A fault diagnosis method, system, device and medium of a power system

Abstract

The application discloses a kind of power system fault diagnosis method, system, equipment and medium, it is characterized in that, including: receiving the alarm event of power system;According to each alarm event is grouped to sliding time window, obtain several to-be-diagnosed alarm event set;Each to-be-diagnosed alarm event set corresponding cause event and each cause event corresponding event time period are obtained by time sequence constraint network;According to each event time period is merged to each cause event, obtain several candidate cause events;Each candidate cause event corresponding expected alarm event is obtained by time sequence constraint network, according to expected alarm event, to-be-diagnosed alarm event set and target cause solving function, obtain several target cause events;The expected alarm event corresponding to each target cause event is compared with each to-be-diagnosed alarm event set, and based on the result of comparison, the fault diagnosis result of power system is output.The application can improve the accuracy of power system alarm processing.

Description

Technical Field

[0001] This invention relates to the field of fault diagnosis, and in particular to a fault diagnosis method, system, equipment and medium for power systems. Background Technology

[0002] With the continuous expansion of power systems and the widespread application of smart substations and dispatch automation systems, after a power system fault occurs, relay protection devices, circuit breakers, and other automated equipment will generate a large number of time-stamped alarm messages within a very short time and upload them to the dispatch center. Dispatchers need to quickly determine the location, cause, and propagation process of the fault based on these alarm messages in order to formulate appropriate fault handling plans in a timely manner. Therefore, how to quickly extract effective fault information from a large number of alarm messages and achieve accurate fault cause diagnosis has become an important research topic in the field of power system dispatch automation.

[0003] Existing power system alarm processing methods mainly include rule-based reasoning methods based on expert systems, learning methods based on artificial intelligence, and fault diagnosis methods based on optimization models. Expert systems require the establishment and maintenance of a large number of diagnostic rules, leading to high maintenance costs and difficulties in rule updates as the power grid expands. Artificial intelligence methods typically rely on a large number of historical fault samples for model training, resulting in high training costs, insufficient model interpretability, and limited generalization ability. While optimization models can transform the fault diagnosis process into a combinatorial optimization problem, they mostly only utilize the correlation between alarm events for analysis, failing to adequately utilize the temporal logical relationships between alarms. They are particularly ill-suited to the uncertainty of alarm timing caused by factors such as GPS time synchronization errors, communication delays, and differences in equipment response times, easily leading to inaccurate fault cause judgments, incorrect alarm correlations, and insufficient abnormal alarm identification capabilities, thus affecting the accuracy of power system alarm processing. Summary of the Invention

[0004] This invention provides a method, system, equipment, and medium for fault diagnosis in power systems, which can improve the accuracy of power system alarm processing.

[0005] In a first aspect, embodiments of the present invention provide a fault diagnosis method for a power system, comprising: Receive alarm events generated during the operation of the power system; The alarm events are grouped according to a preset sliding time window to obtain several sets of alarm events to be diagnosed; The cause event corresponding to each set of alarm events to be diagnosed and the event time period corresponding to each cause event are obtained through a preset temporal constraint network. Based on the time periods of each event, the causal events are merged to obtain several candidate causal events; The expected alarm event corresponding to each candidate cause event is obtained through a preset time-series constraint network, and several target cause events are obtained based on the expected alarm event, the set of alarm events to be diagnosed, and the preset target cause solution function. The expected alarm events corresponding to each of the target cause events are compared with each of the sets of alarm events to be diagnosed, and the fault diagnosis results of the power system are output based on the comparison results.

[0006] This invention improves the accuracy of power system alarm processing by acquiring the original alarm information required for fault diagnosis, providing a unified data foundation for subsequent alarm correlation analysis; by centrally processing alarm events that may be related within the same time period, it avoids isolated analysis that may lead to omissions in alarm correlation, thus improving the accuracy of power system alarm processing; by combining the temporal constraints between alarm events to infer the possible fault causes and occurrence time ranges corresponding to each alarm event, it narrows the search range for fault causes, thus improving the accuracy of power system alarm processing; by fusing cause events corresponding to the same fault cause and occurring at the same time, it reduces duplicate and invalid cause assumptions, thus improving the accuracy of power system alarm processing; by acquiring the expected alarm event corresponding to each candidate cause event, and combining the expected alarm event, the set of alarm events to be diagnosed, and the target cause solution function to determine the target cause event, it can comprehensively evaluate the explanatory power of different candidate cause events for actual alarm events, and screen out the fault cause that best matches the actual alarm, thus improving the accuracy of power system alarm processing; by comparing the expected alarm events corresponding to each target cause event with each set of alarm events to be diagnosed, it can identify normal alarms, missed alarms, and abnormal alarms, thus improving the accuracy of power system alarm processing.

[0007] Furthermore, the construction process of the temporal constraint network includes: Obtain the protection configuration diagram of the power system, the circuit breaker operating logic, and the relay protection principle; Based on the protection configuration diagram, the circuit breaker operation logic, and the relay protection principle, the temporal logic relationship between each cause event and each alarm event, as well as between each alarm event, is obtained, so as to generate alarm configuration rules based on each of the temporal logic relationships. The set of cause event nodes and the set of alarm event nodes are determined based on the alarm configuration rules; The first directed edge from each cause event node to each alarm event node is determined based on the temporal logical relationship between each cause event and each alarm event; The second directed edge between each alarm event node is determined based on the temporal logical relationship between each alarm event; The temporal constraint network is constructed based on the set of cause event nodes, the set of alarm event nodes, the first directed edge, the second directed edge, and the preset time distance constraint.

[0008] The above steps, by acquiring the power system's protection configuration diagram, circuit breaker operating logic, and relay protection principles, construct a timing constraint network that includes causal events, alarm events, and their timing constraints. This network can fully describe the event correlations during the power system fault propagation process, providing a reliable basis for subsequent fault cause reasoning and alarm correlation analysis, thereby improving the accuracy of power system alarm processing.

[0009] Furthermore, the step of obtaining the causal event corresponding to each set of alarm events to be diagnosed and the event time period corresponding to each causal event through a preset temporal constraint network includes: In the time-constraint network, determine the alarm event nodes corresponding to each set of alarm events to be diagnosed; Based on the reverse temporal reasoning along the temporal constraint network for each alarm event node, the causal event corresponding to each alarm event node and the event time period corresponding to each causal event are obtained.

[0010] The above steps, by performing reverse temporal reasoning on alarm events based on a temporal constraint network, determine the causal events corresponding to each alarm event and the possible time periods of the events. This can narrow the scope of fault cause analysis, improve the accuracy of fault cause location, and thus improve the accuracy of power system alarm processing.

[0011] Furthermore, the step of obtaining the expected alarm event corresponding to each candidate cause event through a preset temporal constraint network includes: In the temporal constraint network, determine the causal event node corresponding to each of the candidate causal events; Based on the forward temporal reasoning along the temporal constraint network for each of the aforementioned causal event nodes, the expected alarm events corresponding to each of the aforementioned causal event nodes and the expected occurrence time windows corresponding to each of the aforementioned expected alarm events are obtained.

[0012] The above steps involve forward temporal reasoning of candidate cause events based on a temporal constraint network to obtain the expected alarm event and its expected occurrence time window corresponding to each candidate cause event. This establishes a temporal correspondence between fault causes and alarm events, providing a basis for subsequent alarm matching and cause screening, thereby improving the accuracy of power system alarm processing.

[0013] Furthermore, based on the expected alarm event, the set of alarm events to be diagnosed, and the preset target cause solution function, several target cause events are obtained, including: Set corresponding decision variables for each of the candidate cause events, obtain the target candidate cause events with preset values ​​in each of the decision variables, and count the first number of each of the target candidate cause events; The expected alarm events corresponding to each of the target candidate cause events are matched one by one with each set of alarm events to be diagnosed, and a second number of expected alarm events that do not exist in each set of alarm events to be diagnosed, and a third number of expected alarm events that do not exist in each set of alarm events to be diagnosed; The first quantity, the second quantity, the third quantity, and the preset weight coefficients are weighted and summed to obtain the target cause solution function; Solving the target cause solution function yields several target cause events.

[0014] The above steps involve statistically analyzing the number of target candidate causal events, the number of missed alarms and the number of unmatched alarms between the expected alarm events and the set of alarm events to be diagnosed, and then constructing a comprehensive optimization objective for solution. This allows the target causal events to achieve a comprehensive balance between reducing alarm omissions, reducing false matches, and satisfying the minimum cause assumption principle, thereby improving the accuracy of target causal event identification and, consequently, improving the accuracy of power system alarm processing.

[0015] Furthermore, the solution function for the target cause is solved to obtain several target cause events, including: An initial solution vector is constructed based on the decision variables corresponding to each of the candidate causal events. The value of each decision variable in the initial solution vector is changed sequentially to obtain several neighborhood solution vectors. Substitute each of the neighborhood solution vectors into the target cause solution function for calculation to obtain the target function value corresponding to each of the neighborhood solution vectors; Based on the objective function values ​​and the preset tabu search rules, the current solution vector is selected from the neighborhood solution vectors. The current solution vector is used as the initial solution vector for the next iteration to continue iterative solution until the preset iteration conditions are met, and then the target solution vector is output. The candidate causal events corresponding to the decision variables that take preset values ​​in the target solution vector are determined as the target causal events.

[0016] The above steps iteratively optimize the target cause solution function by using a tabu search algorithm. While ensuring the search range, it avoids repeatedly searching for local solutions and continuously updates the solution vector with better objective function values. Finally, it obtains target cause events that match the set of alarm events to be diagnosed more closely, thereby improving the accuracy of the target cause event solution and thus improving the accuracy of power system alarm processing.

[0017] Furthermore, the step of comparing the expected alarm events corresponding to each of the target cause events with each of the sets of alarm events to be diagnosed, and outputting the fault diagnosis results of the power system based on the comparison results, includes: Determine whether there exists an actual alarm event in each set of alarm events to be diagnosed that is consistent with the expected alarm event corresponding to each of the target cause events; If the actual occurrence time of the actual alarm event falls within the corresponding expected occurrence time window, the actual alarm event is determined to be a normal alarm. If the actual occurrence time of the actual alarm event is not within the corresponding expected occurrence time window, the actual alarm event will be identified as an abnormal alarm. When there is no corresponding actual alarm event, the corresponding target expected alarm event is identified as a missed alarm; The normal alarms, abnormal alarms, and missed alarms are summarized to obtain the fault diagnosis results of the power system.

[0018] The above steps compare the expected alarm events corresponding to the target cause event with the set of alarm events to be diagnosed one by one, and classify, identify and summarize normal alarms, abnormal alarms and missed alarms based on the alarm matching situation and the correspondence between the actual occurrence time and the expected occurrence time window. This can accurately distinguish different types of alarm anomalies, reduce the probability of fault misjudgment and missed judgment, and thus improve the accuracy of power system alarm processing.

[0019] Secondly, embodiments of the present invention provide a stability control system for a power system, the stability control system being used to perform the operation of a power system fault diagnosis method as described in this application.

[0020] Thirdly, embodiments of the present invention provide a terminal device, including: a processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other through the communication bus; The memory is used to store at least one executable instruction that causes the processor to perform the operation of a power system fault diagnosis method as described in this application.

[0021] Fourthly, embodiments of the present invention provide a computer-readable storage medium comprising a stored computer program, wherein, when the computer program is executed, it controls the device or system where the computer-readable storage medium is located to perform a fault diagnosis method for a power system as described in this application.

[0022] The above description is merely an overview of the technical solutions of the embodiments of the present invention. In order to better understand the technical means of the embodiments of the present invention and to implement them in accordance with the contents of the specification, and to make the above and other objects, features and advantages of the embodiments of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0023] To more clearly illustrate the technical solution of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0024] Figure 1 This is a flowchart illustrating one embodiment of a fault diagnosis method for a power system provided in this application; Figure 2 This is a schematic diagram of the process for constructing a temporal constraint network provided in this application; Figure 3 This is a flowchart illustrating the process of determining the target cause event based on the target cause solution function provided in this application; Figure 4 This is a flowchart illustrating the process of solving for the target cause event based on the tabu search algorithm provided in this application. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0026] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application; the terms “comprising” and “having”, and any variations thereof, in the specification, claims, and foregoing description of the drawings are intended to cover non-exclusive inclusion.

[0027] In the description of the embodiments of this application, technical terms such as "first" and "second" are used only to distinguish different objects and should not be construed as indicating or implying relative importance or implicitly specifying the number, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this application, "multiple" means two or more, unless otherwise explicitly defined.

[0028] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0029] In the description of the embodiments in this application, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this document generally indicates that the preceding and following related objects have an "or" relationship.

[0030] In the description of the embodiments of this application, the term "multiple" refers to two or more (including two), similarly, "multiple sets" refers to two or more (including two sets), and "multiple pieces" refers to two or more (including two pieces).

[0031] In the description of the embodiments of this application, unless otherwise expressly specified and limited, technical terms such as "installation," "connection," "joining," and "fixing" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. For those skilled in the art, the specific meaning of the above terms in the embodiments of this application can be understood according to the specific circumstances.

[0032] With the expansion of power systems and the widespread application of smart substations and dispatch automation systems, a large number of time-stamped alarm messages are generated during fault processes. How to quickly and accurately extract the cause of the fault from the massive amount of alarms has become an important problem in the field of power system dispatch. Existing methods based on expert systems, artificial intelligence, and optimization models have problems such as difficulty in maintaining knowledge bases, reliance on a large number of samples with insufficient interpretability, and insufficient utilization of alarm timing logic. They are also unable to effectively cope with the time uncertainty caused by GPS time synchronization errors, communication delays, and differences in equipment response, resulting in insufficient accuracy in fault diagnosis and abnormal alarm identification capabilities.

[0033] See Figure 1 In order to improve the accuracy of power system alarm processing, an embodiment of the present invention provides a power system fault diagnosis method, including steps S101 to S106; Step S101: Receive alarm events generated by the power system during operation; In some embodiments, data connections are established with the power dispatch master station, substation monitoring system, protection information substation, and communication server to receive alarm events generated by the power system during operation in real time. These alarm events include, but are not limited to, circuit breaker trip alarms, relay protection operation alarms, reclosing operation alarms, busbar abnormality alarms, line fault alarms, equipment alarms, and protection device self-test alarms. Each alarm event includes at least an alarm identifier, the identifier of the corresponding equipment, the alarm type, and an alarm timestamp, where the alarm timestamp is the actual occurrence time of the corresponding alarm event. To ensure the reliability of subsequent timing inference, a unified clock source is used to time-align the received alarm events, and the original alarm timestamp information is retained to account for factors such as GPS time synchronization errors, communication delays, sampling errors, and device response differences in subsequent processing. Alarm events are cached according to their arrival order to form a continuous alarm stream data, providing input data for subsequent alarm grouping based on sliding time windows, timing-constrained network inference, and fault diagnosis. A single received alarm event can be represented as: ,in, Indicates the alarm event identifier. This indicates the alarm time stamp corresponding to the alarm event.

[0034] Through the above steps, real-time acquisition and unified time alignment of multi-source alarm events during power system operation are achieved, ensuring the integrity and temporal consistency of alarm data. This provides an accurate and reliable data foundation for subsequent alarm correlation analysis, temporal constraint reasoning, and fault cause diagnosis based on sliding time windows, thereby improving the accuracy and real-time performance of power system alarm processing.

[0035] Step S102: Group the alarm events according to a preset sliding time window to obtain several sets of alarm events to be diagnosed; In some embodiments, the alarm events are grouped according to a preset sliding time window to obtain several sets of alarm events to be diagnosed. Specifically, after completing the offline initialization of the timing constraint network and related mapping relationships, the system enters an online operating state and receives continuous alarm streams sent by the communication server in real time. A sliding time window is preset, including a window length and a window sliding step. As the power system operating time progresses, the arrival time of each alarm event is continuously monitored. When the system time advances to the end time of the current sliding time window, the alarm processing procedure corresponding to the current sliding time window is triggered. All alarm events received within the current sliding time window are extracted and summarized according to the occurrence time of the alarm events. All alarm events belonging to the same sliding time window are grouped into a set of alarm events to be diagnosed. The set of alarm events to be diagnosed can be represented as follows: ,in, Indicates the first One alarm event, Indicates the actual time of occurrence of the corresponding alarm event. This indicates the number of alarm events contained within the current sliding time window. As the sliding time window continues to move forward according to a preset step size, the alarm extraction and grouping process described above is repeated, sequentially obtaining multiple sets of alarm events to be diagnosed from the continuous alarm stream.

[0036] Through the above steps, the spatiotemporally related alarm events in the continuous alarm stream are effectively aggregated, avoiding misjudgments caused by isolated analysis of individual alarms, improving the integrity of associated alarms, and providing a more accurate alarm set for subsequent fault cause reasoning based on time-constrained networks.

[0037] Step S103: Obtain the causal event corresponding to each set of alarm events to be diagnosed and the event time period corresponding to each causal event through a preset temporal constraint network; Please refer to Figure 2 In some embodiments, the construction process of the temporal constraint network includes: steps S201 to S206; Step S201: Obtain the protection configuration diagram of the power system, the circuit breaker operating logic, and the relay protection principle; Step S202: Based on the protection configuration diagram, the circuit breaker operation logic, and the relay protection principle, obtain the temporal logic relationship between each cause event and each alarm event, as well as between each alarm event, and generate alarm configuration rules based on each of the temporal logic relationships; Step S203: Determine the set of cause event nodes and the set of alarm event nodes based on the alarm configuration rules; Step S204: Determine the first directed edge from each cause event node to each alarm event node based on the temporal logical relationship between each cause event and each alarm event; Step S205: Determine the second directed edge between each alarm event node according to the temporal logical relationship between each alarm event; Step S206: The temporal constraint network is constructed based on the set of cause event nodes, the set of alarm event nodes, the first directed edge, the second directed edge, and the preset time distance constraint.

[0038] In this first embodiment, before the system is put into online operation, offline initialization is performed to build the knowledge base and reasoning foundation for the time-constraint network. Basic data such as the power system's protection configuration diagram, circuit breaker operating logic, and relay protection principles are acquired. The protection configuration diagram is used to determine the connection relationships between lines, busbars, transformers, circuit breakers, and relay protection devices in the power system; the circuit breaker operating logic is used to determine the corresponding operating conditions, operating sequence, and linkage relationships of the circuit breaker; and the relay protection principles are used to determine the protection operating logic and the order of protection actions for different protection devices. The protection configuration diagram, circuit breaker operating logic, and relay protection principles are uniformly analyzed to serve as the foundational data for constructing the time-constraint network.

[0039] In this first embodiment, the protection actions and alarm events that can be triggered corresponding to each fault cause are determined according to the protection configuration diagram. The sequential relationship between circuit breaker action and each alarm event is determined according to the circuit breaker action logic. The timing relationship between protection initiation, protection output, circuit breaker tripping, and subsequent alarms is determined according to the relay protection principle. This extracts the timing logic relationships between each cause event and each alarm event, as well as between each alarm event. Alarm configuration rules are generated based on these timing logic relationships. The alarm configuration rules include cause-event-alarm-event rules and alarm-event-alarm-event rules. The cause-event-alarm-event rules describe the timing relationship between a cause event and its triggering alarm event, and their representation is as follows: ,in, Indicates the first One cause event, Indicates the first One alarm event, Indicates the cause event Time of occurrence and alarm event Time distance constraints between occurrence times. Alarm event—Alarm event rules are used to describe the temporal relationship between different alarm events triggered by the same cause event, and their representation is as follows: ,in, Indicates the first One alarm event, Indicates the first One alarm event, This represents the time distance constraint between two alarm events. The time distance constraint is used to represent the minimum and maximum allowed time interval between two related events.

[0040] In this first embodiment, a set of causal event nodes is established based on all causal events in the alarm configuration rules, and a set of alarm event nodes is established based on all alarm events in the alarm configuration rules. Let the time-series constraint network be represented as: ,in, Representing a time-constrained network, Represents a set of nodes. Represents a set of directed edges. Represents a set of time points. Represents a set of unary constraints. Let represent a set of binary constraints. The set of nodes satisfies: ,in, Represents the set of all nodes. Represents the set of cause event nodes. Represents the set of alarm event nodes. This represents the union of sets. The set of cause event nodes is represented as: ,in, Indicates the first One cause event node This represents the total number of cause event nodes. The set of alarm event nodes is represented as: ,in, Indicates the first One alarm event node, This indicates the total number of alarm event nodes.

[0041] In this first embodiment, the cause-event rule in the alarm configuration rules is traversed. When a cause event can trigger a corresponding alarm event, a first directed edge is established between the corresponding cause event node and the alarm event node. The first directed edge is represented as follows: ,in, Indicates the cause event node, Indicates the alarm event node, This represents a set of directed edges; the direction of the directed edges indicates the chronological order in which events occur, i.e., the order in which events occur. Event triggered after occurrence It occurs. Simultaneously, the time distance constraint in the corresponding alarm configuration rule is saved as an edge attribute of this first directed edge.

[0042] In this first embodiment, each alarm event-alarm event rule in the alarm configuration rules is traversed. When the first alarm event can trigger the second alarm event, a second directed edge is established between the corresponding two alarm event nodes. Similarly, the following method is used... Describe the second directed edge, where, Indicates the preceding alarm event node, This indicates the node of the subsequent alarm event. The second directed edge also stores the corresponding time distance constraint, which is used to represent the minimum and maximum allowed time interval between two alarm events.

[0043] In this first embodiment, the set of cause event nodes, the set of alarm event nodes, the first directed edge, and the second directed edge are uniformly organized to form a time-constrained directed acyclic graph. The event time point set is established as follows: ,in, Indicates the first The time of occurrence of each cause event. Indicates the first The time of occurrence of each alarm event This represents the set of all event time points. For any given time point... Establish the corresponding univariate time constraint: ,in, Indicates an event The corresponding time interval, Indicates the earliest possible time of the event. This represents the latest possible time of the event. For causal events, the initial time interval is unknown or a large range; for alert events, the time interval will be determined gradually during subsequent temporal reasoning. For any directed edge... Establish the corresponding binary time distance constraint: ,in, Indicates an event With the event Time distance constraints between them Indicates an event The time of occurrence, Indicates an event The time of occurrence, Indicates an event Post-event The minimum time interval between occurrences Indicates an event Post-event The maximum time interval between occurrences. The binary time distance constraint is directly derived from the alarm configuration rules and stored as an edge attribute of each directed edge in the time-series constraint network, thereby completing the construction of the time-series constraint network.

[0044] In some embodiments, obtaining the causal event corresponding to each set of alarm events to be diagnosed and the event time period corresponding to each causal event through a preset temporal constraint network includes: determining the alarm event node corresponding to each set of alarm events to be diagnosed in the temporal constraint network; and performing reverse temporal reasoning along the temporal constraint network based on each alarm event node to obtain the causal event corresponding to each alarm event node and the event time period corresponding to each causal event.

[0045] In some embodiments, determining the alarm event node corresponding to each set of alarm events to be diagnosed in the time-series constraint network specifically involves: obtaining the set of alarm events to be diagnosed within the current sliding time window. ,in, This represents the set of alarm events to be diagnosed. Indicates the first One alarm event, Indicates an alarm event The corresponding actual time of occurrence, This represents the number of alarm events in the set of alarm events to be diagnosed. It iterates through each alarm event in the set of alarm events to be diagnosed, and identifies the alarm event node set in the time-series constraint network based on the alarm event identifier. Search for the corresponding alarm event node in the middle; among which, Represents the set of alarm event nodes. This represents the total number of alarm event nodes in the temporal constraint network. When an alarm event node corresponding to the current alarm event exists in the set of alarm event nodes, that alarm event node is identified as the target alarm event node, and its actual occurrence time is recorded. Node matching is performed sequentially for all alarm events in the set of alarm events to be diagnosed, resulting in the target alarm event node set corresponding to the set of alarm events to be diagnosed, providing input nodes for subsequent reverse temporal reasoning.

[0046] In some embodiments, reverse temporal reasoning is performed along the temporal constraint network based on each alarm event node to obtain the causal event corresponding to each alarm event node and the event time period corresponding to each causal event. Specifically, for each alarm event node in the target alarm event node set, the observation time interval of the alarm event is constructed according to the actual occurrence time of the corresponding alarm event. ,in, Indicates the first The observation time interval for each alarm event; Indicates the first The actual time of occurrence of each alarm event; This represents the time tolerance, used to compensate for GPS time synchronization errors, communication delays, sampling errors, and protection device response errors. Starting with the current alarm event node as the starting point for reverse reasoning, the system traverses the directed edges in the time-series constraint network connected to the current alarm event node in the reverse direction, searching for causal event nodes that can reach the current alarm event node. Assume there exists a directed path from causal event node c to the current alarm event node a. ,in, This represents a directed path between the cause event node and the alarm event node. Indicates the first position on the path Each event node This indicates the number of edges contained in the path. Let the i-th edge on the path be... The time distance constraint corresponding to each directed edge is ,in, This represents the time distance constraint between two adjacent event nodes. Indicates an event Post-event The minimum time interval between occurrences Indicates an event Post-event The maximum time interval between occurrences. Based on the time distance constraint, calculate the possible time intervals of the causal event node: ,in, This indicates the time range during which the causal event corresponding to the current path may occur. Indicates the lower bound of the observation time interval. This indicates the upper bound of the observation time interval. This applies when there are multiple directed paths between the cause event node and the current alarm event node. At that time, calculate the possible time intervals for each path. Then, the overlapping parts in each time interval are merged to obtain the final possible time window corresponding to the cause event node: ,in, This indicates the time window in which the triggering event corresponding to the alarm event may ultimately occur. This indicates an operation to merge overlapping time intervals. It maps the cause event node and its corresponding potential time window into an alarm event-to-potential-cause mapping relationship. ,in, This involves a mapping table representing alarm events to possible causal events. It iterates through all alarm event nodes in the set of alarm events to be diagnosed, queries the corresponding alarm event-to-possible-cause mapping table, and summarizes the causal events corresponding to all alarm events and their corresponding final possible time windows to obtain an initial set of candidate causal events. When multiple candidate causal events in the initial set correspond to the same causal event and have overlapping final possible time windows, the intersection of these time windows is calculated, and the intersection is used as the corrected occurrence time period of the causal event. The merged causal event and its corrected occurrence time period are then used as the final candidate causal events.

[0047] By using the above steps, we can leverage the pre-built event relationships and time distance constraints in the time-constrained network to perform reverse time-series reasoning on alarm events, accurately locate possible fault causes and their time ranges by integrating multiple alarm information, avoid misjudgments caused by a single alarm, and improve the accuracy and reliability of power system fault diagnosis.

[0048] Step S104: Merge the causal events according to the time periods of each event to obtain several candidate causal events; In some embodiments, the causal events are merged according to the event time periods to obtain several candidate causal events. Specifically, this involves traversing each causal event, filtering causal events with the same causal event identifier, and determining whether there is an overlapping area between the corresponding event time periods. Let's assume that two causal events are obtained through reverse temporal reasoning from different alarm events. as well as ,in, and These represent two causal event hypotheses. and These represent the corresponding causal events. and These represent the time periods of the corresponding causal events. Determine if the two causal events satisfy the condition. ,in, This indicates that two causal events correspond to the same causal event. When two causal events correspond to the same causal event, further judgment is made as to whether the time period of the corresponding events meets the condition. ,in, This represents the intersection of the time periods of two events. This represents the empty set. When the event time periods overlap, the two causal events are merged, and the intersection is used as the merged event time period to obtain a new causal event. ,in, Indicates the causal event after the merger. This represents the corrected event time period. The new causal event replaces the original two corresponding causal events, and the process continues to iterate through the remaining causal events, repeatedly performing the causal event identifier consistency check and the event time period overlap check, until no causal events satisfy the merging conditions are found. Finally, all merged causal events are combined into a candidate causal event set. ,in, Represents the set of candidate cause events. Indicates the first One candidate cause event, This indicates the number of candidate cause events.

[0049] Through the above steps, the same cause event obtained from multiple alarm events can be fused, and the intersection of event time periods can be used to narrow down the occurrence range of cause events, reduce duplicate candidate causes, improve the accuracy of cause event location, and provide a more accurate candidate set for subsequent cause screening and fault diagnosis.

[0050] Step S105: Obtain the expected alarm event corresponding to each candidate cause event through a preset time-series constraint network, so as to obtain a number of target cause events based on the expected alarm events, the set of alarm events to be diagnosed and the preset target cause solution function; In some embodiments, obtaining the expected alarm event corresponding to each candidate causal event through a preset temporal constraint network includes: determining the causal event node corresponding to each candidate causal event in the temporal constraint network; and performing forward temporal reasoning along the temporal constraint network based on each causal event node to obtain the expected alarm event corresponding to each causal event node and the expected occurrence time window corresponding to each expected alarm event.

[0051] In some embodiments, determining the causal event node corresponding to each of the candidate causal events in the temporal constraint network specifically involves: obtaining a set of candidate causal events. ,in, Represents the set of candidate cause events. Indicates the first One candidate cause event, This indicates the number of candidate causal events. Iterate through each candidate causal event in the candidate causal event set, extract the causal event identifier corresponding to each candidate causal event, and then add it to the causal event node set of the time-series constraint network. Search for cause event nodes with the same cause event identifier; among them, Represents the set of cause event nodes. Indicates the first One cause event node This represents the total number of causal event nodes. When a causal event node corresponding to the current candidate causal event exists in the set of causal event nodes, that causal event node is determined as the target causal event node, and the event time period of the corresponding candidate causal event is read. ,in, This indicates the time period of the current causal event. Indicates the earliest time when the causative event could have occurred. This represents the latest possible time of the causal event. Repeat the above process to establish the correspondence between all candidate causal events and causal event nodes, resulting in a set of causal event nodes corresponding to each candidate causal event.

[0052] In some embodiments, forward temporal reasoning is performed along the temporal constraint network based on each of the causal event nodes to obtain the expected alarm events corresponding to each causal event node and the expected occurrence time window corresponding to each expected alarm event. Specifically, for each causal event node in the set of causal event nodes, a forward traversal is performed along the directed edge direction in the temporal constraint network, using the causal event node as the starting node, to search for alarm event nodes that can be reached by the current causal event node. Let the current causal event node be... To the alarm event node There exists a directed path ,in, This represents a directed path between the cause event node and the alarm event node. Indicates the first position on the path Each event node This indicates the number of edges contained in the path. Let the i-th edge on the path be... The time distance constraint corresponding to each directed edge is ,in, Represents event nodes With event nodes Time distance constraints between them Indicates an event Post-event The minimum time interval between occurrences Indicates an event Post-event The maximum time interval between occurrences. The event duration based on the causative event. Perform forward temporal reasoning along the directed path to calculate the expected occurrence time interval of the alarm event corresponding to the current path: ,in, This indicates the expected time interval for the alarm event corresponding to the current path. Indicates the earliest time when the causative event could have occurred. Indicates the latest possible time of the causative event. This applies when there are multiple directed paths between the current causative event node and the same alarm event node. At that time, calculate the expected occurrence time interval for each directed path. ,in, Indicates the first A directed path, Indicates the first The expected occurrence time intervals for each directed path are identified. Overlapping intervals within each expected occurrence time interval are merged to obtain the final expected occurrence time window for the alarm event corresponding to the current causal event. ,in, This represents the expected time window from the causal event to the alarm event. This represents an operation that merges multiple overlapping time intervals. It maps the alarm events corresponding to the current cause event node to their expected occurrence time windows. ,in, A mapping table representing the cause event to the expected alarm event. This indicates an alarm event that can be triggered by the current cause event. This represents the expected time window for the corresponding alarm event. Repeat the forward sequential reasoning process described above until all target cause event nodes are in forward reasoning, obtaining the expected alarm event for each candidate cause event and the expected time window for each expected alarm event.

[0053] Please refer to Figure 3 In some embodiments, obtaining several target cause events based on the expected alarm event, the set of alarm events to be diagnosed, and a preset target cause solution function includes: steps S301 to S304. Step S301: Set corresponding decision variables for each of the candidate causal events, obtain the target candidate causal events with preset values ​​in each of the decision variables, and count the first number of each of the target candidate causal events; Step S302: Match the expected alarm events corresponding to each of the target candidate cause events with each of the sets of alarm events to be diagnosed one by one, and determine the second number of expected alarm events that do not exist in each set of alarm events to be diagnosed, and the third number of expected alarm events that do not exist in each set of alarm events to be diagnosed; Step S303: The first quantity, the second quantity, the third quantity, and the preset weight coefficient are weighted and summed to obtain the target cause solution function; Step S304: Solve the target cause solution function to obtain several target cause events.

[0054] In this first embodiment, after merging the candidate cause events, a set of candidate cause events is obtained: ,in, Indicates the number of candidate cause events. Indicates the first There are 10 candidate causal events. For each candidate causal event in the set of candidate causal events, a corresponding binary decision variable is set, constructing a decision variable vector: ,in, This represents the vector of decision variables corresponding to the process of solving the cause of the objective; Indicates candidate cause event The corresponding decision variables; when When, it indicates acceptance of the candidate cause event. As the real cause of the event; when When, indicate the reason for rejecting a candidate event. Traverse the decision variable vector Select all that satisfy The candidate cause events are used as the target candidate cause event set: ,in, This represents the target candidate cause event set. The number of candidate cause events in this set is counted to obtain the first count: ,in, Indicates the first number of candidate cause events for the target. This represents the total number of candidate causal events. The first quantity is used to characterize the number of causal events included in the current combination of causal hypotheses, in order to satisfy the principle of minimum causal hypothesis.

[0055] In this first embodiment, for each target candidate cause event in the target candidate cause event set, the mapping table from cause events generated during the initialization phase to expected alarms is queried: ,in, This represents a mapping table from causal events to expected alarms. Indicates the causal event, Indicates the expected alarm event. This represents the expected occurrence time window of the alarm event corresponding to the causal event. The expected alarm events corresponding to all candidate causal events are unioned to obtain the total expected alarm set corresponding to the current combination of causal assumptions. ,in, Represents a vector of decision variables The corresponding total expected alarm set, Indicates candidate cause event The corresponding expected alarm event set. The total expected alarm set. With the set of alarm events to be diagnosed Match them one by one; among them, This represents the set of alarm events to be diagnosed. Indicates the first One actual alarm event, Indicates the actual time of occurrence of the corresponding alarm event. This represents the number of alert events awaiting diagnosis. The second number is obtained by counting the number of alerts that exist in the expected alert set but not in the pending alert event set. ,in, Indicates the number of missed alarms. This indicates alarms that exist only in the expected alarm set. Simultaneously, the number of alarms that exist in the alarm event set to be diagnosed but not in the total expected alarm set is counted, yielding a third quantity: ,in, This indicates the number of alarms is unexplained. This refers to alerts that actually occurred but cannot be explained by any assumed cause. Ultimately, the number of missed alerts is used as the second number, and the number of unexplained alerts is used as the third number.

[0056] In this first embodiment, the first quantity obtained in step S301 The second quantity obtained in step S302 And the third quantity Construct the target cause solution function: ,in, This represents the function for solving the objective cause. Represents a vector of decision variables. This represents the weighting coefficient corresponding to the missed alarm. This indicates the weighting coefficient corresponding to the unexplained alarm. This represents the weighting coefficient corresponding to the number of causal events. Indicates the second quantity. Indicates the third quantity. Let represent the first quantity; where the first term is used to reduce the number of missed alarms, the second term is used to reduce the number of unexplainable alarms, and the third term is used to limit the number of causal events, so that the objective function satisfies the principle of least cause assumption while ensuring explanatory power.

[0057] In this first embodiment, the target cause solution function constructed in step S303 is modeled as an unconstrained 0-1 integer programming problem and solved using a tabu search algorithm. An initial decision variable vector is randomly generated. The initial decision variable vector is used as the current solution vector. and initialize the optimal solution vector. ; Calculate the current objective function value and the optimal objective function value A taboo list is established to record the indices of decision variables corresponding to the most recent variable flipping operations, and a preset taboo period is set. Based on the current solution vector, each decision variable in the current solution vector is flipped sequentially to generate a corresponding neighborhood solution vector. Each neighborhood solution vector is then substituted into the objective function to calculate the corresponding objective function value. Among all neighborhood solutions, the one with the smallest objective function value is selected as a candidate solution. If the corresponding variable index does not appear in the taboo list, this candidate solution is accepted as the new current solution; if the corresponding variable index exists in the taboo list but satisfies the condition... If the desired outcome is met, the aspiration criterion is triggered, and the candidate solution is still accepted; otherwise, the non-taboo neighborhood solution with the smallest objective function value is selected as the new current solution. The variable index corresponding to the current move is added to the taboo list; when the taboo list length exceeds the preset taboo period, the variable index that was added to the taboo list earliest is deleted. If the current objective function value satisfies... Then update the optimal solution. At the same time, update the optimal objective function value. Repeat the neighborhood search process described above until the preset maximum number of iterations is reached, or until the optimal objective function value has not been updated during a preset number of consecutive iterations, at which point the tabu search ends. Finally, based on the optimal solution vector... Select one that satisfies The corresponding candidate cause events form the final target cause event set: ,in, Represents the set of events that lead to the final goal. This indicates the corresponding candidate cause event; the target cause event set is output as the cause event set for subsequent alarm diagnostic analysis.

[0058] Please refer to Figure 4 In some embodiments, solving the target cause solution function to obtain several target cause events includes: steps S401 to S404; Step S401: Construct an initial solution vector based on the decision variables corresponding to each of the candidate causal events, and sequentially change the value of each decision variable in the initial solution vector to obtain several neighborhood solution vectors; Step S402: Substitute each of the neighborhood solution vectors into the target cause solution function for calculation to obtain the target function value corresponding to each of the neighborhood solution vectors; Step S403: Based on the objective function values ​​and the preset tabu search rules, select the current solution vector from the neighborhood solution vectors, and use the current solution vector as the initial solution vector for the next iteration to continue iterative solution until the preset iteration conditions are met, and then output the target solution vector. Step S404: The candidate causal events corresponding to the decision variables with preset values ​​in the target solution vector are determined as the target causal events.

[0059] In this first embodiment, the set of candidate cause events is set as follows: ,in, Represents the set of candidate cause events. Indicates the first One candidate cause event, This represents the total number of candidate causal events. A binary decision variable is set for each candidate causal event, and a value of length [value missing] is constructed. Initial solution vector: ,in, Represents the initial solution vector; Indicates the first The decision variables corresponding to each candidate causal event; when When, it indicates acceptance of the corresponding candidate cause event; when When the value is set to zero, the corresponding candidate cause event is rejected. The initial solution vector can be generated randomly, where each decision variable takes the value 0 or 1 with a preset probability; or it can be initialized using a greedy strategy, where the initial solution vector can be set to an all-zero vector. The initial solution vector is then used as the current solution vector. Meanwhile, the initial solution vector is used as the current optimal solution vector: Based on the current solution vector, a neighborhood structure is constructed. The value of each decision variable in the current solution vector is changed sequentially: when the original value of the current decision variable is 0, it is changed to 1; when the original value of the current decision variable is 1, it is changed to 0. Each time a decision variable is flipped, a neighborhood solution vector is obtained. Let the th... If each decision variable is flipped, the corresponding neighborhood solution vector is obtained: ,in, Indicates the first Neighborhood solution vectors Indicates the first The values ​​of each decision variable after the flip operation is performed. After flipping all decision variables in sequence, a total of [number] values ​​are generated. The neighborhood solution vectors are used as candidate search solutions for the current iteration process.

[0060] In this embodiment, for each neighborhood solution vector obtained in step S401, the corresponding target candidate cause event is first determined based on the decision variable with a value of 1 in the neighborhood solution vector. The cause event to expected alarm mapping table is then queried to obtain the expected alarm event set corresponding to each target candidate cause event. Finally, the union of all expected alarm event sets is calculated to obtain the total expected alarm set corresponding to the current neighborhood solution. ,in, This represents the total expected alarm set corresponding to the current neighborhood solution. Indicates candidate cause event The corresponding expected alarm event set. The total expected alarm set. With the actual set of alarm events to be diagnosed Compare the numbers. Count the number of alarms that are expected to exist in the alarm set but not in the actual alarm set: ,in, Indicates the number of missed alarms. This represents alerts that exist in the expected alert set but not in the actual alert set. It counts the number of alerts that exist in the actual alert set but not in the expected alert set. ,in, This indicates the number of alarms is unexplained. This represents alarms that exist in the actual alarm set but not in the expected alarm set. Count the number of causal events corresponding to the current neighborhood solution: ,in, This represents the number of causal events corresponding to the current neighborhood solution. Substituting the above indicators into the target cause solution function: ,in, This represents the objective function value corresponding to the neighborhood solution. This represents the weighting coefficient corresponding to the missed alarm. This indicates the weighting coefficient corresponding to the unexplained alarm. This represents the weighting coefficient corresponding to the number of causal events. Indicates the number of missed alarms. This indicates the number of alarms is unexplained. This indicates the number of causal events. After calculating all neighborhood solutions, the objective function values ​​corresponding to each neighborhood solution are obtained.

[0061] In this first embodiment, a taboo list is established during the initialization phase to record the index of the decision variable corresponding to the most recently executed flip operation, and a taboo period is set. The taboo period is set to... or ,in, Indicates the period of prohibition. This represents the number of decision variables. Among all neighborhood solutions, the neighborhood solution with the smallest objective function value is selected as the candidate move. If the variable index corresponding to the candidate move does not exist in the taboo list, then that neighborhood solution is accepted as the current solution vector. If the variable index corresponding to the candidate move exists in the taboo list, then it is determined whether the desire criterion is satisfied: ,in, This represents the objective function value corresponding to the current candidate neighborhood solution. This represents the current optimal objective function value. If the above conditions are met, the tabu restriction is lifted, and the neighborhood solution is accepted as the current solution vector; if the above conditions are not met, the current candidate neighborhood solution is discarded, and the non-taboo neighborhood solution with the smallest objective function value is selected as the current solution vector. The variable index corresponding to this flip operation is added to the tabu list; when the tabu list length exceeds the tabu time limit, the variable index that entered the tabu list earliest is deleted. The current solution vector is updated to the initial solution vector for the next iteration, and steps S401 and S402 are re-executed. If the conditions are met... Then update the optimal solution. Simultaneously update ;in, This represents the current optimal solution vector. Represents the current solution vector. This represents the current objective function value. This represents the current optimal objective function value. Repeat the above iterative process until the preset maximum number of iterations is reached, or until the optimal objective function value remains unchanged during a preset number of consecutive iterations. Then, end the iteration and output the objective solution vector. .

[0062] In this embodiment, the target solution vector output in step S403 is obtained: ,in, This represents the optimal solution vector obtained by the tabu search algorithm. It iterates through all decision variables in the objective solution vector, and when a certain decision variable satisfies... When that happens, the corresponding candidate cause event is determined. These are the target cause events. The final set of target cause events is: ,in, Represents the set of events that cause the target. Indicates the cause of the target event. This indicates that the corresponding decision variable takes a value of 1.

[0063] Through the above steps, an optimized solution model can be constructed based on the temporal reasoning results of candidate cause events. The tabu search algorithm is then used to select the optimal combination of causes from multiple candidate causes, thereby achieving accurate analysis of complex alarm correlations, reducing invalid cause assumptions, and improving the accuracy and reliability of fault cause location.

[0064] Step S106: Compare the expected alarm events corresponding to each of the target cause events with each set of alarm events to be diagnosed, and output the fault diagnosis results of the power system based on the comparison results.

[0065] In some embodiments, comparing the expected alarm events corresponding to each of the target cause events with each set of alarm events to be diagnosed, and outputting the fault diagnosis result of the power system based on the comparison result, includes: determining whether there is an actual alarm event in each set of alarm events to be diagnosed that is consistent with the expected alarm event corresponding to each of the target cause events; if the actual occurrence time of the actual alarm event is within the corresponding expected occurrence time window, then the actual alarm event is determined as a normal alarm; if the actual occurrence time of the actual alarm event is not within the corresponding expected occurrence time window, then the actual alarm event is determined as an abnormal alarm; if there is no corresponding actual alarm event, then the corresponding target expected alarm event is determined as a missed alarm; and summarizing the normal alarms, the abnormal alarms, and the missed alarms to obtain the fault diagnosis result of the power system.

[0066] In some embodiments, determining whether there exists an actual alarm event in each set of alarm events to be diagnosed that is consistent with the expected alarm event corresponding to each target cause event specifically involves: after obtaining the target solution vector through a tabu search algorithm, determining the final set of target cause events based on the decision variables in the target solution vector that take preset values, and querying the cause event to expected alarm mapping table based on the target cause event set to obtain the expected alarm event corresponding to each target cause event and the expected occurrence time window corresponding to each expected alarm event; wherein, the cause event to expected alarm mapping table is obtained during the system initialization phase based on forward temporal reasoning using a temporal constraint network, and is used to record the alarm events that may be triggered by each cause event and the corresponding time constraint relationships. For any target cause event, an expected alarm set is obtained according to its corresponding mapping relationship. The expected alarm set includes multiple expected alarm events, and each expected alarm event is associated with an expected occurrence time window. For any target cause event... Obtain the corresponding expected alarm set: ,in, Indicates the cause of the target event. The corresponding expected alarm set, Indicates the first An expected alarm event, Indicates expected alarm event The corresponding expected occurrence time window. Merge the expected alarm sets corresponding to all target cause events in the target cause event set to obtain the total expected alarm set under the current fault scenario. Iterate through each actual alarm event in the set of alarm events to be diagnosed, matching the alarm type information of each actual alarm event with each expected alarm event in the total expected alarm set; where each actual alarm event includes at least an alarm identifier and an actual occurrence time; each expected alarm event includes at least an alarm identifier and an expected occurrence time window. When the alarm identifier of an actual alarm event is the same as the alarm identifier of an expected alarm event, a correspondence is established between the actual alarm event and the expected alarm event, and the status of the actual alarm event is further determined based on the positional relationship between the actual occurrence time and the expected occurrence time window. When the alarm identifier of an actual alarm event cannot match any expected alarm event corresponding to a target cause event, it indicates that the actual alarm event cannot be explained by the currently determined target cause event, and the actual alarm event is treated as an abnormal alarm event to be further analyzed.

[0067] In some embodiments, when the actual occurrence time of the actual alarm event falls within the corresponding expected occurrence time window, the actual alarm event is determined to be a normal alarm. Specifically, for an actual alarm event that has already established a correspondence with a target expected alarm event, the actual occurrence time of the actual alarm event is obtained, and the expected occurrence time window corresponding to its target expected alarm event is obtained. The expected occurrence time window is obtained by forward reasoning along the timing constraint network from the target cause event, and is used to describe the theoretical time range within which the corresponding alarm event should occur after the occurrence of the target cause event. It is then determined whether the actual occurrence time falls within the expected occurrence time window. If the actual occurrence time falls within the expected occurrence time window, it indicates that the occurrence time of the actual alarm event satisfies the timing constraint relationship corresponding to the target cause event, that is, the actual alarm event conforms to the power system protection action logic and alarm propagation timing. Therefore, the actual alarm event is determined to be a normal alarm, and a correlation is established between the normal alarm and the corresponding target cause event to support the fault diagnosis results of the target cause event. For example, when a line fault is the target cause event, the expected occurrence time window of the corresponding circuit breaker trip alarm is obtained by time-constrained network reasoning and is within the preset time range after the fault occurs. If the actual received circuit breaker trip alarm occurs within this time range, the circuit breaker trip alarm is considered to conform to the protection action logic and is a normal alarm.

[0068] In some embodiments, when the actual occurrence time of the actual alarm event is not within the corresponding expected occurrence time window, the actual alarm event is determined to be an abnormal alarm. Specifically, for an actual alarm event that has been matched with a corresponding expected alarm event, when it is detected that the actual occurrence time is not within the corresponding expected occurrence time window, the time offset of the actual alarm event is further determined. The actual occurrence time is compared with the corresponding expected occurrence time window: if the actual occurrence time is earlier than the start time of the expected occurrence time window or later than the end time of the expected occurrence time window, it indicates that although the actual alarm event has a type correspondence with the target cause event, its occurrence time does not meet the time distance constraint in the timing constraint network. Since the alarm time in the power system may be affected by factors such as GPS timing error, communication transmission delay, sampling error of the acquisition device, and differences in equipment action response, this type of alarm cannot be directly determined as a false alarm, but is determined to be an abnormal alarm. Abnormal alarms are classified according to the degree of deviation between the actual occurrence time and the expected occurrence time window. If the actual occurrence time deviates only slightly from the expected occurrence time window, the abnormal alarm is considered to be caused by time synchronization error or communication delay; if the actual occurrence time deviates significantly from the expected occurrence time window, the abnormal alarm is considered to indicate abnormal protection action sequence, abnormal device response, or abnormal alarm configuration logic.

[0069] In some embodiments, when no corresponding actual alarm event exists, the corresponding target expected alarm event is identified as a missed alarm. Specifically, for each expected alarm event corresponding to the target cause event, all actual alarm events in the set of alarm events to be diagnosed are traversed. If no actual alarm event has the same alarm identifier as the expected alarm event, it indicates that the alarm that should theoretically be triggered by the target cause event has not appeared in the actual alarm stream. Further, it is determined whether the expected occurrence time window corresponding to the target expected alarm event has ended. If the current time has exceeded the end time of the expected occurrence time window, and no corresponding actual alarm event has been detected, the target expected alarm event is identified as a missed alarm. Missed alarms may be caused by communication link interruption, alarm acquisition device failure, incorrect operation of the protection device, or abnormal alarm upload.

[0070] In some embodiments, the normal alarms, abnormal alarms, and missed alarms are summarized to obtain the fault diagnosis results of the power system. Specifically, after completing the matching analysis between all actual alarm events and expected alarm events, the alarm states are classified and summarized. All actual alarm events that meet the time constraints are assigned to the normal alarm set, which represents valid alarm information that conforms to the protection logic under the current fault scenario. Actual alarm events whose actual occurrence time does not meet the corresponding expected occurrence time window constraints are assigned to the abnormal alarm set, and the alarm identifier, actual occurrence time, corresponding target cause event, and time deviation of the abnormal alarm are recorded. Target expected alarm events that do not appear in the set of alarm events to be diagnosed are assigned to the missed alarm set, and the target cause event and theoretical expected occurrence time window of the missed alarm are recorded. The power system fault diagnosis results are generated based on the target cause event set, normal alarm set, abnormal alarm set, and missing alarm set. The fault diagnosis results include at least: (1) a target cause event list, which represents the most likely fault cause determined by the time-series constraint network and optimization solution, and records the possible time intervals corresponding to each target cause event; (2) an abnormal alarm list, which records alarm events that actually occur but do not meet the time-series constraint relationship, and identifies their abnormal type; (3) a missing alarm list, which records alarm events that should theoretically occur but were not actually received; and (4) inference association information, which shows the correspondence between each actual alarm event and the target cause event, as well as the supporting alarm information corresponding to each target cause event.

[0071] Through the above steps, based on the matching relationship between the expected alarm and the actual alarm corresponding to the target cause event, it is possible to accurately identify normal alarms, abnormal alarms, and missed alarms. It fully considers the impact of factors such as time synchronization error, communication delay, and equipment response differences on alarm timing, thereby improving the accuracy and reliability of alarm analysis. At the same time, by classifying and summarizing multiple types of alarm information, a comprehensive diagnostic result containing fault causes, abnormal alarms, and reasoning basis is formed, providing a reliable basis for power dispatchers to quickly locate fault causes and judge the effectiveness of alarms.

[0072] It is understood that the above-described device embodiments correspond to the method embodiments of the present invention, and can implement the fault diagnosis method for a power system provided by any of the above-described method embodiments of the present invention. More detailed workflows and principles of this system can be found, but are not limited to, the relevant descriptions of the above methods.

[0073] It should be noted that the device embodiments described above are merely illustrative, and some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can specifically be implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0074] Based on the above-described embodiment of a power system fault diagnosis method, another embodiment of the present invention provides a terminal device, which includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements a power system fault diagnosis method according to any embodiment of the present invention.

[0075] For example, in this embodiment, the computer program can be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the terminal device.

[0076] The terminal device may be a desktop computer, laptop, handheld computer, or cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0077] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the terminal device, connecting all parts of the terminal device via various interfaces and lines.

[0078] Based on the above-described method embodiments, another embodiment of the present invention provides a computer-readable storage medium including a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to execute a power system fault diagnosis method according to any of the above-described method embodiments of the present invention.

[0079] The modules / units integrated in the device / terminal equipment, if implemented as software functional units and sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0080] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A fault diagnosis method for a power system, characterized in that, include: Receive alarm events generated during the operation of the power system; The alarm events are grouped according to a preset sliding time window to obtain several sets of alarm events to be diagnosed; The cause event corresponding to each set of alarm events to be diagnosed and the event time period corresponding to each cause event are obtained through a preset temporal constraint network. Based on the time periods of each event, the causal events are merged to obtain several candidate causal events; The expected alarm event corresponding to each candidate cause event is obtained through a preset time-series constraint network, and several target cause events are obtained based on the expected alarm event, the set of alarm events to be diagnosed, and the preset target cause solution function. The expected alarm events corresponding to each of the target cause events are compared with each of the sets of alarm events to be diagnosed, and the fault diagnosis results of the power system are output based on the comparison results.

2. The fault diagnosis method for a power system as described in claim 1, characterized in that, The construction process of the temporal constraint network includes: Obtain the protection configuration diagram of the power system, the circuit breaker operating logic, and the relay protection principle; Based on the protection configuration diagram, the circuit breaker operation logic, and the relay protection principle, the temporal logic relationship between each cause event and each alarm event, as well as between each alarm event, is obtained, so as to generate alarm configuration rules based on each of the temporal logic relationships. The set of cause event nodes and the set of alarm event nodes are determined based on the alarm configuration rules; The first directed edge from each cause event node to each alarm event node is determined based on the temporal logical relationship between each cause event and each alarm event; The second directed edge between each alarm event node is determined based on the temporal logical relationship between each alarm event; The temporal constraint network is constructed based on the set of cause event nodes, the set of alarm event nodes, the first directed edge, the second directed edge, and the preset time distance constraint.

3. The fault diagnosis method for a power system as described in claim 2, characterized in that, The step of obtaining the causal event corresponding to each set of alarm events to be diagnosed and the event time period corresponding to each causal event through a preset temporal constraint network includes: In the time-constraint network, determine the alarm event nodes corresponding to each set of alarm events to be diagnosed; Based on the reverse temporal reasoning along the temporal constraint network for each alarm event node, the causal event corresponding to each alarm event node and the event time period corresponding to each causal event are obtained.

4. The fault diagnosis method for a power system as described in claim 2, characterized in that, The step of obtaining the expected alarm event corresponding to each candidate cause event through a preset temporal constraint network includes: In the temporal constraint network, determine the causal event node corresponding to each of the candidate causal events; Based on the forward temporal reasoning along the temporal constraint network for each of the aforementioned causal event nodes, the expected alarm events corresponding to each of the aforementioned causal event nodes and the expected occurrence time windows corresponding to each of the aforementioned expected alarm events are obtained.

5. A fault diagnosis method for a power system as described in claim 1, characterized in that, The process involves obtaining several target cause events based on the expected alarm events, the set of alarm events to be diagnosed, and a preset target cause solution function, including: Set corresponding decision variables for each of the candidate cause events, obtain the target candidate cause events with preset values ​​in each of the decision variables, and count the first number of each of the target candidate cause events; The expected alarm events corresponding to each of the target candidate cause events are matched one by one with each set of alarm events to be diagnosed, and a second number of expected alarm events that do not exist in each set of alarm events to be diagnosed, and a third number of expected alarm events that do not exist in each set of alarm events to be diagnosed; The first quantity, the second quantity, the third quantity, and the preset weight coefficients are weighted and summed to obtain the target cause solution function; Solving the target cause solution function yields several target cause events.

6. A fault diagnosis method for a power system as described in claim 5, characterized in that, Solving the target cause solution function yields several target cause events, including: An initial solution vector is constructed based on the decision variables corresponding to each of the candidate causal events. The value of each decision variable in the initial solution vector is changed sequentially to obtain several neighborhood solution vectors. Substitute each of the neighborhood solution vectors into the target cause solution function for calculation to obtain the target function value corresponding to each of the neighborhood solution vectors; Based on the objective function values ​​and the preset tabu search rules, the current solution vector is selected from the neighborhood solution vectors. The current solution vector is used as the initial solution vector for the next iteration to continue iterative solution until the preset iteration conditions are met, and then the target solution vector is output. The candidate causal events corresponding to the decision variables that take preset values ​​in the target solution vector are determined as the target causal events.

7. A fault diagnosis method for a power system as described in claim 4, characterized in that, The step of comparing the expected alarm events corresponding to each of the target cause events with each of the sets of alarm events to be diagnosed, and outputting the fault diagnosis results of the power system based on the comparison results, includes: Determine whether there exists an actual alarm event in each set of alarm events to be diagnosed that is consistent with the expected alarm event corresponding to each of the target cause events; If the actual occurrence time of the actual alarm event falls within the corresponding expected occurrence time window, the actual alarm event is determined to be a normal alarm. If the actual occurrence time of the actual alarm event is not within the corresponding expected occurrence time window, the actual alarm event will be identified as an abnormal alarm. When there is no corresponding actual alarm event, the corresponding target expected alarm event is identified as a missed alarm; The normal alarms, abnormal alarms, and missed alarms are summarized to obtain the fault diagnosis results of the power system.

8. A fault diagnosis system for a power system, characterized in that, The fault diagnosis system is used to perform a fault diagnosis method for a power system as described in any one of claims 1-7.

9. A terminal device, characterized in that, The system includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, it implements a fault diagnosis method for a power system as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, include: A stored computer program, wherein, when the computer program is executed, it controls the device containing the computer-readable storage medium to perform a fault diagnosis method for a power system as described in any one of claims 1-7.

Images