A method for determining causal relationships in power grid monitoring information based on artificial intelligence

By using an AI-based power grid monitoring system, the system can acquire switch status in real time and dynamically define connectivity ranges, constructing a dynamic alarm causal topology graph. This solves the problem of misjudgment caused by changes in switch status under a fixed topology graph, enabling accurate correlation analysis and timely processing of power grid alarm events.

CN123116077APending Publication Date: 2026-11-17GUIZHOU POWER GRID CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611325646.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-28
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

In existing power grid monitoring systems, alarm event correlation analysis based on fixed equipment ledger topology diagrams is easily affected by changes in switch status, leading to misjudgments or omissions, especially in ring networks, dual power supplies, and tie switch scenarios, resulting in inaccurate alarm event correlation judgments.

Method used

Using an artificial intelligence-based approach, the system acquires the real-time status of power grid switches, dynamically defines the actual connectivity range, constructs a real-time inference task between alarm signal pairs, generates a dynamic alarm causal topology graph through multi-dimensional connectivity verification, performs joint root cause and independent root cause analysis, and finally outputs a list of alarm events.

Benefits of technology

It improves the accuracy of alarm event correlation analysis, reduces false positives and false negatives, enhances the timeliness of task processing, and ensures the completeness and accuracy of power grid fault analysis.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123116077A_ABST
    Figure CN123116077A_ABST
Patent Text Reader

Abstract

This invention discloses an artificial intelligence-based method for determining causal relationships in power grid monitoring information, relating to the field of power grid monitoring information processing. It acquires the real-time status of power grid switches corresponding to alarm signals to be analyzed and dynamically defines the current actual connectivity range. When preset re-deduction conditions are met, a real-time deduction task is constructed between alarm signal pairs with actual connectivity relationships. Multi-dimensional connectivity verification is performed based on the immediate deduction results of the real-time deduction task. When all verifications pass, a dynamic alarm causal topology graph is constructed. Joint root cause and independent root cause analyses are performed on the dynamic alarm causal topology graph, and corresponding events are categorized to obtain joint root cause alarm events and independent root cause alarm events. Historical records of similar events are obtained to verify the signal aggregation integrity of joint root cause alarm events and independent root cause alarm events. The corresponding alarm event information is then adjusted to output an alarm event list. This method improves the accuracy of alarm event correlation analysis.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the technical field of power grid monitoring information processing, and in particular to a method for determining the causal relationship of power grid monitoring information based on artificial intelligence. Background Technology

[0002] Currently, with the continuous expansion of power grid equipment scale and the continuous improvement of intelligent monitoring level, equipment fault alarm information is showing a geometric growth trend. The existing power grid monitoring business relies on the monitoring method of responding to alarm information one by one. It is necessary to pay attention to the judgment, analysis and feedback of each piece of information. When faced with the ever-increasing number of monitoring devices and the exponential growth of monitoring information, higher requirements are placed on the identification and judgment of alarm events in the monitoring information.

[0003] Existing alarm event analysis methods mainly rely on the independent identification, analysis, and feedback of individual alarm messages, neglecting the interrelationships or influences between alarm events in the monitoring data. This can easily lead to missed or incorrect judgments, and also puts significant pressure on subsequent alarm event verification. When determining whether two alarm events are related or have an impact, a fixed equipment ledger topology diagram is usually used for topological correlation judgment. As long as there are connected physical lines between the equipment related to the alarm events on the ledger topology diagram, the two alarm events are considered to have a causal relationship. However, in the actual operation of the power grid, the switching states are constantly changing, especially in ring networks, dual power supplies, and tie-line scenarios. Physical lines that are connected in a fixed ledger topology diagram may not be able to be energized or transmit signals due to the constantly changing switching states. This can easily lead to changes in the operation mode of physical lines due to random changes in switching states, resulting in unrelated alarm events being misjudged as related or related alarm events being misjudged as unrelated. Summary of the Invention

[0004] To address the problem that existing technologies using fixed equipment ledger topology diagrams for alarm event correlation analysis are prone to misjudgment or omission due to changes in switch status, this application provides an artificial intelligence-based method for determining causal relationships in power grid monitoring information. This method can calculate the actual connectivity status between devices in real time based on the current actual operating status of the power grid, and perform causal analysis between alarm events based on the actual connectivity status to determine correlations, thereby improving the accuracy of alarm event correlation analysis.

[0005] Firstly, the aforementioned inventive objective of this application is achieved through the following technical solution: A method for determining causal relationships in power grid monitoring information based on artificial intelligence, the method comprising: The grid switch status of the alarm signal to be analyzed is acquired in real time and the current actual connection range is dynamically defined. When the preset re-deduction conditions are met, a real-time deduction task is constructed between alarm signal pairs with actual connection relationships. Based on the real-time simulation results of the real-time simulation task, a multi-dimensional connectivity verification is performed on the alarm signal pairs. When all multi-dimensional connectivity verifications pass, the alarm causal relationship between the alarm signal pairs is constructed, and a dynamic alarm causal topology graph is generated. The dynamic alarm cause-effect topology graph is subjected to joint root cause analysis and independent root cause analysis respectively, and the corresponding events are classified based on the analysis results to obtain joint root cause alarm events and independent root cause alarm events. Obtain historical records of similar events, perform signal aggregation integrity verification on the joint root cause alarm events and independent root cause alarm events respectively, adjust the corresponding alarm event information based on the verification results, and output an alarm event list.

[0006] In a preferred embodiment, this application can be further configured as follows: the real-time acquisition of the grid switch status of the alarm signal to be analyzed and the dynamic delineation of the current actual connectivity range, and the construction of a real-time deduction task between alarm signal pairs with actual connectivity when the preset re-deduction conditions are met, specifically includes: Obtain the grid switch status corresponding to the alarm signal, and trigger the re-deduction condition when the grid switch status changes; A connectivity review is performed on the current real connectivity range. If the timeliness or state credibility review result in the connectivity review fails, the re-inference condition is triggered. The power grid system operating time is obtained, and the re-deduction condition is triggered when the power grid system operating time reaches the preset expiration check time. When the re-analysis condition is triggered, a real-time simulation task is constructed between alarm signal pairs that have a real connectivity relationship.

[0007] In a preferred embodiment, this application can be further configured as follows: the real-time acquisition of the grid switch status of the alarm signal to be analyzed and the dynamic delineation of the current actual connectivity range, and the construction of a real-time deduction task between alarm signal pairs with actual connectivity when the preset re-deduction conditions are met, further includes: Obtain existing simulation tasks for other alarm signal pairs, and determine whether there is an overlap in simulation range between the existing simulation tasks and the real-time simulation tasks; If they exist, obtain the current simulation status of existing simulation tasks with overlapping simulation ranges, and predict the actual simulation time of the existing simulation tasks based on the current simulation status. Determine whether the actual simulation time exceeds the reasonable waiting time limit of the real-time simulation task and simultaneously obtain the regional simulation results of the existing simulation tasks; If not, the regional simulation results are reused in the real-time simulation task to obtain the instant simulation results of the real-time simulation task. If so, the real-time simulation task is transferred to the verification queue. When the regional simulation result includes the equipment involved in the real-time simulation task, the real-time simulation task is removed from the verification queue and the regional simulation result is reused as the instant simulation result.

[0008] In a preferred embodiment, this application can be further configured as follows: The step of transferring the real-time simulation task to a verification queue, and when the regional simulation result includes devices involved in the real-time simulation task, removing the real-time simulation task from the verification queue and reusing the regional simulation result as an immediate simulation result, further includes: When the queue waiting time of a real-time simulation task to be verified exceeds the preset maximum retention limit and no simulation result for the region is obtained, the real-time simulation task will be manually judged and removed from the queue to be verified. The result of the manual judgment will be used as the immediate simulation result of the real-time simulation task.

[0009] In a preferred embodiment, this application can be further configured as follows: Based on the real-time inference results of the real-time inference task, a multi-dimensional connectivity check is performed on the alarm signal pairs; when all multi-dimensional connectivity checks pass, an alarm causal relationship is constructed between the alarm signal pairs, generating a dynamic alarm causal topology graph, specifically including: A power grid topology map is constructed by acquiring power grid equipment ledger data. The real-time simulation task is then simulated using the power grid topology map to obtain the simulation results for this round of the real-time simulation task. Based on the results of this round of inference, the connectivity timeliness, connectivity credibility, connectivity properties, and connectivity propagation time are checked in sequence to obtain multi-dimensional connectivity verification results. When all the multidimensional connectivity verification results pass, the alarm causal relationship between the corresponding alarm signal pairs is constructed, and a dynamic alarm causal topology graph is generated.

[0010] In a preferred embodiment, this application can be further configured such that: when all the multidimensional connectivity verification results pass, the step of constructing the alarm causal relationship between the corresponding alarm signal pairs and generating a dynamic alarm causal topology graph further includes: When there are multiple alarm causal relationship candidates for the same alarm signal pair, the connectivity confidence score, timeliness score and timing proximity score of each alarm causal relationship candidate are calculated and weighted summed to obtain the comprehensive confidence score. Based on the comprehensive confidence score, the optimal alarm causal relationship candidates are selected to construct a dynamic alarm causal topology graph.

[0011] In a preferred embodiment, this application can be further configured as follows: performing joint root cause analysis and independent root cause analysis on the dynamic alarm causal topology graph, and classifying corresponding events based on the analysis results to obtain joint root cause alarm events and independent root cause alarm events, specifically including: Based on the dynamic alarm causal topology graph, the root cause analysis of the alarm signal pairs is performed, an alarm causal topology path between the signal pair to be analyzed and the signal root cause is constructed, and the associated signals of the alarm causal topology path are marked based on the current real connectivity range. When the signal root cause is an independent root cause, all signals and associated signals in the alarm causal topology path are classified as independent root cause alarm events. When the signal root cause is a joint root cause, all signals and associated signals of the alarm causal topology path corresponding to the joint root cause are classified as joint root cause alarm events. The joint root cause alarm events include independent signal subsets and shared signal subsets.

[0012] In a preferred embodiment, this application can be further configured as follows: obtaining historical records of similar events, performing signal aggregation integrity verification on the joint root cause alarm events and independent root cause alarm events respectively, adjusting the corresponding alarm event information based on the verification results, and outputting an alarm event list, specifically including: Obtain the historical records of similar events for the independent root cause alarm events to verify the signal integrity of protection actions and linkage relationships. Based on the verification results, adjust the alarm event information of the independent root cause alarm events to obtain a list of independent root cause alarm events. Obtain the historical records of similar events of the joint root cause alarm event to verify the signal integrity of protection actions and linkage relationships. Based on the verification results, adjust the alarm event information of the joint root cause alarm event to obtain a list of joint root cause alarm events. The list of alarm events is obtained by combining the list of independent root cause alarm events and the list of combined root cause alarm events.

[0013] In a preferred embodiment, this application can be further configured as follows: the step of obtaining historical records of similar events, performing signal aggregation integrity verification on the joint root cause alarm events and independent root cause alarm events respectively, adjusting the corresponding alarm event information based on the verification results, and outputting an alarm event list, further includes: Obtain the actual signal reception data corresponding to the alarm event list, perform timeliness judgment on the missing signal pairs in the actual signal reception data, and determine whether the missing signal pairs were missed based on the timeliness judgment result; When the data entry corresponding to the actual signal received data is not in the alarm event list, it is determined that there is a falsely transmitted redundant signal in the actual signal received data.

[0014] Secondly, the above-mentioned inventive objective of this application is achieved through the following technical solutions: An artificial intelligence-based system for determining causal relationships in power grid monitoring information, the system comprising: The task deduction and judgment module is used to obtain the grid switch status of the alarm signal to be analyzed in real time and dynamically define the current real connection range. When the preset re-deduction conditions are met, a real-time deduction task is constructed between alarm signal pairs with real connection relationships. The alarm causality judgment module is used to perform multi-dimensional connectivity verification on the alarm signal pairs based on the real-time inference results of the real-time inference task. When the multi-dimensional connectivity verification passes, the alarm causal relationship between the alarm signal pairs is constructed, and a dynamic alarm causal topology graph is generated. The alarm event classification module is used to perform joint root cause analysis and independent root cause analysis on the dynamic alarm cause-effect topology graph, and classify the corresponding events based on the analysis results to obtain joint root cause alarm events and independent root cause alarm events. The event list adjustment module is used to obtain historical records of similar events, perform signal aggregation integrity verification on the joint root cause alarm events and independent root cause alarm events respectively, adjust the corresponding alarm event information based on the verification results, and output the alarm event list.

[0015] In summary, this application includes at least one of the following beneficial technical effects: 1. This application uses changes in switch states, the timeliness of connectivity results, and time-driven periodic checks as conditions for task re-inference to dynamically update changes in grid switch states and the actual connectivity range. After performing multi-dimensional connectivity verification on alarm signal pairs, it constructs alarm causal relationships between signal pairs, forming a dynamic alarm causal topology graph. This transforms the causal association judgment from a fixed desktop topology to a dynamic alarm causal topology graph that is dynamically related to the actual operating state of the grid. Based on the dynamic alarm causal topology graph, it performs joint root cause event and independent root cause event analysis and classification on alarm signal pairs, and performs integrity verification with reference to historical records of similar events. Finally, it outputs a corresponding alarm event list, which improves the accuracy of alarm event correlation analysis. 2. This application incorporates real-time simulation tasks into the silent waiting phase or computation phase of existing simulation tasks that are already in progress. It determines whether to incorporate the tasks based on a reasonable upper limit of the waiting time and the actual simulation time. After the silent waiting period is extended after incorporation, the application still independently judges whether the waiting time has expired based on its own trigger time. This avoids the response speed being dragged down by irrelevant external extension processes in fault propagation scenarios. Therefore, real-time simulation tasks that have timed out are transferred to the verification queue, reducing invalid waiting and resource consumption. When the existing simulation tasks obtain the simulation results in real time, a valid new trigger condition is generated in the verification queue to retrieve the waiting real-time simulation tasks and reuse the simulation results. This avoids the signal being passively delayed until the maximum time limit before being processed, thus improving the timeliness of task processing. 3. This application also supplements the directional description and dynamic expansion rules of the affected area delineation process, distinguishes between joint root cause alarm events and independent root cause alarm events, and specifies the adjustment rules for the alarm event list of the corresponding events. It incorporates the adjudication authority of shared member signals in the joint triggering rules and the default handling of unmarked historical rules into the signal collection integrity verification, making the data in the alarm event list more complete, clarifying the asymmetry of the timeliness threshold and propagation time window in self-statistics, unifying the waiting time window into the same parameter, aligning the configuration judgment when the joint root cause involves multiple devices to the same reference time, and the fallback method when there are no historical results before the reference time, so that the logic chain of the whole solution is completely closed in the links that are prone to interface gaps, such as cross-link reference, concurrent triggering, and multi-device joint judgment. Attached Figure Description

[0016] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. The elements or parts in the drawings are not necessarily drawn to scale.

[0017] Figure 1 This is a flowchart illustrating the implementation of the AI-based causal correlation judgment method for power grid monitoring information in this embodiment.

[0018] Figure 2 This is a structural block diagram of the AI-based power grid monitoring information causal correlation judgment system in this embodiment. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] It should be understood that, when used in this specification, the terms “comprising” and “including” indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0021] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification, unless the context clearly indicates otherwise, the singular forms “a,” “an,” and “the” are intended to include the plural forms.

[0022] It should also be further understood that the term "and / or" as used in this specification refers to any combination of one or more of the associated listed items, as well as all possible combinations, and includes such combinations.

[0023] In one embodiment, such as Figure 1 As shown, this application discloses a method for determining causal relationships in power grid monitoring information based on artificial intelligence, which specifically includes the following steps: S1: Real-time acquisition of the grid switch status of the alarm signal to be analyzed and dynamic delineation of the current actual connectivity range. When the preset re-deduction conditions are met, a real-time deduction task is constructed between alarm signal pairs with actual connectivity.

[0024] Specifically, in this embodiment, the grid switch status of the alarm signal pair is acquired. When all grid switches are in the "on" state, the signal pair is considered connected; when all grid switches are in the "off" state, the signal pair is considered disconnected. Based on the connection result determination, the current actual connection range is defined. The re-analysis conditions in this embodiment include switch status change, unreliable connection conclusion, and reaching the waiting time limit. Moreover, the triggering conditions are mutually exclusive, and the re-analysis process is triggered only once at a time, as detailed below: S11: Obtain the grid switch status corresponding to the alarm signal. When the grid switch status changes, trigger the re-deduction condition.

[0025] Specifically, the system continuously collects the real-time opening and closing status of each power grid switch, the signal quality identifier for each status data, and the operation mode arrangement information issued by the dispatcher. When the system is first put into operation, it first performs a complete simulation based on the current state of the entire network as an initial benchmark, and then performs simulation analysis on alarm signal pairs in incremental mode. Specifically, it obtains the power grid switch status of alarm signal pairs, and triggers the re-simulation condition when the power grid switch status changes.

[0026] S12: Perform connectivity verification on the current real connectivity range. If the timeliness or state credibility verification result in the connectivity verification result fails, trigger the re-inference condition.

[0027] Specifically, the current actual connectivity range is checked for timeliness and status reliability. If the timeliness or status reliability fails, the corresponding device connectivity result is deemed unreliable, triggering a re-analysis. Device ranges involved in the re-analysis that overlap with existing simulation tasks can be merged into existing simulation tasks for processing, reusing the simulation results of existing simulation tasks. The time difference is obtained by subtracting the current actual connectivity range's connectivity result determination time from the current time. The corresponding timeliness threshold is obtained based on the simulation category to which the actual connectivity determination result belongs. If the calculated time difference is greater than the timeliness threshold, the timeliness determination is deemed to have failed. If the time difference is less than or equal to the timeliness threshold, the timeliness determination is deemed to have passed. In this embodiment, the simulation category to which the connectivity determination result belongs includes two main categories: regional triggering and instant triggering. In this embodiment, the timeliness threshold is obtained by adding a preset multiple of the standard deviation of the historical time consumption samples of the corresponding category. In this embodiment, the state credibility is also judged by directly reading the state credibility identifier attached to the device connectivity determination result. When the state credibility identifier is credible, it is determined to pass; when the state credibility identifier is questionable, it is determined to fail.

[0028] S13: Obtain the power grid system operating time. When the power grid system operating time reaches the preset expiration check time, trigger the re-deduction condition.

[0029] Specifically, the system obtains the operating time of the power grid system and calculates the expiration check time based on the timer built into the power grid system. When the operating time of the power grid system reaches the preset expiration check time, the re-deduction condition is triggered.

[0030] S14: When the re-analysis condition is triggered, construct a real-time simulation task between alarm signal pairs that have a real connectivity relationship.

[0031] Specifically, when the re-deduction condition is triggered, a real-time deduction task is constructed between alarm signal pairs that have a real connectivity relationship. In this embodiment, the existence of a real connectivity relationship is determined by the switch state between alarm signal pairs being open.

[0032] Specifically, step S1 also includes: S15: Obtain the existing simulation tasks of other alarm signal pairs, and determine whether there is an overlap in the simulation range between the existing simulation tasks and the real-time simulation tasks. Specifically, the existing simulation tasks of other alarm signal pairs are acquired, including those in the ongoing stage and those in the silent waiting stage. In this embodiment, the silent waiting stage is when the current round of simulation has not yet started and the simulation begins after the waiting time window ends. The simulation range of the real-time simulation task and the existing simulation task is compared to determine whether there are overlapping devices. In this embodiment, the simulation range is the set of devices involved in the influence range of the alarm signal pair.

[0033] In this embodiment, the silent waiting period of the existing simulation task refers to the entire simulation action itself. When a power grid fault occurs, the protection level difference will cause the switching actions to occur continuously and intensively in a short period of time. If there is no silent period, each change of the switching action will be calculated separately immediately. This can easily lead to the current simulation task not being completed before the next simulation task starts immediately due to the change of the switching action. This causes the simulation task to evolve repeatedly towards the latest state indefinitely, wasting a lot of computing power. At the same time, it will also cut multiple related actions that should belong to the same fault handling process into unrelated rounds of isolated results, which is not conducive to power grid fault analysis.

[0034] In this embodiment, the independent waiting limit for the real-time simulation task is used to determine the timeliness of the response to the current specific judgment request. If no independent waiting limit is set, merging the real-time simulation task into the quiescent period of the existing simulation task will result in severe fault propagation and directional errors in the cause-effect judgment being prolonged.

[0035] S16: If it exists, obtain the current simulation status of existing simulation tasks with overlapping simulation ranges, and predict the actual simulation time of existing simulation tasks based on the current simulation status.

[0036] Specifically, if there are overlapping devices, the current simulation status of existing simulation tasks with overlapping simulation ranges is obtained, including the silent waiting state and the calculation in progress state. In the silent waiting state, the real-time simulation time is the sum of the silent waiting time and the simulation calculation time. In the calculation in progress state, the real-time simulation time is the remaining calculation time.

[0037] S17: Determine whether the actual simulation time exceeds the reasonable waiting time limit for real-time simulation tasks and simultaneously obtain the regional simulation results of existing simulation tasks.

[0038] Specifically, it is determined whether the real-time simulation time exceeds the upper limit of the reasonable waiting time for the real-time simulation task. In this embodiment, the upper limit of the reasonable waiting time for the real-time simulation task is started independently from the time the real-time simulation task is triggered. At the same time as the waiting time is determined, the regional simulation results of the existing simulation tasks are obtained. The simulation results include the connectivity determination results, overlap determination results, confidence identifiers and time identifiers between alarm signal pairs.

[0039] In this embodiment, reasonable waiting time is calculated only for re-analysis tasks triggered by connectivity result verification and periodic check cycles. In re-analysis tasks triggered by switch state changes, it is necessary to wait for the state of the corresponding area to stabilize before re-analyzing once. Therefore, the corresponding re-analysis task is allowed to be merged into the existing simulation task by default and waited in silence until the regional simulation results of the existing simulation task are obtained and reused as the instant simulation results of the real-time simulation task.

[0040] The deduction process in this embodiment is as follows: Based on the deduction task, the starting range is defined. According to the topology results before the deduction task is triggered, all devices covered by all existing paths passing through the starting switch are identified, and the affected area is delineated. In this embodiment, the existing paths passing through the starting switch do not distinguish directions. All power points that may reach the affected area through other paths are re-traversed to avoid missing the backup power supply paths in the ring network, thereby obtaining the deduction result.

[0041] S18: If not, the regional simulation results will be reused in the real-time simulation task to obtain the real-time simulation results of the real-time simulation task.

[0042] Specifically, if the real-time simulation time does not exceed the reasonable waiting time limit, the real-time simulation task will be merged into the silent zone and the simulation results of the region where the existing simulation task is waiting will be reused in the real-time simulation task as the instant simulation result. The simulation structure includes connectivity determination, overlap determination, confidence identifier and time identifier.

[0043] S19: If so, the real-time simulation task will be transferred to the verification queue. When the regional simulation result includes the equipment involved in the real-time simulation task, the real-time simulation task will be removed from the verification queue and the regional simulation result will be reused as the instant simulation result.

[0044] Specifically, if the real-time simulation time exceeds the reasonable waiting time limit, the real-time simulation task will be transferred to the verification queue for waiting. At this time, the existing simulation tasks will continue to be executed until the regional simulation results are obtained. It will be determined whether there are any devices in the regional simulation results that overlap with the waiting simulation tasks in the verification queue. That is, the regional simulation results will be written back to the common topology data of the power grid. If the devices involved in the write-back action include devices within the simulation range of the real-time simulation task in the verification queue, it will be determined as a hit. Thus, the real-time simulation task will be removed from the verification queue, and the regional simulation results of the involved devices will be reused as the real-time simulation results of the real-time simulation task.

[0045] In this embodiment, new switching actions may occur during the simulation process of existing simulation tasks, which may lead to an extension of the silent waiting time or simulation calculation time. If the real-time simulation task is not removed in time, and the waiting time of the real-time simulation task is passively extended along with the existing simulation task, directional errors such as more severe fault expansion, more intensive subsequent actions, and longer waiting time for subsequent results will occur. Therefore, it is necessary to remove the real-time simulation task in time and place it in the verification queue for waiting by judging a reasonable upper limit of waiting time.

[0046] This embodiment also includes: when the queue waiting time of the real-time simulation task to be verified exceeds the preset maximum retention limit and no regional simulation result is obtained, the real-time simulation task is manually judged and removed from the queue to be verified, and the manual judgment result is used as the instant simulation result of the real-time simulation task.

[0047] Specifically, in this embodiment, the simulation tasks in the queue to be verified do not wait indefinitely. Each simulation task has a maximum maximum retention time limit. When the queue waiting time of a real-time simulation task to be verified exceeds the preset maximum retention time limit and the regional simulation result has not yet been obtained, manual judgment is required to make the real-time simulation task jump out of the waiting period. The data related to the real-time simulation task is sent to the manual end for manual judgment and the real-time simulation task is removed from the queue to be verified. The manual judgment result is used as the manual simulation result of the real-time simulation task.

[0048] S2: Based on the real-time simulation results of the real-time simulation task, perform multi-dimensional connectivity verification on the alarm signal pairs. When all multi-dimensional connectivity verifications pass, construct the alarm causal relationship between the alarm signal pairs and generate a dynamic alarm causal topology graph. Specifically, step S2 includes: S21: Obtain power grid equipment ledger data to construct a power grid topology map, and use the power grid topology map to perform real-time simulation tasks to obtain the simulation results of this round of real-time simulation tasks.

[0049] Specifically, in this embodiment, when the real-time simulation task overlaps with an existing simulation task, the regional simulation results are directly reused. When the real-time simulation task does not overlap with an existing simulation task, the real-time simulation task needs to be simulated independently. Specifically, by acquiring power grid equipment ledger data and constructing a power grid topology map according to the fixed physical connection relationships in the ledger, starting from the equipment in the affected area determined by the real-time simulation task, all possible topology links with topology connection relationships are traversed, and the switch status information and connectivity information on each topology link are obtained. When the switch quality is marked as abnormal, the relevant topology link is marked as questionable. By using the switch status information and connectivity information on the topology link, it is determined whether there is a path completely connected by a closed switch between each power source point and the corresponding equipment, thereby obtaining the simulation result of this round of the real-time simulation task. The simulation result in this embodiment includes connectivity determination result, overlap determination result, credibility identifier and time identifier, as well as the actual time consumed from triggering to completion of this round of simulation, which is classified into independent time consumption sample sequences of regional triggering or instant triggering according to the triggering source. It should be noted that the overlap judgment is to determine whether there are more than two overlapping devices in all the topology links in the simulation. If so, it is determined that there is an overlap. Further, the operation mode arrangement information of the overlapping devices is checked to determine whether the switching operation of the overlapping devices is maintained closed and forms a closed loop operation. If so, it is determined to be a continuous closed loop overlap. Otherwise, it is determined to be a timed switching overlap during the switching process. At the same time, a waiting upper limit determined based on the actual switching operation time is started. In this embodiment, the waiting upper limit started by the timed switching overlap judgment is actively verified by the timed expiration check. When the verification result is that the overlap has disappeared, the overlap judgment result is confirmed. If the verification result is that the overlap has not disappeared, it is classified as a continuous closed loop overlap.

[0050] In this embodiment, the switch quality is confirmed based on the standardized remote signaling quality code mechanism in the power monitoring system. The switch quality identifier is a status word composed of multiple flag bits. Each flag bit corresponds to different abnormal types such as manual setting, blocked, suspicious, exceeding limit, and communication interruption. When the status word hits any abnormal flag bit, it is determined to be a quality abnormality. When all flag bits are not hit, it is determined to be a quality normality.

[0051] S22: Based on the results of this round of simulation, sequentially perform connectivity timeliness verification, connectivity credibility verification, connectivity property verification, and connectivity propagation time rationality verification to obtain multi-dimensional connectivity verification results.

[0052] Specifically, the simulation results in this embodiment include the regional simulation results of reused existing simulation tasks or the simulation results obtained independently from real-time simulation tasks. The simulation results in this embodiment are sequentially subjected to connectivity timeliness verification, connectivity credibility verification, connectivity property verification, and connectivity propagation time rationality verification. The connectivity timeliness verification in this embodiment is performed separately for regional triggering and real-time triggering. For the time-consuming sample sequences accumulated independently by regional triggering simulation and real-time triggering simulation, the sample mean and standard deviation are calculated respectively. The mean plus a preset multiple of the standard deviation is used as a dynamic cumulative threshold to calculate the time difference between the current time and the simulation time identifier of the current connectivity result. When the time difference is less than or equal to the dynamic cumulative threshold, the connectivity timeliness verification is determined to be passed.

[0053] When the time difference is greater than the dynamic cumulative threshold, an instant simulation is triggered for the device to which the signal belongs. If the instant simulation fails to obtain the experience result within a reasonable time, it is marked as pending verification.

[0054] The connectivity reliability verification is determined by reading the reliability identifier in the connectivity judgment result. If the connectivity judgment result carries a doubtful identifier, and the real-time quality identifier changes to normal after a reread, the new identifier is used as the basis for connectivity reliability verification. If the real-time quality identifier is an abnormal type such as doubtful after a reread, an instant simulation is triggered to find an independent path. The reliability identifier of the previous round is replaced by the reliability simulation result of the independent path. If the independent path cannot be found or the reliability simulation result of the independent path is still an abnormal type, it is judged as doubtful and marked as pending verification.

[0055] Connectivity verification prioritizes excluding disconnected paths before judging the connectivity of connected paths. If there is no overlap, timing and rule matching is performed directly. If there is instantaneous switching overlap, independent timing and rule matching is performed on each topology path of the covered devices to generate multiple candidate results. After confirming that the overlap has been restored to a single source through a timed expiration check, the candidate results that match the subsequently restored single source are selected as the connectivity results. If there is persistent loop overlap, the upstream of each path of the covered devices is first checked for the existence of protection action signals with directional indications. The direction definition of each directional protection device and its direction correspondence in the topology path are maintained first to determine the final causal path. If no directional indication is available upstream of each path, the candidate causal relationships corresponding to each path are retained and finally confirmed by manual judgment.

[0056] The connection conduction time rationality verification checks whether the connection conduction time difference falls within the reasonable range of the conduction window. The range of the conduction window is initially determined by the time range calibrated in the relay protection operation. After the causal chain time difference samples accumulated in subsequent operation and supported by multiple independent evidence reach a preset number, the mean and standard deviation of the samples are calculated. Under the constraint of the preset upper limit amplitude of change, the window center of the conduction window is updated with the mean and the window width of the conduction window is updated with the standard deviation. The connection results that fall within the reasonable range of the conduction window are compared with the preset directional rules. If the signal type combination conforms to the power grid causal sequence pattern set by the directional rules, the causal relationship is confirmed to be established.

[0057] S23: When the multi-dimensional connectivity verification results all pass, construct the alarm causal relationship between the corresponding alarm signal pairs and generate a dynamic alarm causal topology graph.

[0058] Specifically, when the multi-dimensional connectivity verification results all pass, the connectivity links between the corresponding alarm signal pairs are obtained to construct the alarm causal relationship, and a dynamic alarm causal topology graph is generated that is dynamically updated according to each inference result.

[0059] Step S23 in this embodiment further includes: S231: When there are multiple alarm causal relationship candidates for the same alarm signal pair, calculate the connectivity confidence score, timeliness score and timing proximity score of each alarm causal relationship candidate and perform a weighted summation operation to obtain the comprehensive confidence score.

[0060] Specifically, when multiple causal relationship candidates exist for the same alarm signal pair, i.e., the same signal is simultaneously pointed to by multiple candidate results and they contradict each other, the candidates are filtered by calculating a comprehensive confidence score. In this embodiment, the comprehensive confidence score is calculated by weighted summation of connectivity confidence score, timeliness score, and temporal proximity score. Specifically, the connectivity confidence judgment result, timeliness judgment result, and temporal proximity judgment result are first normalized and mapped to the 0-1 interval, and then weighted summation is performed. In this embodiment, the connectivity confidence score is 1 when there is no overlap and the connectivity judgment does not have a doubt mark. After being recalculated due to doubt, the middle value is taken for those confirmed as reliable. Parallel candidates that are in an overlapping state and have not yet converged into a unique path are given a low score close to 0. The timeliness score is calculated by the ratio between the time difference interval between the inference time and the signal occurrence time and the corresponding category timeliness threshold, and then subtracting the corresponding ratio from 1 to obtain the timeliness score. The timeliness score ranges from 0 to 1. The temporal proximity score is obtained by calculating the ratio between the absolute value of the difference between the actual time difference and the window center value and the half width of the window, and then subtracting the corresponding ratio from 1. The temporal proximity score ranges from 0 to 1.

[0061] In this embodiment, the weighted summation weights are initialized to be equally distributed. If the difference between the highest score and the second highest score is less than a preset threshold, it is marked as pending verification. The initialized weighted summation weights are adjusted based on historical confirmation data that has been manually verified. The adjustment range is constrained by a preset upper limit. If the correlation of a dimension is strong in the historical confirmation data, the corresponding weight is increased. If the historical confirmation results show that the proportion of options with high temporal proximity scores that are ultimately confirmed as correct candidates is high, the corresponding weight value is increased.

[0062] S232: Construct a dynamic alarm causal topology graph by selecting the optimal alarm causal relationship candidates based on the comprehensive confidence score.

[0063] Specifically, the candidate with the highest overall confidence score is selected as the optimal alarm causal relationship candidate, and events are spliced ​​and root causes are labeled according to the optimal alarm causal relationship candidate to construct an alarm causal topology graph.

[0064] S3: Perform joint root cause analysis and independent root cause analysis on the dynamic alarm cause-effect topology graph, and classify the corresponding events based on the analysis results to obtain joint root cause alarm events and independent root cause alarm events.

[0065] Specifically, step S3 includes: S31: Perform root cause analysis on alarm signal pairs based on dynamic alarm causal topology graph, construct alarm causal topology path between the signal pair to be analyzed and the signal root cause, and label the associated signals of the alarm causal topology path based on the current real connectivity range.

[0066] Specifically, root cause analysis is performed on alarm signal pairs based on the dynamic alarm causal topology graph. All devices without upstream signals are identified as event origins in the dynamic alarm causal topology graph. Event origins include joint root causes and independent root causes. A joint root cause refers to an alarm signal pair having two or more event origins and conforming to preset joint triggering rules. An independent root cause refers to an alarm signal pair having only one event origin. The alarm causal topology link between the event origin and the corresponding device of the alarm signal pair is marked as the alarm causal topology path. The associated signals of the alarm causal topology path are marked based on the current real connectivity range. The associated signals are all signals that may have a topological association with the alarm causal topology path within the current real connectivity range, such as signals on the branches of the alarm causal topology path.

[0067] S32: When the root cause of a signal is an independent root cause, all signals and associated signals in the alarm causal topology path will be classified as independent root cause alarm events.

[0068] Specifically, when the root cause of a signal is an independent root cause, all signals and related signals on the alarm causal topology path are classified as independent root cause alarm events.

[0069] S33: When the signal root cause is a joint root cause, all signals and associated signals of the alarm causal topology path corresponding to the joint root cause are classified as joint root cause alarm events. Joint root cause alarm events include independent signal subsets and shared signal subsets.

[0070] Specifically, when the signal root cause is a joint root cause, all signals and associated signals along the alarm causal topology path corresponding to the joint root cause are classified as joint root cause alarm events. Joint root cause alarm events include independent signal subsets and shared signal subsets. The upstream signals in the independent signal subsets have no overlap. If there is overlap, the joint triggering rules for the overlapping signals are used to determine whether the rules indicate that shared member signals are allowed. If so, they are marked as shared signal subsets. For historical joint triggering rules that do not indicate sharing attributes, they are treated as not allowing sharing by default, or they are treated as sharing only after the annotation is completed and explicitly allowed.

[0071] The joint triggering rules in this embodiment involve a list of signal types and the allowed time relationships between signal types.

[0072] In this embodiment, for candidate causal relationships that are continuously overlapping and have no available directional indication information, and are retained in parallel, the upstream signal of the candidate causal relationship is used to determine whether the event type has been determined. If so, the corresponding candidate causal relationship is extended as a possible branch of the determined event. If the upstream signal does not determine the event type, the event type is re-inferred in all candidate causal relationships. The signal with no upstream cause and the earliest occurrence time in the re-inferencing result is marked as the root cause. If it is a jointly triggered and merged event, all signals involved in the joint triggering are marked as joint root causes. The final confirmation is made by manual judgment.

[0073] S4: Obtain historical records of similar events, perform signal aggregation integrity verification on joint root cause alarm events and independent root cause alarm events respectively, adjust the corresponding alarm event information based on the verification results, and output the alarm event list.

[0074] Specifically, step S4 includes: S41: Obtain the historical records of similar events for independent root cause alarm events to verify the signal integrity of protection actions and linkage relationships. Based on the verification results, adjust the alarm event information of independent root cause alarm events to obtain a list of independent root cause alarm events.

[0075] Specifically, the system retrieves historical records of similar events for independent root cause alarm events, verifies the signal integrity of protection actions and linkage relationships for these events, and removes the corresponding protection actions and linkage relationships if the historical records show that the logically expected verification signal does not appear in the actual simulation results. If the historical records show the protection actions and linkage relationships of the verification signal in the simulation results multiple times, the integrity verification is considered successful. The system retains the relevant alarm event information that has passed verification and removes the alarm event information that has failed verification, resulting in a list of independent root cause alarm events. The list includes the signal type, reasonable occurrence time window, and the branch to which it belongs.

[0076] S42: Obtain the historical records of similar events of the joint root cause alarm events to verify the signal integrity of protection actions and linkage relationships. Based on the verification results, adjust the alarm event information of the joint root cause alarm events to obtain a list of joint root cause alarm events.

[0077] Specifically, the system retrieves historical records of similar events related to the joint root cause alarm events, performs signal integrity verification on the inference results of the joint root cause alarm events for protection actions and linkage relationships. When the historical records show multiple instances of a logically expected verification signal that does not appear in the actual inference results, the corresponding protection actions and linkage relationships are removed and a removal reminder is triggered. When the historical records show multiple instances of protection actions and linkage relationships for verification signals in the inference results, the integrity verification is deemed successful. The system retains the relevant alarm event information that has passed verification and removes the alarm event information that has failed verification, thus obtaining a list of joint root cause alarm events.

[0078] S43: Combine the lists of independent root cause alarm events and the combined root cause alarm event lists to obtain an alarm event list.

[0079] This embodiment also includes: S5: Obtain the corresponding real signal reception data of the alarm event list, perform timeliness judgment on the missing signal pairs in the real signal reception data, and determine whether the missing signal pairs have been missed based on the timeliness judgment result.

[0080] Specifically, the system obtains the actual signal reception data corresponding to the alarm event list, filters out missing signal pairs in the alarm event list, and judges the timeliness of the missing signal pairs. For missing signal pairs with independent root causes, the waiting time is calculated from the time the root cause signal occurs. For missing signal pairs with joint root causes, the waiting time is calculated from the reference time. When the window expires, the timed expiration check actively initiates verification. If the missing signal pair still does not appear when the waiting window ends, it is determined to be a missed signal.

[0081] S6: When the data entry corresponding to the actual signal received data is not in the alarm event list, it is determined that there is a falsely transmitted redundant signal in the actual signal received data.

[0082] Specifically, if the data entry corresponding to the actual signal received data is not in the alarm event list composed of the independent root cause alarm event list and the joint root cause alarm event list of the current security event, and does not belong to any independent event, then it is determined that there is a falsely transmitted redundant signal in the actual signal received data.

[0083] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0084] In one embodiment, an artificial intelligence-based causal correlation judgment system for power grid monitoring information is provided, which corresponds one-to-one with the artificial intelligence-based causal correlation judgment method for power grid monitoring information in the above embodiments. For example... Figure 2 As shown, this AI-based power grid monitoring information causal correlation judgment system includes a task deduction and judgment module, an alarm causal judgment module, an alarm event classification module, and an event list adjustment module. Detailed descriptions of each functional module are as follows: The task deduction and judgment module is used to obtain the grid switch status of the alarm signal to be analyzed in real time and dynamically define the current real connection range. When the preset re-deduction conditions are met, a real-time deduction task is constructed between alarm signal pairs with real connection relationships. The alarm causality judgment module is used to perform multi-dimensional connectivity verification on alarm signal pairs based on the real-time simulation results of the real-time simulation task. When all multi-dimensional connectivity verifications pass, the alarm causal relationship between alarm signal pairs is constructed, and a dynamic alarm causal topology graph is generated.

[0085] The alarm event classification module is used to perform joint root cause analysis and independent root cause analysis on the dynamic alarm cause-effect topology graph, and classify the corresponding events based on the analysis results to obtain joint root cause alarm events and independent root cause alarm events.

[0086] The event list adjustment module is used to obtain historical records of similar events, perform signal aggregation integrity verification on joint root cause alarm events and independent root cause alarm events respectively, adjust the corresponding alarm event information based on the verification results, and output the alarm event list.

[0087] Specific limitations regarding the AI-based power grid monitoring information causal correlation judgment system can be found in the limitations of the AI-based power grid monitoring information causal correlation judgment method described above, and will not be repeated here. Each module in the aforementioned AI-based power grid monitoring information causal correlation judgment system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0088] Those skilled in the art will recognize that the units of the various examples described in connection with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application of the technical solution and the constraints involved. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of the invention.

[0089] In the embodiments provided by the present invention, it should be understood that the division of units is only a logical functional division. In actual implementation, there may be other division methods, such as multiple units can be combined into one unit, one unit can be split into multiple units, or some features can be ignored.

[0090] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0091] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0092] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the present invention specification.

Claims

1. A method for determining causal relationships in power grid monitoring information based on artificial intelligence, characterized in that, The method includes: The grid switch status of the alarm signal to be analyzed is acquired in real time and the current actual connection range is dynamically defined. When the preset re-deduction conditions are met, a real-time deduction task is constructed between alarm signal pairs with actual connection relationships. Based on the real-time simulation results of the real-time simulation task, a multi-dimensional connectivity verification is performed on the alarm signal pairs. When all multi-dimensional connectivity verifications pass, the alarm causal relationship between the alarm signal pairs is constructed, and a dynamic alarm causal topology graph is generated. The dynamic alarm cause-effect topology graph is subjected to joint root cause analysis and independent root cause analysis respectively, and the corresponding events are classified based on the analysis results to obtain joint root cause alarm events and independent root cause alarm events. Obtain historical records of similar events, perform signal aggregation integrity verification on the joint root cause alarm events and independent root cause alarm events respectively, adjust the corresponding alarm event information based on the verification results, and output an alarm event list.

2. The method for determining causal relationships in power grid monitoring information based on artificial intelligence according to claim 1, characterized in that, The process of constructing the real-time simulation task, which involves acquiring the grid switch status of the alarm signal to be analyzed in real time and dynamically defining the current actual connectivity range, and constructing a real-time simulation task between alarm signal pairs with actual connectivity when the preset re-simulation conditions are met, specifically includes: Obtain the grid switch status corresponding to the alarm signal, and trigger the re-deduction condition when the grid switch status changes; A connectivity review is performed on the current real connectivity range. If the timeliness or state credibility review result in the connectivity review fails, the re-inference condition is triggered. The power grid system operating time is obtained, and the re-deduction condition is triggered when the power grid system operating time reaches the preset expiration check time. When the re-analysis condition is triggered, a real-time simulation task is constructed between alarm signal pairs that have a real connectivity relationship.

3. The method for determining causal relationships in power grid monitoring information based on artificial intelligence according to claim 2, characterized in that, The real-time acquisition of the grid switch status of the alarm signal to be analyzed and the dynamic delineation of the current actual connectivity range, and the construction of a real-time simulation task between alarm signal pairs with actual connectivity when the preset re-simulation conditions are met, further includes: Obtain existing simulation tasks for other alarm signal pairs, and determine whether there is an overlap in simulation range between the existing simulation tasks and the real-time simulation tasks; If they exist, the current simulation status of existing simulation tasks with overlapping simulation ranges is obtained, and the actual simulation time of the existing simulation tasks is predicted based on the current simulation status. Determine whether the actual simulation time exceeds the reasonable waiting time limit of the real-time simulation task and simultaneously obtain the regional simulation results of the existing simulation task; If not, the regional simulation results are reused in the real-time simulation task to obtain the instant simulation results of the real-time simulation task. If so, the real-time simulation task is transferred to the verification queue. When the regional simulation result includes the equipment involved in the real-time simulation task, the real-time simulation task is removed from the verification queue and the regional simulation result is reused as the instant simulation result.

4. The method for determining causal relationships in power grid monitoring information based on artificial intelligence according to claim 3, characterized in that, The step of transferring the real-time simulation task to a verification queue, and when the regional simulation result includes the equipment involved in the real-time simulation task, removing the real-time simulation task from the verification queue and reusing the regional simulation result as the immediate simulation result, further includes: When the queue waiting time of a real-time simulation task to be verified exceeds the preset maximum retention limit and no simulation result for the region is obtained, the real-time simulation task will be manually judged and removed from the queue to be verified. The result of the manual judgment will be used as the immediate simulation result of the real-time simulation task.

5. The method for determining causal relationships in power grid monitoring information based on artificial intelligence according to claim 1, characterized in that, The alarm signal pairs are subjected to multi-dimensional connectivity verification based on the real-time simulation results of the real-time simulation task. When all multi-dimensional connectivity verifications pass, the alarm causal relationship between the alarm signal pairs is constructed, and a dynamic alarm causal topology graph is generated, specifically including: A power grid topology map is constructed by acquiring power grid equipment ledger data. The real-time simulation task is then simulated using the power grid topology map to obtain the simulation results for this round of the real-time simulation task. Based on the results of this round of inference, the connectivity timeliness, connectivity credibility, connectivity properties, and connectivity propagation time are checked in sequence to obtain multi-dimensional connectivity verification results. When all the multidimensional connectivity verification results pass, the alarm causal relationship between the corresponding alarm signal pairs is constructed, and a dynamic alarm causal topology graph is generated.

6. The method for determining causal relationships in power grid monitoring information based on artificial intelligence according to claim 5, characterized in that, The step of constructing the alarm causal relationship between corresponding alarm signal pairs and generating a dynamic alarm causal topology graph when all the multidimensional connectivity verification results pass also includes: When there are multiple alarm causal relationship candidates for the same alarm signal pair, the connectivity confidence score, timeliness score and timing proximity score of each alarm causal relationship candidate are calculated and weighted summed to obtain the comprehensive confidence score. Based on the comprehensive confidence score, the optimal alarm causal relationship candidates are selected to construct a dynamic alarm causal topology graph.

7. The method for determining causal relationships in power grid monitoring information based on artificial intelligence according to claim 1, characterized in that, The process involves performing joint root cause analysis and independent root cause analysis on the dynamic alarm cause-effect topology graph, and classifying the corresponding events based on the analysis results to obtain joint root cause alarm events and independent root cause alarm events, specifically including: Based on the dynamic alarm causal topology graph, the root cause analysis of the alarm signal pairs is performed, an alarm causal topology path between the signal pair to be analyzed and the signal root cause is constructed, and the associated signals of the alarm causal topology path are marked based on the current real connectivity range. When the signal root cause is an independent root cause, all signals and associated signals in the alarm causal topology path are classified as independent root cause alarm events. When the signal root cause is a joint root cause, all signals and associated signals of the alarm causal topology path corresponding to the joint root cause are classified as joint root cause alarm events. The joint root cause alarm events include independent signal subsets and shared signal subsets.

8. The method for determining causal relationships in power grid monitoring information based on artificial intelligence according to claim 1, characterized in that, The process of obtaining historical records of similar events, performing signal aggregation integrity verification on the joint root cause alarm events and independent root cause alarm events respectively, adjusting the corresponding alarm event information based on the verification results, and outputting an alarm event list specifically includes: Obtain the historical records of similar events for the independent root cause alarm events to verify the signal integrity of protection actions and linkage relationships. Based on the verification results, adjust the alarm event information of the independent root cause alarm events to obtain a list of independent root cause alarm events. Obtain the historical records of similar events of the joint root cause alarm event to verify the signal integrity of protection actions and linkage relationships. Based on the verification results, adjust the alarm event information of the joint root cause alarm event to obtain a list of joint root cause alarm events. The list of alarm events is obtained by combining the list of independent root cause alarm events and the list of combined root cause alarm events.

9. The method for determining causal relationships in power grid monitoring information based on artificial intelligence according to claim 1, characterized in that, The step of obtaining historical records of similar events, performing signal aggregation integrity verification on the joint root cause alarm events and independent root cause alarm events respectively, adjusting the corresponding alarm event information based on the verification results, and outputting an alarm event list also includes: Obtain the actual signal reception data corresponding to the alarm event list, perform timeliness judgment on the missing signal pairs in the actual signal reception data, and determine whether the missing signal pairs were missed based on the timeliness judgment result; When the data entry corresponding to the actual signal received data is not in the alarm event list, it is determined that there is a falsely transmitted redundant signal in the actual signal received data.

10. A system for determining causal relationships in power grid monitoring information based on artificial intelligence, characterized in that, The system includes: The task deduction and judgment module is used to obtain the grid switch status of the alarm signal to be analyzed in real time and dynamically define the current real connection range. When the preset re-deduction conditions are met, a real-time deduction task is constructed between alarm signal pairs with real connection relationships. The alarm causality judgment module is used to perform multi-dimensional connectivity verification on the alarm signal pairs based on the real-time inference results of the real-time inference task. When the multi-dimensional connectivity verification passes, the alarm causal relationship between the alarm signal pairs is constructed, and a dynamic alarm causal topology graph is generated. The alarm event classification module is used to perform joint root cause analysis and independent root cause analysis on the dynamic alarm cause-effect topology graph, and classify the corresponding events based on the analysis results to obtain joint root cause alarm events and independent root cause alarm events. The event list adjustment module is used to obtain historical records of similar events, perform signal aggregation integrity verification on the joint root cause alarm events and independent root cause alarm events respectively, adjust the corresponding alarm event information based on the verification results, and output the alarm event list.