An operating safety warning method and system applicable to waste power plants

Abstract

The present invention discloses an operation safety warning method and system applicable to a waste power plant, relating to the technical field of safety warning. The method includes: performing operation monitoring on a target waste power plant to obtain P operation monitoring data sets; performing outlier data identification to obtain P outlier operation monitoring data sets, and performing anomaly authentication to obtain Q abnormal outlier operation monitoring data sets; performing prediction on the types of operation anomalies to determine Q predicted identification node anomaly types; performing retrieval in a pre-constructed node association network to determine Q associated node sets and Q associated node anomaly type sets; performing safety anomaly identification on the Q predicted identification nodes, and obtaining operation safety warning information according to the safety anomaly results. The technical problem of insufficient accuracy and real-time performance of operation safety monitoring in the prior art for waste power plants is solved, and the technical effect of improving the accuracy and real-time performance of operation safety warning for waste power plants is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of safety early warning, and in particular to an operation safety early warning method and system applicable to a garbage power plant. Background Art

[0002] With the widespread application of waste incineration power generation technology, waste power plants play an important role in the energy production process. However, the operating environment of waste power plants is complex and susceptible to multiple factors, such as equipment failure, operational errors or changes in the external environment. These factors may cause abnormal equipment operation and even cause safety accidents. Therefore, how to effectively monitor and predict the operating status of waste power plants and identify potential safety risks in a timely manner has become a technical problem that needs to be solved urgently in the industry. Existing safety monitoring systems for waste power plants usually rely on traditional monitoring methods, such as regular inspections and manual inspections, but these methods often cannot provide real-time and effective early warnings when faced with complex operating conditions and sudden safety incidents. Summary of the invention

[0003] The present application provides an operation safety early warning method and system applicable to a waste power plant, which solves the technical problem of insufficient accuracy and real-time performance of operation safety monitoring of a waste power plant in the prior art.

[0004] In view of the above problems, the present application provides an operation safety early warning method and system applicable to a waste power plant.

[0005] In a first aspect of the present application, there is provided an operation safety early warning method applicable to a waste power plant, the method comprising:

[0006] In a preset monitoring window, the sensor topology network is used to monitor the operation of the target waste power plant to obtain P operation monitoring data sets, where P is an integer greater than or equal to 1; the P operation monitoring data sets are traversed to identify outlier data to obtain P outlier operation monitoring data sets, and abnormal authentication is performed to obtain Q abnormal outlier operation monitoring data sets, where Q is a positive integer less than or equal to P; based on the Q abnormal outlier operation monitoring data sets, operation abnormality types are predicted to determine the abnormal types of Q prediction and identification nodes; with the Q prediction and identification node abnormal types and the corresponding Q prediction and identification nodes as indexes, a search is performed in the pre-constructed node association network to determine the Q associated node sets and the Q associated node abnormal type sets; the Q associated node sets, the Q associated node abnormal type sets and the Q prediction and identification node abnormal types are combined to perform safety abnormality identification on the Q prediction and identification nodes, and operation safety warning information is obtained according to the safety abnormality results.

[0007] The second aspect of the present application provides an operation safety early warning system applicable to a waste power plant, the system comprising:

[0008] A monitoring module, configured to monitor the operation of a target waste power plant within a preset monitoring window by using a sensor topology network, and obtain P sets of operation monitoring data, where P is an integer greater than or equal to 1; a data identification module, configured to traverse the P sets of operation monitoring data to identify outlier data, obtain P sets of outlier operation monitoring data, perform anomaly authentication, and obtain Q sets of abnormal outlier operation monitoring data, where Q is a positive integer less than or equal to P; an anomaly type prediction module, configured to predict the operation anomaly type based on the Q sets of abnormal outlier operation monitoring data to determine Q predicted identification node anomaly types; a retrieval module, configured to retrieve in a pre-constructed node association network by using the Q predicted identification node anomaly types and the corresponding Q predicted identification nodes as indexes to determine Q sets of associated nodes and Q sets of associated node anomaly types; an anomaly identification module, configured to perform security anomaly identification on the Q predicted identification nodes by combining the Q sets of associated nodes, the Q sets of associated node anomaly types, and the Q predicted identification node anomaly types, and obtain operation security warning information according to the security anomaly result.

[0009] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0010] First, within a preset monitoring window, the operation of a target waste power plant is monitored by using a sensor topology network to obtain P sets of operation monitoring data, where P is an integer greater than or equal to 1. Then, the P sets of operation monitoring data are traversed to identify outlier data, obtain P sets of outlier operation monitoring data, perform anomaly authentication, and obtain Q sets of abnormal outlier operation monitoring data, where Q is a positive integer less than or equal to P. Further, the operation anomaly type is predicted based on the Q sets of abnormal outlier operation monitoring data to determine Q predicted identification node anomaly types. Then, by using the Q predicted identification node anomaly types and the corresponding Q predicted identification nodes as indexes, a retrieval is performed in a pre-constructed node association network to determine Q sets of associated nodes and Q sets of associated node anomaly types. Finally, the Q predicted identification nodes are subjected to security anomaly identification by combining the Q sets of associated nodes, the Q sets of associated node anomaly types, and the Q predicted identification node anomaly types, and operation security warning information is obtained according to the security anomaly result. This solves the technical problem of insufficient accuracy and real-time performance in the operation safety monitoring of waste power plants in the prior art, and achieves the technical effect of improving the accuracy and real-time performance of the operation safety warning of waste power plants. Description of the Drawings

[0011] To more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for use in the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0012] Figure 1 Schematic flow diagram of an operation safety early warning method applicable to a waste power plant provided by an embodiment of the present application;

[0013] Figure 2 Schematic structural diagram of an operation safety early warning system applicable to a waste power plant provided by an embodiment of the present application.

[0014] Explanation of reference numerals: monitoring module 11, data recognition module 12, abnormal type prediction module 13, retrieval module 14, abnormal recognition module 15. Detailed implementation manners

[0015] By providing an operation safety early warning method and system applicable to a waste power plant, the present application solves the technical problems of insufficient accuracy and real-time performance in the operation safety monitoring of waste power plants in the prior art.

[0016] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0017] It should be noted that the terms "include" and "have" are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units does not necessarily have to be limited to those clearly listed steps or units, but may include other steps or modules that are not clearly listed or are inherent to these processes, methods, products, or devices.

[0018] Embodiment 1, as Figure 1 shown, the present application provides an operation safety early warning method applicable to a waste power plant, where the method includes:

[0019] Within a preset monitoring window, use a sensor topology network to monitor the operation of the target waste power plant to obtain P operation monitoring data sets, where P is an integer greater than or equal to 1.

[0020] Within a preset monitoring window, real-time operation monitoring of a waste power plant is carried out through a sensor topology network, which refers to a network composed of multiple sensor nodes responsible for real-time monitoring of various operation data of the waste power plant; the data collected by these sensor nodes form P operation monitoring data sets, where P is the number of sensor nodes in the sensor topology network. That is to say, each sensor node will collect and transmit certain operation data, and the data sets of all sensor nodes constitute P data sets.

[0021] Traverse the P operation monitoring data sets for outlier data identification to obtain P outlier operation monitoring data sets, and perform anomaly authentication to obtain Q outlier operation monitoring data sets with anomalies, where Q is a positive integer less than or equal to P.

[0022] During the monitoring process, the operation data of the waste power plant usually has a certain fluctuation range, and this kind of fluctuation is normal. For example, sensors may generate different measurement values according to different working conditions or environmental factors. Normally, the data should fluctuate up and down within a specific range, and these fluctuations reflect the normal operation state of the equipment. Identifying outlier data from the P operation monitoring data sets is to find data points that exceed the normal fluctuation range from these data, that is, to obtain P outlier operation monitoring data sets. These data points that exceed the range may represent potential anomalies, such as equipment failures, sensor problems, etc. Then, perform anomaly authentication on the P outlier operation monitoring data sets, that is, confirm whether these abnormal data really represent a safety issue. Usually, the anomaly authentication of outlier data can be verified by setting a certain threshold to check whether these data points exceed the normal range. After anomaly authentication, Q outlier operation monitoring data sets with anomalies are obtained, representing the outlier data points determined to be abnormal after authentication.

[0023] Furthermore, traversing the P operation monitoring data sets for outlier data identification and obtaining P outlier operation monitoring data sets includes:

[0024] Traverse the P sets of operation monitoring data to calculate the mean values, obtaining P mean values of operation monitoring data; take the P mean values of operation monitoring data as P aggregation starting points, take the P aggregation starting points as neighborhood centers, and take a preset aggregation bandwidth as the radius to construct P neighborhoods of aggregation starting points in the P sets of operation monitoring data; iterate the P aggregation starting points in the P sets of operation monitoring data in the first direction and the second direction respectively according to the preset aggregation bandwidth, and combine the P neighborhoods of aggregation starting points to determine P updated iteration directions and P updated iteration starting points; based on the P updated iteration directions, iterate the P updated iteration starting points in the P sets of operation monitoring data according to the preset aggregation bandwidth until the preset number of iterations is satisfied, obtaining P target aggregation points and P neighborhoods of target aggregation points; take the P target aggregation points as neighborhood centers and take the preset aggregation bandwidth as the radius to construct the P neighborhoods of target aggregation points; diffuse outward the P neighborhoods of target aggregation points according to the preset aggregation bandwidth to obtain P diffusion stop neighborhoods of target aggregation points; add the data in the P sets of operation monitoring data that are outside the P diffusion stop neighborhoods of target aggregation points into P sets of outlier operation monitoring data respectively.

[0025] Specifically, traverse the P sets of operation monitoring data, and calculate the mean value for each data set, obtaining P mean values of operation monitoring data, which represent the average operation state of the node within the preset monitoring window and serve as the aggregation starting points for each sensor node; take these aggregation starting points as neighborhood centers and set a preset aggregation bandwidth as the radius to construct P neighborhoods of aggregation starting points in the operation monitoring data sets. At this time, each data point will be divided into the corresponding neighborhood according to its distance from the aggregation starting point; iterate and update in the first direction and the second direction according to the preset aggregation bandwidth, where the first direction and the second direction represent two different directions, left and right. By adjusting the position and direction of the aggregation starting point, the definition of the neighborhood is gradually optimized to ensure more accurate aggregation of relevant data. Each iteration will combine the data in the current neighborhood of the aggregation starting point to calculate and update the new aggregation direction and position until the preset number of iterations is satisfied, obtaining P target aggregation points and P neighborhoods of target aggregation points. With the determination of the target aggregation points, then based on these aggregation points, continue to expand the neighborhoods of the target aggregation points with the same preset aggregation bandwidth to form the diffusion stop neighborhoods of the target aggregation points. Finally, all the data points outside these diffusion stop neighborhoods, that is, the data points that deviate significantly from the normal operation mode, will be identified as outlier data and added to the P sets of outlier operation monitoring data respectively.

[0026] Furthermore, iterating the P aggregation starting points in the P sets of operation monitoring data in the first direction and the second direction respectively according to the preset aggregation bandwidth, and combining the P neighborhoods of aggregation starting points to determine P updated iteration directions and P updated iteration starting points includes:

[0027] Iteratively process the P clustering starting points in the P sets of running monitoring data in the first direction and the second direction respectively according to a preset clustering bandwidth, to obtain P first-direction clustering iterative starting points and P second-direction clustering iterative starting points; construct P first-direction clustering iterative starting point neighborhoods and P second-direction clustering iterative starting point neighborhoods, and calculate the densities of the P first-direction clustering iterative starting point neighborhoods and the densities of the P second-direction clustering iterative starting point neighborhoods; calculate the neighborhood densities of the P clustering starting point neighborhoods to obtain P clustering starting point neighborhood densities; respectively determine whether the densities of the P first-direction clustering iterative starting point neighborhoods and the densities of the P second-direction clustering iterative starting point neighborhoods are both greater than the P clustering starting point neighborhood densities. If not, use the P clustering starting points as P updated iterative starting points, and use iteration in any direction as P updated iterative directions.

[0028] Specifically, perform iterative updates in the first direction and the second direction respectively according to a preset clustering bandwidth. During each iteration, for each clustering starting point in the P sets of running monitoring data, move along these two directions respectively, and calculate P first-direction clustering iterative starting points and P second-direction clustering iterative starting points. These new iterative starting points represent the new positions of each clustering starting point after one iteration update in the two directions. Then, based on the new clustering iterative starting points, construct P first-direction clustering iterative starting point neighborhoods and P second-direction clustering iterative starting point neighborhoods respectively; through these neighborhoods, the density of each clustering iterative starting point neighborhood can be further calculated, that is, the densities of the P first-direction clustering iterative starting point neighborhoods and the densities of the P second-direction clustering iterative starting point neighborhoods. These density values reflect the degree of data concentration in each neighborhood. A neighborhood with a larger density means that the data points in this area are more clustered, which may represent the normal pattern of the data. At the same time, it is also necessary to calculate the neighborhood density of each clustering starting point neighborhood and obtain P clustering starting point neighborhood densities. These density values are used to measure the normal data distribution near the clustering starting points and provide a basis for subsequent outlier data judgment. Then, compare the densities of the P first-direction clustering iterative starting point neighborhoods and the densities of the P second-direction clustering iterative starting point neighborhoods, and determine whether they are both greater than the P clustering starting point neighborhood densities; if the neighborhood density in a certain direction does not meet this condition, it means that this direction may not effectively cluster reasonable data points, then the clustering starting point needs to remain in its original position, and choose to iterate in other directions. Therefore, these original clustering starting points will be used as P updated iterative starting points, and choose iteration in any direction as P updated iterative directions to continue to optimize the positions of the clustering starting points.

[0029] Furthermore, respectively determining whether the densities of the P first-direction clustering iterative starting point neighborhoods and the densities of the P second-direction clustering iterative starting point neighborhoods are both greater than the P clustering starting point neighborhood densities further includes:

[0030] If so, when the neighborhood density of the P first - direction aggregation iteration starting points is greater than the neighborhood density of the P second - direction aggregation iteration starting points, take the P first - direction aggregation iteration starting points as the P updated iteration starting points, and take the P first directions as the P updated iteration directions; when the neighborhood density of the P first - direction aggregation iteration starting points is less than the neighborhood density of the P second - direction aggregation iteration starting points, take the P second - direction aggregation iteration starting points as the P updated iteration starting points, and take the P second directions as the P updated iteration directions; when the neighborhood density of the P first - direction aggregation iteration starting points is equal to the neighborhood density of the P second - direction aggregation iteration starting points, take the P aggregation starting points as the P updated iteration starting points, and take iteration in any direction as the P updated iteration directions.

[0031] In addition to determining whether the neighborhood densities of the P first - direction aggregation iteration starting points and the P second - direction aggregation iteration starting points are both greater than the neighborhood density of the P aggregation starting points, it is also necessary to further compare these two density values to determine the update direction and starting point of the iteration. Specifically, when the neighborhood density of the P first - direction aggregation iteration starting points is greater than the neighborhood density of the P second - direction aggregation iteration starting points, it indicates that the aggregation update along the first direction is more in line with the normal distribution of the data. Therefore, the P first - direction aggregation iteration starting points will be selected as the starting points of the updated iteration, and the first direction will be taken as the iteration direction to continue the subsequent iteration along the first direction. On the contrary, when the neighborhood density of the P first - direction aggregation iteration starting points is less than the neighborhood density of the P second - direction aggregation iteration starting points, it means that the aggregation effect in the second direction is better. Therefore, the P second - direction aggregation iteration starting points will be selected as the starting points of the updated iteration, and the second direction will be taken as the iteration direction to continue the update along the second direction. In the case where the neighborhood densities of the two are equal, it indicates that the aggregation effects in the two directions are not very different. At this time, it is no longer limited to a certain direction, but any direction can be flexibly selected for iterative update. Therefore, the P aggregation starting points will be used as the starting points of the updated iteration, and the iteration direction can be any direction.

[0032] Furthermore, expand outward from the neighborhoods of the P target aggregation points according to a preset aggregation bandwidth until a preset diffusion stop condition is met, and obtain the neighborhoods where the diffusion of the P target aggregation points stops, including:

[0033] Diffuse outward from the neighborhoods of the P target aggregation points according to a preset aggregation bandwidth to obtain P target aggregation point diffusion neighborhoods; determine whether the densities of the P target aggregation point diffusion neighborhoods of the P target aggregation point diffusion neighborhoods are greater than or equal to the densities of the P target aggregation point neighborhoods of the P target aggregation point neighborhoods. If so, continue to diffuse outward from the P target aggregation point diffusion neighborhoods according to the preset aggregation bandwidth until the preset diffusion stop condition is met. Take the diffusion neighborhood obtained from the last diffusion as the P target aggregation point stop diffusion neighborhood, where the preset diffusion stop condition is that the difference between the densities of two target aggregation point diffusion neighborhoods obtained from two adjacent diffusions is less than or equal to a preset density difference threshold.

[0034] Specifically, according to the preset aggregation bandwidth, diffuse the neighborhoods of the P target aggregation points to extend the target aggregation point neighborhoods outward, so as to include more running data by increasing the range of the neighborhoods, thereby accurately determining which data points conform to the normal running mode and which may be abnormal; Next, it is necessary to determine whether the density of the P target aggregation point diffusion neighborhoods after diffusion is greater than or equal to the density of the target aggregation point neighborhoods; If the density of the diffused neighborhood is large, it means that the extended data area still maintains a high degree of aggregation, indicating that the diffusion is effective and the aggregation pattern of the data has not changed significantly. In this case, continue to expand the diffusion neighborhood of the target aggregation point outward according to the preset aggregation bandwidth until the preset diffusion stop condition is met. The diffusion stop condition is that when the density difference of the target aggregation point diffusion neighborhoods between two adjacent diffusion results is less than or equal to the preset density difference threshold, stop the diffusion, that is, when the diffusion effect tends to be stable and the density difference of the neighborhoods becomes small, it means that the neighborhood of the target aggregation point has stabilized and no further diffusion is required. Finally, when the diffusion stop condition is met, the diffusion neighborhood obtained from the last diffusion is used as the P target aggregation point stop diffusion neighborhood. At this time, the neighborhood range of the target aggregation point has been determined, and the points included can be regarded as normal running data, while the data points not within the diffusion stop neighborhood range are identified as outlier data.

[0035] Predict the types of running anomalies based on Q sets of abnormal outlier running monitoring data, and determine Q predicted identification node anomaly types.

[0036] The Q sets of abnormal outlier operation monitoring data are obtained through outlier data identification and anomaly authentication. These data points represent abnormal data that deviates from the normal operation mode. Using data analysis methods (such as machine learning, statistical analysis, pattern recognition, etc.), in-depth analysis is carried out on the Q sets of abnormal outlier data. These data sets contain multiple possible abnormal patterns, and each data point may be related to different types of anomalies, such as equipment overload, temperature anomaly, pressure anomaly, vibration anomaly, etc. By analyzing the characteristics of these abnormal data, the anomaly type corresponding to each data point can be predicted. Specifically, relevant eigenvalue are extracted from the Q sets of abnormal outlier data, such as vibration frequency, temperature fluctuation, pressure change, etc. These eigenvalue can reflect the operating state of the equipment; according to historical operation data and known anomaly types, a prediction model is trained. Common models include Support Vector Machine (SVM), Random Forest, Decision Tree, etc.; the Q sets of abnormal outlier data are predicted through the model to determine the anomaly type of each data point; for each abnormal data point, the model outputs its corresponding anomaly type, such as equipment failure, over-temperature alarm, too low pressure, etc. Finally, based on these prediction results, the anomaly types of the Q prediction identification nodes can be clarified, thus providing clear guidance for subsequent safety warning, fault location and repair.

[0037] Taking the anomaly types of the Q prediction identification nodes and the corresponding Q prediction identification nodes as indexes, retrieve in the pre-constructed node association network to determine the Q sets of associated nodes and the Q sets of associated node anomaly types.

[0038] By indexing the anomaly types of the Q prediction identification nodes with the corresponding Q prediction identification nodes, retrieval can be carried out in the pre-constructed node association network to deeply explore the relevance and propagation path of abnormal data. The node association network is a structured graph model. The nodes in the network represent sensors, and the edges between the nodes represent their functional dependencies. The retrieval results include the Q sets of associated nodes and the Q sets of associated node anomaly types. The set of associated nodes is obtained by querying the node association network, which is a set of nodes directly or indirectly related to the prediction identification nodes; the set of associated node anomaly types is the set of anomaly types of these corresponding associated nodes, indicating whether they also have anomalies and the specific types of anomalies.

[0039] Furthermore, taking the anomaly types of the Q prediction identification nodes and the corresponding Q prediction identification nodes as indexes, retrieve in the pre-constructed node association network to determine the Q sets of associated nodes and the Q sets of associated node anomaly types, including:

[0040] Obtain M historical operation safety exception logs of the target waste power plant, where M is an integer greater than or equal to 1; traverse the M historical operation safety exception logs to extract exception nodes, and obtain K historical exception node sets and K historical exception node type sets, where K is an integer greater than or equal to 1; based on the K historical exception node sets and K historical exception node type sets, perform exception association on P nodes in the sensor topology network to obtain the node association network.

[0041] Obtain M historical operation safety exception logs from the historical data of the target waste power plant, where M is an integer greater than or equal to 1, representing the number of logs. These logs record past operation safety exception events, including key information such as the type of exception, the exception node, and the occurrence time. Next, traverse these M historical operation safety exception logs to extract the exception node information. By analyzing the exception situations recorded in the logs, K historical exception node sets and K historical exception node type sets can be obtained, where K is an integer greater than or equal to 1, representing the number of exception nodes extracted from the logs. These historical exception nodes represent the key locations where exceptions occurred during the operation of the waste power plant in the past, and the corresponding exception node types describe the specific failure types of these nodes, such as sensor failure, equipment failure, etc. After having these historical exception data, based on the K historical exception node sets and K historical exception node type sets, perform exception association analysis on P nodes in the sensor topology network. The P nodes in the sensor topology network represent the various sensors in the current monitoring system, and they have clear association relationships and functional dependencies during normal operation. Through the association analysis with historical exception nodes, the potential relationships between the current nodes and historical exception nodes can be identified, thereby constructing a node association network. This network reveals which nodes have potential exception propagation paths and correlations, helping the system better understand the mutual influences between different nodes.

[0042] Combine Q associated node sets, Q associated node exception type sets, and Q predicted recognition node exception types to perform safety exception recognition on Q predicted recognition nodes, and obtain operation safety warning information according to the safety exception results.

[0043] The set of Q associated nodes represents other nodes associated with the prediction recognition nodes. These nodes may have been affected by the anomalies of the prediction recognition nodes during operation, or there may be some functional dependencies between them. The set of Q associated node anomaly types describes whether these associated nodes also have anomalies and the specific types of their anomalies. The Q prediction recognition node anomaly types describe the anomalies of each prediction recognition node. Combining these anomaly types with the anomaly types of the associated nodes forms a more comprehensive fault recognition framework. Based on the set of Q associated nodes, the set of Q associated node anomaly types, and the Q prediction recognition node anomaly types, security anomaly recognition is performed to analyze whether there are potential security risks or fault chains. If the spread of anomaly patterns or the accumulation of potential faults is detected, corresponding operation security warning messages will be issued. These warning messages usually indicate the nodes where the faults occur, the types of faults, and the possible scope of influence, helping operation and maintenance personnel take timely measures to prevent the expansion of faults or the occurrence of more serious security problems.

[0044] Furthermore, combining the set of Q associated nodes, the set of Q associated node anomaly types, and the Q prediction recognition node anomaly types to perform security anomaly recognition on the Q prediction recognition nodes, and obtaining operation security warning messages according to the security anomaly results, including:

[0045] Extract the first associated node set, the first associated node anomaly type set, and the first prediction recognition node from the set of Q associated nodes, the set of Q associated node anomaly types, and the Q prediction recognition nodes;

[0046] Judge whether the Q prediction recognition nodes exist in the first associated node set. If so, when the number of the Q prediction recognition node anomaly types that are the same as the corresponding first associated node anomaly type set exceeds the preset quantity threshold, the first prediction recognition node is regarded as the first abnormal node;

[0047] And so on, traverse the Q prediction recognition nodes to perform security anomaly recognition, obtain the set of abnormal nodes, use the set of abnormal nodes as the security anomaly result, and obtain operation security warning messages according to the security anomaly result.

[0048] Specifically, the first set of associated nodes, the first set of abnormal types of associated nodes, and the first set of predicted recognition nodes are extracted from Q sets of associated nodes, Q sets of abnormal types of associated nodes, and Q predicted recognition nodes. Then, it is determined whether there are Q predicted recognition nodes in the first set of associated nodes. If so, the matching situation between the abnormal types of the Q predicted recognition nodes and the first set of abnormal types of associated nodes will be further analyzed. If the number of the same abnormal types between the predicted recognition nodes and the first set of associated nodes exceeds the preset quantity threshold, it indicates that there is a strong abnormal correlation between this part of the predicted recognition nodes and the associated nodes, and they may be part of the same fault chain. Therefore, the first predicted recognition nodes are marked as the first abnormal nodes, indicating that these nodes are one of the key points where the abnormality occurs. Subsequently, the same security anomaly recognition process is performed by traversing all Q predicted recognition nodes in sequence. If the matching of the abnormal types between a certain predicted recognition node and its associated nodes exceeds the threshold, this node will be marked as an abnormal node, and the recognition of other nodes will continue until all predicted recognition nodes have undergone the corresponding matching judgment. Finally, through this traversal and one-by-one judgment method, a set of abnormal nodes is obtained, and these nodes are considered to be the key nodes that may have security risks in the entire device. Based on these identified abnormal nodes, a security anomaly result is generated, and based on this security anomaly result, an operation security warning message is sent in a timely manner to prompt the operation and maintenance personnel to conduct key monitoring on these nodes or perform fault troubleshooting in a timely manner.

[0049] In summary, the embodiments of the present application have at least the following technical effects:

[0050] First, within a preset monitoring window, the operation of the target waste power plant is monitored by using the sensor topology network to obtain P sets of operation monitoring data, where P is an integer greater than or equal to 1. Then, the P sets of operation monitoring data are traversed to identify outlier data, obtaining P sets of outlier operation monitoring data, and abnormal authentication is performed to obtain Q sets of abnormal outlier operation monitoring data, where Q is a positive integer less than or equal to P. Further, based on the Q sets of abnormal outlier operation monitoring data, the prediction of the operation abnormal type is performed to determine the abnormal types of the Q predicted recognition nodes. Then, using the abnormal types of the Q predicted recognition nodes and the corresponding Q predicted recognition nodes as indexes, a search is performed in the pre-constructed node association network to determine Q sets of associated nodes and Q sets of abnormal types of associated nodes. Finally, the security anomaly recognition of the Q predicted recognition nodes is performed by combining the Q sets of associated nodes, the Q sets of abnormal types of associated nodes, and the abnormal types of the Q predicted recognition nodes, and an operation security warning message is obtained according to the security anomaly result. This solves the technical problem of insufficient accuracy and real-time performance in the operation security monitoring of waste power plants in the prior art, and achieves the technical effect of improving the accuracy and real-time performance of the operation security warning of waste power plants.

[0051] Embodiment 2. Based on the same inventive concept as the operation safety warning method for a waste power plant in the foregoing embodiment, as Figure 2 shown, the present application provides an operation safety warning system for a waste power plant, wherein the system includes:

[0052] A monitoring module 11, configured to perform operation monitoring on a target waste power plant within a preset monitoring window by using a sensor topology network, and obtain P operation monitoring data sets, where P is an integer greater than or equal to 1;

[0053] A data identification module 12, configured to traverse the P operation monitoring data sets to identify outlier data, obtain P outlier operation monitoring data sets, perform anomaly authentication, and obtain Q abnormal outlier operation monitoring data sets, where Q is a positive integer less than or equal to P;

[0054] An abnormal type prediction module 13, configured to predict the operation abnormal type based on the Q abnormal outlier operation monitoring data sets, and determine Q predicted identification node abnormal types;

[0055] A retrieval module 14, configured to retrieve in a pre-constructed node association network by using the Q predicted identification node abnormal types and the corresponding Q predicted identification nodes as indexes, and determine Q associated node sets and Q associated node abnormal type sets;

[0056] An abnormal identification module 15, configured to perform safety anomaly identification on the Q predicted identification nodes by combining the Q associated node sets, the Q associated node abnormal type sets, and the Q predicted identification node abnormal types, and obtain operation safety warning information according to the safety anomaly results.

[0057] Further, the data identification module 12 is configured to execute the following method:

[0058] Traverse the P sets of operation monitoring data for mean calculation to obtain P means of operation monitoring data; use the P means of operation monitoring data as P aggregation starting points, use the P aggregation starting points as neighborhood centers, and use a preset aggregation bandwidth as the radius to construct P neighborhoods of the aggregation starting points in the P sets of operation monitoring data; iterate the P aggregation starting points in the P sets of operation monitoring data in the first direction and the second direction respectively according to the preset aggregation bandwidth, and combine the P neighborhoods of the aggregation starting points to determine P updated iteration directions and P updated iteration starting points; based on the P updated iteration directions, iterate the P updated iteration starting points in the P sets of operation monitoring data according to the preset aggregation bandwidth until the preset number of iterations is satisfied, to obtain P target aggregation points and P neighborhoods of the target aggregation points; use the P target aggregation points as neighborhood centers and use the preset aggregation bandwidth as the radius to construct the P neighborhoods of the target aggregation points; diffuse outward the P neighborhoods of the target aggregation points according to the preset aggregation bandwidth to obtain P diffusion stop neighborhoods of the target aggregation points; add the data in the P sets of operation monitoring data that are outside the P diffusion stop neighborhoods of the target aggregation points into P sets of outlier operation monitoring data respectively.

[0059] Further, the data recognition module 12 is used to execute the following method:

[0060] Iterate the P aggregation starting points in the P sets of operation monitoring data in the first direction and the second direction respectively according to the preset aggregation bandwidth to obtain P first-direction aggregation iteration starting points and P second-direction aggregation iteration starting points; construct P neighborhoods of the first-direction aggregation iteration starting points and P neighborhoods of the second-direction aggregation iteration starting points, and calculate the neighborhood densities of the P neighborhoods of the first-direction aggregation iteration starting points and the neighborhood densities of the P neighborhoods of the second-direction aggregation iteration starting points; calculate the neighborhood densities of the P neighborhoods of the aggregation starting points to obtain P neighborhood densities of the aggregation starting points; respectively judge whether the neighborhood densities of the P first-direction aggregation iteration starting points and the neighborhood densities of the P second-direction aggregation iteration starting points are both greater than the P neighborhood densities of the aggregation starting points. If not, use the P aggregation starting points as P updated iteration starting points, and use iteration in any direction as P updated iteration directions.

[0061] Further, the data recognition module 12 is used to execute the following method:

[0062] If so, when the neighborhood density of the P first-direction aggregation iteration starting points is greater than that of the P second-direction aggregation iteration starting points, the P first-direction aggregation iteration starting points are used as the P updated iteration starting points, and the P first directions are used as the P updated iteration directions; when the neighborhood density of the P first-direction aggregation iteration starting points is less than that of the P second-direction aggregation iteration starting points, the P second-direction aggregation iteration starting points are used as the P updated iteration starting points, and the P second directions are used as the P updated iteration directions; when the neighborhood density of the P first-direction aggregation iteration starting points is equal to that of the P second-direction aggregation iteration starting points, the P aggregation starting points are used as the P updated iteration starting points, and the iteration in any direction is used as the P updated iteration directions.

[0063] Further, the data recognition module 12 is used to execute the following method:

[0064] Diffuse outward from the neighborhoods of the P target aggregation points according to a preset aggregation bandwidth to obtain P target aggregation point diffusion neighborhoods; determine whether the P target aggregation point diffusion neighborhood densities of the P target aggregation point diffusion neighborhoods are greater than or equal to the P target aggregation point neighborhood densities of the P target aggregation point neighborhoods. If so, continue to diffuse outward from the P target aggregation point diffusion neighborhoods according to the preset aggregation bandwidth until the preset diffusion stop condition is met. The diffusion neighborhood obtained in the last diffusion is used as the P target aggregation point stop diffusion neighborhood, where the preset diffusion stop condition is that the difference between the neighborhood densities of two target aggregation point diffusion neighborhoods obtained in two adjacent diffusions is less than or equal to a preset density difference threshold.

[0065] Further, the retrieval module 14 is used to execute the following method:

[0066] Obtain M historical operation safety anomaly logs of the target waste power plant, where M is an integer greater than or equal to 1; traverse the M historical operation safety anomaly logs to extract anomaly nodes, and obtain K historical anomaly node sets and K historical anomaly node type sets, where K is an integer greater than or equal to 1; based on the K historical anomaly node sets and K historical anomaly node type sets, perform anomaly association on the P nodes in the sensor topology network to obtain the node association network.

[0067] Further, the anomaly recognition module 15 is used to execute the following method:

[0068] Extract the first associated node set, the first associated node anomaly type set, and the first prediction and identification node from the Q associated node sets, the Q associated node anomaly type sets, and the Q prediction and identification nodes; determine whether the Q prediction and identification nodes exist in the first associated node set. If so, when the number of the Q prediction and identification node anomaly types that are the same as the corresponding first associated node anomaly type set exceeds a preset quantity threshold, regard the first prediction and identification node as the first anomaly node; and so on, traverse the Q prediction and identification nodes for security anomaly identification, obtain the anomaly node set, regard the anomaly node set as the security anomaly result, and obtain the operation security warning information according to the security anomaly result.

[0069] It should be noted that the above sequence of the embodiments of the present application is only for description and does not represent the superiority or inferiority of the embodiments. And the above describes specific embodiments of the present specification. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0070] The above are only the preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0071] This specification and the drawings are only exemplary descriptions of the present application and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of the present application. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalent technologies, the present application is intended to include these changes and modifications.

Claims

1. An operating safety warning method applicable to waste power plants, characterized in that, The method includes: Within a preset monitoring window, using a sensor topology network to monitor the operation of a target waste power plant, obtaining P operation monitoring data sets, where P is an integer greater than or equal to 1; Traverse the P operation monitoring data sets for outlier data identification, obtain P outlier operation monitoring data sets, and perform anomaly authentication to obtain Q outlier operation monitoring data sets with anomalies, where Q is a positive integer less than or equal to P; Based on the Q outlier operation monitoring data sets with anomalies, predict the types of operation anomalies to determine Q predicted identification node anomaly types; Using the Q predicted identification node anomaly types and the corresponding Q predicted identification nodes as indexes, retrieve in a pre-constructed node association network to determine Q associated node sets and Q associated node anomaly type sets, where the pre-constructed node association network is obtained by processing the historical operation safety anomaly logs of the target waste power plant; Combining the Q associated node sets, the Q associated node anomaly type sets, and the Q predicted identification node anomaly types, perform safety anomaly identification on the Q predicted identification nodes, and obtain operation safety warning information according to the safety anomaly results; Traverse the P operation monitoring data sets for outlier data identification and obtain P outlier operation monitoring data sets, including: Traverse the P operation monitoring data sets to calculate the means, obtaining P operation monitoring data means; Using the P operation monitoring data means as P aggregation starting points, taking the P aggregation starting points as neighborhood centers, and using a preset aggregation bandwidth as the radius, construct P aggregation starting point neighborhoods in the P operation monitoring data sets; Iterate the P aggregation starting points in the P operation monitoring data sets in the first direction and the second direction respectively according to the preset aggregation bandwidth, and combine the P aggregation starting point neighborhoods to determine P update iteration directions and P update iteration starting points; Based on the P update iteration directions, in the P operation monitoring data sets, iterate the P update iteration starting points according to the preset aggregation bandwidth until the preset number of iterations is satisfied, obtaining P target aggregation points and P target aggregation point neighborhoods; Using the P target aggregation points as neighborhood centers and the preset aggregation bandwidth as the radius, construct the P target aggregation point neighborhoods; Diffuse outward from the P target aggregation point neighborhoods according to the preset aggregation bandwidth to obtain P target aggregation point diffusion stop neighborhoods; Add the data in the P operation monitoring data sets that are outside the P target aggregation point diffusion stop neighborhoods into the P outlier operation monitoring data sets respectively.

2. The operation safety warning method applicable to a waste power plant according to claim 1, wherein Iterate the P aggregation starting points in the P operation monitoring data sets in the first direction and the second direction respectively according to the preset aggregation bandwidth, and combine the P aggregation starting point neighborhoods to determine P update iteration directions and P update iteration starting points, including: Iterate the P aggregation starting points in the P operation monitoring data sets in the first direction and the second direction respectively according to the preset aggregation bandwidth to obtain P first direction aggregation iteration starting points and P second direction aggregation iteration starting points; Construct P neighborhood areas of the first - direction aggregation iteration starting points and P neighborhood areas of the second - direction aggregation iteration starting points, and calculate the densities of the P neighborhood areas of the first - direction aggregation iteration starting points and the densities of the P neighborhood areas of the second - direction aggregation iteration starting points; Calculate the neighborhood densities of the P aggregation starting - point neighborhood areas to obtain P neighborhood densities of the aggregation starting - point neighborhood areas; Respectively determine whether the densities of the P neighborhood areas of the first - direction aggregation iteration starting points and the densities of the P neighborhood areas of the second - direction aggregation iteration starting points are both greater than the densities of the P aggregation starting - point neighborhood areas. If not, use the P aggregation starting points as P updated iteration starting points, and use iteration in any direction as P updated iteration directions.

3. The operation safety warning method applicable to a waste power plant according to claim 2, characterized in that, Respectively determining whether the densities of the P neighborhood areas of the first - direction aggregation iteration starting points and the densities of the P neighborhood areas of the second - direction aggregation iteration starting points are both greater than the densities of the P aggregation starting - point neighborhood areas also includes: If so, when the densities of the P neighborhood areas of the first - direction aggregation iteration starting points are greater than the densities of the P neighborhood areas of the second - direction aggregation iteration starting points, use the P neighborhood areas of the first - direction aggregation iteration starting points as P updated iteration starting points, and use the P first directions as P updated iteration directions; When the densities of the P neighborhood areas of the first - direction aggregation iteration starting points are less than the densities of the P neighborhood areas of the second - direction aggregation iteration starting points, use the P neighborhood areas of the second - direction aggregation iteration starting points as P updated iteration starting points, and use the P second directions as P updated iteration directions; When the densities of the P neighborhood areas of the first - direction aggregation iteration starting points are equal to the densities of the P neighborhood areas of the second - direction aggregation iteration starting points, use the P aggregation starting points as P updated iteration starting points, and use iteration in any direction as P updated iteration directions.

4. The operation safety warning method applicable to a waste power plant according to claim 3, wherein, Diffuse the P target - aggregation - point neighborhood areas outward according to a preset aggregation bandwidth until a preset diffusion - stop condition is met, to obtain P target - aggregation - point diffusion - stop neighborhood areas, including: Diffuse the P target - aggregation - point neighborhood areas outward according to a preset aggregation bandwidth to obtain P target - aggregation - point diffusion neighborhood areas; Judge whether the densities of the P target - aggregation - point diffusion neighborhood areas of the P target - aggregation - point diffusion neighborhood areas are greater than or equal to the densities of the P target - aggregation - point neighborhood areas of the P target - aggregation - point neighborhood areas. If so, continue to diffuse the P target - aggregation - point diffusion neighborhood areas outward according to the preset aggregation bandwidth until the preset diffusion - stop condition is met, and use the diffusion neighborhood area obtained in the last diffusion as the P target - aggregation - point stop - diffusion neighborhood area, where the preset diffusion - stop condition is that the difference between the densities of two target - aggregation - point diffusion neighborhood areas obtained from two adjacent diffusions is less than or equal to a preset density - difference threshold.

5. The operation safety early warning method applicable to a waste power plant according to claim 1, wherein, Using the Q predicted - recognition - node abnormal types and the corresponding Q predicted - recognition nodes as indexes, retrieve in a pre - constructed node - association network to determine Q associated - node sets and Q associated - node abnormal - type sets, including: Obtain M historical operation - safety abnormal logs of the target waste - to - energy power plant, where M is an integer greater than or equal to 1; Traverse the M historical operation - safety abnormal logs to extract abnormal nodes, to obtain K historical - abnormal - node sets and K historical - abnormal - node type sets, where K is an integer greater than or equal to 1; Based on the K historical abnormal node sets and the K historical abnormal node type sets, perform abnormal association on P nodes in the sensor topology network to obtain the node association network.

6. The operation safety warning method applicable to a waste power plant according to claim 1, characterized in that Combine the Q associated node sets, the Q associated node abnormal type sets, and the Q predicted and identified node abnormal types to perform security abnormal identification on the Q predicted and identified nodes, and obtain the operation security warning information according to the security abnormal results, including: Extract the first associated node set, the first associated node abnormal type set, and the first predicted and identified node from the Q associated node sets, the Q associated node abnormal type sets, and the Q predicted and identified nodes; Judge whether the Q predicted and identified nodes exist in the first associated node set. If so, when the number of the same types between the Q predicted and identified node abnormal types and the corresponding first associated node abnormal type set exceeds the preset quantity threshold, regard the first predicted and identified node as the first abnormal node; And so on, traverse the Q predicted and identified nodes to perform security abnormal identification, obtain the abnormal node set, use the abnormal node set as the security abnormal result, and obtain the operation security warning information according to the security abnormal result.

7. An operation safety early warning system applicable to waste power plants, characterized in that, For implementing an operation security warning method applicable to a waste power plant according to any one of claims 1-6, the system includes: A monitoring module, configured to use the sensor topology network to perform operation monitoring on the target waste power plant within a preset monitoring window, and obtain P operation monitoring data sets, where P is an integer greater than or equal to 1; A data identification module, configured to traverse the P operation monitoring data sets to perform outlier data identification, obtain P outlier operation monitoring data sets, and perform abnormal authentication to obtain Q abnormal outlier operation monitoring data sets, where Q is a positive integer less than or equal to P; An abnormal type prediction module, configured to perform operation abnormal type prediction based on the Q abnormal outlier operation monitoring data sets to determine the Q predicted and identified node abnormal types; A retrieval module, configured to perform retrieval in the pre-constructed node association network by using the Q predicted and identified node abnormal types and the corresponding Q predicted and identified nodes as indexes to determine the Q associated node sets and the Q associated node abnormal type sets; An abnormal identification module, configured to combine the Q associated node sets, the Q associated node abnormal type sets, and the Q predicted and identified node abnormal types to perform security abnormal identification on the Q predicted and identified nodes, and obtain the operation security warning information according to the security abnormal results.

Images