Knowledge graph generation method and system applied to safety monitoring of thermal power plant

Abstract

The present application relates to a kind of knowledge graph generation method and system applied to thermal power plant safety monitoring, the present application can improve data quality and reliability by obtaining original data set and carrying out cleaning and standardization processing, unified time stamp mark and semantic label break the isolated state between data, improve the availability and comparability of data.Feature extraction link comprehensively captures entity and relationship features in thermal power plant safety monitoring scene, provides rich information for in-depth understanding of system operation.Knowledge association analysis creatively incorporates semantic association and time sequence continuity, deeply mines potential relationships behind data, and the generated associated knowledge set has high logic and practicality.Finally, the thermal power plant safety monitoring knowledge graph is constructed, which presents complex information in an intuitive knowledge network structure, can more efficiently support the analysis, decision and early warning of thermal power plant safety monitoring, and improve the intelligent and accurate level of thermal power plant safety monitoring.

Description

Technical Field

[0001] This invention relates to the field of data analysis technology, specifically to a knowledge graph generation method and system for safety monitoring in thermal power plants. Background Technology

[0002] In the current field of thermal power plant safety monitoring, traditional methods are mostly limited to simple analysis of single data sources, lacking systematic integration and in-depth mining of massive and complex data. For raw data from different devices and at different times, they fail to effectively address issues such as inconsistent data formats and semantic inconsistencies, resulting in low data utilization efficiency. When analyzing thermal power plant safety monitoring information, traditional methods often focus only on the isolated characteristics of the entities themselves, ignoring the complex relationships between entities and the dynamic changes of these relationships over time, making it difficult to fully grasp the inherent logic of safe operation in thermal power plants.

[0003] Therefore, how to present various information in the safety monitoring of thermal power plants in an intuitive and efficient way to meet the needs of real-time and accurate decision-making is a technical problem that needs to be solved at present. Summary of the Invention

[0004] In view of this, embodiments of the present invention provide a knowledge graph generation method and system for safety monitoring in thermal power plants.

[0005] This invention provides a method for generating a knowledge graph for safety monitoring in thermal power plants. The method, applied to a knowledge graph generation system, includes at least the following steps: acquiring an original data set corresponding to a thermal power plant safety monitoring scenario; performing data cleaning and standardization on the original data set to obtain a standardized data set with unified timestamps and semantic tags; performing feature extraction on the standardized data set to obtain entity features and inter-entity relationship features in the thermal power plant safety monitoring scenario; performing knowledge association analysis on the entity features and inter-entity relationship features to generate an associated knowledge set containing semantic association and temporal continuity; and generating a thermal power plant safety monitoring knowledge graph based on the associated knowledge set, wherein the thermal power plant safety monitoring knowledge graph constructs a knowledge network structure with entities as nodes and relationship features as edges.

[0006] This invention also provides a knowledge graph generation system, characterized in that it includes a processor, a network module, and a memory; the processor and the memory communicate through the network module, and the processor reads a computer program from the memory and runs it to execute the above-described method.

[0007] This invention also provides a computer-readable storage medium having a computer program stored thereon, the computer program implementing the above-described method when running.

[0008] This invention also provides a computer storage medium storing a computer program that implements the above-described method when running.

[0009] This invention improves data quality and reliability by merging and cleaning the original datasets, and by using unified timestamps and semantic tags to break down data silos and enhance usability and comparability. The feature extraction process comprehensively captures entity and relationship features within the power plant safety monitoring scenario, providing rich information for a deeper understanding of system operation. Knowledge association analysis creatively integrates semantic relevance and temporal continuity, deeply mining the potential connections behind the data, and generating a highly logical and practical set of related knowledge. The final constructed knowledge graph for power plant safety monitoring presents complex information in an intuitive knowledge network structure, more efficiently supporting the analysis, decision-making, and early warning of power plant safety monitoring, and improving the intelligence and precision of power plant safety monitoring. Attached Figure Description

[0010] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a block diagram illustrating a knowledge graph generation system provided in an embodiment of the present invention.

[0012] Figure 2 This is a flowchart of a knowledge graph generation method for safety monitoring in thermal power plants, provided as an embodiment of the present invention. Detailed Implementation

[0013] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0014] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0015] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0016] Figure 1 The diagram shows a block illustration of a knowledge graph generation system 10 provided in an embodiment of the present invention. The knowledge graph generation system 10 in this embodiment can be a server with data storage, transmission, and processing functions, such as... Figure 1 As shown, the knowledge graph generation system 10 includes: a memory 11, a processor 12, a network module 13, and a knowledge graph generation device 20.

[0017] The memory 11, processor 12, and network module 13 are electrically connected directly or indirectly to enable data transmission or interaction. For example, these components can be electrically connected to each other via one or more communication buses or signal lines. The memory 11 stores a knowledge graph generation device 20, which includes at least one software function module that can be stored in the memory 11 in the form of software or firmware. The processor 12 executes various functional applications and data processing by running the software program and modules stored in the memory 11, such as the knowledge graph generation device 20 in this embodiment of the invention, thereby implementing the knowledge graph generation method for safety monitoring of thermal power plants in this embodiment of the invention.

[0018] The memory 11 may be, but is not limited to, Random Access Memory (RAM), Read Only Memory (ROM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), etc. The memory 11 stores programs, which are executed by the processor 12 upon receiving an execution instruction.

[0019] The processor 12 may be an integrated circuit chip with data processing capabilities. The processor 12 described above can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this invention. The general-purpose processor can be a microprocessor or any conventional processor.

[0020] The network module 13 is used to establish a communication connection between the knowledge graph generation system 10 and other communication terminal devices via the network, and to realize the transmission and reception of network signals and data. The aforementioned network signals may include wireless signals or wired signals.

[0021] Understandable. Figure 1 The structure shown is for illustrative purposes only; the knowledge graph generation system 10 may also include more... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown. Figure 1 The components shown can be implemented using hardware, software, or a combination thereof.

[0022] This invention also provides a computer storage medium storing a computer program that implements the above-described method when running.

[0023] Figure 2 A flowchart of a knowledge graph generation method for safety monitoring in thermal power plants, provided by an embodiment of the present invention, is shown. The method steps defined in the process are applied to the knowledge graph generation system 10 and can be implemented by the processor 12, namely steps 110-150.

[0024] Step 110: Obtain the raw data set corresponding to the safety monitoring scenario of the thermal power plant.

[0025] In this embodiment of the invention, to comprehensively understand the relevant information on safety monitoring of thermal power plants, it is necessary to obtain the corresponding raw data set. The raw data set is the foundation for subsequent analysis and processing; the more comprehensive the information it covers, the more accurately the generated knowledge graph can reflect the safety monitoring status of the thermal power plant.

[0026] Optionally, the raw data set includes multiple data units such as equipment operating status records, operation log information, environmental monitoring data, and historical accident reports. The process of acquiring the raw data set corresponding to the safety monitoring scenario of a thermal power plant includes:

[0027] Step 111: Collect equipment operation status records from the distributed control system of the thermal power plant. The equipment operation status records include start-up and shutdown times of different types of equipment, operation mode switching records, and real-time feedback data from sensors.

[0028] In this thermal power plant, a distributed control system records the operational dynamics of various equipment. Collecting equipment operating status records from this system provides a detailed understanding of the equipment's operation. For example, the boiler, as one of the key pieces of equipment in the thermal power plant, directly affects power generation efficiency and safety. By collecting the boiler's operating status logs, its start-up and shutdown times can be obtained, understanding when it starts and stops, which is crucial for rationally planning power generation and equipment maintenance. Simultaneously, operating mode switching records reflect the boiler's transitions under different operating conditions, such as the time and conditions for switching from low-load to high-load operation. Real-time sensor feedback data monitors various boiler parameters, such as temperature and pressure, and can promptly issue alarms if any parameters become abnormal.

[0029] Specifically, the acquisition of equipment operating status records from the distributed control system of the thermal power plant includes:

[0030] Step 1111: Access the historical database interface of the distributed control system of the thermal power plant to obtain the operating status logs of the boiler equipment, turbine equipment, generator equipment, and auxiliary system equipment; perform field parsing processing on the operating status logs to extract the equipment identification field, timestamp field, operating status field, and sensor data field from each log record; classify the operating status fields into status types, dividing the operating status into start-up status, stable operation status, mode switching status, and shutdown status, and record the start and end times of each status as start-up and shutdown times; perform difference comparison processing on the operating status fields of adjacent timestamps to identify the time points when the operating status changes and the type of the changed status, and generate equipment operating mode switching records; perform data extraction processing on the sensor data fields to collect real-time feedback values ​​from temperature sensors, pressure sensors, speed sensors, and vibration sensors, and generate real-time sensor feedback data reflecting the real-time operating status of the equipment; and perform correlation and integration processing on the start-up and shutdown times, the operating mode switching records, and the real-time sensor feedback data to generate equipment operating status records.

[0031] In this thermal power plant, rich equipment operation status logs can be obtained by accessing the historical database interface of the distributed control system. Taking steam turbine equipment as an example, its operation status logs are parsed. The equipment identification field clearly identifies which steam turbine it is, the timestamp field records the specific time the data was generated, and the operation status field indicates the turbine's operating state. The operation status is categorized, such as the turbine's journey from startup to stable operation, then to mode switching, and finally shutdown, with the start and end times of each stage accurately recorded. Comparing operation statuses with adjacent timestamps reveals the switching points in the turbine's operating mode, such as the time when it switches from conventional power generation mode to peak shaving mode. For sensor data fields, temperature sensor readings reflect the turbine's operating temperature, pressure sensor data monitors internal pressure, speed sensors record rotational speed, and vibration sensors detect vibration. Integrating and correlating these data forms a complete record of the steam turbine equipment's operation status.

[0032] Step 112: Extract operation log information from the operation management system. The operation log information includes the operator's identity, operation timestamp, operation object identity, and operation instruction content.

[0033] In the daily operation of a thermal power plant, the operation management system records every operational action. For example, when an operator operates a piece of equipment, the system records the operator's identification, clearly identifying who performed the operation. The operation timestamp records the specific time the operation occurred. The operation object identifier indicates which equipment or system was targeted. The operation instruction details what action the operator performed, such as starting the equipment, adjusting parameters, or shutting it down. By extracting this operation log information, the operational procedures and historical operation data within the thermal power plant can be clearly understood.

[0034] Step 113: Collect environmental monitoring data from the environmental monitoring sensor network. The environmental monitoring data includes temperature field distribution records, humidity change sequences, gas concentration fluctuation data, and vibration frequency sampling values.

[0035] An environmental monitoring sensor network constantly monitors the environmental conditions within a thermal power plant. Temperature field distribution records reflect temperature changes in different areas of the plant, such as temperature differences in different locations within the generator room, which is crucial for assessing equipment heat dissipation and operational stability. Humidity change sequences reveal dynamic changes in air humidity; excessively high or low humidity levels can affect equipment operation. Gas concentration fluctuation data monitors the concentration of harmful gases, such as sulfur dioxide and carbon monoxide, ensuring the health of workers and the safety of equipment. Vibration frequency sampling values ​​detect vibrations in the environment; excessive vibration can damage equipment.

[0036] Step 114: Retrieve historical accident reports from the accident management database. The historical accident reports include the time of the accident, the scope of equipment affected by the accident, a description of the accident phenomenon, and a conclusion of the post-accident cause analysis.

[0037] The accident management database stores information on various accidents that have occurred in the thermal power plant in the past. By retrieving historical accident reports, lessons can be learned from past experiences. For example, in the case of a past equipment overheating accident, the exact time of the accident is clearly stated, the affected equipment is identified, and the description of the accident phenomena is detailed, such as smoke from equipment and a rapid rise in temperature. The post-accident cause analysis conclusions provide an in-depth analysis of the causes of the accident, whether it was due to equipment aging, operational errors, or other factors.

[0038] Step 115: Perform unified data format processing on the equipment operation status record, the operation log information, the environmental monitoring data, and the historical accident report to generate an original data set containing text data units, numerical data units, and time series data units.

[0039] Since data from different sources may have different formats, it is necessary to standardize these data formats to facilitate subsequent processing and analysis. For example, the time format in equipment operation status records may differ from that in operation log information; standardization converts them all to a consistent format. Text data units may originate from operation instructions, accident descriptions, etc.; numerical data units include sensor data, equipment parameters, etc.; and time series data units are related to equipment start-up and shutdown times, operation timestamps, etc. Integrating these data forms a complete raw dataset.

[0040] Step 120: Perform data cleaning and standardization on the original dataset to obtain a standardized dataset with unified timestamps and semantic labels.

[0041] The original dataset may contain some noisy and non-standard data. To improve data quality, cleaning and standardization are necessary. For example, in the data processing of this thermal power plant, cleaning and standardization ensure the accuracy and consistency of the data, providing reliable data support for generating an accurate knowledge graph.

[0042] Optionally, the step of performing data cleaning and standardization on the original dataset to obtain a standardized dataset with unified timestamps and semantic labels includes:

[0043] Step 121: Perform noise filtering on the text data units in the original dataset, deleting duplicate records, invalid data with missing fields exceeding a preset proportion, and garbled content with non-standard format.

[0044] Within the text data units of the original dataset, there may be duplicate records. These records consume storage space and affect data processing efficiency, and therefore need to be deleted. For example, operation log information may contain multiple records of the same operation. Invalid data with missing fields exceeding a preset proportion has no practical value. For instance, if too many key fields are missing in a historical incident report, it is impossible to accurately understand the incident situation, and these should be deleted. Irregularly formatted garbled content may be due to errors in data transmission or storage, and cannot be parsed correctly; it also needs to be removed.

[0045] Step 122: Perform outlier detection processing on the numerical data units in the original dataset. Identify outlier numerical points that deviate from the normal range using the sliding window statistical method, and correct them using the linear interpolation method of adjacent time points.

[0046] Outliers in numerical data units can mislead subsequent analysis and processing, thus requiring detection and correction. In the operational data of thermal power plant equipment, such as temperature sensor data, if the temperature value at a certain point in time deviates significantly from the normal range, it may be due to sensor malfunction or other abnormalities. A sliding window statistical method is used, setting a fixed-length sliding window and traversing it across the time series data array. The mean and standard deviation of data points within each sliding window are calculated, and a normal value range is constructed with the mean as the center and twice the standard deviation as the boundary. If any data point within the window exceeds this range, it is considered an outlier. For these outliers, the values ​​and timestamps of the preceding and following normal data points are extracted. Based on the difference in values ​​and time between these two normal data points, linear interpolation coefficients are calculated. Then, a correction value for the outlier is calculated using linear interpolation, and the corrected value replaces the original outlier.

[0047] In detail, the outlier detection processing of the numerical data units in the original dataset, identifying outlier values ​​that deviate from the normal range using a sliding window statistical method, and correcting them using a linear interpolation method of adjacent time points, includes:

[0048] Step 1221: Arrange the numerical data units into a time series data array according to the timestamp order; set a sliding window of fixed length and perform sliding traversal processing on the time series data array, with each sliding window containing a preset number of consecutive data points; calculate the mean and standard deviation of the data points in each sliding window, and construct a normal value range with the mean as the center and twice the standard deviation as the boundary; detect data points in the sliding window that exceed the normal value range as abnormal value points, and record the timestamp position of the abnormal value points; for each abnormal value point, extract the values ​​and timestamp information of its preceding and following normal data points; calculate the linear interpolation coefficients based on the value difference and time difference between the preceding and following normal data points, and calculate the correction value of the abnormal value point using the linear interpolation formula; replace the original value of the abnormal value point with the correction value to generate the corrected numerical data unit.

[0049] Taking generator speed data as an example, the speed data units are arranged into a time series data array according to timestamp order. A sliding window containing several consecutive data points is set and slid across the array. The mean and standard deviation of the data points within each window are calculated. If a data point in the window exceeds the normal range, the timestamp position of this abnormal value point is recorded. Then, the values ​​and timestamp information of the preceding and following normal data points are extracted. Based on the numerical difference and time difference, linear interpolation coefficients are calculated, and then the correction value for the abnormal value point is calculated. This correction value is used to replace the abnormal value point, resulting in the corrected data unit.

[0050] Step 123: Perform timestamp alignment processing on the time series data units in the original data set, and convert the timestamps of data from different sources into the standard time format of the power plant's global clock system.

[0051] Data from different sources may have different timestamp formats, which can complicate data analysis. For example, the timestamps for equipment operating status records might be from a local clock, while the timestamps for environmental monitoring data might be in a different format. By performing timestamp alignment processing, they are all converted to the standard time format of the power plant's global clock system, ensuring consistency across all data in the time dimension.

[0052] Step 124: Perform semantic tagging on the cleaned text data units, numerical data units, and time series data units, and add semantic tags that reflect the theme of each data unit. The semantic tags include equipment type tags, operation type tags, environmental parameter tags, and accident type tags.

[0053] Semantic tagging makes data more readable and understandable. For data units in equipment operation status records, add equipment type tags such as "boiler equipment" and "steam turbine equipment" according to the equipment type; for operation log information, add operation type tags such as "startup operation" and "parameter adjustment operation" according to the operation instruction content; add environmental parameter tags such as "temperature parameter" and "humidity parameter" to environmental monitoring data; and add accident type tags such as "equipment overheating accident" and "pressure over-limit accident" to historical accident reports.

[0054] Step 125: Associate and integrate the labeled data units according to the timestamp order to generate a standardized data set with temporal continuity and semantic consistency.

[0055] By associating and integrating semantically labeled data units according to their timestamps, different types of data can be linked together over time. For example, at a certain point in time, the equipment operation status record shows that the boiler has started, while the operation log contains corresponding start-up records, and environmental monitoring data also records changes in environmental parameters at that time. Through association and integration, these data are combined to form a standardized dataset with temporal continuity and semantic consistency.

[0056] Step 130: Perform feature extraction processing on the standardized dataset to obtain entity features and inter-entity relationship features in the power plant safety monitoring scenario.

[0057] After acquiring a standardized dataset, valuable features need to be extracted from it. These features will be used to construct a knowledge graph. Feature extraction condenses large amounts of data into key information, facilitating subsequent analysis and processing. For example, in the safety monitoring scenario of a thermal power plant, extracting entity features and inter-entity relationship features allows for a better understanding of the intrinsic connections between equipment, operations, environment, and accidents.

[0058] Preferably, the entity features include equipment entity attribute features, operational behavior attribute features, and environmental parameter attribute features; the entity relationship features include the association features between equipment and operation, the influence features between equipment and the environment, and the causal features between operation and accident; the feature extraction processing of the standardized dataset to obtain the entity features and entity relationship features in the thermal power plant safety monitoring scenario includes:

[0059] Step 131: Perform entity recognition processing on the equipment type label data in the standardized data set, and extract the identification information, model parameters, installation location information and design operation indicators of the equipment entities as equipment entity attribute features.

[0060] In standardized datasets, equipment type label data is crucial for identifying equipment entities. Taking boilers as an example, by performing entity recognition processing on relevant data, the unique identifier of the boiler is extracted from the equipment identification field to determine which specific boiler it is. The boiler's model parameters are obtained from the equipment model field to understand its specifications and performance characteristics. Installation location information determines the boiler's specific installation location within the power plant. Design operating indicators, including rated pressure, temperature, and evaporation capacity, reflect the boiler's designed operating capabilities.

[0061] Step 132: Perform behavioral pattern parsing on the operation type label data in the standardized dataset, and extract the execution subject information, operation object information, operation triggering conditions, and operation duration as operation behavior attribute features.

[0062] For operation type label data, behavioral pattern analysis is performed. For example, for an operation of "starting a generator", by parsing the data, we can extract the execution subject information, that is, which operator performed the operation; the operation object information, which generator was started; the operation triggering condition may be due to increased power demand in the power grid; and the operation duration records the time spent from the start of the operation to the generator's stable operation.

[0063] Step 133: Perform parameter feature extraction processing on the environmental parameter label data in the standardized dataset, and extract the monitoring location information, change cycle characteristics, and synchronous fluctuation relationship with the equipment operating status of the environmental parameters as environmental parameter attribute features.

[0064] Regarding environmental parameter labeling data, taking temperature as an example, we can extract monitoring location information to determine which area of ​​the power plant the temperature was monitored in. The periodicity characteristics of temperature changes can reveal the pattern of temperature variation over time, whether it is periodic or random. The synchronous fluctuation relationship with equipment operating status studies the connection between temperature changes and equipment operating status, such as whether the temperature rises during generator operation.

[0065] Step 134: Perform correlation analysis on the device entity attribute features and the operation behavior attribute features, and extract the description of the degree of influence of operation behavior on the device operating status and the frequency of the co-occurrence of operation behavior and device failure as the correlation features between the device and operation.

[0066] This involves correlation analysis between equipment entity attributes and operational behavior attributes. For example, frequent start-up and shutdown operations cause significant wear and tear on equipment. Data analysis can describe the impact of such operational behaviors on equipment operating status, such as shortened equipment lifespan. Simultaneously, statistical analysis of the frequency of co-occurrence between operational behaviors and equipment failures reveals which operations are more likely to cause equipment malfunctions.

[0067] Step 135: Perform correlation analysis on the physical attributes of the equipment and the attributes of the environmental parameters, and extract the direction and intensity of the influence of environmental parameter changes on equipment performance indicators as the influence features of equipment and environment.

[0068] Analyze the correlation between the physical attributes of the equipment and the attributes of environmental parameters. For example, excessively high ambient temperature may affect the power generation efficiency of a generator. Through correlation analysis, determine whether the impact of temperature changes on generator performance indicators is to reduce efficiency, and describe the intensity of the impact.

[0069] Step 136: Perform causal relationship mining on the operational behavior attribute features and the historical accident report data, and extract the time interval features between abnormal operational behavior and accident occurrence, as well as the correspondence between operational error type and accident type as causal features between operation and accident.

[0070] In the following steps, the causal relationship mining process is performed on the operational behavior attribute features and the historical accident report data to extract the time interval features between abnormal operational behavior and accident occurrences, as well as the correspondence between operational error types and accident types, as causal features between operations and accidents. This includes:

[0071] Step 1361: Extract operation error type information from the operation behavior attribute features, including accidental touch operation, timeout operation, parameter setting error, and process execution omission; extract accident type information from the historical accident report data, including equipment overheating accident, pressure over-limit accident, vibration abnormality accident, and system shutdown accident; perform time alignment processing on the operation error type information and the accident type information, matching the timestamp of each operation error event with the timestamp of subsequent accident events, and calculate the time interval between operation error events and accident events; count the number of times the same operation error type and the same accident type combination occurs, and calculate the frequency of occurrence of the combination as the correlation strength between operation error type and accident type; extract operation error events and accident event combinations with time intervals within a preset range, count the time interval distribution characteristics of the combinations, and generate time interval characteristics between operation behavior anomalies and accident occurrences; integrate the correlation strength and the time interval characteristics to generate a correspondence description between operation error type and accident type, as the causal characteristics between operation and accident.

[0072] In this real-world case study of a thermal power plant, when extracting operational error type information from operational behavior attribute characteristics, for example, during equipment maintenance, a misoperation occurred. The operation that should have shut down auxiliary system A mistakenly shut down auxiliary system B; this falls under the category of misoperation. Timeout operations might occur when starting up large equipment, where the operator fails to complete a step in the startup process within the prescribed time, resulting in a timeout. Parameter setting errors, such as setting generator excitation parameters outside the reasonable range, could occur. Omissions in process execution might happen during equipment maintenance procedures, where the operator overlooks a critical inspection step.

[0073] When extracting accident type information from historical accident report data, equipment overheating accidents may be caused by cooling system failures, leading to a continuous rise in equipment temperature and ultimately causing an accident. Pressure over-limit accidents may be caused by pipe blockages, causing internal pressure to rise continuously and exceed safety thresholds. Abnormal vibration accidents may be caused by a loose component in the equipment, resulting in abnormal vibration during operation. System shutdown accidents may be a chain reaction of multiple equipment failures, ultimately causing the entire power generation system to stop operating.

[0074] When aligning operation error type information with incident type information by time, if an accidental touch occurs at 10:00 AM, and a system shutdown occurs at 10:15 AM, then the time interval between the operation error event and the incident event is 15 minutes. By processing a large amount of similar data, the frequency of the same operation error type and the same incident type combination is statistically analyzed. For example, if the combination of accidental touch and system shutdown occurs 5 times, and a total of 100 records are recorded for this combination, then the frequency of this combination is 5%. This represents the correlation strength between the operation error type and the incident type.

[0075] Extract combinations of operational error events and accident events with time intervals within a preset range (e.g., 0-30 minutes), and statistically analyze the time interval distribution characteristics of these combinations. If there are 10 combinations that meet the preset range, with 3 combinations having time intervals between 0-10 minutes, 4 combinations between 10-20 minutes, and 3 combinations between 20-30 minutes, this generates the time interval characteristics of operational abnormalities and accidents.

[0076] Finally, the correlation strength and time interval characteristics are integrated. For example, it is found that there is a strong correlation between accidental touch operations and system downtime accidents, with a correlation strength of 5%, and the time intervals are mostly concentrated within 0-30 minutes. This generates a correspondence description between operation error types and accident types, serving as a causal feature between operations and accidents.

[0077] Step 140: Perform knowledge association analysis on the entity features and the relationship features between entities to generate a set of associated knowledge that includes semantic association and temporal continuity.

[0078] In the context of safety monitoring in thermal power plants, entity features and inter-entity relationship features have been extracted. The next step is to conduct in-depth analysis of these features to generate a set of related knowledge. Through knowledge association analysis, hidden relationships between equipment, operations, environment, and accidents can be discovered.

[0079] Alternatively, the step of performing knowledge association analysis on the entity features and the relationship features between entities to generate a set of associated knowledge containing semantic relevance and temporal continuity includes:

[0080] Step 141: Perform semantic similarity calculation on the device entity attribute features, the operation behavior attribute features and the environmental parameter attribute features to identify entity attributes with the same or similar semantic descriptions and generate semantic association relationships between entity attributes.

[0081] Data from thermal power plants contains rich semantic information in equipment entity attributes, operational behavior attributes, and environmental parameter attributes. For example, in equipment entity attributes, the "rated power" attribute, although having different specific values, is semantically related to the power capacity of different types of equipment. Semantic similarity calculations can identify these semantically similar attributes. Taking operational behavior attributes as an example, "start-up operation" and "start-up operation" are semantically similar, and their semantic relationship can be determined through calculation. Generating this semantic relationship helps to integrate and understand the connections between different entity attributes.

[0082] Step 142: Perform time-series analysis on the correlation characteristics of the equipment and operation, the influence characteristics of the equipment and environment, and the causal characteristics of the operation and accident, extract the sequential occurrence pattern and duration of the relationship characteristics in the time dimension, and generate the time-series correlation relationship between the relationship characteristics.

[0083] For analyzing the relationships between equipment and operation, equipment and environment, and operation and accidents, the time dimension is crucial. For example, in the relationship between equipment and operation, an operational command is typically issued first, followed by the corresponding change in the equipment's operating state. Time-series analysis determines the specific time point of the equipment's state change after the operational action, as well as the duration of this change. Regarding the impact of equipment on the environment, changes in environmental parameters may affect equipment performance after a certain period; analyzing the temporal sequence and duration of this impact is essential. For the causal characteristics of operation and accidents, it clarifies how long after an operational error it might lead to an accident, and how long the accident's impact would last. Extracting these temporal relationships allows for a better understanding of the evolution of various events and relationships within a thermal power plant over time.

[0084] Step 143: Perform cross-validation on the semantic association and the temporal association, and select the associations that simultaneously satisfy the semantic similarity condition and the temporal regularity condition as valid associations.

[0085] Semantic and temporal associations reveal the connections between entity attributes and relational features from different perspectives, but a single association may be inaccurate or incomplete. Cross-validation combines both. For example, semantically, "equipment maintenance operations" are associated with "reduced equipment failures," but temporally, if no reduction in equipment failures occurs for a long period after each maintenance operation, this association may not be valid. Only associations that simultaneously satisfy both semantic similarity and temporal regularity conditions—such as semantically, "temperature increase operations" are associated with "increased risk of equipment overheating," and temporally, the probability of equipment overheating significantly increases within a certain timeframe after a temperature increase operation—are considered valid associations.

[0086] Step 144: Perform knowledge abstraction processing on the effective association relationship, convert the specific entity attribute features and relationship features into a generalizable knowledge representation form, and generate a knowledge tuple containing entity category, attribute type and relationship type.

[0087] Effective relationships are derived from specific data and analysis, but knowledge abstraction is necessary for easier application and management of this knowledge. For example, the specific relationship "the operating efficiency of a certain boiler equipment decreases under high-temperature conditions" can be transformed into a knowledge tuple like "the relationship type of the equipment class whose operating efficiency attribute decreases under the high-temperature environment attribute." Here, "equipment class" is the entity category, "high-temperature environment attribute" and "operating efficiency attribute" are attribute types, and "decreases" is the relationship type. In this way, specific relationships are transformed into a more general, easier-to-understand, and easier-to-process form of knowledge.

[0088] Step 145: Organize the knowledge tuples hierarchically according to semantic and temporal relationships to generate a hierarchical set of related knowledge.

[0089] Knowledge tuples are fragmented units of knowledge. To make them more organized and easier to use, they need to be hierarchically organized according to semantic and temporal relationships. For example, semantically, knowledge tuples related to equipment can be placed under a large hierarchy, and then further subdivided according to different equipment types. Under each equipment type hierarchy, further subdivisions can be made based on the relationship between the equipment and its operation, environment, etc. Simultaneously, considering temporal relationships, knowledge tuples with a chronological order are arranged in chronological order. This forms a hierarchical set of related knowledge, clearly demonstrating the various knowledge relationships in the field of thermal power plant safety monitoring, from macro to micro, from whole to part.

[0090] Step 150: Generate a knowledge graph for safety monitoring of thermal power plants based on the associated knowledge set. The knowledge graph for safety monitoring of thermal power plants is constructed with entities as nodes and relational features as edges to build a knowledge network structure.

[0091] The associated knowledge set provides rich material for generating knowledge graphs. By visualizing entity and relation features, it is possible to understand various information and relationships in the safety monitoring of thermal power plants more intuitively.

[0092] In an optional embodiment, the step of generating a knowledge graph for power plant safety monitoring based on the associated knowledge set, wherein the knowledge graph for power plant safety monitoring constructs a knowledge network structure with entities as nodes and relational features as edges, includes:

[0093] Step 151: Perform entity node mapping processing on the knowledge tuples in the associated knowledge set, mapping the entity categories and attribute types in the knowledge tuples to nodes in the knowledge graph. Each node contains corresponding entity identification information and attribute feature description.

[0094] In knowledge tuples within a related knowledge set, entity category and attribute type are key information for constructing knowledge graph nodes. For example, the entity category "Equipment" is mapped to a node in the knowledge graph, with the node name "Equipment". Then, corresponding entity identification information, such as the equipment number and model, and attribute descriptions, such as the equipment's function and rated parameters, are added to this node. For the knowledge tuple "Operation Behavior", it is mapped to an "Operation Behavior" node, with identification information such as operator and operation time, and attribute descriptions such as operation type and purpose. Through this mapping process, abstract knowledge tuples are transformed into concrete knowledge graph nodes.

[0095] Step 152: Perform relation edge mapping processing on the knowledge tuples in the associated knowledge set, mapping the relation types in the knowledge tuples to edges in the knowledge graph. Each edge contains the corresponding relation feature description and association strength information.

[0096] The relationship types within knowledge tuples determine the connection methods and meanings of edges in the knowledge graph. For example, in the knowledge tuple "Association between devices and operations," the "startup" relationship type is mapped to an edge in the knowledge graph, and the edge's name could be "Startup Relationship." Relationship feature descriptions are added to this edge, such as the specific steps of the operation, its impact on the device, and association strength information, such as the degree of influence of the operation on the device's startup success rate derived from data analysis. For the knowledge tuple "Influence of devices on the environment," the "temperature influence" relationship type is mapped to an edge, adding information such as the direction and intensity of the impact of temperature changes on device performance. In this way, through relation edge mapping, the relationships in the knowledge tuples are transformed into edges connecting nodes in the knowledge graph, demonstrating the associations between entities.

[0097] Step 153: Perform conflict detection processing on the nodes and edges, identify contradictory content between different attribute descriptions of the same entity and deviation values ​​between different strength descriptions of the same relationship, correct contradictory content using majority voting rules, and calibrate deviation values ​​using a weighted average method.

[0098] During the construction of a knowledge graph, conflicts may arise in node attributes and edge strengths due to the wide range of data sources. For example, for the same device, different data records may contain different descriptions of its rated power, necessitating conflict detection. A majority voting rule is used to count the frequency of different descriptions, selecting the most frequent description as the corrected attribute value. For different strength descriptions of the same relationship, such as varying association strengths between devices and operations across different time periods, a weighted average method is employed. Weights are assigned to different strength values ​​based on factors such as data reliability, calculating the calibrated association strength value. This ensures the accuracy and consistency of the knowledge graph.

[0099] Step 154: Perform knowledge fusion processing on the corrected nodes and edges, merge and integrate the attribute features of the same entity scattered in different knowledge tuples, comprehensively describe the multiple relation features involving the same relationship, and generate a knowledge graph infrastructure with information integrity.

[0100] After conflict detection and correction, the information of nodes and edges needs further fusion. For the same entity, such as a generator, different knowledge tuples may contain different attribute features regarding its operating status, maintenance records, etc. These attribute features are merged and integrated to make the generator node contain more comprehensive information. For multiple relational features involving the same relationship, such as the relationship between equipment and the environment, different knowledge tuples may contain different environmental factors such as the impact of temperature and humidity on the equipment. These relational features are comprehensively described to form a more complete description of the relationship between equipment and the environment. Through knowledge fusion processing, a knowledge graph infrastructure with information integrity is generated.

[0101] Step 155: Perform topology optimization on the basic structure of the thermal power plant safety monitoring knowledge graph, adjust the spatial layout of nodes in the graph according to the association strength information between nodes, so that nodes that reach the preset strength value present a dense spatial distribution relationship in the graph, and generate the thermal power plant safety monitoring knowledge graph.

[0102] The basic structure of the knowledge graph has been established, but topology optimization is needed for better display and analysis. Based on the strength of associations between nodes, the spatial layout of nodes in the graph is adjusted. For example, if certain equipment nodes have strong associations with multiple operational and environmental nodes, these nodes are placed in the central or relatively concentrated areas of the graph, creating a dense spatial distribution. Nodes with weaker associations are placed at the edges. Through this topology optimization, the generated knowledge graph for power plant safety monitoring can more intuitively display the closeness and importance of relationships between entities, facilitating observation and analysis by staff.

[0103] In a standalone embodiment, the method further includes:

[0104] Step 210: Obtain real-time incremental data for the safety monitoring scenario of the thermal power plant. The real-time incremental data includes real-time equipment operating status update records, the latest operation log fragments, and real-time sampling data from environmental monitoring sensors.

[0105] The operation of a thermal power plant is a dynamic process, and real-time incremental data can reflect the latest operating status. For example, real-time equipment operating status update records can promptly provide feedback on the current operating parameters of the equipment, such as the real-time power output of the generator and the real-time temperature of the boiler. The latest operation log fragment records the most recent operational actions, including the operator, operation time, and operation content. Real-time sampling data from environmental monitoring sensors monitors changes in environmental parameters in real time, such as current humidity and gas concentration.

[0106] Step 220: Perform rapid cleaning and label alignment on the real-time incremental data, and add device type labels, operation type labels, and environmental parameter labels to the cleaned data using a semantic labeling system consistent with the standardized dataset.

[0107] Real-time incremental data may contain noise and formatting issues, requiring rapid cleaning. This includes removing invalid and duplicate values. Then, a semantic tagging system consistent with the previously standardized dataset is used for tag alignment. For real-time equipment operation status updates, appropriate equipment type tags are added based on the equipment type, such as "steam turbine equipment." For the latest operation log fragments, operation type tags are added based on the operation content, such as "parameter adjustment operation." For real-time sampling data from environmental monitoring sensors, environmental parameter tags are added, such as "temperature parameter." This ensures that the real-time incremental data maintains semantic consistency with the existing standardized data.

[0108] Step 230: Perform entity matching processing between the real-time incremental data after label alignment and the knowledge graph of power plant safety monitoring, identify existing entities and their attribute features in the graph, and extract unmatched new data units as candidate data for potential new entities or new relationships.

[0109] Matching the tagged, real-time incremental data with existing knowledge graphs is crucial for determining which data already has corresponding entities in the graph and which is new. For example, for a device in a real-time operational status update record, if a node for that device already exists in the knowledge graph, the node and its attribute features are identified. However, if a new device type or new device attribute appears, these unmatched data units become candidate data for potential new entities. The same applies to operational and environmental data; if new operational types or environmental parameters do not match relationships in the existing graph, these data are extracted as candidate data for potential new relationships.

[0110] Step 240: Perform semantic relevance verification on the candidate data. By comparing the semantic labels of the candidate data with the semantic labels of existing entities in the graph, select valid candidate data that are relevant to the security monitoring task.

[0111] Candidate data may contain information unrelated to the safety monitoring tasks of thermal power plants, requiring semantic relevance verification. Relevance is determined by comparing the semantic labels of candidate data with those of existing entities in the graph. For example, a new operation in the candidate data is considered valid candidate data if its semantic label has a clear semantic relationship with existing operation type labels in the graph and is related to safety monitoring tasks, such as "emergency shutdown operation." Conversely, candidate data is excluded if its semantic label differs significantly from existing entity semantic labels and is unrelated to safety monitoring, such as operation records related to administrative affairs.

[0112] Step 250: Update the knowledge graph of power plant safety monitoring based on valid candidate data, add new entities as new nodes, add new relation features as new edges, and update the attribute feature descriptions of relevant nodes in a timely manner to generate an updated knowledge graph.

[0113] Based on the selected valid candidate data, the knowledge graph is updated. If a new entity is added, such as a new environmental monitoring device, it is added as a new node in the knowledge graph, and corresponding identification information and attribute descriptions are added to the node. For new relationship features, such as the association between the environmental monitoring device and other devices, new edges are added, and the relationship features are described. Simultaneously, the addition of new data may affect the attribute features of related nodes; for example, if the operating parameters of a device change due to new environmental factors, the attribute feature descriptions of the relevant nodes are updated accordingly. Through this update process, a timely updated knowledge graph is generated, which can promptly reflect the latest situation in the safety monitoring scenario of thermal power plants.

[0114] This invention improves data quality and reliability by merging and cleaning the original datasets, and by using unified timestamps and semantic tags to break down data silos and enhance usability and comparability. The feature extraction process comprehensively captures entity and relationship features within the power plant safety monitoring scenario, providing rich information for a deeper understanding of system operation. Knowledge association analysis creatively integrates semantic relevance and temporal continuity, deeply mining the potential connections behind the data, and generating a highly logical and practical set of related knowledge. The final constructed knowledge graph for power plant safety monitoring presents complex information in an intuitive knowledge network structure, more efficiently supporting the analysis, decision-making, and early warning of power plant safety monitoring, and improving the intelligence and precision of power plant safety monitoring.

[0115] In the several embodiments provided in this invention, it should be understood that the disclosed apparatus and methods can also be implemented in other ways. The apparatus and method embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0116] In addition, the functional modules in the various embodiments of the present invention can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0117] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a knowledge graph generation system 10, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks. It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. In the absence of further restrictions, an element defined by the phrase "comprising a..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0118] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A knowledge graph generation method applied to safety monitoring of a thermal power plant, characterized in that, The method comprises: acquiring a raw data set corresponding to a thermal power plant safety monitoring scene; performing data cleaning and standardization processing on the raw data set to obtain a standardized data set with unified timestamp markers and semantic labels; performing feature extraction processing on the standardized data set to obtain entity features and inter-entity relationship features in the thermal power plant safety monitoring scene; performing knowledge association analysis on the entity features and the inter-entity relationship features to generate an associated knowledge set containing semantic association and time sequence continuity; generating a thermal power plant safety monitoring knowledge graph based on the associated knowledge set, the thermal power plant safety monitoring knowledge graph constructing a knowledge network structure with entities as nodes and relationship features as edges; The raw data set contains equipment operation state records, operation log information, environmental monitoring data, and historical accident reports. The acquisition of the raw data set corresponding to the thermal power plant safety monitoring scene comprises: collecting equipment operation state records from a thermal power plant distributed control system, the equipment operation state records containing start and stop time points of different types of equipment, operation mode switching records, and real-time sensor feedback data; extracting operation log information from an operation management system, the operation log information containing operator identity, operation timestamp, operation object identification, and operation instruction content; collecting environmental monitoring data from an environmental monitoring sensor network, the environmental monitoring data containing temperature field distribution records, humidity change sequences, gas concentration fluctuation data, and vibration frequency sampling values; retrieving historical accident reports from an accident management database, the historical accident reports containing accident occurrence time, accident affected equipment range, accident phenomenon description, and post-accident cause analysis conclusion; performing data format unification processing on the equipment operation state records, the operation log information, the environmental monitoring data, and the historical accident reports to generate a raw data set containing text data units, numerical data units, and time sequence data units; The knowledge association analysis on the entity features and the inter-entity relationship features to generate an associated knowledge set containing semantic association and time sequence continuity comprises: performing semantic similarity calculation processing on equipment entity attribute features, operation behavior attribute features, and environmental parameter attribute features to identify entity attributes with the same or similar semantic descriptions and generate semantic association relationships between entity attributes; performing time sequence order analysis processing on the association features of equipment and operation, the influence features of equipment and environment, and the causal features of operation and accident to extract the occurrence rules and continuous action periods of relationship features in the time dimension and generate time sequence association relationships between relationship features; performing cross-validation processing on the semantic association relationships and the time sequence association relationships to filter out the association relationships that simultaneously satisfy the semantic similarity condition and the time sequence regularity condition as effective association relationships; performing knowledge abstraction processing on the effective association relationships to convert specific entity attribute features and relationship features into generalized knowledge representation forms, and generating knowledge tuples containing entity categories, attribute types, and relationship types; The knowledge tuples are hierarchically organized according to semantic association relationships and time sequence association relationships, and an associated knowledge set with a hierarchical structure is generated.

2. The knowledge graph generation method for safety monitoring of a thermal power plant according to claim 1, characterized in that, The original data set is subjected to data cleaning and standardization processing to obtain a standardized data set with uniform timestamp labels and semantic labels, including: The text data units in the original data set are subjected to noise filtering processing, and invalid data with repeated records, missing fields exceeding a preset proportion, and non-standard format random code contents are deleted; The numerical data units in the original data set are subjected to outlier detection processing, and abnormal numerical points deviating from the normal range are identified by a sliding window statistical method, and are corrected by a linear interpolation method of adjacent time point values; The time sequence data units in the original data set are subjected to timestamp alignment processing, and the timestamps of different source data are uniformly converted into the standard time format of the global clock system of the thermal power plant; The text data units, the numerical data units and the time sequence data units after cleaning are subjected to semantic label annotation processing, and a semantic label reflecting the content theme of each data unit is added, and the semantic label includes an equipment type label, an operation type label, an environment parameter label and an accident type label; The labeled multi-class data units are subjected to association integration processing according to the timestamp order, and a standardized data set with time continuity and semantic consistency is generated.

3. The knowledge graph generation method for safety monitoring of a thermal power plant according to claim 1, characterized in that, The entity features include equipment entity attribute features, operation behavior attribute features and environment parameter attribute features, the inter-entity relationship features include equipment and operation association features, equipment and environment influence features and operation and accident causality features, the feature extraction processing of the standardized data set obtains entity features and inter-entity relationship features in the thermal power plant safety monitoring scene, including: The equipment type label data in the standardized data set is subjected to entity recognition processing, and the identification information, model parameters, installation location information and design running indicators of the equipment entity are extracted as equipment entity attribute features; The operation type label data in the standardized data set is subjected to behavior mode analysis processing, and the execution subject information, operation object information, operation trigger condition and operation duration of the operation behavior are extracted as operation behavior attribute features; The environment parameter label data in the standardized data set is subjected to parameter feature extraction processing, and the monitoring location information, change period feature and synchronous fluctuation relationship with the equipment running state of the environment parameter are extracted as environment parameter attribute features; The equipment entity attribute features and the operation behavior attribute features are subjected to association analysis processing, and the influence degree description of the operation behavior on the equipment running state and the accompanying appearance frequency of the operation behavior and the equipment failure are extracted as equipment and operation association features; The equipment entity attribute features and the environment parameter attribute features are subjected to correlation analysis processing, and the influence direction and influence intensity description of the environment parameter change on the equipment performance indicators are extracted as equipment and environment influence features; The operation behavior attribute characteristics are subjected to causal relationship mining processing with the historical accident report data, and operation behavior abnormalities and time interval characteristics of accident occurrence and corresponding relationships between operation error types and accident types are extracted as causal characteristics of operation and accidents.

4. The knowledge graph generation method for safety monitoring of a thermal power plant according to claim 1, characterized in that, The power plant safety monitoring knowledge graph is generated based on the association knowledge set, and the power plant safety monitoring knowledge graph constructs a knowledge network structure with entities as nodes and relationship characteristics as edges, and includes: The knowledge tuples in the association knowledge set are subjected to entity node mapping processing, and the entity categories and attribute types in the knowledge tuples are mapped to the nodes of the knowledge graph, and each node contains corresponding entity identification information and attribute characteristic description; The knowledge tuples in the association knowledge set are subjected to relationship edge mapping processing, and the relationship types in the knowledge tuples are mapped to the edges of the knowledge graph, and each edge contains corresponding relationship characteristic description and association strength information; The nodes and edges are subjected to conflict detection processing, and contradictions between different attribute descriptions of the same entity and deviation values between different strength descriptions of the same relationship are identified, and the contradictions are corrected using majority voting rules, and the deviation values are calibrated using weighted average method; The modified nodes and edges are subjected to knowledge fusion processing, and the attribute characteristics of the same entity scattered in different knowledge tuples are merged and integrated, and multiple relationship characteristics involving the same relationship are comprehensively described, and a knowledge graph basic structure with complete information is generated; The power plant safety monitoring knowledge graph basic structure is subjected to topological optimization processing, and the spatial layout of the nodes in the graph is adjusted according to the association strength information between the nodes, so that the nodes reaching the preset strength value present a dense spatial distribution relationship in the graph, and the power plant safety monitoring knowledge graph is generated.

5. The knowledge graph generation method for safety monitoring of a thermal power plant according to claim 1, characterized in that, The equipment operation state records are collected from the power plant distributed control system, including: Accessing the historical database interface of the power plant distributed control system to obtain the operation state logs of the boiler equipment, turbine equipment, generator equipment and auxiliary system equipment; The operation state logs are subjected to field analysis processing, and the equipment identification field, timestamp field, operation state field and sensor data field in each log record are extracted; The operation state field is subjected to state type classification processing, and the operation state is divided into start state, stable operation state, mode switching state and shutdown state, and the start time point and end time point of each state are recorded as start-stop time points; The operation state fields of adjacent timestamps are subjected to difference comparison processing, and the time points at which the operation state changes and the changed state types are identified, and equipment operation mode switching records are generated; The sensor data field is subjected to data extraction processing, and the real-time feedback values of temperature sensors, pressure sensors, speed sensors and vibration sensors are collected, and sensor real-time feedback data reflecting the real-time operation state of the equipment are generated; The start-stop time points, operation mode switching records and sensor real-time feedback data are subjected to association integration processing, and equipment operation state records are generated.

6. The knowledge graph generation method for safety monitoring of a thermal power plant according to claim 2, characterized in that, The abnormal value detection processing on the numerical data units in the original data set identifies abnormal numerical points deviating from the normal range through a sliding window statistical method, and corrects them using a linear interpolation method of adjacent time point values, comprising: Arranging the numerical data units into a time series data array in timestamp order; Setting a fixed length sliding window, and performing sliding traversal processing on the time series data array, with each sliding window containing a preset number of continuous data points; Calculating the average value and standard deviation of the data points in each sliding window, and constructing a normal numerical range with the average value as the center and twice the standard deviation as the boundary; Detecting data points in the sliding window that exceed the normal numerical range as abnormal numerical points, and recording the timestamp positions of the abnormal numerical points; For each abnormal numerical point, extracting the numerical value and timestamp information of the previous and next normal data points; According to the numerical difference and time difference between the previous and next normal data points, calculating the linear interpolation coefficient, and calculating the correction value of the abnormal numerical point through the linear interpolation formula; Using the correction value to replace the numerical value of the original abnormal numerical point to generate a corrected numerical data unit.

7. The knowledge graph generation method for safety monitoring of a thermal power plant according to claim 3, characterized in that, The causal relationship mining processing of the operation behavior attribute features and the historical accident report data extracts the time interval feature of operation behavior abnormalities and accident occurrence and the corresponding relationship of operation error types and accident types as the causal features of operation and accident, comprising: Extracting operation error type information from the operation behavior attribute features, which includes accidental touch operation, timeout operation, parameter setting error and process execution omission; Extracting accident type information from the historical accident report data, which includes equipment overheating accident, pressure overrun accident, vibration anomaly accident and system shutdown accident; Performing time alignment processing on the operation error type information and the accident type information, matching the timestamp of each operation error event with the timestamp of the subsequent accident event, and calculating the time interval between the operation error event and the accident event; Statistically counting the number of occurrences of the same operation error type and the same accident type combination, and calculating the occurrence frequency of the combination as the association strength of the operation error type and the accident type; Extracting the operation error event and accident event combination with a time interval within a preset range, and statistically counting the time interval distribution characteristics of the combination to generate the time interval feature of operation behavior abnormalities and accident occurrence; Integrating the association strength and the time interval feature to generate a corresponding relationship description of the operation error type and the accident type as the causal features of operation and accident.

8. A knowledge graph generation system, characterized by, A device comprising a processor, a network module and a memory; the processor and the memory communicate through the network module, the processor reads the computer program from the memory and runs to execute the method of any one of claims 1-7.

Images