations on decision suggestions based on the visualization interface.

[0084] In this embodiment, the feature information refers to the data set extracted from the results of fault prediction, root cause localization and situational analysis, which is used to describe the key attributes of the current fault scenario.

[0085] In this embodiment, the case number refers to the unique identifier assigned to each historical fault case, which is used to accurately locate and associate relevant information in the operation and maintenance knowledge graph.

[0086] In this embodiment, the target emergency plan refers to a pre-defined standardized emergency handling and recovery plan associated with a specific historical failure case through a case number.

[0087] In this embodiment, a structured task order refers to a standardized task document that is converted from decision recommendations according to a preset template and contains structured fields such as fault description, recommended handling plan, responsible personnel and handling time limit.

[0088] In this embodiment, user roles refer to different personnel identity categories in the operation and maintenance system based on responsibilities and permissions, such as administrators, engineers, etc.

[0089] In this embodiment, permission configuration refers to a set of rules set according to user roles to control system functions, data access, and operation scope.

[0090] In this embodiment, the visual interface refers to a graphical display interface used to present decision suggestions, structured work orders, and provide human-computer interaction for feedback and correction operations.

[0091] The beneficial effects of the above technical solution are as follows: by extracting the feature information of the current fault scenario, intelligently matching similar historical cases and related emergency plans in the operation and maintenance knowledge graph, automatically generating decision suggestions containing specific operation steps, and converting them into standardized structured work orders, which are accurately distributed to the corresponding interfaces according to personnel roles and permissions, supporting rapid feedback and correction, thereby realizing the automation of fault response, the precision of decision support, and the efficiency of task collaboration, and greatly improving the speed and quality of operation and maintenance processing.

[0092] Example 8: Based on Example 7, this example provides an intelligent operation and maintenance decision-making system, including a feature information determination unit, comprising: The target event graph construction subunit is used to construct a target event graph containing devices, services, applications, and topology relationships, using the root cause association nodes determined by the root cause localization results as the central nodes. The mapping subunit is used to map the fault prediction probability to the corresponding nodes and edges of the target event graph, thereby generating a target influence network containing the fault prediction probability. The macro-environment feature vector generation subunit is used to extract raw indicator data representing resource load, business traffic and security status based on the situation analysis results, and process the raw indicator data to generate macro-environment feature vectors. The encoding subunit is used to encode the target influence network into a graph structure, generate graph embedding features, and encode the graph embedding features and macro-environment feature vectors according to the feature encoding layer to generate a scene feature embedding representation with a unified dimension. The calibration subunit is used to obtain the feature recognition format of the operation and maintenance knowledge graph, calibrate the scene feature embedding representation according to the feature recognition format, and generate feature information of the current fault scene based on the calibration results.

[0093] In this embodiment, the target event graph refers to a network structure graph centered on the root cause association node, which includes related devices, services, applications and their connection relationships.

[0094] In this embodiment, the central node refers to the node that occupies a core position in the target event graph and represents the root cause of the failure.

[0095] In this embodiment, the target influence network refers to a network with quantified influence degree formed by mapping fault prediction probability information to its nodes and edges based on the target event graph.

[0096] In this embodiment, raw indicator data refers to unprocessed initial data that is directly extracted from the situation analysis results and reflects the macro environment such as resource load, business traffic, and security status.

[0097] In this embodiment, the macro-environment feature vector refers to a numerical vector generated after processing the original indicator data, which can comprehensively characterize the overall operating environment status of the system.

[0098] In this embodiment, graph structure encoding refers to a technique that converts graph data, such as target influence networks, into a numerical vector form that can be processed by a computer.

[0099] In this embodiment, graph embedding features refer to numerical vectors that can express the structural information of the target influence network after processing it through graph structure coding technology.

[0100] In this embodiment, the feature encoding layer refers to a processing layer (usually a layer in a neural network) used to transform and fuse features from different sources and forms (such as embedded features and macro-environment feature vectors) into a unified dimensional representation.

[0101] In this embodiment, the scene feature embedding representation refers to a numerical vector with a unified dimension that is output by the feature encoding layer and integrates graph structure information and macroscopic environmental information, which is used to comprehensively characterize the current fault scene.

[0102] In this embodiment, the feature recognition format refers to the specific structure or specification required for the feature data that the operation and maintenance knowledge graph can receive and match.

[0103] The beneficial effects of the above technical solution are: by constructing a network graph centered on the root cause of the fault and incorporating multi-dimensional information such as fault probability and environmental state, a unified and standardized scene feature representation is generated, enabling the system to describe the fault scenario more comprehensively and accurately. This lays a high-quality data foundation for intelligent matching of similar cases in the knowledge base, thereby improving the accuracy of fault diagnosis and decision-making.

[0104] Example 9: Based on Example 1, this example provides an intelligent operation and maintenance decision-making system, including an operation and maintenance optimization module, comprising: The information receiving unit is used to receive operation and maintenance personnel's operation and maintenance handling information in response to the issued decision suggestions based on a visual interface. The operation and maintenance handling information includes confirmation of the suggestions, parameter adjustments, or supplements to the handling plan. Information parsing unit, used for: The operation and maintenance information is analyzed to obtain the final handling steps after confirmation or correction by the operation and maintenance personnel. The final handling steps are then compared with the original decision recommendations to obtain feedback logs of the differences between the final handling steps and the handling steps in the original decision recommendations. Simultaneously, the final handling steps are executed, and the recovery data of the system under maintenance is obtained in real time under the final handling steps; The optimization unit is used to determine the operational effectiveness of the decision path based on feedback logs and recovery data of the system to be maintained under the final disposal steps, and to update the disposal suggestions and associated weights of the corresponding cases in the operational knowledge graph based on the operational effectiveness, thereby completing the reverse optimization of the operational decision mechanism.

[0105] In this embodiment, the operation and maintenance handling information refers to the feedback content submitted by operation and maintenance personnel after confirming the decision suggestions issued by the system, modifying parameters, or supplementing the solution through a visual interface.

[0106] In this embodiment, the feedback log refers to a comparison file that records the specific differences between the final handling steps confirmed or corrected by the operation and maintenance personnel and the handling steps in the original decision suggestions of the system.

[0107] In this embodiment, recovery data refers to the indicator data that reflects the recovery status of the system being maintained, which is obtained in real time from the system during and after the maintenance personnel perform the final disposal steps.

[0108] In this embodiment, the operational effectiveness of the decision path refers to the evaluation of the actual effect of the entire decision-making and action process from the occurrence of a fault to the completion of its handling, based on a comprehensive assessment of feedback logs and recovery data.

[0109] In this embodiment, the association weight refers to an adjustable parameter in the operation and maintenance knowledge graph used to measure the closeness or importance of the relationship between historical cases or between cases and emergency plans.

[0110] The beneficial effects of the above technical solution are: by collecting actual feedback and operational corrections from personnel regarding decision-making suggestions, and by monitoring the system recovery status after handling in real time, practical data is transformed into optimization basis, thereby dynamically correcting case suggestions in the knowledge base and adjusting their importance, realizing continuous iteration and self-improvement of the decision-making mechanism based on real feedback, and forming a closed-loop improvement of operation and maintenance capabilities.

[0111] Example 10: This example provides an intelligent operation and maintenance decision-making method, such as... Figure 3 As shown, it includes: Step 1: Collect operation and maintenance data in real time from multiple source device terminals based on a distributed acquisition mechanism, and clean and standardize the collected operation and maintenance data to obtain multi-source time series data; Step 2: Fusion of multi-source time series data, and fault prediction, root cause localization and situation analysis based on the fusion results; Step 3: Based on the analysis results, match historical cases from the operation and maintenance knowledge graph and generate decision suggestions, and distribute the decision suggestions to the terminals of personnel at all levels for display through a visual interface; Step 4: Based on the distributed results, receive the operation and maintenance information submitted by the operation and maintenance personnel through the visualization interface, and optimize the operation and maintenance decision-making mechanism in reverse based on the operation and maintenance information.

[0112] The beneficial effects of the above technical solution are as follows: By collecting and cleaning operation and maintenance data from various devices in real time and converting it into standardized time series for fusion analysis, it is possible to achieve early prediction of faults, rapid location of root causes, and accurate assessment of the overall situation. Secondly, based on the analysis results, targeted suggestions are generated by matching historical experience and promptly distributed through a visual interface, which improves the speed of operation and maintenance response and the accuracy of decision-making. Finally, by receiving actual handling feedback from operation and maintenance personnel, the decision-making logic is continuously optimized in reverse, thereby continuously enhancing operation and maintenance efficiency and the system's adaptability, and improving the accuracy and reliability of operation and maintenance decisions.

[0113] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. An intelligent operation and maintenance decision-making system, characterized in that, include: The data acquisition and processing module is used to collect operation and maintenance data from multiple source device terminals in real time based on a distributed acquisition mechanism, and to clean and standardize the collected operation and maintenance data to obtain multi-source time series data. The data fusion module is used to fuse multi-source time series data and perform fault prediction, root cause localization, and situation analysis based on the fusion results. The decision management module is used to match historical cases from the operation and maintenance knowledge graph based on the analysis results and generate decision suggestions, and then distribute the decision suggestions to the terminals of personnel at all levels for display through a visual interface. The operation and maintenance optimization module is used to receive operation and maintenance handling information submitted by operation and maintenance personnel through a visual interface based on the distributed results, and to perform reverse optimization of the operation and maintenance decision-making mechanism based on the operation and maintenance handling information.

2. The intelligent operation and maintenance decision-making system according to claim 1, characterized in that, The data acquisition and processing module includes: The interface adaptation unit is used to determine the multi-source device terminals to be collected based on the operation and maintenance decision requirements, allocate communication interfaces to the multi-source device terminals to be collected respectively, and adapt the corresponding communication interfaces according to the transmission protocol of the multi-source device terminals based on the multi-protocol adapter. The data acquisition unit is used for: Access to multi-source device terminals is performed based on protocol adaptation results and communication interfaces, and the operation and maintenance database of multi-source device terminals is logged in based on the access results. The system collects operation and maintenance data from the operation and maintenance database sequentially based on time series data. Simultaneously, it generates timestamp labels based on the time series data and returns the collected operation and maintenance data after marking it with the timestamp labels.

3. The intelligent operation and maintenance decision-making system according to claim 1, characterized in that, The data acquisition and processing module includes: The data receiving unit is used to collect operation and maintenance data from multi-source device terminals in real time based on a distributed acquisition mechanism, and to receive the collected operation and maintenance data stream from the multi-source device terminals. It also determines the statistical characteristics of the numerical indicator sequence in the operation and maintenance data stream based on a sliding time window, and constructs a dynamic normal data pattern. Data processing unit, used for: The values ​​of the current operation and maintenance data points are compared with the normal data patterns. When the deviation exceeds the adaptive threshold set according to the historical distribution, the current operation and maintenance data points are identified as candidate anomalies. The candidate anomalies are then analyzed based on an unsupervised clustering algorithm to distinguish between occasional noise and correlated anomaly patterns. Candidate outliers that belong to sporadic noise are marked as invalid data, and consecutive candidate outliers that represent correlated outlier patterns are marked as outlier data. Candidate outliers marked as invalid data are deleted, and consecutive candidate outliers marked as anomalous data are smoothly interpolated and replaced with normal data within the preceding and following sliding time windows to obtain an effective operation and maintenance data stream. Based on a unified time benchmark, effective operation and maintenance data streams are time-series aligned and resampled to obtain time series data. Numerical indicators in the time series data are normalized, and textual information is segmented and vectorized to generate standardized multi-source time series data.

4. The intelligent operation and maintenance decision-making system according to claim 3, characterized in that, The data processing unit includes: The information splitting subunit is used to extract textual information from time series data and to segment the textual information into continuous word units based on a preset word segmentation dictionary and character sequence rules. Vectorized representation defines the unit, used for: Based on the segmented set of word units, the frequency of each word unit in the text information is determined, and the corresponding statistical features are determined based on the frequency of occurrence. Based on statistical features, each word unit is mapped to a numerical vector in a high-dimensional space, and the vectorized representation of each word unit in the text information is obtained based on the mapping result.

5. The intelligent operation and maintenance decision-making system according to claim 1, characterized in that, The data fusion module includes: The data acquisition unit is used to receive multi-source time series data and extract features from the multi-source time series data to obtain time series feature vectors of different data source characteristics in the multi-source device terminal. The data fusion unit is used for: A neural network model based on the attention mechanism performs dynamic weighting and correlation analysis on temporal feature vectors, and completes the fusion of temporal feature vectors at the feature layer to obtain fused data; The fused data is input into an ensemble learning framework, which includes multiple base models pre-trained based on historical operation and maintenance data. The fused data is analyzed independently based on the base models, and preliminary prediction results of each base model for the fused data are obtained based on the results of the independent analysis. At the same time, the performance characteristics of each base model in historical scenarios are obtained, and weights are dynamically assigned to each base model based on the performance characteristics. Based on the results of dynamic weight allocation, the preliminary prediction results of each base model on the fused data are weighted and decided to obtain the unified judgment characteristics of the fused data. The data analysis unit is used for fault prediction, root cause localization, and situation analysis based on unified judgment characteristics.

6. The intelligent operation and maintenance decision-making system according to claim 5, characterized in that, The data analysis unit includes: The fault prediction subunit is used to input the time series data in the unified judgment features into the pre-trained prediction model for analysis, obtain the predicted values ​​of key performance indicators at multiple future time points, compare the predicted values ​​with the actual observed values ​​collected in real time, and generate and output a fault warning signal when the actual observed values ​​continuously exceed the preset deviation threshold of the predicted values. Root cause localization subunit, used for: When a fault warning signal is present, the current abnormal indicators and timestamps are extracted from the unified judgment features, and a causal reasoning graph is constructed based on the association rules between historical alarms and performance indicator data, with system components as nodes and causal relationships between indicators as directed edges. Map the current abnormal indicators to the corresponding nodes in the causal inference graph, and perform backpropagation traversal based on the directed edges of the causal inference graph to determine the abnormal contribution of each upstream node. The upstream node with the highest contribution to the anomaly is identified as the root cause node of this anomaly. The situation analysis subunit is used for: Based on the root cause node, multiple dimensions of indicators such as resource utilization, service response time and error rate are extracted from the unified judgment features, and the multidimensional indicators are clustered based on the unsupervised clustering algorithm to obtain data groups of different situation categories. Based on the correspondence between the target values ​​of each dimension indicator in each situation category data group and the preset reference table, the health level and load level of each situation category data group are obtained, and the corresponding situation analysis results are obtained based on the health level and load level.

7. The intelligent operation and maintenance decision-making system according to claim 1, characterized in that, The decision management module includes: The feature information determination unit is used to extract feature information of the current fault scenario based on the results of fault prediction, root cause localization and situation analysis; The retrieval unit is used to search the operation and maintenance knowledge graph based on feature information to identify historical fault cases similar to the current fault. The target emergency response plan retrieval unit is used to obtain the case number of historical failure cases and retrieve them from the operation and maintenance knowledge graph based on the case number to determine the target emergency response plan associated with the historical failure case number. The decision recommendation determination unit is used to generate decision recommendations based on historical failure cases and target emergency plans. The decision recommendations include handling suggestions, operating procedures and precautions. The structured work order conversion unit is used to convert decision suggestions into structured work orders based on a preset IT service management system. The structured work order includes a fault description, recommended handling solution, responsible personnel and handling time limit. The distribution unit is used to obtain user roles, configure permissions based on user roles, distribute the corresponding structured orders to the corresponding level of visualization interface based on the permission configuration, and provide feedback and correction operations on decision suggestions based on the visualization interface.

8. The intelligent operation and maintenance decision-making system according to claim 7, characterized in that, Feature information determination unit, including: The target event graph construction subunit is used to construct a target event graph containing devices, services, applications, and topology relationships, using the root cause association nodes determined by the root cause localization results as the central nodes. The mapping subunit is used to map the fault prediction probability to the corresponding nodes and edges of the target event graph, thereby generating a target influence network containing the fault prediction probability. The macro-environment feature vector generation subunit is used to extract raw indicator data representing resource load, business traffic and security status based on the situation analysis results, and process the raw indicator data to generate macro-environment feature vectors. The encoding subunit is used to encode the target influence network into a graph structure, generate graph embedding features, and encode the graph embedding features and macro-environment feature vectors according to the feature encoding layer to generate a scene feature embedding representation with a unified dimension. The calibration subunit is used to obtain the feature recognition format of the operation and maintenance knowledge graph, calibrate the scene feature embedding representation according to the feature recognition format, and generate feature information of the current fault scene based on the calibration results.

9. The intelligent operation and maintenance decision-making system according to claim 1, characterized in that, The operation and maintenance optimization module includes: The information receiving unit is used to receive operation and maintenance personnel's operation and maintenance handling information in response to the issued decision suggestions based on a visual interface. The operation and maintenance handling information includes confirmation of the suggestions, parameter adjustments, or supplements to the handling plan. Information parsing unit, used for: The operation and maintenance information is analyzed to obtain the final handling steps after confirmation or correction by the operation and maintenance personnel. The final handling steps are then compared with the original decision recommendations to obtain feedback logs of the differences between the final handling steps and the handling steps in the original decision recommendations. Simultaneously, the final handling steps are executed, and the recovery data of the system under maintenance is obtained in real time under the final handling steps; The optimization unit is used to determine the operational effectiveness of the decision path based on feedback logs and recovery data of the system to be maintained under the final disposal steps, and to update the disposal suggestions and associated weights of the corresponding cases in the operational knowledge graph based on the operational effectiveness, thereby completing the reverse optimization of the operational decision mechanism.

10. An intelligent operation and maintenance decision-making method, characterized in that, include: Step 1: Collect operation and maintenance data in real time from multiple source device terminals based on a distributed acquisition mechanism, and clean and standardize the collected operation and maintenance data to obtain multi-source time series data; Step 2: Fusion of multi-source time series data, and fault prediction, root cause localization and situation analysis based on the fusion results; Step 3: Based on the analysis results, match historical cases from the operation and maintenance knowledge graph and generate decision suggestions, and distribute the decision suggestions to the terminals of personnel at all levels for display through a visual interface; Step 4: Based on the distributed results, receive the operation and maintenance information submitted by the operation and maintenance personnel through the visualization interface, and optimize the operation and maintenance decision-making mechanism in reverse based on the operation and maintenance information.

Images