A smart power plant safety early warning method based on multi-data coupling

Abstract

This invention relates to the field of data early warning technology, and in particular to a smart power plant safety early warning method based on multi-data coupling. The method includes the following steps: acquiring multi-source coupled data including power plant production and operation data, key equipment status monitoring data, environmental monitoring data, and historical accident records; calculating the acquisition time interval and synchronization delay between each data source; defining the coupling propagation rate as the ratio of the acquisition time interval to the topological distance between corresponding nodes of key equipment; constructing a time-series-topology correlation window based on the coupling propagation rate, and determining the status type of key equipment according to the time-series-topology correlation window; obtaining the risk parameters of key equipment by quantifying the status type. This invention, through multi-source coupled data and time-series-topology correlation analysis, realizes the quantification of risk parameters and potential risk marking of key equipment, thereby supporting dynamic hierarchical early warning and improving the safety and real-time performance of power plant operations.

Description

Technical Field

[0001] This invention relates to the field of data early warning technology, and in particular to a smart power plant safety early warning method based on multi-data coupling. Background Technology

[0002] Traditional power plant safety early warning methods often rely on a single data source, such as equipment condition monitoring or environmental monitoring, and typically identify risks through simple threshold judgments or rule matching. These methods have limitations in real-time performance and accuracy, and are difficult to effectively handle the coupling relationships between multiple devices and parameters under complex operating conditions.

[0003] Existing research attempts to integrate production operation data, critical equipment condition monitoring data, environmental monitoring data, and historical accident records, and to conduct risk assessments through time-series analysis or topological relationships to achieve early fault prediction and warning. However, existing technologies still have several shortcomings: First, there are delays in the acquisition time and synchronization of multi-source data, which may lead to a lag in risk assessment; second, there is a lack of precise quantification of the coupling and propagation characteristics between different data sources, making it difficult to establish dynamic condition discrimination models; third, there is insufficient sensitivity to detect minor anomalies such as gradient changes and amplitude consistency, which may easily overlook potentially risky equipment, thus affecting the accuracy and timeliness of early warnings. Summary of the Invention

[0004] Therefore, it is necessary to provide a smart power plant safety early warning method based on multi-data coupling to solve at least one of the above-mentioned technical problems.

[0005] To achieve the above objectives, this invention provides a smart power plant safety early warning method based on multi-data coupling, comprising the following steps:

[0006] Step S1: Acquire multi-source coupled data including power plant production and operation data, key equipment status monitoring data, environmental monitoring data and historical accident records, and calculate the acquisition time interval and synchronization delay between each data source;

[0007] Step S2: Define the ratio of the acquisition time interval to the topological distance between the corresponding nodes of the key equipment as the coupling propagation rate;

[0008] Step S3: Construct a time-series-topology correlation window based on the coupling propagation rate, and determine the state type of key equipment based on the time-series-topology correlation window; obtain the risk parameters of key equipment by quantifying the state type;

[0009] Step S4: Calculate the gradient vector and amplitude consistency index of the risk parameters of key equipment based on the synchronization delay. When the gradient direction angle exceeds the preset angle threshold or the relative difference in amplitude exceeds the preset proportion threshold, it is marked as a potential risk equipment.

[0010] Step S5: Set the equipment monitoring and early warning timestamps in the status monitoring data of key equipment through potential risk equipment to perform dynamic power plant equipment safety early warning operations.

[0011] Preferably, the method for obtaining the topological distance between the corresponding nodes of the key equipment in step S2 includes:

[0012] Nodes are extracted from power plant production and operation data and key equipment status monitoring data to generate key equipment node set data;

[0013] A node connectivity graph is established based on the key equipment node set data to generate preliminary topology connection data;

[0014] Path weights are assigned to the initial topology connection data to generate weighted topology path data;

[0015] Optimal path search is performed based on weighted topology path data to calculate the shortest weighted path length between key equipment node pairs, thereby obtaining the topology distance between corresponding nodes of key equipment.

[0016] Preferably, establishing a node connectivity graph based on key equipment node set data includes:

[0017] Classify the key equipment node set data to generate node function category data;

[0018] Based on the node function category data, the direct coupling relationship between nodes is analyzed to generate initial connected edge data;

[0019] Redundancy and conflict detection is performed on the initial connected edge data to generate edge conflict correction data;

[0020] Topology consistency correction is performed based on edge conflict correction data to generate topology connectivity data;

[0021] A graph of node connectivity is established using topological connection data.

[0022] Preferably, step S3 includes the following steps:

[0023] Step S31: Divide the coupled propagation rate data into time scales to generate multi-level time segment data;

[0024] Step S32: Based on the topological path lengths in the multi-level time segment data and topological connection data, construct a time-series-topological association window and generate window boundary data;

[0025] Step S33: Perform device state sequence analysis on the window boundary data to generate preliminary device state classification data; perform disturbance sensitivity analysis on the preliminary device state classification data to generate state correction data;

[0026] Step S34: Perform quantitative calculations based on the condition correction data to generate risk parameters for key equipment.

[0027] Preferably, step S33 includes:

[0028] The window boundary data is split into time series segments, and device status sequence fragments are generated according to the sampling interval to obtain segmented status sequence data.

[0029] Local change features are extracted from equipment state sequence segments, including state switching frequency, duration and abnormal fluctuation amplitude, to generate local state feature data;

[0030] Preliminary cluster analysis is performed using local state feature data to classify equipment state sequences according to state change patterns and generate preliminary state category labels.

[0031] The consistency of the initial classification is verified by combining the initial status category labels with window boundary data, and abnormal or incomplete sequences are corrected to generate initial classification data of device status.

[0032] Disturbance sensitivity analysis is performed on the preliminary classification data of equipment status to generate status correction data.

[0033] Preferably, the disturbance sensitivity analysis of the preliminary equipment status classification data includes:

[0034] For each preliminary classification state, a disturbance signal of preset amplitude is applied, and the change response of the equipment status data is observed;

[0035] Analyze the magnitude and trend of change of each preliminary classification state under different perturbation conditions, and generate the perturbation sensitivity characteristics of each state;

[0036] Based on the perturbation sensitivity characteristics, the bias or misjudgment area in the preliminary classification state is determined;

[0037] Based on the deviation or misjudgment area, the preliminary classification status data is corrected to generate corrected equipment status data.

[0038] Preferably, step S34 includes the following steps:

[0039] Step S341: Based on the status correction data, for each key device, statistically analyze the frequency and magnitude of its status data anomalies within a preset time period;

[0040] Step S342: Combine the frequency and magnitude of the anomalies to calculate the risk score for each critical device;

[0041] Step S343: Generate risk parameters for key equipment based on the risk score.

[0042] Preferably, step S4 includes the following steps:

[0043] Step S41: Perform synchronous delay correction on the risk parameter sequence of key equipment to generate delay-corrected risk parameter data;

[0044] Step S42: Based on the risk parameter data after delay correction, calculate the gradient vector between adjacent sampling points and generate a gradient vector sequence;

[0045] Step S43: Perform angle calculation on the gradient vector sequence to obtain the gradient direction angle distribution data;

[0046] Step S44: Based on the risk parameter data after delay correction, calculate the amplitude consistency index and perform joint discrimination with the gradient direction angle distribution data to generate potential risk equipment labeling data.

[0047] Preferably, the method for confirming adjacent sampling points includes:

[0048] Signals from key equipment are sampled to obtain basic sampling points;

[0049] The basic sampling points are arranged according to the risk parameter sequence of key equipment to obtain an ordered set of sampling points;

[0050] In the ordered set of sampling points, with each target sampling point as the center, its previous sampling point and its next sampling point are extracted to form candidate adjacent point pairs;

[0051] Perform interval consistency detection on candidate adjacent point pairs. When the time interval between adjacent sampling points is within the preset time threshold range, the point pair is confirmed as a valid adjacent sampling point.

[0052] When the time interval between adjacent sampling points exceeds the preset time threshold, a proximity compensation strategy is adopted to use the effective sampling point closest to the target sampling point as a substitute neighbor, thereby generating the final pair of adjacent sampling points.

[0053] Preferably, step S44 includes:

[0054] Based on the risk parameter data after delay correction, the amplitude change ratio of adjacent data segments is calculated point by point to obtain the amplitude consistency index sequence.

[0055] Divide the gradient direction angle distribution data into intervals, extract the angle exceeding the limit interval and the normal interval, and generate angle discrimination data;

[0056] The amplitude consistency index sequence and the angle discrimination data are cross-mapped to obtain the joint discrimination result data;

[0057] Based on the joint discrimination results, key equipment is marked one by one, and the potentially risky equipment marking data is output.

[0058] The beneficial effects of this invention lie in its acquisition of multi-source coupled data, including power plant production and operation data, critical equipment status monitoring data, environmental monitoring data, and historical accident records. This enables quantitative analysis of the acquisition time intervals and synchronization delays between various data sources. Furthermore, by introducing the concept of coupling propagation rate, the time-series information of critical equipment is correlated with the topology, constructing a time-series-topology correlation window to accurately determine the status type of critical equipment. Risk parameters are generated by quantifying status types, and gradient vectors and amplitude consistency indices are calculated based on synchronization delays. Equipment with abnormal gradient direction angles or amplitude differences is marked as potentially at risk. Ultimately, this method enables the setting of hierarchical early warning timestamps for critical equipment status monitoring data based on potentially risky equipment, supporting dynamic power plant equipment safety early warning operations. This process not only enhances the ability to identify potential risks in critical equipment and improves power plant operation safety and management accuracy, but also achieves real-time and targeted risk assessment and early warning through multi-source data coupling analysis and time-series-topology correlation judgment, demonstrating significant application value and practicality. Attached Figure Description

[0059] Figure 1 This is a flowchart illustrating the steps of a smart power plant safety early warning method based on multi-data coupling.

[0060] Figure 2 for Figure 1 A detailed flowchart illustrating the implementation steps of step S3.

[0061] Figure 3 for Figure 1 A detailed flowchart illustrating the implementation steps of step S4.

[0062] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0063] The technical method of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of this invention.

[0064] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0065] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0066] To achieve the above objectives, please refer to Figures 1 to 3 A smart power plant safety early warning method based on multi-data coupling, the method comprising the following steps:

[0067] Step S1: Acquire multi-source coupled data including power plant production and operation data, key equipment status monitoring data, environmental monitoring data and historical accident records, and calculate the acquisition time interval and synchronization delay between each data source;

[0068] Step S2: Define the ratio of the acquisition time interval to the topological distance between the corresponding nodes of the key equipment as the coupling propagation rate;

[0069] Step S3: Construct a time-series-topology correlation window based on the coupling propagation rate, and determine the state type of key equipment based on the time-series-topology correlation window; obtain the risk parameters of key equipment by quantifying the state type;

[0070] Step S4: Calculate the gradient vector and amplitude consistency index of the risk parameters of key equipment based on the synchronization delay. When the gradient direction angle exceeds the preset angle threshold or the relative difference in amplitude exceeds the preset proportion threshold, it is marked as a potential risk equipment.

[0071] Step S5: Set the equipment monitoring and early warning timestamps in the status monitoring data of key equipment through potential risk equipment to perform dynamic power plant equipment safety early warning operations.

[0072] In this embodiment of the invention, multi-source coupled data is acquired by sensors deployed in key equipment and the plant environment. This multi-source coupled data includes power plant production and operation data, key equipment status monitoring data, environmental monitoring data, and historical accident records. Specifically, production and operation data includes generator output, load rate, voltage, current, and fuel consumption; key equipment status monitoring data includes equipment vibration, temperature, pressure, oil level, lubrication status, and insulation status; environmental monitoring data includes temperature, humidity, dust concentration, electromagnetic interference intensity, wind speed and direction, and noise levels inside and outside the plant area; and historical accident records include equipment failure time, type, maintenance records, and downtime information. While acquiring various types of data, the acquisition time interval and data transmission delay between different data sources are calculated to form a complete multi-source coupled dataset and data synchronization delay information, ensuring accurate and reliable temporal correspondence between various data types during subsequent analysis.

[0073] After data acquisition, an equipment topology diagram is established based on the electrical or mechanical coupling relationships between key equipment in the power plant, determining the location of each key piece of equipment within the plant and its association with other equipment. Then, the data collection time intervals from each data source are correlated with the topological distances between equipment to determine the propagation speed of risk information among key equipment. Faster propagation speed indicates timely response to risk information between equipment, while slower propagation speed indicates a higher potential delay risk. Based on this propagation speed, a time-series and topology correlation window can be constructed for each key piece of equipment. This window integrates multi-source data within a specific time range to determine the status type of the key equipment, such as normal, warning, or abnormal status. The status type is quantified by comprehensively combining historical accident records and real-time monitoring data, thereby generating risk parameters for each key piece of equipment for subsequent risk assessment.

[0074] The data sources mentioned in this application embodiment are of four types: production operation data, key equipment status monitoring data, environmental monitoring data, and historical accident records. The collection frequencies of different data sources vary. Therefore, when performing corresponding analysis, it is first necessary to calculate the difference in collection time intervals between each data source. For example, the interval difference between production operation data and key equipment status monitoring data is 4 seconds, between production operation data and environmental monitoring data is 9 seconds, between production operation data and historical accident records is 29 seconds, between key equipment status monitoring data and environmental monitoring data is 5 seconds, between key equipment status monitoring data and historical accident records is 25 seconds, and between environmental monitoring data and historical accident records is 20 seconds. This yields six collection time interval differences. Simultaneously, the electrical or mechanical coupling relationship between key power plant equipment can determine a unique topological distance. For example, the topological distance between a generator and a transformer is 100 topological units. Each collection time interval difference can then be compared with this topological distance to obtain six coupling propagation rates, such as 25, 11.1, 3.4, 20, 4, and 5. Therefore, the final result is not a single propagation rate, but a set of propagation rates formed by combinations of multiple data sources. In subsequent processing, different representation methods can be selected according to different needs. For example, the average value can be calculated to reflect the overall propagation speed of risk information between devices; the maximum and minimum values ​​can be selected to reflect the propagation performance under best and worst conditions, respectively; or a weighted calculation can be performed combining the importance of different data sources to obtain a weighted propagation rate. This rate is used to construct the time-series and topology correlation window of key devices and ultimately support the risk assessment process.

[0075] After the risk parameters of key equipment are determined, potential risk equipment is identified by analyzing the changing trends of these risk parameters and the consistency of their amplitudes among adjacent equipment. Specifically, when the direction of change of the risk parameter of a certain key equipment differs significantly from that of adjacent equipment, or when the amplitude of its risk parameter deviates markedly from that of adjacent equipment, that equipment is marked as a potential risk equipment. This marking method can effectively detect single-point anomalies or localized risk concentrations, providing a basis for early warning in power plants.

[0076] Subsequently, based on the identified potentially risky equipment, tiered early warning systems are implemented for the status monitoring data of critical equipment. These systems typically include Level 1, Level 2, and Level 3 warnings. Level 1 warnings indicate that high-risk equipment requires immediate attention or shutdown for inspection; Level 2 warnings indicate that medium-risk equipment requires continuous monitoring for a short period; and Level 3 warnings indicate that low-risk equipment can continue with routine monitoring. This tiered early warning strategy enables dynamic and real-time safety management of power plant equipment while ensuring overall safety.

[0077] Ultimately, this implementation method generates a dynamic power plant equipment safety early warning strategy by combining the monitoring results of potentially risky equipment with early warning levels and processing feedback. This strategy can transmit risk information to the power plant monitoring system and operation and maintenance terminals in real time, automatically triggering audible and visual alarms, alarm information push notifications, and necessary equipment shutdown or load adjustment operations. It also records and analyzes the processing results to optimize the risk parameter assessment model, achieving comprehensive dynamic monitoring and safety assurance for key power plant equipment. Through this method, potential equipment risks can be identified in a timely manner even with highly coupled multi-source data, preventing accidents and improving the safety and reliability of power plant operation. Specifically, potentially risky equipment includes boiler main steam pump A with abnormal vibration, turbine B bearing with abnormal temperature, and generator C with abnormal winding temperature, etc. These devices continuously collect key status monitoring data, such as vibration acceleration, bearing temperature, winding temperature, and current, and send the data to the monitoring center in real time. Based on the amplitude and trend of monitoring data for potentially risky equipment, the system classifies equipment risks into three warning levels: Level 1, Level 2, and Level 3. Level 1 warnings indicate high risk requiring immediate action or shutdown for inspection; Level 2 warnings indicate medium risk requiring continuous monitoring within a short period; and Level 3 warnings indicate low-risk equipment that can continue with routine monitoring. Different warning levels correspond to different timestamp generation frequencies. For example, Level 1 warnings generate a timestamp every second for high-frequency monitoring and triggering audible and visual alarms; Level 2 warnings generate a timestamp every 5 seconds for trend monitoring; and Level 3 warnings generate a timestamp every minute for routine recording and statistical analysis. Through this timestamp-based classification strategy, the system can dynamically adjust the monitoring frequency according to the risk level, enabling real-time response to high-risk equipment while simultaneously managing low-risk equipment through routine procedures, thus ensuring the rational utilization of monitoring resources. The generated timestamp data constitutes a complete time-series database, providing a basis for dynamic power plant equipment safety early warning operations. It supports automated alarm triggering, equipment operation control, and maintenance decision analysis. By recording and processing results, it optimizes risk parameter assessment models, achieving timely identification and comprehensive dynamic monitoring of potential risks in key equipment under multi-source coupled data conditions, significantly improving the safety and reliability of power plant operations.

[0078] Preferably, the method for obtaining the topological distance between the corresponding nodes of the key equipment in step S2 includes:

[0079] Nodes are extracted from power plant production and operation data and key equipment status monitoring data to generate key equipment node set data;

[0080] A node connectivity graph is established based on the key equipment node set data to generate preliminary topology connection data;

[0081] Path weights are assigned to the initial topology connection data to generate weighted topology path data;

[0082] Optimal path search is performed based on weighted topology path data to calculate the shortest weighted path length between key equipment node pairs, thereby obtaining the topology distance between corresponding nodes of key equipment.

[0083] In this embodiment of the invention, key equipment nodes are extracted from the power plant's production operation data and key equipment status monitoring data. Specifically, by analyzing the operating status information, equipment number, and operating parameters of various types of equipment in the power plant, each key piece of equipment is identified as an independent node, and a node set data containing all key equipment is generated. Each node contains the equipment's unique identifier, equipment type, and location attribute information within the power plant, providing basic data for subsequent topology analysis. For example, in a power plant, there are three key pieces of equipment: generator A, transformer B, and turbine C. These three pieces of equipment are identified as node A, node B, and node C, respectively. Then: the topological distance between A and B is 50; the topological distance between B and C is 80; and the topological distance between A and C is 100. Next, based on the above key equipment node set data, a connectivity graph between nodes is established. During the establishment process, the electrical connection relationships, mechanical transmission relationships, and process coupling relationships between equipment are analyzed, and the interconnected equipment nodes are connected by lines to form preliminary topology connection data. This topology map can reflect the spatial and functional interdependencies of key equipment within the power plant, providing a reference for subsequent path analysis.

[0084] Subsequently, path weights are assigned to the preliminary topology connection data. In this embodiment, path weights can be determined comprehensively based on factors such as the physical distance between devices, information transmission delay, current transmission ratio, and mechanical coupling strength. Each connection path is assigned a corresponding weight value, thus forming weighted topology path data. The smaller the weight, the tighter the coupling relationship between devices and the more timely the risk information transmission; the larger the weight, the weaker the coupling or the higher the transmission delay. Finally, an optimal path search is performed based on the weighted topology path data. By analyzing all possible connection paths between nodes, the shortest weighted path between each pair of key device nodes is determined. This step compares the cumulative weights of different paths and selects the path with the smallest total weight as the optimal path, thereby obtaining the topology distance between pairs of key device nodes. Through the above method, the relative distances between key devices within the power plant in terms of function and information transmission can be accurately quantified, providing basic data support for risk propagation analysis and dynamic monitoring.

[0085] Preferably, establishing a node connectivity graph based on key equipment node set data includes:

[0086] Classify the key equipment node set data to generate node function category data;

[0087] Based on the node function category data, the direct coupling relationship between nodes is analyzed to generate initial connected edge data;

[0088] Redundancy and conflict detection is performed on the initial connected edge data to generate edge conflict correction data;

[0089] Topology consistency correction is performed based on edge conflict correction data to generate topology connectivity data;

[0090] A graph of node connectivity is established using topological connection data.

[0091] In this embodiment of the invention, node classification processing is performed on the key equipment node set data. Specifically, each key equipment node in the power plant is divided into different functional categories according to equipment type, functional attributes, and process characteristics, such as power generation equipment, transmission equipment, substation equipment, and control equipment, and corresponding node functional category data is generated. The node classification results not only reflect the functional role of the equipment in the power plant system, but also provide a basis for subsequent judgment of the coupling relationship between equipment. Subsequently, the direct coupling relationship between key equipment nodes is analyzed based on the node functional category data. During the analysis process, by analyzing the electrical connection relationship, mechanical transmission relationship, and process flow dependency relationship of the equipment, it is identified which equipment nodes have direct influence or dependency, and initial connected edge data is generated. Each initial connected edge represents a direct coupling or information transmission path between two nodes, which is the preliminary basis for establishing the topology of key equipment in the power plant.

[0092] After generating the initial connectivity edge data, redundancy and conflict detection is performed. This includes checking for duplicate edges, logical conflicts, or unreasonable connections, such as two edges repeatedly representing the same coupling relationship, or connections violating equipment functionality or process constraints. Edge conflict correction data is generated based on the detection results, and unreasonable connections are adjusted or deleted to eliminate redundancy and conflicts, ensuring the accuracy and reliability of the topology. Next, the topology is consistent with the edge conflict correction data. During the correction process, the connectivity of each node is checked holistically to ensure that the connectivity of all nodes conforms to the actual layout and process requirements of the power plant, while also ensuring the rationality of information or risk propagation paths between critical equipment. After topology consistency correction, complete topology connectivity data is generated, providing an accurate foundation for constructing a node connectivity graph. Finally, a connectivity graph of critical equipment nodes is established using the aforementioned topology connectivity data. In this graph, each node represents a critical piece of equipment in the power plant, and the connections between nodes reflect the direct coupling relationships and information transmission channels between the equipment. The connectivity graph of this node can intuitively display the structured relationships between key equipment within the power plant, providing important basic data support for subsequent topology distance calculations, coupling propagation analysis, and risk assessment.

[0093] As an example of the present invention, reference is made to... Figure 2 As shown, step S3 in this example includes:

[0094] Step S31: Divide the coupled propagation rate data into time scales to generate multi-level time segment data;

[0095] Step S32: Based on the topological path lengths in the multi-level time segment data and topological connection data, construct a time-series-topological association window and generate window boundary data;

[0096] Step S33: Perform device state sequence analysis on the window boundary data to generate preliminary device state classification data; perform disturbance sensitivity analysis on the preliminary device state classification data to generate state correction data;

[0097] Step S34: Perform quantitative calculations based on the condition correction data to generate risk parameters for key equipment.

[0098] In this embodiment of the invention, by dividing the coupled propagation rate data into time scales, continuous propagation rate information is divided into multiple time segments according to different time granularities, thereby generating multi-level time segment data. Each time segment corresponds to a specific time interval, which can reflect the changes in the propagation of risk information between key equipment at different time scales. This multi-level time division can fully capture the differences in short-term fluctuations and long-term trends of equipment risk information, providing a flexible time reference for subsequent analysis. Subsequently, based on the multi-level time segment data and the path length information recorded in the topological connection data between key equipment, a time-series-topology association window is constructed. During the construction process, by analyzing the topological distance and propagation rate between key equipment nodes in different time segments, the boundaries of the window are determined, forming time-topology interval data for comprehensive analysis. These windows can associate the state changes of each equipment node within a specific time range, reflecting the propagation characteristics of risk information along the topological path within the power plant.

[0099] Next, equipment status sequence analysis is performed on the time-series-topology association window boundary data. Specifically, this involves organizing and initially classifying the status change sequences of key equipment within each window, identifying the operating status of equipment in different time periods, such as normal, slightly abnormal, or significantly abnormal, thereby generating preliminary equipment status classification data. Subsequently, disturbance sensitivity analysis is performed on the preliminary classification results. By simulating or evaluating the impact of various disturbances (such as load fluctuations, environmental changes, or abnormal status of neighboring equipment) on the equipment status, the status classification results are corrected and optimized, generating status correction data. This step eliminates the interference of occasional fluctuations or atypical anomalies on status judgment, ensuring the reliability and stability of the classification results. Finally, quantitative calculations are performed based on the status correction data to generate key equipment risk parameters. During the quantification process, the corrected equipment status information is transformed into parameters that can be used for risk assessment, such as through level classification or risk index assignment, so that each key piece of equipment has a corresponding risk value in each time window. These risk parameters reflect the current risk level and potential anomaly trends of the equipment, providing basic data support for subsequent potential risk identification and dynamic early warning, and realizing accurate monitoring and risk management of the operating status of key power plant equipment. In the quantitative calculation process, the generation method of risk parameters for transformer A in a power plant can be illustrated using an example: Within a certain time-topology correlation window, after state correction, the state data of the equipment are obtained as follows: voltage deviation 2%, temperature 85℃, and vibration value 4 mm / s. The system converts this state information into parameters usable for risk assessment according to preset quantification rules. For example, voltage deviation between 1% and 3% is classified as risk level 2, temperature between 81 and 90 degrees Celsius corresponds to risk index 2, and vibration value between 3.1 and 5 mm / s corresponds to risk index 2. Subsequently, the system performs a weighted summation of these three types of indicators according to preset weight ratios, where voltage deviation has a weight of 40%, temperature has a weight of 30%, and vibration has a weight of 30%. The final calculation result is 2.0. Therefore, within this time window, the risk parameter of transformer A is 2.0, corresponding to a medium-risk state. Through a similar processing method, each key piece of equipment can generate a clear risk value for each time window, thereby transforming the equipment status from raw monitoring data into quantifiable risk parameters. This not only facilitates subsequent horizontal comparisons and vertical trend analysis, but also provides a quantitative basis for the overall risk assessment, early warning triggering, and operational decision-making of the power plant.

[0100] Preferably, step S33 includes:

[0101] The window boundary data is split into time series segments, and device status sequence fragments are generated according to the sampling interval to obtain segmented status sequence data.

[0102] Local change features are extracted from equipment state sequence segments, including state switching frequency, duration and abnormal fluctuation amplitude, to generate local state feature data;

[0103] Preliminary cluster analysis is performed using local state feature data to classify equipment state sequences according to state change patterns and generate preliminary state category labels.

[0104] The consistency of the initial classification is verified by combining the initial status category labels with window boundary data, and abnormal or incomplete sequences are corrected to generate initial classification data of device status.

[0105] Disturbance sensitivity analysis is performed on the preliminary classification data of equipment status to generate status correction data.

[0106] In this embodiment of the invention, the boundary data of the time-series-topology association window is split into time series segments. Specifically, the device status data within each window is divided into several continuous state sequence segments according to a preset sampling interval, thereby obtaining segmented state sequence data. Each segment reflects the state changes of the device within a specific time interval, providing basic data for subsequent feature analysis. Local change features are extracted based on the segmented state sequence data. The extracted content includes the switching frequency of device states, the duration of states, and the amplitude of abnormal fluctuations, forming local state feature data. By analyzing these features, the operational fluctuations and potential anomalies of the device in a short time scale can be revealed, providing a basis for device state classification.

[0107] Preliminary cluster analysis is performed using local state feature data. Cluster analysis identifies patterns in equipment state changes, grouping state sequences with similar fluctuation characteristics into the same category, thus generating preliminary state category labels. This step effectively groups equipment states according to their change trends and fluctuation characteristics, providing a preliminary classification basis for further risk assessment. The preliminary state category labels are compared and verified with window boundary data to check the consistency of the preliminary classification. For sequences with anomalies, missing, or incomplete data, necessary corrections are made to ensure the accuracy and completeness of the classification results, thereby generating preliminary equipment state classification data. This step eliminates biases caused by data anomalies or missing sampling, ensuring the reliability of the classification results. Disturbance sensitivity analysis is performed on the preliminary equipment state classification data. Specifically, this includes assessing the impact of external disturbances (such as load changes, environmental fluctuations, or anomalies in neighboring equipment) on equipment state classification, and adjusting and optimizing the preliminary classification based on the analysis results to generate state correction data. Through the above methods, corrected equipment state classification results can be obtained, providing stable and reliable input for the quantification of key equipment risk parameters, enabling accurate monitoring and dynamic risk assessment of power plant equipment states.

[0108] Preferably, the disturbance sensitivity analysis of the preliminary equipment status classification data includes:

[0109] For each preliminary classification state, a disturbance signal of preset amplitude is applied, and the change response of the equipment status data is observed;

[0110] Analyze the magnitude and trend of change of each preliminary classification state under different perturbation conditions, and generate the perturbation sensitivity characteristics of each state;

[0111] Based on the perturbation sensitivity characteristics, the bias or misjudgment area in the preliminary classification state is determined;

[0112] Based on the deviation or misjudgment area, the preliminary classification status data is corrected to generate corrected equipment status data.

[0113] In this embodiment of the invention, a disturbance signal of preset amplitude is applied to each category of the preliminary equipment status classification data. The disturbance signal can simulate load fluctuations, environmental changes, or abnormal status of neighboring equipment that occur during actual operation. By applying the disturbance, the response of each key piece of equipment's status data is observed, thereby assessing the stability and reliability of the equipment status under external disturbances. The amplitude and trend of change of each preliminary category under different disturbance conditions are analyzed. By statistically analyzing the changes in equipment status under disturbance, the disturbance sensitivity characteristics of each status can be obtained, including the intensity of status fluctuations, the rate of change, and the frequency of abnormal status occurrences. These characteristics reflect the sensitivity of the preliminary category status to disturbances, providing a basis for identifying existing classification biases.

[0114] Next, based on disturbance sensitivity characteristics, biased or misjudged regions in the initial classification are identified. Specifically, state sequences exhibiting significant changes or unstable responses under disturbances are marked as potential biased or misjudged regions, indicating that these states failed to accurately reflect the actual operating status of the equipment in the initial classification. The initial classification data is then corrected based on these biased or misjudged regions. During the correction process, state sequences significantly affected by disturbances are re-evaluated and adjusted according to their true response characteristics to generate corrected equipment status data. This method improves the accuracy and robustness of equipment status classification, providing reliable foundational data for the subsequent quantification of risk parameters for critical equipment, and enabling precise monitoring and dynamic risk assessment of the operating status of key power plant equipment.

[0115] Preferably, step S34 includes the following steps:

[0116] Step S341: Based on the status correction data, for each key device, statistically analyze the frequency and magnitude of its status data anomalies within a preset time period;

[0117] Step S342: Combine the frequency and magnitude of the anomalies to calculate the risk score for each critical device;

[0118] Step S343: Generate risk parameters for key equipment based on the risk score.

[0119] In this embodiment of the invention, statistical analysis is performed on the state changes of each key device within a preset time period based on the device state data after disturbance sensitivity analysis and correction. Specifically, the number of abnormal states that occur in each device within this time period is recorded, and the magnitude and severity of each abnormal state are statistically analyzed to obtain the frequency and magnitude information of abnormal occurrences for each key device. This step can comprehensively reflect the potential risk behaviors and abnormal characteristics of the equipment during actual operation. By combining the abnormal frequency and abnormal magnitude, a risk assessment is performed on each key device. By comprehensively considering the number of abnormal occurrences and the magnitude of abnormalities, a risk score can be obtained for each device, reflecting the level of its potential risk and the severity of abnormal states. The risk score calculation method makes full use of historical abnormal data and real-time monitoring information, enabling the risk assessment to dynamically reflect the current safety status of the equipment. Among them, historical abnormal data refers to the abnormal records accumulated by the equipment during past operation, including the number of abnormal occurrences, the magnitude of abnormalities, and historical fault or alarm information, mainly used to statistically analyze the frequency of abnormalities and identify long-term potential risks; real-time monitoring information refers to the state data collected by the equipment in the current or most recent time window, such as voltage, temperature, vibration, and other indicators, mainly used to measure the magnitude of abnormalities and the current state of the equipment.

[0120] Specifically, risk assessment requires a comprehensive analysis combining the frequency and magnitude of anomalies in critical equipment. Anomaly frequency refers to the number of times equipment malfunctions within a given time window, while anomaly magnitude refers to the degree to which each anomaly deviates from the normal value. By considering both indicators simultaneously, a risk score can be obtained for each critical piece of equipment, reflecting the level of potential risk and the severity of the anomaly. The risk score can be calculated by weighting the anomaly frequency and magnitude with certain weights, for example, 50% for both. Then, the weighted sum of the anomaly data for each piece of equipment within the time window yields the risk score. For instance, suppose transformer A in a power plant experiences three anomalies within a certain hour, with magnitudes of 2%, 4%, and 3% respectively; equipment B experiences one anomaly within the same time window, with a magnitude of 5%. After standardizing the frequency and amplitude of anomalies, the comprehensive anomaly index for transformer A is: a standardized value of 0.6 for 3 occurrences of frequency, a standardized value of 0.4 for an average amplitude of 3%, and a risk score calculated by weighting: 0.5 × 0.6 + 0.5 × 0.4 = 0.5. For equipment B, the standardized value for 1 occurrence of anomaly is 0.2, the standardized value for 5% amplitude is 0.7, and the risk score is 0.5 × 0.2 + 0.5 × 0.7 = 0.45. This shows that although transformer A has smaller amplitudes per anomaly, it experiences more anomalies, resulting in a slightly higher risk score than equipment B. Conversely, equipment B has larger amplitude anomalies but fewer anomalies, resulting in a slightly lower risk score. This calculation method can simultaneously reflect the frequency and severity of anomalies, providing a basis for equipment risk ranking and management.

[0121] Based on the risk scores of each key piece of equipment, corresponding risk parameters are generated. These risk parameters quantify the operational risk level of each key piece of equipment, reflecting both the frequency of equipment anomalies and the magnitude and intensity of potential threats. The generated risk parameters serve as the foundational data for subsequent equipment early warning and safety management, enabling precise monitoring, risk identification, and dynamic safety management of key power plant equipment. The threshold setting method can be a fixed threshold method: directly setting the threshold corresponding to the risk score based on industry standards or equipment operating experience. For example: a risk score ≤ 1.0 indicates low risk; a risk score of 1.0–2.0 indicates medium risk; and a risk score > 2.0 indicates high risk.

[0122] As an example of the present invention, reference is made to... Figure 3 As shown, step S4 in this example includes:

[0123] Step S41: Perform synchronous delay correction on the risk parameter sequence of key equipment to generate delay-corrected risk parameter data;

[0124] Step S42: Based on the risk parameter data after delay correction, calculate the gradient vector between adjacent sampling points and generate a gradient vector sequence;

[0125] Step S43: Perform angle calculation on the gradient vector sequence to obtain the gradient direction angle distribution data;

[0126] Step S44: Based on the risk parameter data after delay correction, calculate the amplitude consistency index and perform joint discrimination with the gradient direction angle distribution data to generate potential risk equipment labeling data.

[0127] In this embodiment of the invention, a synchronization delay correction is performed on the risk parameter sequence of key equipment. Specifically, this involves time alignment and delay correction of the risk parameter sequence of each key device based on the acquisition time interval and transmission delay of different data sources. This eliminates errors caused by asynchronous data acquisition or transmission delays, thereby generating delay-corrected risk parameter data. This step ensures the temporal consistency of risk parameters for each device in subsequent analysis, improving the accuracy of risk assessment. For example, when performing synchronization delay correction on the risk parameter sequence of key equipment, the acquisition time interval and data transmission delay of different data sources must first be considered. For instance, the production operation data of device A is acquired once per second, the status monitoring data is acquired once every five seconds, and there is a two-second delay in the transmission of the status monitoring data. To ensure that risk parameters from different data sources can be compared and synthesized within the same time window, time alignment and delay correction are required for each risk parameter sequence. For example, suppose the risk parameters of transformer A's production operation data are 1.2, 1.3, and 1.4 at 0, 1, 2, ... within a certain time window, and the risk parameters of its condition monitoring data are 1.5, 1.6, and 1.7 at 0, 5, and 10 seconds, respectively. However, due to transmission delay, the actual arrival times of this sequence are 2, 7, and 12 seconds. By using synchronous delay correction, the condition monitoring data sequence is shifted forward by two seconds, and the data at different time points are interpolated or aligned. This ensures that the risk parameters corresponding to each data source can match at the same time point, forming a unified risk parameter sequence. For example, at the 2nd second, both the production operation data risk parameter 1.4 and the condition monitoring data risk parameter 1.5 are available, guaranteeing the accuracy of subsequent comprehensive risk analysis and assessment. Subsequently, based on the delay-corrected risk parameter data, the risk change trend between adjacent sampling points is analyzed, and a gradient vector sequence is calculated. Each gradient vector reflects the direction and magnitude of change of the equipment risk parameters between adjacent time points, providing a reference for identifying potential abnormal trends. By generating a sequence of gradient vectors, one can intuitively understand the changing characteristics of risk parameters over time.

[0128] Next, directional analysis is performed on the gradient vector sequence to calculate the angle between adjacent gradient vectors and generate gradient direction angle distribution data. This data reflects the continuity and consistency of changes in equipment risk parameters. Larger angles indicate abnormal risk change directions or sudden local anomalies, providing a basis for identifying potentially risky equipment. Based on the delay-corrected risk parameter data, an amplitude consistency index is calculated, which compares the relative differences in the amplitudes of risk parameters at adjacent sampling points. The amplitude consistency index is jointly judged with the gradient direction angle distribution data to comprehensively determine the changing trend and amplitude deviation of equipment risk parameters, thereby generating potential risky equipment labeling data. This method can accurately identify key equipment exhibiting abnormal trends or amplitude anomalies during risk parameter changes, providing a reliable foundation for subsequent dynamic early warning and safety management.

[0129] Preferably, the method for confirming adjacent sampling points includes:

[0130] Signals from key equipment are sampled to obtain basic sampling points;

[0131] The basic sampling points are arranged according to the risk parameter sequence of key equipment to obtain an ordered set of sampling points;

[0132] In the ordered set of sampling points, with each target sampling point as the center, its previous sampling point and its next sampling point are extracted to form candidate adjacent point pairs;

[0133] Perform interval consistency detection on candidate adjacent point pairs. When the time interval between adjacent sampling points is within the preset time threshold range, the point pair is confirmed as a valid adjacent sampling point.

[0134] When the time interval between adjacent sampling points exceeds the preset time threshold, a proximity compensation strategy is adopted to use the effective sampling point closest to the target sampling point as a substitute neighbor, thereby generating the final pair of adjacent sampling points.

[0135] In this embodiment of the invention, signal sampling is performed on key equipment in a power plant to obtain the equipment's operating status information at different points in time, forming basic sampling points. Each basic sampling point records the equipment's status parameters and related monitoring information at the sampling time, providing a data foundation for the confirmation of subsequent adjacent sampling points. Based on the risk parameter sequence of the key equipment, the basic sampling points are arranged in chronological order to generate an ordered sampling point set. This chronological sorting ensures that the sampling points reflect the continuous changes in equipment status within the sequence, providing accurate temporal basis for analyzing the changing trends of equipment risk parameters. Within the ordered sampling point set, the preceding and following sampling points are extracted from each target sampling point to form candidate adjacent point pairs. These candidate adjacent point pairs represent the changing relationship of equipment status within a continuous time interval and are the basis for gradient analysis and risk trend determination.

[0136] Interval consistency checks are performed on candidate adjacent point pairs. Specifically, the time interval between two points is checked to see if it falls within a preset time threshold. When the time interval meets the preset requirement, the point pair is confirmed as a valid adjacent sampling point for subsequent gradient calculation and risk analysis. When the time interval of a candidate adjacent point pair exceeds the preset time threshold, a proximity compensation strategy is adopted. This strategy selects the valid sampling point closest to the target sampling point as a substitute adjacent point, thereby ensuring that the adjacent sampling point pairs of all critical equipment meet the requirements of continuity and time consistency. Through the above method, the final adjacent sampling point pairs can be generated, providing reliable basic data for risk parameter gradient analysis, angle discrimination, and identification of potential risk equipment.

[0137] Preferably, step S44 includes:

[0138] Based on the risk parameter data after delay correction, the amplitude change ratio of adjacent data segments is calculated point by point to obtain the amplitude consistency index sequence.

[0139] Divide the gradient direction angle distribution data into intervals, extract the angle exceeding the limit interval and the normal interval, and generate angle discrimination data;

[0140] The amplitude consistency index sequence and the angle discrimination data are cross-mapped to obtain the joint discrimination result data;

[0141] Based on the joint discrimination results, key equipment is marked one by one, and the potentially risky equipment marking data is output.

[0142] In this embodiment of the invention, the amplitude changes of each data point and its adjacent data segments are analyzed based on the risk parameter data of key equipment after synchronous delay correction. Specifically, the amplitude change ratio of adjacent data segments is calculated point by point to obtain an amplitude consistency index sequence. This sequence reflects the amplitude change characteristics of the risk parameters of each key piece of equipment between continuous sampling points, providing basic information for identifying abnormal fluctuations. Subsequently, the aforementioned gradient direction angle distribution data is divided into intervals. By analyzing the numerical range of the gradient direction angle, it is divided into normal intervals and out-of-limit intervals, generating angle discrimination data. The normal interval indicates that the risk parameter change trend is continuous and stable, while the out-of-limit interval indicates that the risk parameter change direction is abnormal, with the existence of sudden risks or local anomalies. For example, assuming that in the gradient direction angle measurement of equipment in a power plant, historical operating data shows that under normal conditions, the angle is generally between 0° and 30°, while angles exceeding 30° are often related to abnormal vibrations or equipment offset. Based on this experience, the included angle can be divided as follows: Normal range: 0°~30°, corresponding to "equipment operating normally"; Exceeding limit range: greater than 30°, corresponding to "potentially abnormal or risky state". For example, if the gradient direction included angle data sequence collected within a certain time window is: 12°, 25°, 32°, 28°, 35°, then after interval division: 12°, 25°, and 28° belong to the normal range; 32° and 35° belong to the exceeding limit range. Based on this division, included angle discrimination data can be generated, that is, marking whether each included angle exceeds the limit, providing a basis for subsequent risk assessment and anomaly judgment.

[0143] A cross-mapping analysis is performed between the amplitude consistency index sequence and the gradient direction discrimination data. By jointly comparing the amplitude change characteristics and the gradient direction change characteristics, a joint discrimination result is obtained. This result can simultaneously consider the cases of abnormal amplitude and abnormal change direction of risk parameters, improving the accuracy and reliability of identifying potential risk equipment. For example, suppose that within a certain time window, the amplitude consistency index sequence collected by transformer A is: 0.95, 0.88, 0.76, 0.80, 0.92, where a threshold of 0.85 is set, and values ​​below 0.85 are considered abnormal amplitudes; the gradient direction discrimination data are: normal, normal, exceeding limit, normal, exceeding limit. The two sets of data are cross-mapped according to time correspondence, as shown in the table below:

[0144]

[0145] Cross-mapping analysis reveals that the simultaneous occurrence of amplitude anomalies and angle exceeding limits at time point t3 indicates a significantly abnormal equipment state at that moment; while the occurrence of only one type of anomaly at time points t4 and t5 suggests potential risks requiring attention. Based on the joint discrimination results, each critical piece of equipment is individually marked. When the joint discrimination results show anomalies in both amplitude and gradient direction of a device, that device is marked as a potential risk device, thus outputting complete potential risk device marking data. This method can accurately identify critical equipment exhibiting abnormal trends or amplitude anomalies during risk parameter changes, providing reliable basic data support for dynamic early warning and safety management of power plant equipment.

[0146] More specifically, when marking critical equipment for risk assessment, a joint discrimination method can be used, combining two indicators: gradient direction angle and relative amplitude difference. Specific discrimination conditions include whether the gradient direction angle exceeds a preset angle threshold, such as 30°, and whether the relative amplitude difference exceeds a preset percentage threshold, such as 15%. Joint discrimination typically uses an "OR" logic, meaning that if either indicator exceeds the threshold, the equipment is marked as potentially risky; only when both indicators do not exceed the threshold is it marked as normal equipment. For example, equipment A has a gradient direction angle of 32° (exceeding the threshold) and a relative amplitude difference of 10% (not exceeding the threshold) within a certain time window; because the angle exceeds the threshold, it is marked as potentially risky equipment. Equipment B has a gradient direction angle of 28° (not exceeding the threshold) and a relative amplitude difference of 18% (exceeding the threshold), with the amplitude abnormally exceeding the threshold, and is therefore also marked as potentially risky equipment. Equipment C has a gradient direction angle of 25° and a relative amplitude difference of 12%; since neither indicator exceeds the threshold, it is marked as normal equipment. This joint discrimination method can simultaneously consider the directionality and amplitude changes of the equipment's operating status. As long as any one of them is abnormal, potential risks can be identified in a timely manner, thereby improving the coverage of anomaly detection and the sensitivity of risk identification, and providing a reliable basis for subsequent risk assessment and management.

[0147] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0148] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A smart power plant safety early warning method based on multi-data coupling, characterized in that, Includes the following steps: Step S1: Acquire multi-source coupled data including power plant production and operation data, key equipment status monitoring data, environmental monitoring data and historical accident records, and calculate the acquisition time interval and synchronization delay between each data source; Step S2: Define the ratio of the acquisition time interval to the topological distance between the corresponding nodes of the key equipment as the coupling propagation rate. The method for obtaining the topological distance between the corresponding nodes of the key equipment includes: Nodes are extracted from power plant production and operation data and key equipment status monitoring data to generate key equipment node set data; A node connectivity graph is established based on the key equipment node set data to generate preliminary topology connection data; Path weights are assigned to the initial topology connection data to generate weighted topology path data; Optimal path search is performed based on weighted topology path data to calculate the shortest weighted path length between key equipment node pairs, thereby obtaining the topology distance between corresponding nodes of key equipment. Step S3: Construct a time-series-topology correlation window based on the coupling propagation rate, and determine the state type of critical equipment based on the time-series-topology correlation window; obtain the risk parameters of critical equipment by quantifying the state type; Step S3 includes the following steps: Step S31: Divide the coupled propagation rate data into time scales to generate multi-level time segment data; Step S32: Based on the topological path lengths in the multi-level time segment data and topological connection data, construct a time-series-topological association window and generate window boundary data; Step S33: Perform device state sequence analysis on the window boundary data to generate preliminary device state classification data; perform disturbance sensitivity analysis on the preliminary device state classification data to generate state correction data. Specifically, step S33 involves: The window boundary data is split into time series segments, and device status sequence fragments are generated according to the sampling interval to obtain segmented status sequence data. Local change features are extracted from equipment state sequence segments, including state switching frequency, duration and abnormal fluctuation amplitude, to generate local state feature data; Preliminary cluster analysis is performed using local state feature data to classify equipment state sequences according to state change patterns and generate preliminary state category labels. The consistency of the initial classification is verified by combining the initial status category labels with window boundary data, and abnormal or incomplete sequences are corrected to generate initial classification data of device status. Disturbance sensitivity analysis is performed on the preliminary equipment condition classification data to generate condition correction data, specifically: For each preliminary classification state, a disturbance signal of preset amplitude is applied, and the change response of the equipment status data is observed; Analyze the magnitude and trend of change of each preliminary classification state under different perturbation conditions, and generate the perturbation sensitivity characteristics of each state; Based on the perturbation sensitivity characteristics, the bias or misjudgment area in the preliminary classification state is determined; Based on the deviation or misjudgment area, the preliminary classification status data is corrected to generate corrected equipment status data; Step S34: Perform quantitative calculations based on the condition correction data to generate risk parameters for key equipment; Step S4: Calculate the gradient vector and amplitude consistency index of the risk parameters of key equipment based on the synchronization delay. When the gradient direction angle exceeds the preset angle threshold or the relative difference in amplitude exceeds the preset proportion threshold, it is marked as a potential risk equipment. Step S5: Set the equipment monitoring and early warning timestamps in the status monitoring data of key equipment through potential risk equipment to perform dynamic power plant equipment safety early warning operations.

2. The smart power plant safety early warning method based on multi-data coupling according to claim 1, characterized in that, Establishing a node connectivity graph based on key equipment node set data includes: Classify the key equipment node set data to generate node function category data; Based on the node function category data, the direct coupling relationship between nodes is analyzed to generate initial connected edge data; Redundancy and conflict detection is performed on the initial connected edge data to generate edge conflict correction data; Topology consistency correction is performed based on edge conflict correction data to generate topology connectivity data; A graph of node connectivity is established using topological connection data.

3. The smart power plant safety early warning method based on multi-data coupling according to claim 1, characterized in that, Step S34 includes the following steps: Step S341: Based on the status correction data, for each key device, statistically analyze the frequency and magnitude of its status data anomalies within a preset time period; Step S342: Combine the frequency and magnitude of the anomalies to calculate the risk score for each critical device; Step S343: Generate risk parameters for key equipment based on the risk score.

4. The smart power plant safety early warning method based on multi-data coupling according to claim 1, characterized in that, Step S4 includes the following steps: Step S41: Perform synchronous delay correction on the risk parameter sequence of key equipment to generate delay-corrected risk parameter data; Step S42: Based on the risk parameter data after delay correction, calculate the gradient vector between adjacent sampling points and generate a gradient vector sequence; Step S43: Perform angle calculation on the gradient vector sequence to obtain the gradient direction angle distribution data; Step S44: Based on the risk parameter data after delay correction, calculate the amplitude consistency index and perform joint discrimination with the gradient direction angle distribution data to generate potential risk equipment labeling data.

5. The smart power plant safety early warning method based on multi-data coupling according to claim 4, characterized in that, The methods for confirming adjacent sampling points include: Signals from key equipment are sampled to obtain basic sampling points; The basic sampling points are arranged according to the risk parameter sequence of key equipment to obtain an ordered set of sampling points; In the ordered set of sampling points, with each target sampling point as the center, its previous sampling point and its next sampling point are extracted to form candidate adjacent point pairs; Perform interval consistency detection on candidate adjacent point pairs. When the time interval between adjacent sampling points is within the preset time threshold range, the point pair is confirmed as a valid adjacent sampling point. When the time interval between adjacent sampling points exceeds the preset time threshold, a proximity compensation strategy is adopted to use the effective sampling point closest to the target sampling point as a substitute neighbor, thereby generating the final pair of adjacent sampling points.

6. The smart power plant safety early warning method based on multi-data coupling according to claim 4, characterized in that, Step S44 includes: Based on the risk parameter data after delay correction, the amplitude change ratio of adjacent data segments is calculated point by point to obtain the amplitude consistency index sequence. Divide the gradient direction angle distribution data into intervals, extract the angle exceeding the limit interval and the normal interval, and generate angle discrimination data; The amplitude consistency index sequence and the angle discrimination data are cross-mapped to obtain the joint discrimination result data; Based on the joint discrimination results, key equipment is marked one by one, and the potentially risky equipment marking data is output.

Images