Automatic operation and maintenance method and system of power plant monitoring system

Abstract

The invention discloses an automatic operation and maintenance method and system of a power plant monitoring system, and belongs to the technical field of power plant monitoring and operation and maintenance. The method comprises the following steps: firstly, configuring a monitoring object list, and setting log acquisition, detection and early warning rules; secondly, preprocessing the log data, including cleaning, segmenting and grouping; extracting features by using a deep learning model and carrying out anomaly detection; analyzing the exception log, positioning an exception root and mapping the exception root to a source code; and finally, generating an optimization patch and dynamically loading the optimization patch to the monitoring object through a Java hot update technology to realize repair during operation. According to the method, the AI technology and the Java hot update mechanism are integrated, closed-loop operation and maintenance from anomaly detection to automatic repair are achieved, the operation and maintenance efficiency and stability of the power plant monitoring system are remarkably improved, and the problems that a traditional operation and maintenance mode is slow in response and depends on manpower are solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of power plant monitoring and operation and maintenance, in particular to an automatic operation and maintenance method and system of a power plant monitoring system. BACKGROUND

[0002] The power plant plant-level monitoring system needs to collect, monitor and analyze the real-time operation data of each control system such as the boiler, steam turbine and auxiliary system. The traditional operation and maintenance method relies on manual experience and predefined rules, which is inefficient in troubleshooting and difficult to respond to complex abnormalities in a timely manner. With the growth of business scale and the increase of system complexity, simple rules or script-based automation has been difficult to meet the needs.

[0003] In recent years, the concept of AIOps (Artificial Intelligence for IT Operations) has gradually emerged, introducing machine learning into the field of operation and maintenance to automatically learn operation and maintenance rules from large-scale operation and maintenance data (including monitoring indicators, logs and manual processing records). Unlike traditional expert systems that rely on manual rules, AIOps advocates using algorithms to analyze and detect abnormalities in large amounts of operation and maintenance data in real time. Under the AIOps framework, log analysis is a key means: logs record system operation details and can be used for fault detection and root cause analysis. However, the amount of logs from industrial monitoring systems is huge, diverse and complex, making manual analysis difficult. Deep learning has shown good results in log anomaly detection. The core idea is to use neural networks to automatically analyze log sequences and detect and locate faults. The general process of log anomaly detection includes log partitioning, feature extraction, model training and online updating. Recent research shows that sequence-based deep learning models (such as LSTM / CNN) can effectively capture the timing patterns in logs and detect abnormal events. Although existing technologies can detect anomalies at the log level and assist in locating problems, there is still a lack of automated technology to map anomalies to source code and perform real-time repairs. In addition, the hot update technology available in Java applications provides the possibility for such online repairs: hot deployment updates Java classes during application runtime without the need for a restart, making code changes take effect immediately. Using Java Agent and bytecode enhancement technology, class bytecode can be modified during runtime to achieve hot replacement. Based on this, if deep learning log analysis is combined with Java hot patching, intelligent fault location and automatic repair of monitoring systems can be achieved, bringing innovation to operation and maintenance. However, existing literature has not yet provided a customizable AI operation and maintenance method for power plant-level monitoring systems. SUMMARY

[0004] In view of the problems in the prior art, the present application provides an automatic operation and maintenance method and system of a power plant monitoring system, which is developed in Java language, emphasizes Java hot patch mechanism, realizes runtime exception automatic detection and source code repair through deep analysis on source code and logs of a monitoring system such as SIS, and guarantees stable operation and continuous update of the system.

[0005] The present application is realized by the following technical solutions: An automatic operation and maintenance method of a power plant monitoring system, comprising the following steps: Step 1, configuring a monitoring object list for the power plant monitoring system, and configuring log collection rules, detection rules and early warning rules for each monitoring object; Step 2, pre-processing log data of the monitoring object based on the log collection rules thereof; The log data preprocessing comprises data cleaning, and the cleaned log is segmented and grouped in the time dimension; Step 3, extracting features from the preprocessed log data based on a deep learning model, and detecting abnormal events of the monitoring object according to the extracted features; Step 4, analyzing abnormal logs according to the abnormal events, obtaining key features of the abnormal logs, determining abnormal information according to the key features, mapping the abnormal information to corresponding source code, combining context information and dependency analysis, and finally obtaining an abnormal root cause of the monitoring object; Step 5, generating an optimized patch for the monitoring object according to the abnormal root cause, and loading the optimized patch into the monitoring object based on Java hot update technology.

[0006] Preferably, the monitoring object list in step 1 comprises at least one of a power plant level monitoring information system, a distributed control system, an auxiliary control system and a device online diagnosis system.

[0007] Preferably, the log collection rules, detection rules and early warning rules of the monitoring object in step 1 are as follows: The log collection rules comprise collecting log files or / and event sources of the monitoring object. The detection rules comprise selecting an abnormal detection model, setting a feature range of the model type, adjusting algorithm parameters of the abnormal detection model or / and setting an online update function of the abnormal detection model.

[0008] Preferably, the pre-processing of the log data of the monitoring object based on the log collection rules thereof in step 2 comprises: Configuring a Java Agent or a listening component in the power plant monitoring system to obtain source logs of the monitoring object in real time; The source log is converted into a structured log, and then the structured log is data cleaned, including filtering repeated or irrelevant records, filling or removing missing or abnormal fields, and desensitizing sensitive information; The cleaned log is segmented based on the log level or trace identifier, and the log is grouped by timestamp and log identifier.

[0009] Preferably, the deep learning model of step 3 is a sequence-based neural network model for automatically mining time sequence features in logs and detecting anomalies through a multi-layer network.

[0010] Preferably, step 4 analyzes the abnormal log according to the abnormal event, obtains the key features of the abnormal log, determines the abnormal information according to the key features, maps the abnormal information to the corresponding source code, and obtains the abnormal root cause of the monitored object, including: Determine the key features according to the log information of the abnormal event, and determine the abnormal type according to the key features; According to the abnormal information corresponding to the abnormal type, the label information of the abnormal information is retrieved to determine the monitored object that has an abnormality; According to the preset source code mapping relationship, the abnormal information is mapped into the source code, and the calling context and stack information are combined to determine the source code corresponding to the abnormal information; Analyze the source code to determine whether the abnormal information is historical abnormal information; According to the historical abnormal information, a positioning log report containing the abnormal occurrence time, the affected component, the code or the configuration position is generated, and the abnormal root cause of the monitored object is obtained according to the positioning log report.

[0011] Preferably, step 5 generates an optimized patch for the monitored object according to the abnormal root cause, including: According to the abnormal root cause, the corresponding patch template is retrieved, the specific context information in the abnormal root cause is bound to the placeholder in the patch template, a repair code segment is generated, and a Java optimization patch is generated according to all repair code segments.

[0012] Preferably, step 5 loads the optimized patch into the monitored object based on the Java hot update technology, including: Java Agent and bytecode enhancement technology are used to dynamically load Java optimization patches during system runtime to realize online repair, and the target class bytecode is modified in the system running state.

[0013] An automatic operation and maintenance system of a power plant monitoring system, comprising: A configuration module is configured to configure a monitored object list for the power plant monitoring system, and configure log collection rules, detection rules and early warning rules for each monitored object. The log processing module is configured to preprocess log data of the monitoring object based on a log collection rule of the monitoring object. The log data preprocessing includes data cleaning, segmenting the cleaned log, and grouping the segmented log in the time dimension. The detection module is configured to extract features from the preprocessed log data based on a deep learning model and detect abnormal events of the monitoring object according to the extracted features. The abnormal positioning module is configured to analyze abnormal logs according to abnormal events, obtain key features of the abnormal logs, determine abnormal information according to the key features, map the abnormal information to corresponding source code, and finally obtain an abnormal root cause of the monitoring object by combining context information and dependency analysis. The patch module is configured to generate an optimized patch for the monitoring object according to the abnormal root cause and load the optimized patch into the monitoring object based on a Java hot update technology.

[0014] An electronic device comprises: A memory for storing a computer program; A processor for executing the computer program to implement the steps of the power plant monitoring system automatic operation and maintenance method.

[0015] Compared with the prior art, the present application has the following beneficial technical effects: The power plant monitoring system automatic operation and maintenance method provided by the present application integrates log collection rule configuration, log data preprocessing, deep learning model anomaly detection, abnormal root cause positioning, and Java hot update optimized patch deployment, and forms an efficient, intelligent, and automatic operation and maintenance system. First, the power plant monitoring system is finely configured to clearly define the monitoring object and log collection, detection, and early warning rules, ensuring the comprehensiveness and pertinence of monitoring. Then, through data cleaning, segmentation, and grouping preprocessing steps, the quality and usability of log data are improved, laying a solid foundation for subsequent analysis. Deep learning model is used for feature extraction and anomaly detection, which fully utilizes the advantages of AI algorithms in processing complex data patterns, significantly improving the accuracy and efficiency of anomaly detection. Further, by deeply analyzing abnormal logs, the abnormal root cause is accurately located and mapped to the source code level, providing a scientific basis for rapid repair. Finally, with the help of Java hot update technology, the optimized patch is dynamically loaded to realize real-time repair during system operation without the need for downtime maintenance, greatly improving the usability and operation flexibility of the system. This method not only solves the problems of low efficiency and slow response in traditional operation and maintenance methods, but also improves the accuracy and speed of fault handling through intelligent means, providing a strong guarantee for the stable operation and continuous optimization of the power plant monitoring system, and showing significant technical advantages and application value.

[0016] The application also provides an automatic operation and maintenance system of a power plant monitoring system, an electronic device and a computer storage medium, which have all the advantages of the automatic operation and maintenance method of the power plant monitoring system. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some embodiments of the present application, and therefore should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can also be obtained without creative labor on the basis of these drawings.

[0018] Figure 1 Flow chart of the automatic operation and maintenance method of the power plant monitoring system of the present application; Figure 2 Flow chart of the abnormality detection of the present application; Figure 3 Flow chart of the log processing and source code repair of the present application; Figure 4 Flow chart of the automatic deployment of the present application. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solutions and advantages of the embodiments of the present application more clear, the following will combine the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all. The components of the embodiments of the present application described and shown in the drawings can be arranged and designed in various different configurations.

[0020] Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0021] An automatic operation and maintenance method of a power plant monitoring system, comprising the following steps: Step 1, configuring a monitoring object list for the power plant monitoring system, and configuring log collection rules, detection rules and early warning rules for each monitoring object; This step specifies the specific monitoring range and monitoring details of the monitoring system. By configuring the monitoring object list and its corresponding log collection, detection and early warning rules, it is ensured that the system can collect key data in a targeted manner, and timely issue an alarm according to the preset conditions, improve the efficiency and accuracy of monitoring, so that the operation and maintenance personnel can quickly locate the problem and reduce the response time.

[0022] Step 2, pre-process the log data of the monitoring object based on its log collection rules; Log data preprocessing includes data cleaning, segmenting the cleaned logs, and grouping the segmented logs in time dimension; Log data preprocessing is a key link to ensure data quality and accuracy of subsequent analysis. Through data cleaning to remove noise and irrelevant information, segmentation and grouping help to organize data in time and logical order, facilitating subsequent feature extraction and anomaly detection. It improves the efficiency and accuracy of data analysis, providing a more reliable data foundation for anomaly detection.

[0023] Step 3, feature extraction of pre-processed log data based on deep learning model, anomaly event detection of monitoring object according to extracted features; Utilizing the powerful feature extraction capability of deep learning model, key patterns and abnormal features are automatically identified from log data, and then abnormal events are detected. Deep learning model can handle complex nonlinear relationships and capture abnormal patterns that traditional methods cannot find. It improves the sensitivity and accuracy of anomaly detection, reducing false positives and false negatives.

[0024] Step 4, analysis of abnormal logs according to abnormal events, acquisition of key features of abnormal logs, determination of abnormal information according to key features, mapping of abnormal information to corresponding source code, and obtaining of abnormal root cause of monitoring object.

[0025] Through in-depth analysis of abnormal logs, key features are extracted and located to specific abnormal information, and further mapping of abnormalities to source code level is realized, so as to find the root cause of the problem. This process is crucial for quickly fixing problems, realizes accurate positioning of problems, provides clear direction for subsequent repair work, and shortens troubleshooting time.

[0026] Step 5, generating optimization patches for monitoring object according to abnormal root cause, and loading optimization patches into monitoring object based on Java hot update technology.

[0027] This step automatically generates optimization patches according to the located abnormal root cause, and uses Java hot update technology to dynamically load the patches into the running system, without the need to restart the application to realize instant repair of problems. The advantage of this process is that it greatly improves the usability and operation efficiency of the system, reduces the risk of business interruption caused by downtime maintenance. At the same time, Java hot update technology ensures the smoothness and safety of patch loading, ensuring the stable operation of the system.

[0028] Embodiment 1 Referring to Figures 1-4 An automatic operation and maintenance method of a power plant monitoring system, comprising the following steps: Step 1: Data source and access method description.

[0029] Determine the data sources required for the operation and maintenance system and describe the access methods for each data source. Data sources include power plant SIS (Supervisory Information System), DCS (Distributed Control System), auxiliary control system, equipment online diagnosis system, etc.

[0030] Among them, the SIS system as the plant-level real-time monitoring center, through the high-speed network platform from each production control system to collect real-time operation data, including the performance parameters, running status and environmental protection monitoring data of boiler, steam turbine and other units. Specifically, the data acquisition node of SIS is usually deployed with interface machine to obtain real-time data from DCS, DEH (electronic speed regulation), NCS (network control system), power collection system, flue gas emission monitoring system and other equipment through secure isolation gateway. The data collected by the interface machine is transmitted to the real-time database server of SIS through firewall, gateway and other isolation devices. In addition, the system can read PLC and sensor data in real time through OPC, Modbus, IEC61850 and other industrial communication protocols; through SSH / HTTPS and other secure channels to regularly pull or receive control system logs, application logs and alarm information.

[0031] All accessed data are subjected to primary and secondary isolation and encryption authentication mechanism deployed by the system, ensuring consistency with the security protection requirements of power plants (i.e. "safety partition, network dedicated, horizontal isolation, vertical authentication" requirements). Through the above-mentioned way, the operation and maintenance system can comprehensively cover all kinds of operation data of SIS system and related production control system, providing basic data guarantee for subsequent analysis.

[0032] Step 2: Define monitoring requirements and alarm strategies through interface or configuration file, including monitored modules, log collection rules and abnormal threshold, etc. User customization process is described in detail, and the system provides configuration interface or console for operation and maintenance personnel to customize monitoring strategies according to requirements.

[0033] Firstly, the user logs in to the operation and maintenance platform and selects the monitoring object list, including specific units (such as No. 1 unit, No. 2 unit), equipment (such as boiler main steam pump, condenser) or control software modules, etc. Users can filter based on asset directory or topology structure tree to flexibly define the equipment and system to be monitored.

[0034] Secondly, configure the log collection and alarm rules corresponding to the monitoring object. For each monitoring object, the user can specify the log files or event sources to be collected, define the matching rules (such as keywords, regular expressions, log levels, event types, etc.) and alarm conditions (such as threshold, frequency or abnormal pattern). At the same time, the user can also set the data collection frequency and quality requirements. For example, higher sampling rate can be set for critical equipment, and lower frequency can be used for general monitoring points.

[0035] Again, users can adjust the parameters and strategies of the deep learning model. Specifically, it includes: selecting the model type for anomaly detection (such as LSTM-based time series model, CNN-based log sequence model or other custom models); setting the feature range of model input (for example, selecting key operating condition parameters, performance indicators or log event sequences as input); adjusting algorithm parameters (such as window size, anomaly score threshold, detection granularity, etc.); enabling or disabling the online update function of the model, etc. The system saves the above user-configured information to the configuration database as the basis for subsequent log processing and analysis, and supports users to modify and extend the configuration items at any time. Through this customization process, users can customize the monitoring range and monitoring strategy, so that the operation and maintenance system can provide flexible monitoring and early warning functions according to the specific power plant operation characteristics and user needs.

[0036] Step 3: Deploy Java Agent or listener components in the running monitoring system to collect system source logs in real time; Log collection, parsing, structuring, cleaning, partitioning and archiving process. For specified monitoring objects and log rules, the system automatically performs log data collection and processing.

[0037] First, deploy the log collection module or agent at the monitoring object to subscribe or pull the log files of the target device and application in real time, and send the data to the log collection server of the operation and maintenance center. The collected logs can be transmitted to the center in multiple protocols (such as Syslog, Filebeat, Kafka, etc.).

[0038] Then, the original log is parsed. In the parsing process, according to the log format predefined rule or using the log template extraction algorithm, the log text is converted into structured fields (for example: timestamp, level, module name, event ID, parameter value, etc.). At the same time, data cleaning is performed: filtering out duplicate or irrelevant records, filling or removing missing or abnormal fields, and desensitizing sensitive information. To improve the query efficiency of the subsequent, the system performs partitioning on the structured log: usually based on timestamp, log level or trace identifier to segment the log. For example, the log can be time-partitioned by hour or day, or the logs of the same request or the same device can be classified into the same partition. The above partitioning operation can be similar to the "log partitioning" technology described in Huawei Cloud and other log management systems, which divides the log into different groups through timestamp and log identifier.

[0039] In addition, real-time filtering and structuring of logs can be performed in the pre-processing stage of the log pipeline. In the data processing flow, the system uses a filtering module to remove unnecessary logs to reduce storage and indexing overhead, and a structuring module to convert logs into key-value pair format. Finally, the processed log data is archived according to the regulations. Recent logs are stored in high-performance databases or time-series storage for real-time queries; historical logs beyond a certain time period are archived using compression strategies and stored in cold storage (such as distributed file systems or archival databases) for future verification.

[0040] Throughout the process, efficient transmission is achieved through data pipelines and message queues, ensuring that log data can be quickly structured and stored for subsequent analysis and queries.

[0041] Step 4: Use deep learning models to automatically extract and learn features from pre-processed log data, and detect potential abnormal patterns in the data.

[0042] Deep learning model loading, inference, and anomaly detection process. The system inputs pre-processed structured log sequences and real-time sensor data into the deep learning model for analysis.

[0043] In the initialization phase, the system loads pre-trained or configured deep learning models into memory, which can include LSTM, GRU, CNN, or custom neural network structures. After the model is loaded, the system inputs real-time data in batches or streaming to the model. Deep learning models extract features from the original input layer by layer through multi-layer neural network architecture, with different layers responsible for capturing different levels of abstraction in the data. For example, LSTM or GRU models can capture long-term dependencies for time series data; models based on Transformer or convolutional networks can extract context and semantic features for log sequences. The system calculates anomaly scores (or classification results) based on the forward inference results of the model. Specifically, for each time period or log sequence segment to be detected, the model outputs an anomaly confidence or probability, which is compared with a pre-set threshold. When the score exceeds the threshold, it is determined as an abnormal event and an alarm is triggered. In addition to offline inference, online learning mode can also be used: during detection, model parameters are fine-tuned based on new data to adapt to changes. The deep learning module follows the general process of log anomaly detection, including log partitioning, feature extraction, model inference, and online updating. Through the above methods, the system can automatically identify implicit anomalies in complex logs and monitoring data using AI algorithms, improving the sensitivity and accuracy of fault detection. Step 5: When an anomaly is detected, automatically analyze the abnormal log information, extract key features and determine the abnormal information, and map the problem to the corresponding source code file and code line number; Detailed flow of anomaly localization and source code analysis. When the deep learning model or rule detects an anomaly, the system starts the anomaly localization process to determine the source of the fault.

[0044] First, according to the label information of the abnormal event (such as the monitoring object identifier, log file name, process ID, etc.), the corresponding monitoring object and module are retrieved to determine the device or software range where the anomaly occurs. If the anomaly is related to an application program, further analysis of the anomaly log content and call context is performed. For example, the stack trace or error code recorded in the anomaly log can be parsed to associate the anomaly event to a specific service module and code. The system uses the pre-defined source code mapping relationship (such as debug symbols, version control information, etc.) to map the log entries and anomaly traces to the class name, method name, and line number in the source code. Through control flow analysis and dependency graph, the fault localization range is further narrowed down.

[0045] At this time, in combination with the change record or historical fault library, it can be judged whether it is a known problem. After localization is completed, the system generates a localization report, including the time of anomaly occurrence, affected components, corresponding code fragments or configuration file locations, etc. key information to provide a basis for subsequent automatic repair. Through the above process, precise localization from high-level anomaly signals to specific code locations is achieved.

[0046] Step 6: Generate an optimized patch based on the localization results, including modified Java classes or configurations, and verify its correctness.

[0047] Automatic patch generation logic. According to the anomaly root cause and error pattern obtained in step 5, the system automatically calls the patch generation engine.

[0048] The patch generation engine first retrieves the matching patch template in the pre-defined repair rule library. The rule library contains repair strategies and code fragment templates corresponding to common fault types, such as inserting non-null checks for null pointer exceptions (NullPointerException), adding resource closing code for resource leaks, and modifying corresponding parameter values for configuration errors. The system binds the specific context (such as variable name, class name, method signature) in the localization results to the placeholders in the patch template to generate a specific repair code fragment. When multiple repair operations may be required for an anomaly, the system can combine the corresponding patch fragments in the order specified by the rules.

[0049] Finally, the automatic patch generator outputs these code snippets in the form of a patch package, for example, generating a Patch package containing the modified class file or configuration file differences. The basic composition of the patch package is to replace the modified new function or class into the target program. As described in the prior art, the basic principle of hot patching is to load the new function body (or class definition) into the memory of the target process, and modify the entry of the old function to jump to the new function. In this step, the patch generation logic automatically generates a complete patch file that can be used to fix the positioning error according to the above principle.

[0050] Step 7: Based on the Java hot update technology (such as the Instrumentation interface), dynamically load the generated patch during system runtime without restarting the application. Verify the deployment effect and feed the new log data back to the AI model to update the model parameters and continuously optimize the operation and maintenance analysis capability.

[0051] Java hot patch execution flow. For the generated Java patch package, the system uses Java dynamic loading technology for online deployment.

[0052] Specifically, the patch code is compiled and packaged into a JAR file containing the modified class, and the Java Agent is declared in the JAR package manifest (MANIFEST). The system can use the JVM startup loading (javaagent parameter) or runtime attachment (Attach API) method to load the Agent. After the Agent starts, it will obtain the Instrumentation instance passed by the Java virtual machine. Through the interface provided by Instrumentation, the Agent can redefine the class without stopping the JVM: for example, use the redefineClasses(ClassDefinition…) method to replace the original class with the new class definition in the patch. In the actual process, the Agent first locates the target class to be replaced, and then passes the bytecode or new version Class definition in the patch to the Instrumentation.redefineClasses interface to achieve online updating. After replacement, the new code takes effect immediately and fixes the defect.

[0053] After deployment, the monitoring system continues to run and observes the update effect. The system can automatically regression test key functions to confirm whether the repair is successful. The new logs and running data generated after repair will be used to update the deep learning model, continuously iterating and optimizing the model precision, thereby realizing the self-evolution of operation and maintenance intelligence.

[0054] The whole hot patching process includes the steps of starting the Agent to load the patch package, calling Instrumentation to redefine the class, verifying the consistency of the new and old versions, and recording the update results, so as to ensure that the patch is correctly applied and the system remains stable. Thus, the method of the application completes the closed-loop AI operation process from data access to automatic repair deployment, and the abnormality can be located and automatically repaired without shutdown.

[0055] Figure 2 The AI model flowchart of the application is shown, which shows the process of abnormality detection after the log data is partitioned, feature extracted, model trained and online inferences. Figure 4 The system automatic deployment flowchart is shown, which illustrates the process of patch packaging, deployment (which can be real-time or timed) and post-deployment verification.

[0056] The application realizes the full-process AI operation closed loop based on user customization requirements: from log analysis to code repair to automatic deployment, forming an automatic intelligent operation and maintenance system. The method combines the powerful pattern recognition ability of deep learning with Java hot update technology, can identify system abnormalities and locate source code errors in real time, automatically generate and load patches, greatly improving the fault response speed and operation and maintenance efficiency. At the same time, users can configure monitoring strategies and model parameters according to their own needs, improving the flexibility and customizability of the system.

[0057] Embodiment 2 Correspondingly, the application also provides an automatic operation and maintenance system of a power plant monitoring system, comprising: A configuration module is configured to configure a monitoring object list for the power plant monitoring system, and configure log collection rules, detection rules and early warning rules for each monitoring object; A log processing module is configured to preprocess log data of the monitoring object based on the log collection rules of the monitoring object; The log data preprocessing includes data cleaning, and the cleaned log is segmented and grouped in the time dimension; A detection module is configured to extract features from the preprocessed log data based on a deep learning model, and detect abnormal events of the monitoring object according to the extracted features; An abnormality positioning module is configured to analyze abnormal logs according to abnormal events, obtain key features of the abnormal logs, determine abnormal information according to the key features, map the abnormal information to corresponding source code, and finally obtain the abnormal root cause of the monitoring object by combining context information and dependency analysis; A patch module is configured to generate an optimized patch for the monitoring object according to the abnormal root cause, and load the optimized patch into the monitoring object based on Java hot update technology.

[0058] The related parts in the power plant monitoring system automation operation system, the electronic device and the computer readable storage medium provided in the embodiments of the present application can refer to the detailed description of the corresponding parts in the power plant monitoring system automation operation method provided in the embodiments of the present application, which will not be described here. In addition, the parts in the above technical solutions provided in the embodiments of the present application that are consistent with the implementation principles of the corresponding technical solutions in the prior art are not described in detail, so as not to be too redundant.

[0059] It should be noted that in several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of each module is only a logical function division, and actual implementation can have another division manner, for example, multiple modules can be combined or integrated into another device, or some features can be ignored or not executed. The modules described as separate components can be or can not be physically separated, and the components displayed as modules can be one physical unit or multiple physical units, that is, they can be located in one place or distributed to multiple different places. According to actual needs, part or all of the modules can be selected to achieve the purpose of the embodiments of the present application.

[0060] In addition, each module in each embodiment of the present application can be integrated in one processing unit, or each module can exist physically, or two or more modules can be integrated in one unit. The above integrated unit can be realized in the form of hardware or in the form of a software function unit.

[0061] The electronic device provided in the embodiments of the present application includes a memory and a processor, the memory stores a computer program, and the processor implements the steps of the power plant monitoring system automation operation method described in any of the above embodiments when executing the computer program.

[0062] The electronic device provided in another embodiment of the present application can further include: an input port connected with the processor, used for transmitting the multi-modal data collected by the external collection device to the processor; and a display unit connected with the processor, used for displaying the processing result of the processor to the outside world; and a communication module connected with the processor, used for realizing the communication between the electronic device and the outside world. The display unit can be a display panel, a laser scanning display, etc.; the communication mode adopted by the communication module includes but is not limited to mobile high-definition link technology (HML), universal serial bus (USB), high-definition multimedia interface (HDMI), wireless connection (including wireless fidelity technology (WiFi), Bluetooth communication technology, low-power Bluetooth communication technology, communication technology based on IEEE 802.11s).

[0063] The embodiment of the present application provides a computer readable storage medium, the computer readable storage medium stores a computer program, and the computer program is executed by a processor to realize the steps of the power plant monitoring system automatic operation and maintenance method described in any of the above embodiments.

[0064] The related parts of the power plant monitoring system automatic operation and maintenance system, the electronic device and the computer readable storage medium provided by the embodiment of the present application are described in detail in the corresponding part of the power plant monitoring system automatic operation and maintenance method provided by the embodiment of the present application, and will not be described here. In addition, the part of the above technical solution provided by the embodiment of the present application which is consistent with the implementation principle of the corresponding technical solution in the prior art is not described in detail, so as not to be too much redundant.

[0065] The above content only illustrates the technical idea of the present application, and cannot limit the protection scope of the present application. Any modification made according to the technical idea of the present application on the basis of the technical solution falls within the protection scope of the claims of the present application.

Claims

1. An automated operation and maintenance method of a power plant monitoring system, characterized by, The method comprises the following steps: Step 1, configuring a monitoring object list for a power plant monitoring system, and configuring log collection rules, detection rules and early warning rules for each monitoring object; Step 2, pre-processing log data of the monitoring object based on the log collection rules of the monitoring object; The log data preprocessing comprises data cleaning, segmenting the cleaned log, and grouping the segmented log in the time dimension; Step 3, extracting features from the pre-processed log data based on a deep learning model, and detecting abnormal events of the monitoring object according to the extracted features; Step 4, analyzing abnormal logs according to the abnormal events, obtaining key features of the abnormal logs, determining abnormal information according to the key features, mapping the abnormal information to corresponding source code, combining context information and dependency analysis, and finally obtaining the abnormal root cause of the monitoring object; Step 5, generating an optimized patch for the monitoring object according to the abnormal root cause, and loading the optimized patch into the monitoring object based on Java hot update technology. 2.The method of claim 1, wherein The monitoring object list of step 1 comprises at least one of a power plant level monitoring information system, a distributed control system, an auxiliary control system and a device online diagnosis system. 3.The method of claim 1, wherein, The log collection rules, detection rules and early warning rules of the monitoring object of step 1 are as follows: The log collection rules comprise collecting log files or / and event sources of the monitoring object; The detection rules comprise selecting an abnormal detection model, setting a feature range of the model type, adjusting algorithm parameters of the abnormal detection model or / and setting an online update function of the abnormal detection model.

4. The method of claim 1, wherein the method further comprises: The pre-processing of the log data of the monitoring object based on the log collection rules of step 2 comprises: Configuring a Java Agent or a listening component in the power plant monitoring system to obtain source logs of the monitoring object in real time; Converting the source logs into structured logs, and then performing data cleaning on the structured logs, including filtering duplicate or irrelevant records, filling or removing missing or abnormal fields, and desensitizing sensitive information; Segmenting the cleaned logs based on log levels or trace identifiers, and grouping the logs by timestamps and log identifiers.

5. The method of claim 1, wherein the method further comprises: The deep learning model of step 3 is a sequence-based neural network model for automatically mining time sequence features in logs and detecting abnormalities through multiple layers of networks.

6. The method of claim 1, wherein the method further comprises: The analysis of abnormal logs according to abnormal events, the obtaining of key features of the abnormal logs, the determination of abnormal information according to the key features, the mapping of the abnormal information to corresponding source code, and the obtaining of the abnormal root cause of the monitoring object of step 4 comprise: Determining key features according to log information of the abnormal event, and determining an abnormal type according to the key features; Retrieving the monitoring object that has occurred an abnormality according to the abnormal information corresponding to the abnormal type and the label information of the abnormal information; Mapping the abnormal information to the source code according to a preset source code mapping relationship, and determining the source code corresponding to the abnormal information by combining the calling context and stack information; Determining whether the abnormal information is historical abnormal information by analyzing the source code; Generating a positioning log report containing abnormal occurrence time, affected components, code or configuration location according to the historical abnormal information, and obtaining the abnormal root cause of the monitoring object according to the positioning log report.

7. The method of claim 1, wherein the method further comprises: The step 5 generates an optimized patch for the monitoring object according to the abnormal root, and the method comprises the steps of: According to the abnormal root, the corresponding patch template is retrieved, the specific context information in the abnormal root is bound to the placeholder in the patch template, the repair code segment is generated, and the Java optimized patch is generated according to all repair code segments. 8.The method of Claim 7, wherein, The step 5 loads the optimized patch into the monitoring object based on the Java hot update technology, and the method comprises the steps of: The Java Agent and the bytecode enhancement technology are used to dynamically load the Java optimized patch during system running to realize online repair, and the target class bytecode is modified under the system running state.

9. An automatic operation and maintenance system of a power plant monitoring system, characterized by, Comprise: The configuration module is used for configuring the monitoring object list of the power plant monitoring system, and configuring the log collection rule, detection rule and early warning rule for each monitoring object; The log processing module is used for pre-processing the log data of the monitoring object based on the log collection rule of the monitoring object; The log data preprocessing includes data cleaning, and the cleaned log is segmented, and the segmented log is grouped in time dimension; The detection module is used for extracting features from the preprocessed log data based on a deep learning model, and detecting abnormal events of the monitoring object according to the extracted features; The abnormal positioning module is used for analyzing the abnormal log according to the abnormal event, obtaining the key features of the abnormal log, determining the abnormal information according to the key features, mapping the abnormal information to the corresponding source code, combining the context information and the dependency analysis, and finally obtaining the abnormal root of the monitoring object; The patch module is used for generating an optimized patch for the monitoring object according to the abnormal root, and loading the optimized patch into the monitoring object based on the Java hot update technology.

10. An electronic device, comprising: Comprise: Memory for storing computer programs; The processor is used for executing the computer program to realize the steps of the automatic operation and maintenance method of the power plant monitoring system in any one of claims 1-8.

Images