CN117591560A - An event discovery and tracking method and system based on real-time signals from hydropower stations - Google Patents

Abstract

The invention discloses an event discovery and tracking method and system based on hydropower station real-time signals, comprising the following steps: based on historical signal data and historical operation instruction data of the hydropower station, a typical association relation event library is established, unique identification signals or signal combinations of the events are obtained, and a mapping relation between the events and the signals is obtained; establishing a device-event-signal relationship map according to the mapping relationship between the event and the signal and based on the relationship between the device and the event; acquiring multi-source real-time data, and carrying out data fusion on the acquired multi-source real-time data to obtain a fused real-time data stream; according to the fused real-time data stream and based on the relation graph of the equipment-event-signal, adopting an event finding and tracking algorithm to carry out three processes of event finding, event dynamic tracking and event overtime detection on the fused real-time data stream, and finding an event signal of the real-time data stream; the method improves the operation efficiency, safety and reliability of the hydropower station.

Description

An event discovery and tracking method and system based on real-time signals from hydropower stations

Technical field

The invention relates to the technical field of hydropower stations, and in particular to an event discovery and tracking method and system based on real-time signals of hydropower stations.

Background technique

Centralized and unified monitoring of the signals of the eight stations under the Dadu River faces challenges such as complex and diverse event signal descriptions and extreme reliance on manual judgment and empirical analysis. In addition, multiple events may overlap at the same time, resulting in a large number of signals flooding in a short period of time. enter. However, there are still multiple problems in relying solely on experienced technicians to perform signal analysis and take corresponding measures. First, the tasks have diverse equipment types, a large amount of signals, require complex data processing, frequent alarms, and difficult monitoring work, and The risks are high, so even experienced technicians spend a lot of time and human resources analyzing these signals. Secondly, operational safety faces challenges, and signal monitoring requires manual 24-hour uninterrupted monitoring. This kind of high-intensity and high-pressure work requires technicians to always be highly concentrated and keep a clear mind in order to detect abnormalities in time and take correct response measures. In addition, the signal data is complex, and emergency and non-emergency events are intertwined. Considering the "zero fault tolerance" security requirements of signal monitoring, even experienced technicians need to invest a lot of time to sort and analyze these signals. Finally, the experience of technicians mainly comes from the accumulation of handled incidents, so it usually takes a long time to cultivate experienced technicians. Therefore, adopting the above method will lead to low operating efficiency of hydropower stations, and too high dependence on technical personnel will increase the work intensity of technical personnel.

Contents of the invention

In view of the above-mentioned deficiencies in the prior art, the present invention provides an event discovery and tracking method and system based on real-time signals of hydropower stations. Through the historical signal data and historical operation instruction data of the hydropower station, a relationship diagram of equipment-event-signals is constructed. The system is used to analyze the collected data flow to realize automatic analysis and monitoring of signals, so as to reduce the work intensity of technicians and improve the operating efficiency of hydropower stations.

In order to achieve the above-mentioned object of the invention, the technical solutions adopted by the present invention are:

An event discovery and tracking method based on real-time signals from hydropower stations, including the following steps:

S1. Based on the historical signal data and historical operation instruction data of the hydropower station, establish a typical correlation event library, obtain the unique identification signal or signal combination of the event, and obtain the mapping relationship between events and signals;

S2. Based on the mapping relationship between events and signals obtained in step S1, and based on the relationship between devices and events, establish a device-event-signal relationship map;

S3. Obtain multi-source real-time data, perform data fusion on the obtained multi-source real-time data, and obtain the fused real-time data stream;

S4. Based on the fused real-time data stream obtained in step S3 and based on the device-event-signal relationship map established in step S2, use event discovery and tracking algorithms to perform event search and event dynamics on the fused real-time data stream. The three processes of tracking and event timeout detection are used to discover event signals of real-time data streams.

Further, step S1 specifically includes:

S11. Obtain the events of various scenarios involved in the daily dispatching process of historical hydropower stations, analyze the correlation between events and characteristic signals, and based on the logical description of the correlation between characteristic signals and events, eventize the characteristic signals and establish a typical correlation event library ;

S12. Obtain the operation instructions involved in the daily dispatching process of the historical hydropower station and the instruction events corresponding to the operation instructions, and based on the historical monitoring system signal data, use the frequent pattern mining method to obtain the signal data corresponding to the instruction events, and input the signal data into step S11 From the established typical association event library, a typical association event library is obtained;

S13. Collect each event in the typical correlation event library obtained in step S12, and obtain the signal of each event;

S14. Use the continuous digital encoding method to encode the signal of each event obtained in step S13 to obtain the event-encoded signal;

S15. According to the event coded signal obtained in step S14, perform a full arrangement and combination of the coded signals of each event to obtain various combinations of each event, and traverse them to obtain the number of combinations after event traversal;

S16. Determine whether the number of combinations after event traversal obtained in step S15 has the same number of combinations as other events. If so, execute step S17. Otherwise, stop looking for combinations that exceed the current length and obtain the signal combination length;

S17. Continue to increase the number of combinations of events until the number of combinations of the event is equal to the number of all signals contained in the event, and obtain the signal combination length;

S18. Obtain the unique identification signal or signal combination of the event based on the signal combination length obtained in steps S16 and S17.

Further, in step S12, the frequent pattern mining method is used to obtain the signal data corresponding to the instruction event, and the process of inputting the signal data into the typical association event library established in step S11 is:

Obtain the historical monitoring system signal data of the operation instructions in different time periods, sort the historical monitoring system signal data according to time, obtain the signal data of the same operation instructions in different time periods, and use the frequent pattern mining method to find out the different times Signal data with higher frequency appears in the segment, and these signal data are collected to form a signal subset, and it is judged whether these signal subsets constitute a complete event. If so, the signal data in the signal subset is added to the typical correlation event. library and as a confirmed event.

Furthermore, the frequent pattern mining method used in step S12 is based on the Apriori algorithm, and the algorithm steps are as follows:

First, calculate the support of each item, that is, a single signal, in the data set, retain items whose support is higher than the preset threshold, that is, the minimum support, generate candidate 1-item sets, and construct frequent 1-item sets;

Then, the constructed frequent 1-itemset is used to generate candidate 2-itemsets, and the support of a single signal in the data set is again used to filter to obtain frequent 2-itemsets;

Continue to generate larger candidate sets and filter until no more new candidate sets can be generated.

Further, step S2 specifically includes:

S21. Treat each device as a node, treat each event as a node, and use different identifiers to identify the nodes;

S22. Use undirected edges to define the relationship between main device nodes and sub-device nodes, and use directed edges to define the relationship between sub-device nodes and main device nodes, and the relationship between event nodes and device nodes;

S23. According to the undirected edges and directed edges defined in step S22, add all device nodes and event nodes to the graph, and add directed edges between event nodes and device nodes, and between sub-device nodes and main device nodes. The directed edges between devices, the undirected edges between main device nodes, and the undirected edges between sub-device nodes are used to establish a device-event-signal relationship graph.

Further, step S3 specifically includes:

S31. Obtain the signal data flow and operation instruction data generated during the daily dispatching process of the hydropower station, and obtain multi-source real-time data;

S32. Perform data fusion on the multi-source real-time data obtained in step S31, correlate the signal data flow and operation instruction data in the multi-source real-time data, and unify the data formats of the signal data flow and operation instruction data to obtain the fused real-time data. flow.

Further, the event search process in step S4 is:

Analyze the fused real-time data stream obtained in step S3. If there is operation instruction data in the real-time data stream, identify the confirmed event corresponding to the real-time data stream through the operation instruction data;

If the real-time data stream is in a single form, match the real-time data stream with the events in the device-event-signal relationship map established in step S2, and determine the unique identification signal or signal combination of the matched confirmed event. ;

If the real-time data stream corresponds to multiple events, enable storage space for each event in the cache, and back up the real-time data stream to each cache, while continuing to accumulate the real-time data stream;

Determine whether there is a unique identification signal or signal combination in the currently accumulated real-time data stream based on the matched unique identification signal or signal combination of the confirmed event. If it exists, identify the confirmed event corresponding to the currently accumulated real-time data stream. If not, If exists, mark the currently accumulated real-time data stream as an event to be confirmed and continue to accumulate real-time data streams;

When the accumulated real-time data stream reaches the set quantity threshold, the similarity of the event to be confirmed and the matched multiple confirmed events is calculated, and the event with the highest similarity and exceeding the threshold is obtained, which is identified as the real-time data stream. event.

Further, the process of event dynamic tracking in step S4 is:

When the real-time data stream corresponds to multiple events, create a cache of multiple events, monitor and maintain the cache of multiple events, and back up the accumulation of real-time data streams to the cache corresponding to the matched multiple events. If If a unique event is identified, the cache of the identified event will be retained and the cache of other matching events will be deleted.

Further, event timeout detection in step S4 includes fixed time limit timeout detection and dynamic time limit timeout detection. The specific process is as follows:

Fixed time limit timeout detection sets a fixed time for the determined event based on the identified event in the real-time data stream. When the accumulated time of the real-time data stream exceeds the set fixed time, the current real-time data stream is submitted;

Dynamic time limit timeout detection determines the timeout based on the actual situation of the real-time data flow. After the real-time data flow is identified as a confirmed event and the end signal of the event is identified, the current real-time data flow is no longer submitted according to a fixed time, but ends early. Accumulation of real-time data streams.

An event discovery and tracking system based on real-time signals from hydropower stations, including:

The event database establishment module establishes a typical correlation event database based on the historical signal data and historical operation instruction data of the hydropower station. By obtaining the unique identification signal or signal combination of the event, the mapping relationship between the event and the signal is obtained;

The device event relationship graph module establishes a device-event-signal relationship graph based on the mapping relationship between events and signals in the event library based on the typical association relationship established in the event library establishment module, and based on the relationship between devices and events;

The real-time data docking and pre-processing module obtains multi-source real-time data, and performs fusion processing of multi-source real-time data in time sequence to obtain a fused real-time data stream;

The event discovery and tracking module implements signal event discovery, multi-cache construction and maintenance based on the device-event-signal relationship graph established by the device event relationship graph module based on the fused real-time data stream obtained by the real-time data docking and preprocessing module. with event tracking.

The invention has the following beneficial effects:

The invention proposes an event discovery and tracking method and system based on real-time signals of hydropower stations. By processing the real-time data flow of hydropower stations, events can be discovered and tracked in a timely manner, thereby improving the operating efficiency, safety and reliability of hydropower stations. .

Description of drawings

Figure 1 is a schematic flow chart of an event discovery and tracking method based on real-time signals from hydropower stations proposed by the present invention;

Figure 2 is a schematic structural diagram of an event discovery and tracking system based on real-time signals from hydropower stations.

Detailed ways

The specific embodiments of the present invention are described below to facilitate those skilled in the art to understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific embodiments. For those of ordinary skill in the technical field, as long as various changes These changes are obvious within the spirit and scope of the invention as defined and determined by the appended claims, and all inventions and creations utilizing the concept of the invention are protected.

As shown in Figure 1, an event discovery and tracking method based on real-time signals from hydropower stations includes the following steps S1-S4:

S1. Based on the historical signal data and historical operation instruction data of the hydropower station, establish a typical correlation event library, obtain the unique identification signal or signal combination of the event, and obtain the mapping relationship between events and signals.

In this embodiment, various scenarios involved in the daily dispatching process of historical hydropower stations, such as startup and shutdown, switching operations, accidents, etc., are first studied, the events of each scenario are obtained, and the correlation between the events and characteristic signals is established. The relationship is specifically to establish the mapping relationship between events and signals and to clarify the characteristic signals associated with each event. Among them, the characteristic signals include the cause signal of the event, the trigger signal, the end signal, the key signal, etc. Secondly, obtain the operating instructions involved in the daily dispatching process of historical hydropower stations and the events corresponding to the operating instructions. Specifically, it uses historical monitoring system signal data and applies frequent pattern mining technology to obtain signal data corresponding to operation instructions, and supplements these signal data into the typical correlation event library to further enrich event information in order to provide more accurate event discovery and tracking. ability.

Specifically, step S1 specifically includes S11-S18:

S11. Obtain the events of various scenarios involved in the daily dispatching process of historical hydropower stations, analyze the correlation between events and characteristic signals, and based on the logical description of the correlation between characteristic signals and events, eventize the characteristic signals and establish a typical correlation event library .

S12. Obtain the operation instructions involved in the daily dispatching process of the historical hydropower station and the instruction events corresponding to the operation instructions, and based on the historical monitoring system signal data, use the frequent pattern mining method to obtain the signal data corresponding to the instruction events, and input the signal data into step S11 From the established typical association event library, a typical association event library is obtained.

In this embodiment, the signal data obtained by the historical Kafka platform, that is, the historical monitoring system signal data, is sorted to ensure that the signal data is arranged in chronological order, and the time period in which the operation instruction is issued is determined based on the time information of the operation instruction. , obtain the signal data of the same operation instructions in different time periods, and use the frequent pattern mining method to analyze the sorted signal data to find out the signal data that frequently appears in different time periods, and collect these signal data to form a signal sub- Sets, and hands these signal subsets to technical personnel for verification and confirmation, and technical personnel evaluate the relevance of the signal subsets and the completeness of events based on knowledge in the field. And determine whether the signal subset constitutes a complete event. If so, add the signal data in the signal subset to the typical correlation event library and regard it as a confirmed event. In addition, the frequent pattern mining method used in this embodiment is based on the Apriori algorithm, which can find correlations between data in large-scale data sets, thereby enhancing the accuracy of event matching.

Specifically, in step S12, the frequent pattern mining method is used to obtain the signal data corresponding to the instruction event, and the process of inputting the signal data into the typical association event library established in step S11 is:

Obtain the historical monitoring system signal data of the operation instructions in different time periods, sort the historical monitoring system signal data according to time, obtain the signal data of the same operation instructions in different time periods, and use the frequent pattern mining method to find out the different times Signal data with higher frequency appears in the segment, and these signal data are collected to form a signal subset, and it is judged whether these signal subsets constitute a complete event. If so, the signal data in the signal subset is added to the typical correlation event. library and as a confirmed event.

Specifically, the frequent pattern mining method used in step S12 is based on the Apriori algorithm, and the algorithm steps are as follows:

First, calculate the support of each item, that is, a single signal, in the data set, retain items whose support is higher than the preset threshold, that is, the minimum support, generate candidate 1-item sets, and construct frequent 1-item sets.

Then, the constructed frequent 1-itemset is used to generate candidate 2-itemsets, and the support of a single signal in the data set is again used to screen to obtain frequent 2-itemsets.

Continue to generate larger candidate sets and filter until no more new candidate sets can be generated.

S13. Collect each event in the typical correlation event library obtained in step S12, and obtain the signal of each event.

S14. Use the continuous digital encoding method to encode the signal of each event obtained in step S13 to obtain an event-encoded signal.

S15. According to the event coded signal obtained in step S14, perform a full arrangement and combination of the coded signals of each event to obtain various combinations of each event, and traverse them to obtain the number of combinations after event traversal.

S16. Determine whether the number of combinations after event traversal obtained in step S15 has the same number of combinations as other events. If so, execute step S17. Otherwise, stop looking for combinations that exceed the current length and obtain the signal combination length.

S17. Continue to increase the number of combinations of events until the number of combinations of the event is equal to the number of all signals contained in the event, and obtain the signal combination length.

S18. Obtain the unique identification signal or signal combination of the event based on the signal combination length obtained in steps S16 and S17.

Steps S13-S18 are the process of obtaining the unique identification signal or signal combination of each event. The specific process is: first, collect the signals of each event in the typical correlation event library in this embodiment, and assign a unique digital code to each signal to ensure that the signals have unique identifiers and avoid confusion and conflicts. ; Secondly, the collected signals are standardized and normalized, the signals are mapped to a unified form, and the signals are continuously processed so that the signals can be represented as compact continuous coding forms; finally, for each event, Its signals are fully arranged and combined, and all possible combinations of all signals of each event are performed to explore the combination of different signals and check whether there are other events with the same signal combination. If the current number of signal combinations can uniquely represent the event, stop looking for combinations that exceed the length of the current signal combination, which can be marked as a signal combination belonging to the event. If the current number of signal combinations cannot uniquely represent the event, continue to increase the length of the number of signal combinations until the size of the signal combination is equal to the event. Up to the number of all signals included, if the signal combination length of the event is 1, then the signal combination is the unique identification signal of the current event. At this time, there is no need to traverse all other combinations with a signal combination length exceeding 1; if the signal combination length of the event If the length exceeds 1 and the combination does not exist in other events, the signal combination represents the signal combination of the current event.

S2. Based on the mapping relationship between events and signals obtained in step S1, and based on the relationship between devices and events, establish a device-event-signal relationship map.

In this embodiment, the mapping relationship between events and signals has been obtained in the above steps. Therefore, as long as the association between events and devices is clear, the device-event-signal relationship map can be obtained. The specific process is: first, node definition, consider each device and event as a node, and use different identifiers to identify it, such as: "Device1", "Device2", "Event1", "Event2". Secondly, the definition of edges represents different relationships through the definition of edges. Undirected edges point from one main device to another main device, or from one sub-device to another sub-device, and are used to represent the relationship between devices at the same level; directed edges point from sub-devices to main device nodes, and are used to represent the main device node. The hierarchical relationship between the device and the sub-device is used to indicate that the sub-device is subordinate to the main device; in addition, there is a directed edge from the event node to the device node, which is used to represent the association between the event and the device, indicating that the device is related to the device. Events are associated. Finally, in the component diagram stage, all devices and event nodes are added to the diagram, and directed edges between events and devices are added to represent the relationship between them. At the same time, sub-devices and Directed edges between master devices to establish hierarchies. In order to represent a unified level of devices, undirected edges between main devices and undirected edges between sub-devices are also added. In addition, the signal in the constructed device-event-signal relationship graph represents the attribute of the event node.

Specifically, step S2 specifically includes S21-S23:

S21. Treat each device as a node, treat each event as a node, and use different identifiers to identify the nodes.

S22. Use undirected edges to define the relationship between main device nodes and between sub-device nodes, and use directed edges to define the relationship between sub-device nodes and main device nodes, and the relationship between event nodes and device nodes.

S23. According to the undirected edges and directed edges defined in step S22, add all device nodes and event nodes to the graph, and add directed edges between event nodes and device nodes, and between sub-device nodes and main device nodes. The directed edges between devices, the undirected edges between main device nodes, and the undirected edges between sub-device nodes are used to establish a device-event-signal relationship graph.

S3: Obtain multi-source real-time data, perform data fusion on the obtained multi-source real-time data, and obtain a fused real-time data stream.

In this embodiment, multi-source real-time data is obtained by acquiring the signal data stream and operation instruction data generated during the daily dispatching process of the hydropower station. Since the signal data flow generated in the daily scheduling process is large and involves multiple sensor and device types, the resulting signal data flow includes signals generated by different devices in different systems. "System" as used in this embodiment is a term for different components, elements, parts, parts or assemblies at different levels. At the same time, the obtained multi-source real-time data is fused and processed, and a unified format of real-time data flow is obtained by correlating and unifying the format of the operation instruction data and signal data streams in the multi-source real-time data. The specific process is to parse and clean the signal data stream from the Kafka platform, extract useful information, analyze the operation instructions subordinate to the signal data flow, obtain the time, object and instruction content of the operation instruction data, etc., and then The signal data flow and the instruction data flow are correlated, and the formats of the signal data flow and operation instruction data of the Kafka platform are unified to complete the unified formatting and correlation operations of multi-source real-time data from different sources and types, and obtain the fused real-time data. flow.

Specifically, step S3 specifically includes S31-S32:

S31. Obtain the signal data flow and operation instruction data generated during the daily dispatching process of the hydropower station, and obtain multi-source real-time data.

S32. Perform data fusion on the multi-source real-time data obtained in step S31, correlate the signal data flow and operation instruction data in the multi-source real-time data, and unify the data formats of the signal data flow and operation instruction data to obtain the fused real-time data. flow.

S4. Based on the fused real-time data stream obtained in step S3 and based on the device-event-signal relationship map established in step S2, use event discovery and tracking algorithms to perform event search and event dynamics on the fused real-time data stream. The three processes of tracking and event timeout detection are used to discover event signals of real-time data streams.

In this embodiment, the event discovery and tracking algorithm includes: real-time monitoring, collection and analysis of acquired real-time data streams, converting these real-time data streams into events, and separating them in the form of events. In this process, including tracking events There are three stages: start, continue and end. Therefore, the event discovery and tracking algorithm is used to eventize the fused real-time data stream, including event search, event dynamic tracking, and event timeout monitoring. In addition, in order to further improve the event discovery algorithm, in this embodiment, the algorithm is calculated based on the similarity of the Jaccard similarity coefficient, which can accurately identify and classify the signal flow to determine the degree of consistency with the determined event. The Jaccard similarity coefficient is a calculation method used to compare the similarity between two sets. It is mainly used to measure the proportion of common elements between two sets in the total elements, thereby judging the degree of similarity between sets. The value range of Jaccard similarity coefficient is between 0 and 1, where 0 means that the two sets have no common elements, that is, they are completely dissimilar; 1 means that the elements of the two sets are exactly the same, that is, they are completely similar. Therefore, in practical applications, the closer the similarity coefficient is to 1, the more similar the two sets are. Therefore, using this method, the event to which the current real-time data stream belongs can be determined. Most importantly, the results of the event discovery and tracking algorithm can also be used as preliminary data for hydropower station fault diagnosis, anomaly detection and other scenarios, and provide support for these scenarios.

The specific process of event search is: when the operation instruction data is carried in the real-time data stream, the events corresponding to the current real-time data stream are directly identified through these operation instructions; when only a single real-time data stream arrives, the real-time data stream is combined with the established Match the events in the device-event-signal relationship diagram, and determine the unique identification signal or combined signal of the matched determined event. If a live data signal stream corresponds to multiple events, enable storage for each event in the cache and back it up into each cache, and continue to accumulate the live data stream. Determine whether there is such an identification signal in the current real-time data stream based on the matched unique identification signal or combined signal of the determined event. If it exists, the event corresponding to the current real-time data stream can be determined; if the matched identified event is If there is no unique identification signal or combined signal, the current real-time data stream will be marked as an event to be confirmed, and the accumulation will continue until the accumulated real-time data stream reaches the set threshold, then the event to be confirmed will be matched with the multiple matching events. Perform similarity calculation on known events to determine the event with the highest similarity and exceeding the threshold, and identify it as the event that belongs to the current real-time data stream. In this embodiment, similarity calculation is performed between the event to be confirmed and multiple matched known events to determine the event with the highest similarity and exceeding the threshold. A similarity matching algorithm is used and the calculation process is: from the matched to The signal is extracted from the event as a feature set, and the similarity between the matched signal set in the buffer and each feature set is calculated. Finally, the calculated similarity is compared with the preset threshold.

Specifically, the event search process in step S4 is:

Analyze the fused real-time data stream obtained in step S3. If there is operation instruction data in the real-time data stream, identify the confirmed event corresponding to the real-time data stream through the operation instruction data.

If the real-time data stream is in a single form, match the real-time data stream with the events in the device-event-signal relationship map established in step S2, and determine the unique identification signal or signal combination of the matched confirmed event. .

If the real-time data stream corresponds to multiple events, storage space is enabled in the cache for each event and the real-time data stream is backed up to each cache while continuing to accumulate the real-time data stream.

Determine whether there is a unique identification signal or signal combination in the currently accumulated real-time data stream based on the matched unique identification signal or signal combination of the confirmed event. If it exists, identify the confirmed event corresponding to the currently accumulated real-time data stream. If not, If exists, the currently accumulated real-time data stream will be marked as an event to be confirmed, and the real-time data stream will continue to be accumulated.

When the accumulated real-time data stream reaches the set quantity threshold, the similarity of the event to be confirmed and the matched multiple confirmed events is calculated, and the event with the highest similarity and exceeding the threshold is obtained, which is identified as the real-time data stream. event.

In this embodiment, dynamic tracking monitors and maintains the cache of multiple events, and regularly monitors the cache status and updates the cache content to track the development process of the event while retaining signal data related to the event, through unique event confirmation and cache management Strategy to retain only necessary information and optimize cache usage. The entire dynamic tracking process covers the entire life cycle of signal events from identification, gradual accumulation to final end. Its purpose is to monitor signal flow.

Specifically, the process of event dynamic tracking in step S4 is:

When the real-time data stream corresponds to multiple events, create a cache of multiple events, monitor and maintain the cache of multiple events, and back up the accumulation of real-time data streams to the cache corresponding to the matched multiple events. If If a unique event is identified, the cache of the identified event will be retained and the cache of other matching events will be deleted.

In this embodiment, during the event search process, when the real-time data stream is identified as a confirmed event, but the real-time data stream does not end the signal for a long time, due to real-time limitations, signals cannot be accumulated without limit, so timeout detection is introduced. , once the determined event is confirmed, use the preset time limit information of the event itself to determine when to end the event, and submit the event signal. Timeout detection includes two modes, namely fixed time limit and dynamic time limit.

Specifically, event timeout detection in step S4 includes fixed time limit timeout detection and dynamic time limit timeout detection. The specific process is as follows:

Fixed time limit timeout detection sets a fixed time for the determined event based on the identified event in the real-time data stream. When the accumulated time of the real-time data stream exceeds the set fixed time, the current real-time data stream is submitted.

Dynamic time limit timeout detection determines the timeout based on the actual situation of the real-time data flow. After the real-time data flow is identified as a confirmed event and the end signal of the event is identified, the current real-time data flow is no longer submitted according to a fixed time and ends in advance. Accumulation of data streams.

As shown in Figure 2, an event discovery and tracking system based on real-time signals from hydropower stations includes:

The event library establishment module establishes a typical correlation event library based on the historical signal data and historical operation instruction data of the hydropower station. By obtaining the unique identification signal or signal combination of the event, the mapping relationship between the event and the signal is obtained.

In this embodiment, the event library establishment module is responsible for collecting and storing event signal data of historical hydropower stations, and is used to construct hydropower station event signals. This module includes an event library probability table, which covers all events and corresponding time limits and other information. At the same time, this module also includes an event signal list, which records the key signals, cause signals, end signals, signal timing and other related relationship and attribute information of the event.

The device event relationship graph module establishes a device-event-signal relationship graph based on the mapping relationship between events and signals in the typical association event library established in the event library establishment module, and based on the relationship between devices and events.

In this embodiment, the equipment event relationship map module creates a hydropower station equipment event relationship map based on the corresponding relationship between events and equipment. By analyzing this map, the events associated with each device and the signals associated with the events are determined.

The real-time data docking and preprocessing module acquires multi-source real-time data, and performs fusion processing of multi-source real-time data in time sequence to obtain a fused real-time data stream.

In this embodiment, the real-time data docking and preprocessing module is responsible for obtaining the signal data stream and operation instruction data generated by the relevant equipment from the real-time monitoring system of the hydropower station, and performing the fusion processing of the multi-source signal streams in accordance with the time sequence.

The event discovery and tracking module implements signal event discovery, multi-cache construction and maintenance based on the device-event-signal relationship graph established by the device event relationship graph module based on the fused real-time data stream obtained by the real-time data docking and preprocessing module. with event tracking.

In this embodiment, the event discovery and tracking module implements signal event discovery, multi-cache construction and maintenance, and event tracking based on the device-event-signal relationship graph established by the device event relationship graph module based on the fused real-time data stream, thereby Realize intelligent monitoring of a large number of real-time data streams, quickly discover and track event status, and assist hydropower station technicians in operating the working status of the hydropower station to improve work efficiency, prevent accidents, and make timely emergency responses.

The present invention uses specific embodiments to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for those of ordinary skill in the art, based on this The idea of the invention will be subject to change in the specific implementation and scope of application. In summary, the contents of this description should not be understood as limiting the invention.

Those of ordinary skill in the art will appreciate that the embodiments described here are provided to help readers understand the principles of the present invention, and it should be understood that the scope of the present invention is not limited to such specific statements and embodiments. Those of ordinary skill in the art can make various other specific modifications and combinations based on the technical teachings disclosed in the present invention without departing from the essence of the present invention, and these modifications and combinations are still within the protection scope of the present invention.

Claims (10)

1. An event discovery and tracking method based on real-time signals from hydropower stations, which is characterized by including the following steps: S1. Based on the historical signal data and historical operation instruction data of the hydropower station, establish a typical correlation event library, obtain the unique identification signal or signal combination of the event, and obtain the mapping relationship between events and signals; S2. Based on the mapping relationship between events and signals obtained in step S1, and based on the relationship between devices and events, establish a device-event-signal relationship map; S3. Obtain multi-source real-time data, perform data fusion on the obtained multi-source real-time data, and obtain the fused real-time data stream; S4. Based on the fused real-time data stream obtained in step S3 and based on the device-event-signal relationship map established in step S2, use event discovery and tracking algorithms to perform event search and event dynamics on the fused real-time data stream. The three processes of tracking and event timeout detection are used to discover event signals of real-time data streams. 2. An event discovery and tracking method based on real-time signals of hydropower stations according to claim 1, characterized in that step S1 specifically includes: S11. Obtain the events of various scenarios involved in the daily dispatching process of historical hydropower stations, analyze the correlation between events and characteristic signals, and based on the logical description of the correlation between characteristic signals and events, eventize the characteristic signals and establish a typical correlation event library ; S12. Obtain the operation instructions involved in the daily dispatching process of the historical hydropower station and the instruction events corresponding to the operation instructions, and based on the historical monitoring system signal data, use the frequent pattern mining method to obtain the signal data corresponding to the instruction events, and input the signal data into step S11 From the established typical association event library, a typical association event library is obtained; S13. Collect each event in the typical correlation event library obtained in step S12, and obtain the signal of each event; S14. Use the continuous digital encoding method to encode the signal of each event obtained in step S13 to obtain the event-encoded signal; S15. According to the event coded signal obtained in step S14, perform a full arrangement and combination of the coded signals of each event to obtain various combinations of each event, and traverse them to obtain the number of combinations after event traversal; S16. Determine whether the number of combinations after event traversal obtained in step S15 has the same number of combinations as other events. If so, execute step S17. Otherwise, stop looking for combinations that exceed the current length and obtain the signal combination length; S17. Continue to increase the number of combinations of events until the number of combinations of the event is equal to the number of all signals contained in the event, and obtain the signal combination length; S18. Obtain the unique identification signal or signal combination of the event based on the signal combination length obtained in steps S16 and S17. 3. An event discovery and tracking method based on real-time signals of hydropower stations according to claim 2, characterized in that in step S12, a frequent pattern mining method is used to obtain signal data corresponding to the command event, and the signal data is input into step S11. The process of establishing a typical correlation event library is: Obtain the historical monitoring system signal data of the operation instructions in different time periods, sort the historical monitoring system signal data according to time, obtain the signal data of the same operation instructions in different time periods, and use the frequent pattern mining method to find out the different times Signal data with higher frequency appears in the segment, and these signal data are collected to form a signal subset, and it is judged whether these signal subsets constitute a complete event. If so, the signal data in the signal subset is added to the typical correlation event. library and as a confirmed event. 4. An event discovery and tracking method based on real-time signals of hydropower stations according to claim 2, characterized in that the frequent pattern mining method used in step S12 is based on the Apriori algorithm, and the algorithm steps are as follows: First, calculate the support of each item, that is, a single signal, in the data set, retain items whose support is higher than the preset threshold, that is, the minimum support, generate candidate 1-item sets, and construct frequent 1-item sets; Then, the constructed frequent 1-itemset is used to generate candidate 2-itemsets, and the support of a single signal in the data set is again used to filter to obtain frequent 2-itemsets; Continue to generate larger candidate sets and filter until no more new candidate sets can be generated. 5. An event discovery and tracking method based on real-time signals of hydropower stations according to claim 1, characterized in that step S2 specifically includes: S21. Treat each device as a node, treat each event as a node, and use different identifiers to identify the nodes; S22. Use undirected edges to define the relationship between main device nodes and sub-device nodes, and use directed edges to define the relationship between sub-device nodes and main device nodes, and the relationship between event nodes and device nodes; S23. According to the undirected edges and directed edges defined in step S22, add all device nodes and event nodes to the graph, and add directed edges between event nodes and device nodes, and between sub-device nodes and main device nodes. The directed edges between devices, the undirected edges between main device nodes, and the undirected edges between sub-device nodes are used to establish a device-event-signal relationship graph. 6. An event discovery and tracking method based on real-time signals of hydropower stations according to claim 1, characterized in that step S3 specifically includes: S31. Obtain the signal data flow and operation instruction data generated during the daily dispatching process of the hydropower station, and obtain multi-source real-time data; S32. Perform data fusion on the multi-source real-time data obtained in step S31, correlate the signal data flow and operation instruction data in the multi-source real-time data, and unify the data formats of the signal data flow and operation instruction data to obtain the fused real-time data. flow. 7. An event discovery and tracking method based on real-time signals of hydropower stations according to claim 1, characterized in that the process of event search in step S4 is: Analyze the fused real-time data stream obtained in step S3. If there is operation instruction data in the real-time data stream, identify the confirmed event corresponding to the real-time data stream through the operation instruction data; If the real-time data stream is in a single form, match the real-time data stream with the events in the device-event-signal relationship map established in step S2, and determine the unique identification signal or signal combination of the matched confirmed event. ; If the real-time data stream corresponds to multiple events, enable storage space for each event in the cache, and back up the real-time data stream to each cache, while continuing to accumulate the real-time data stream; Determine whether there is a unique identification signal or signal combination in the currently accumulated real-time data stream based on the matched unique identification signal or signal combination of the confirmed event. If it exists, identify the confirmed event corresponding to the currently accumulated real-time data stream. If not, If exists, mark the currently accumulated real-time data stream as an event to be confirmed and continue to accumulate real-time data streams; When the accumulated real-time data stream reaches the set quantity threshold, the similarity of the event to be confirmed and the matched multiple confirmed events is calculated, and the event with the highest similarity and exceeding the threshold is obtained, which is identified as the real-time data stream. event. 8. An event discovery and tracking method based on real-time signals of hydropower stations according to claim 1, characterized in that the process of event dynamic tracking in step S4 is: When the real-time data stream corresponds to multiple events, create a cache of multiple events, monitor and maintain the cache of multiple events, and back up the accumulation of real-time data streams to the cache corresponding to the matched multiple events. If If a unique event is identified, the cache of the identified event will be retained and the cache of other matching events will be deleted. 9. An event discovery and tracking method based on real-time signals of hydropower stations according to claim 1, characterized in that event timeout detection in step S4 includes fixed time limit timeout detection and dynamic time limit timeout detection. The specific process is as follows: Fixed time limit timeout detection sets a fixed time for the determined event based on the identified event in the real-time data stream. When the accumulated time of the real-time data stream exceeds the set fixed time, the current real-time data stream is submitted; Dynamic time limit timeout detection determines the timeout based on the actual situation of the real-time data flow. After the real-time data flow is identified as a confirmed event and the end signal of the event is identified, the current real-time data flow is no longer submitted according to a fixed time, but ends early. Accumulation of real-time data streams. 10. An event discovery and tracking system based on real-time signals from hydropower stations, characterized by including: The event database establishment module establishes a typical correlation event database based on the historical signal data and historical operation instruction data of the hydropower station. By obtaining the unique identification signal or signal combination of the event, the mapping relationship between the event and the signal is obtained; The device event relationship graph module establishes a device-event-signal relationship graph based on the mapping relationship between events and signals in the event library based on the typical association relationship established in the event library establishment module, and based on the relationship between devices and events; The real-time data docking and pre-processing module obtains multi-source real-time data, and performs fusion processing of multi-source real-time data in time sequence to obtain a fused real-time data stream; The event discovery and tracking module implements signal event discovery, multi-cache construction and maintenance based on the device-event-signal relationship graph established by the device event relationship graph module based on the fused real-time data stream obtained by the real-time data docking and preprocessing module. with event tracking.