Digraph-based power grid abnormal event identification method and device, terminal equipment and storage medium

Abstract

The invention discloses a digraph-based power grid abnormal event identification method and device, terminal equipment and a storage medium, and the method comprises the steps: obtaining current power data and historical fault data in a to-be-detected region, and extracting entities, relationships and attributes from the current power data and historical fault data to construct an initial digraph; acquiring historical power data, and adjusting the weight of a directed edge in the initial directed graph to obtain an optimized directed graph and conditional probability distribution of each node; performing Monte Carlo random sampling on the optimized directed graph to generate a plurality of abnormal event sequences; and determining the actual node state of each node according to the current operation parameter, matching the node state feature of the actual node state with the event state feature of the abnormal event for each abnormal event sequence, determining a target abnormal event sequence, and further identifying the abnormal event occurring in the power grid. According to the invention, the abnormal event of the power grid can be identified.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid anomaly recognition, and in particular, to a method, device, terminal device and storage medium for recognizing power grid anomaly events based on a directed graph. Background Art

[0002] With the continuous access of new energy, the architecture and stability characteristics of the new power system are becoming more and more complex. Timely detecting potential anomalies in the power grid, discovering anomaly signs at the initial stage of a fault, and taking measures in advance to avoid the further expansion of the fault are crucial for ensuring the safe and stable operation of the power grid.

[0003] Accurately identifying power grid anomaly events can help dispatchers comprehensively understand the operation status of the power grid, timely adjust the operation mode of the power grid, ensure that the power grid can maintain stable operation under various working conditions, and improve the power supply quality and reliability of the power grid. If an abnormal fault event occurs, quickly and accurately identifying the type and location of the fault can significantly shorten the fault troubleshooting time and improve the fault repair efficiency. Therefore, how to identify power grid anomaly events is an urgent problem to be solved. Summary of the Invention

[0004] The present invention provides a method, device, terminal device and storage medium for recognizing power grid anomaly events based on a directed graph, which can recognize power grid anomaly events.

[0005] An embodiment of the present invention provides a method for recognizing power grid anomaly events based on a directed graph, including:

[0006] Obtaining current power data and historical fault data within a region to be measured, and constructing an initial directed graph for representing the operation states of various devices in the power grid according to the current power data and the historical fault data; wherein, the above-mentioned power data includes: power grid devices, operation parameters, and topological structures;

[0007] Obtaining historical power data within the region to be measured, adjusting the weights of the directed edges in the initial directed graph according to the historical power data and the above-mentioned historical fault data, obtaining an optimized directed graph, and the conditional probability distributions of each node in the optimized directed graph; wherein, the above-mentioned conditional probability distribution represents the conditional probability that when the parent node is in different node states, its child node is in different node states under different node states of the parent node;

[0008] Performing multiple Monte Carlo random samplings on the above-mentioned optimized directed graph, and for each Monte Carlo random sampling, generating an anomaly event sequence according to the conditional probability distributions of each node, and then obtaining a plurality of anomaly event sequences;

[0009] Determine the actual node states of each node in the optimized directed graph according to the operating parameters in the current power data, and for each abnormal event sequence, match the node state characteristics of the above actual node states with the event state characteristics of the abnormal events in the above abnormal event sequence to determine the target abnormal event sequence;

[0010] Identify the abnormal events occurring in the power grid according to the above target abnormal event sequence.

[0011] Furthermore, construct an initial directed graph for representing the operating states of various devices in the power grid according to the current power data and historical fault data, including:

[0012] Extract the entities, relationships, and attributes of the power data from the current power data and historical fault data according to knowledge extraction techniques;

[0013] Construct the above initial directed graph with the above entities as nodes, the above relationships as directed edges, and the above attributes as the node characteristics of the above nodes.

[0014] Furthermore, adjust the weights of the directed edges in the initial directed graph according to the historical power data and the above historical fault data to obtain an optimized directed graph, and the conditional probability distribution of each node in the optimized directed graph, including:

[0015] Input the initial directed graph, the above historical power data, and the above historical fault data into a preset graph neural network, so that the preset graph neural network performs convolution operations on each node according to the above historical power data and the above historical fault data to obtain the weights of each directed edge, and then obtain the optimized directed graph;

[0016] Statistically obtain the above conditional probability distribution of each node in the above optimized directed graph according to the above historical power data, the above historical fault data, and the Bayesian network.

[0017] Furthermore, every time a Monte Carlo random sampling is performed, generate an abnormal event sequence according to the conditional probability distribution of each node, including:

[0018] Perform multiple random samplings from the above optimized directed graph by the Monte Carlo random sampling method until the current target node obtained by the current random sampling has no child nodes, and generate the above abnormal event sequence according to the node states of all the target nodes obtained currently;

[0019] Wherein, every time a random sampling is performed, determine the node state of the current target node according to the node state of the target node obtained by the random sampling at the previous moment and the corresponding conditional probability distribution.

[0020] Further, matching the node state features of the actual node states with the event state features of the abnormal events in the abnormal event sequence described above to determine the target abnormal event sequence includes:

[0021] Calculating the feature similarity between the node state features and the event state features in the selected abnormal event sequence to obtain the feature similarity corresponding to each abnormal event sequence;

[0022] Taking the abnormal event sequence with the maximum feature similarity as the target abnormal event sequence.

[0023] Further, after identifying the abnormal events that occur in the power grid, it further includes:

[0024] Real-time monitoring of the actual working condition information of the area to be measured; wherein, the actual working condition information includes: weather conditions and construction conditions;

[0025] Correcting the recognition result according to the actual working condition information.

[0026] Based on the above method item embodiments, the present invention correspondingly provides device item embodiments;

[0027] The present invention provides a power grid abnormal event recognition device based on a directed graph, including:

[0028] A data acquisition module, a directed graph optimization module, an abnormal event sequence generation module, a target abnormal event sequence determination module, and a power grid abnormal recognition module;

[0029] The above data acquisition module is used to acquire the current power data and historical fault data in the area to be measured, and construct an initial directed graph representing the operating states of each device in the power grid according to the current power data and historical fault data; wherein, the power data includes: power grid equipment, operating parameters, and topological structure;

[0030] The above directed graph optimization module is used to acquire the historical power data in the area to be measured, adjust the weights of the directed edges in the initial directed graph according to the historical power data and the historical fault data, obtain the optimized directed graph, and the conditional probability distribution of each node in the optimized directed graph; wherein, the conditional probability distribution represents the conditional probability that when the parent node is in different node states, its child node is in different node states under different node states of the parent node;

[0031] The above abnormal event sequence generation module is used to perform multiple Monte Carlo random samplings on the optimized directed graph. Each time a Monte Carlo random sampling is performed, an abnormal event sequence is generated according to the conditional probability distribution of each node, and then several abnormal event sequences are obtained;

[0032] The above-mentioned target abnormal event sequence determination module is used to determine the actual node states of each node in the optimized directed graph according to the operating parameters in the current power data, and for each abnormal event sequence, match the node state characteristics of the above-mentioned actual node states with the event state characteristics of the abnormal events in the above-mentioned abnormal event sequence to determine the target abnormal event sequence;

[0033] The above-mentioned power grid abnormality recognition module is used to recognize the abnormal events occurring in the power grid according to the above-mentioned target abnormal event sequence.

[0034] Further, the above-mentioned data acquisition module includes:

[0035] A data extraction unit and an initial directed graph construction unit;

[0036] The above-mentioned data extraction unit is used to extract the entities, relationships, and attributes of the power data from the current power data and historical fault data according to the knowledge extraction technology;

[0037] The above-mentioned initial directed graph construction unit is used to construct the above-mentioned initial directed graph with the above-mentioned entities as nodes, the above-mentioned relationships as directed edges, and the above-mentioned attributes as the node characteristics of the above-mentioned nodes.

[0038] Based on the above method item embodiment, the present invention correspondingly provides a terminal device item embodiment;

[0039] The present invention provides a terminal device, including a processor, a memory, and a computer program stored in the above-mentioned memory and configured to be executed by the above-mentioned processor. When the above-mentioned processor executes the above-mentioned computer program, it implements the above-mentioned method for identifying abnormal events in a power grid based on a directed graph according to any embodiment of the present invention.

[0040] Based on the above method item embodiment, the present invention correspondingly provides a storage medium item embodiment;

[0041] The present invention provides a storage medium, including a processor, a memory, and a computer program stored in the above-mentioned memory and configured to be executed by the above-mentioned processor. When the above-mentioned processor executes the above-mentioned computer program, it implements the above-mentioned method for identifying abnormal events in a power grid based on a directed graph according to any embodiment of the present invention.

[0042] The embodiments of the present invention have the following beneficial effects:

[0043] The present invention provides a method, apparatus, terminal device and storage medium for identifying power grid abnormal events based on a directed graph. The method includes: First, obtain the current power data and historical fault data in the area to be measured, and construct an initial directed graph representing the operating states of various devices in the power grid according to the current power data and historical fault data; wherein, the above power data includes: power grid devices, operating parameters and topological structure; Subsequently, obtain the historical power data in the area to be measured, and adjust the weights of the directed edges in the initial directed graph according to the historical power data and the above historical fault data to obtain an optimized directed graph and the conditional probability distribution of each node in the optimized directed graph; wherein, the above conditional probability distribution represents the conditional probability that when the parent node is in different node states, its child node is in different node states under different node states of the parent node; Then, perform multiple Monte Carlo random samplings on the above optimized directed graph. Each time a Monte Carlo random sampling is performed, generate an abnormal event sequence according to the conditional probability distribution of each node, and then obtain several abnormal event sequences; Subsequently, determine the actual node states of each node in the optimized directed graph according to the operating parameters in the current power data, and for each abnormal event sequence, match the node state characteristics of the above actual node states with the event state characteristics of the abnormal events in the above abnormal event sequence to determine the target abnormal event sequence; Finally, identify the abnormal events occurring in the power grid according to the above target abnormal event sequence. Therefore, the present invention performs random sampling on each node in the optimized directed graph by the Monte Carlo random sampling method to generate several abnormal event sequences with different abnormal events. Subsequently, based on the current operating parameters, determine the actual node states of each node in the directed graph, and match the actual node state characteristics of the actual node states with the abnormal event characteristics of each abnormal event in the abnormal event sequence to determine the target abnormal event sequence. Then, the abnormal events occurring in the current power grid can be identified according to the abnormal events in the target abnormal event sequence, realizing the identification of abnormal events. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 FIG. is a schematic flowchart of a method for identifying power grid abnormal events based on a directed graph provided by an embodiment of the present invention.

[0045] Figure 2 FIG. is a schematic structural diagram of an initial directed graph provided by an embodiment of the present invention.

[0046] Figure 3 FIG. is a schematic structural diagram of an apparatus for identifying power grid abnormal events based on a directed graph provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0047] The following will clearly and completely describe the technical solutions in the present invention in conjunction with the accompanying drawings in the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present invention without creative efforts belong to the scope of protection of the present invention.

[0048] As Figure 1 shown, a method for identifying abnormal events in a power grid based on a directed graph provided by an embodiment of the present invention includes:

[0049] Step S101: Obtain the current power data and historical fault data in the area to be measured, and construct an initial directed graph for representing the operating states of various devices in the power grid according to the current power data and historical fault data; wherein, the above-mentioned power data includes: power grid devices, operating parameters, and topological structure;

[0050] Specifically, obtain detailed information on various power grid devices from the equipment management system (EMS) of the power grid, including the model, capacity, production date, and manufacturer of transformers in substations, the length, material, and rated voltage of transmission lines, etc.

[0051] Specifically, collect the operating parameters updated every second from the real-time monitoring system (SCADA), such as the incoming and outgoing line voltages, currents, active powers, and reactive powers of each substation, as well as the load rates of transmission lines.

[0052] Specifically, use the geographic information system (GIS) and the power grid topology diagram to obtain the topological structure of the power grid, and clarify the connection relationships between various substations and transmission lines, such as "Substation A is connected to Substation B through transmission line L1".

[0053] Specifically, collect historical fault data for the past ten years, including the time, location, faulty equipment, fault type (such as short circuit, open circuit, overload, etc.) of the fault occurrence, and the description of the operating conditions before the fault occurrence.

[0054] In a preferred embodiment, the above-mentioned construction of the initial directed graph for representing the operating states of various devices in the power grid according to the current power data and historical fault data includes:

[0055] Extract the entities, relationships, and attributes of the power data from the current power data and historical fault data according to knowledge extraction technology;

[0056] Construct the above-mentioned initial directed graph with the above-mentioned entities as nodes, the above-mentioned relationships as directed edges, and the above-mentioned attributes as the node features of the above-mentioned nodes.

[0057] Preferably, before constructing an initial directed graph for representing the operating states of various devices in the power grid based on the current power data, preprocess the power data and historical fault data to remove invalid information.

[0058] Specifically, the above-mentioned knowledge extraction techniques include: natural language processing techniques, rule-based extraction methods, and long short-term memory network (abbreviated as LSTM) extraction methods.

[0059] Preferably, natural language processing techniques mainly perform information extraction on text data for part-of-speech tagging, named entity recognition, and syntactic analysis. Part-of-speech tagging is to quickly locate the words describing key information such as device attributes and operating states; named entity recognition can accurately find power grid entities such as transformers and circuit breakers and their related entities such as models and fault types; syntactic analysis helps to understand the sentence structure, thereby accurately extracting the relationships between entities.

[0060] Preferably, the rule-based extraction method is to directly and accurately extract entities, relationships, and attributes based on pre-set rules for some structured data with a fixed format or following specific industry specifications.

[0061] Preferably, the LSTM extraction method is to train a model using the labeled sample data. Through continuous training of the LSTM algorithm model, it is enabled to have the ability to automatically identify and extract knowledge, and extract entities, relationships, and attributes from similar data. As the training data increases, the extraction accuracy and generalization ability are continuously improved.

[0062] Specifically, for the above-mentioned power grid device type data, natural language processing techniques and rule-based extraction methods are used to extract entity information such as device names to extract attribute information such as device models, specifications, and manufacturers; for the above-mentioned operating parameter type data, natural language processing techniques are used to label the numerical data of operating parameters such as voltage, current, and power and associate them with the corresponding device entities; for the above-mentioned topological structure type data, rule-based extraction methods are used to extract entities such as each node (site, device), and the connection relationships between nodes to construct a relationship network; for the above-mentioned historical fault data, first use natural language processing techniques to process and label the text data, and then use the LSTM algorithm model to extract the entities, relationships, and attributes therein, such as fault components, fault points, and fault types, through multiple trainings, and associate them with the entities, relationships, and attributes extracted from the power grid devices, operating parameters, and topological structures.

[0063] Preferably, after extracting entities, relationships, and attributes, a power knowledge graph can be constructed based on the above-mentioned entities, relationships, and attributes and stored in the form of a graph database. For example, from the fault record "On July 10, 2019, the transformer T2 of substation C had a winding short-circuit fault due to long-term overload, resulting in the tripping of the transmission line L3 connected to it and a power outage in the surrounding area", entities such as "transformer T2", "winding short-circuit fault", and "transmission line L3" are extracted, as well as the causal relationship of "transformer T2 causing the tripping of transmission line L3", so as to construct a comprehensive power knowledge graph of the county power grid and store it in a graph database for convenient subsequent query and use.

[0064] Specifically, the above-mentioned initial directed graph is represented by a binary tuple as:

[0065] G=(V,E)

[0066] In the formula, V represents the set of all nodes in the directed graph, that is, all entities, and E represents the set of all edges in the directed graph, that is, the relationships between all entities, such as site connection, equipment connection, event causality, etc. Each directed edge is an ordered pair (u,v), representing a directed edge pointing from node u to node v.

[0067] Schematically, the constructed initial directed graph is as Figure 2 shown, Figure 2 the l in i is a node, and l (i,j) is the directed edge pointing from l i to l j , and x i is the node attribute. For example, l4 represents the l4 node, and l (1,4) represents the directed edge from the l1 node to the l4 node, and x4 represents the node attribute of the l4 node.

[0068] In this preferred embodiment, entities, relationships, and attributes are extracted from the current power data and historical abnormal events, and then an initial directed graph for representing the operating states of various devices in the power grid is constructed based on the entities, relationships, and attributes.

[0069] Step S102: Obtain historical power data within the area to be measured, and adjust the weights of the directed edges in the initial directed graph according to the historical power data and the above-mentioned historical fault data to obtain an optimized directed graph and the conditional probability distribution of each node in the optimized directed graph; wherein, the above-mentioned conditional probability distribution represents the conditional probability that when the parent node is in different node states, its child node is in different node states under different node states of the parent node.

[0070] In a preferred embodiment, the weights of the directed edges in the initial directed graph are adjusted according to the historical power data and the historical fault data to obtain an optimized directed graph, and the conditional probability distribution of each node in the optimized directed graph includes:

[0071] Input the initial directed graph, the historical power data, and the historical fault data into a preset graph neural network, so that the preset graph neural network performs a convolution operation on each node according to the historical power data and the historical fault data to obtain the weights of each directed edge, and then obtain an optimized directed graph;

[0072] Specifically, the preset graph neural network performs a convolution operation on the feature information (such as equipment operation parameters, historical fault data, etc.) of the grid equipment nodes and their neighbor equipment nodes (determine neighbors according to the topological connection relationship) in the historical power data and the historical abnormal events, aggregates the neighbor node information to update its own node representation, thereby learning the hidden association patterns between nodes and determining the weights of the directed edges. For example, after multiple convolutions, it is found that when the loads of multiple substations in a certain power supply area increase simultaneously, the probability of overload of the main transmission line connected to them increases significantly, and accordingly, the weights of the directed edges between related nodes are adjusted.

[0073] According to the historical power data, the historical fault data, and the Bayesian network, the conditional probability distribution of each node in the optimized directed graph is statistically obtained.

[0074] Specifically, the Bayesian network learns the conditional probability distribution of each node through the historical power data and the historical fault data. If a certain node fails, given different state combinations of its parent nodes, the frequency of the occurrence of this failure under all state combinations is statistically determined to obtain the corresponding conditional probability, and then as the data continues to accumulate, the Bayesian update rule is used to update the conditional probability between nodes.

[0075] Schematically, the Bayesian network uses the historical power data and the historical fault data to establish a conditional probability table for each node. Taking the "transmission line fault" node as an example, given different value combinations of its parent nodes (such as "adjacent substation fault", "severe weather impact", etc.), the frequency of the occurrence of transmission line faults in the historical power data and the historical fault data under these combinations is statistically obtained and filled into the table as the conditional probability. Through the Bayesian update rule, as new data is introduced, these conditional probabilities are adjusted in real time to ensure that the model closely fits the actual operation of the power grid.

[0076] Specifically, the abnormal conditional probability P in the above conditional probability is represented by the following formula:

[0077]

[0078] wherein, (x1, x2, …, x n ) represents all nodes, and a(x i ) represents the parent node of node x i .

[0079] In this preferred embodiment, the initial directed graph is optimized through a graph neural network and a Bayesian network.

[0080] Step S103: Perform multiple Monte Carlo random samplings on the optimized directed graph above. Each time a Monte Carlo random sampling is performed, an abnormal event sequence is generated according to the conditional probability distribution of each node, and then several abnormal event sequences are obtained;

[0081] In a preferred embodiment, each time a Monte Carlo random sampling is performed above, an abnormal event sequence is generated according to the conditional probability distribution of each node, including:

[0082] Perform multiple random samplings on the optimized directed graph above through the Monte Carlo random sampling method until the current target node obtained by the current random sampling has no child nodes. According to the node states of all the target nodes obtained currently, the above abnormal event sequence is generated;

[0083] Wherein, each time a random sampling is performed, according to the node state of the target node obtained by the random sampling at the previous moment and the corresponding conditional probability distribution, the node state of the current target node is determined.

[0084] Specifically, the Monte Carlo simulation is represented in the following manner:

[0085] x t+1 = f(x t , u t , ∈ t )

[0086] wherein, x t represents the node state of the target node at the current t moment, x t+1 represents the node state of the target node at the next moment, u t represents the node states of other nodes at the current t moment, ∈ t represents the random perturbation variable at the current t moment, and the function f represents the process of the change of the power grid state, which is determined in combination with the conditional probability distribution of the target node at the t moment.

[0087] Specifically, since there are directed edges between the nodes of a directed graph, when the state of a certain node is determined, the node state of a child node can be inferred based on the conditional probability of the child node of this node in this state and the result of random sampling. Schematically, if there are three nodes A, B, and C in the currently optimized directed graph, and there are directed edges between them (A→B, B→C). Subsequently, according to historical fault data, the probability that device A operates normally is 0.7, the probability of a minor fault is 0.2, and the probability of a serious fault is 0.1; when node A operates normally, the probability that node B operates normally is 0.8, the probability of a minor fault is 0.1, and the probability of a serious fault is 0.1; when node A has a minor fault, the probability that B operates normally is 0.6, the probability of a minor fault is 0.2, and the probability of a serious fault is 0.2; when node A has a serious fault, the probability that B operates normally is 0.4, the probability of a minor fault is 0.3, and the probability of a serious fault is 0.3; the probability distribution of node C is determined in the same way according to the state of B.

[0088] Subsequently, during the first random sampling, it is assumed that node A is in a minor fault state. Since A→B, according to the conditional probabilities of A and B, it can be inferred that when A is in a minor fault, the probability that B operates normally at this time is 0.6, the probability of a minor fault is 0.2, and the probability of a serious fault is 0.2. Then, another random sampling is carried out, and it is assumed that B is in a minor fault state. Also, because B→C, the operating state of C is determined by random sampling based on the state of B and the conditional probabilities of B and C. It is assumed that C is in a normal operating state. In this way, an abnormal event sequence is obtained: A minor fault → B minor fault → C normal operation.

[0089] Specifically, after multiple Monte Carlo random samplings, a large number of different abnormal event sequences can be generated, including possible abnormal events and their propagation paths.

[0090] Schematically, during Monte Carlo random sampling, during one sampling process, it is possible to draw an abnormal event that due to a certain transmission line operating at a high load for a long time, the aging of its insulation layer accelerates, and partial discharge occurs. Since there is a directed edge from the "partial discharge of transmission line" node to the "short - circuit fault of transmission line" node in the directed graph model, and based on the previously obtained conditional probability, it is determined that there is a 0.2 probability of triggering a short - circuit fault. If a short - circuit fault occurs, then according to the relationship with other nodes connected to this transmission line, such as a sudden drop in voltage of the adjacent substation, overload of other transmission lines due to power flow transfer, etc., the subsequent abnormal event sequence is continued to be generated.

[0091] Subsequently, after several such random samplings, several sequences of abnormal events with different probabilities and credibility are generated. For example: "Transformer overload in substation A in a certain area → Heating of transmission line L1 → Short - circuit fault occurs in L1 → Sudden voltage drop in adjacent substation B → Power outage of users in the surrounding area". The occurrence of each event has a corresponding probability estimate.

[0092] In this preferred embodiment, an abnormal event sequence is generated through Monte Carlo random sampling.

[0093] Step S104: According to the operating parameters in the current power data, determine the actual node states of each node in the optimized directed graph, and for each abnormal event sequence, match the node - state characteristics of the above - mentioned actual node states with the event - state characteristics of the abnormal events in the above - mentioned abnormal event sequence to determine the target abnormal event sequence;

[0094] In a preferred embodiment, the above - mentioned matching of the node - state characteristics of the actual node states with the event - state characteristics of the abnormal events in the abnormal event sequence to determine the target abnormal event sequence includes:

[0095] Calculate the feature similarity between the above - mentioned node - state characteristics and the event - state characteristics in the selected abnormal event sequence to obtain the feature similarity corresponding to each abnormal event sequence;

[0096] Take the abnormal event sequence with the maximum above - mentioned feature similarity as the above - mentioned target abnormal event sequence.

[0097] Illustratively, continuously obtain the current operating parameters from the real - time monitoring system (SCADA), and based on these operating parameters, determine the actual node states of each node in the optimized directed graph, such as a sudden increase in the oil temperature of a certain transformer or a sudden increase in the current of a certain transmission line, and extract these node - state characteristics.

[0098] Subsequently, match the extracted node - state characteristics with the event - state characteristics of the abnormal events in the abnormal event sequence. For example, during the matching process, it is found that the feature of the increase in the oil temperature of the transformer is consistent with the feature description of the "transformer overheat fault" node, and it is preliminarily judged that this abnormal event may exist. Subsequently, by matching with all the generated abnormal event sequences, the characteristic similarity corresponding to each abnormal event sequence can be obtained.

[0099] In this preferred embodiment, the target abnormal event sequence is determined by matching the event - state characteristics.

[0100] Step S106: Identify the abnormal events occurring in the power grid according to the above - mentioned target abnormal event sequence.

[0101] Specifically, based on the abnormal events within the determined target abnormal event sequence, the abnormalities occurring in the current power grid can be identified.

[0102] In a preferred embodiment, after identifying the abnormal events occurring in the power grid, it further includes:

[0103] Real-time monitoring of the actual working condition information of the area to be measured; wherein, the above-mentioned actual working condition information includes: weather conditions and construction conditions;

[0104] Revise the recognition result according to the above-mentioned actual working condition information.

[0105] Specifically, after determining the target abnormal event sequence, the recognition result can be revised by combining the actual working condition information obtained from on-site real-time monitoring, such as the current weather conditions, surrounding construction conditions, and expert experience. If the expert judges based on experience that the current high temperature weather is the main cause of the increase in the oil temperature of the transformer, and there is no substantial risk of equipment failure for the time being, the overly sensitive fault recognition can be revised, and the recognition result can be adjusted to "the transformer is affected by high temperature, the oil temperature is on the high side, it needs to be closely monitored, and no failure has occurred for the time being", ensuring that the recognition result accurately reflects the actual situation of the power grid and provides practical operation suggestions for the operation and maintenance personnel.

[0106] Based on the above method item embodiment, the present invention correspondingly provides a device item embodiment.

[0107] As Figure 3 shown, an embodiment of the present invention provides a power grid abnormal event recognition device based on a directed graph, including:

[0108] A data acquisition module, a directed graph optimization module, an abnormal event sequence generation module, a target abnormal event sequence determination module, and a power grid abnormal recognition module;

[0109] The above-mentioned data acquisition module is used to acquire the current power data and historical fault data within the area to be measured, and construct an initial directed graph for representing the operating states of various devices in the power grid according to the current power data and historical fault data; wherein, the above-mentioned power data includes: power grid equipment, operating parameters, and topological structure;

[0110] The above-mentioned directed graph optimization module is used to acquire the historical power data within the area to be measured, adjust the weights of the directed edges in the initial directed graph according to the historical power data and the above-mentioned historical fault data, obtain an optimized directed graph, and the conditional probability distribution of each node in the optimized directed graph; wherein, the above-mentioned conditional probability distribution represents the conditional probability that when the parent node is in different node states, its child node is in different node states under different node states of the parent node;

[0111] The above abnormal event sequence generation module is used to perform multiple Monte Carlo random samplings on the above optimized directed graph. Each time a Monte Carlo random sampling is performed, an abnormal event sequence is generated according to the conditional probability distribution of each node, and then several abnormal event sequences are obtained;

[0112] The above target abnormal event sequence determination module is used to determine the actual node states of each node in the optimized directed graph according to the operating parameters in the current power data, and for each abnormal event sequence, match the node state features of the above actual node states with the event state features of the abnormal events in the above abnormal event sequence to determine the target abnormal event sequence;

[0113] The above power grid abnormality recognition module is used to recognize the abnormal events occurring in the power grid according to the above target abnormal event sequence.

[0114] In a preferred embodiment, the above data acquisition module includes:

[0115] A data extraction unit and an initial directed graph construction unit;

[0116] The above data extraction unit is used to extract the entities, relationships, and attributes of the power data from the current power data and historical fault data according to the knowledge extraction technology;

[0117] The above initial directed graph construction unit is used to construct the above initial directed graph with the above entities as nodes, the above relationships as directed edges, and the above attributes as the node features of the above nodes.

[0118] It should be noted that the device embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationships between the modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those of ordinary skill in the art can understand and implement without creative efforts. The above schematic diagram is only an example of a power grid abnormal event recognition device based on a directed graph, and does not constitute a limitation on a power grid abnormal event recognition device based on a directed graph, and may include more or fewer components than shown in the figure, or combine some components, or different components.

[0119] Based on the above method item embodiments, the present invention correspondingly provides terminal device item embodiments.

[0120] Another embodiment of the present invention provides a terminal device, including 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, a method for identifying grid abnormal events based on a directed graph in any one of the embodiments of the present invention is implemented.

[0121] Exemplarily, in this embodiment, the computer program may be divided into one or more modules. The one or more modules are stored in the memory and executed by the processor to complete the present invention. The one or more module elements may be a series of computer program instruction segments capable of completing specific functions, and the instruction segments are used to describe the execution process of the computer program in the device.

[0122] The terminal device may be a computing device such as a desktop computer, a notebook, a palm computer, and a cloud server. The device may include, but is not limited to, a processor and a memory.

[0123] The so-called processor may be a central processing module (Central Processing Unit, CPU), or may also be other general-purpose processors, digital signal processors (Digital Signal Processor, DSP), application specific integrated circuits (Application Specific Integrated Circuit, ASIC), off-the-shelf programmable gate arrays (Field-Programmable Gate Array, FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor, or the processor may also be any conventional processor, etc. The processor is the control center of the device, and connects various parts of the entire device through various interfaces and lines.

[0124] The memory may be used to store the computer program and / or module. The processor realizes various functions of the device by running or executing the computer program and / or module stored in the memory, and calling the data stored in the memory. The memory may mainly include a program storage area and a data storage area. Among them, the program storage area may store an operating system, application programs required for at least one function, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart media card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one magnetic disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0125] Based on the above method embodiments, the present invention correspondingly provides storage medium embodiments.

[0126] Another embodiment of the present invention provides a storage medium. The above storage medium includes a stored computer program. When the computer program runs, it controls the device where the storage medium is located to execute the method for identifying power grid abnormal events based on a directed graph according to any one of the embodiments of the present invention.

[0127] In this embodiment, the above storage medium is a computer-readable storage medium. The above computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file or some intermediate form, etc. The above computer-readable medium can include: any entity or device capable of carrying the above computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disc, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signal, telecommunication signal, and software distribution medium, etc.

[0128] Compared with the prior art, by implementing the above various embodiments of the present invention, power grid abnormal events can be identified.

[0129] The above are the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements are also regarded as the protection scope of the present invention.

Claims

1. A method for identifying abnormal events in a power grid based on a directed graph, characterized in that Including: Obtain the current power data and historical fault data within the area to be measured, and construct an initial directed graph representing the operating states of various devices in the power grid based on the current power data and historical fault data; wherein, the power data includes: power grid devices, operating parameters, and topological structures. Obtain the historical power data within the area to be measured, and adjust the weights of the directed edges in the initial directed graph according to the historical power data and the historical fault data to obtain an optimized directed graph and the conditional probability distributions of each node in the optimized directed graph; wherein, the conditional probability distribution represents the conditional probability that when the parent node is in different node states, its child node is in different node states under the different node states of the parent node. Perform multiple Monte Carlo random samplings on the optimized directed graph. Each time a Monte Carlo random sampling is performed, generate an abnormal event sequence according to the conditional probability distributions of each node, and then obtain several abnormal event sequences. Determine the actual node states of each node in the optimized directed graph according to the operating parameters in the current power data, and for each abnormal event sequence, match the node state characteristics of the actual node states with the event state characteristics of the abnormal events in the abnormal event sequence to determine the target abnormal event sequence. Identify the abnormal events occurring in the power grid according to the target abnormal event sequence.

2. The method for identifying abnormal events in a power grid based on a directed graph according to claim 1, wherein The constructing an initial directed graph representing the operating states of various devices in the power grid according to the current power data and historical fault data includes: Extract the entities, relationships, and attributes of the power data from the current power data and historical fault data according to knowledge extraction techniques. Construct the initial directed graph with the entities as nodes, the relationships as directed edges, and the attributes as the node characteristics of the nodes.

3. The method for identifying abnormal events in a power grid based on a directed graph according to claim 2, wherein The adjusting the weights of the directed edges in the initial directed graph according to the historical power data and the historical fault data to obtain an optimized directed graph and the conditional probability distributions of each node in the optimized directed graph includes: Input the initial directed graph, the historical power data, and the historical fault data into a preset graph neural network, so that the preset graph neural network performs convolution operations on each node according to the historical power data and the historical fault data to obtain the weights of each directed edge, and then obtain the optimized directed graph. Statistically obtain the conditional probability distributions of each node in the optimized directed graph according to the historical power data, the historical fault data, and the Bayesian network.

4. The method for identifying abnormal events in a power grid based on a directed graph according to claim 3, characterized in that Each time a Monte Carlo random sampling is performed, generating an abnormal event sequence according to the conditional probability distributions of each node includes: Perform multiple random samplings from the optimized directed graph by the Monte Carlo random sampling method until the current target node obtained by the current random sampling has no child nodes, and generate the abnormal event sequence according to the node states of all the target nodes obtained currently. Among them, each time a random sampling is performed, the node state of the target node obtained by the random sampling at the previous moment and the corresponding conditional probability distribution are used to determine the node state of the current target node.

5. The method for identifying abnormal events in a power grid based on a directed graph according to claim 4, characterized in that, The matching of the node state features of the actual node state with the event state features of the abnormal events in the abnormal event sequence to determine the target abnormal event sequence includes: Calculating the feature similarity between the node state features and the event state features in the selected abnormal event sequence to obtain the feature similarity corresponding to each abnormal event sequence; Taking the abnormal event sequence with the largest feature similarity as the target abnormal event sequence.

6. The method for identifying abnormal events in a power grid based on a directed graph according to claim 5, characterized in that, After identifying the abnormal events occurring in the power grid, it further includes: Real-time monitoring of the actual working condition information of the area to be measured; among them, the actual working condition information includes: weather conditions and construction conditions; Correcting the recognition result according to the actual working condition information.

7. A power grid abnormal event recognition device based on a directed graph, characterized in that, It includes: A data acquisition module, a directed graph optimization module, an abnormal event sequence generation module, a target abnormal event sequence determination module, and a power grid abnormal recognition module; The data acquisition module is used to acquire the current power data and historical fault data in the area to be measured, and construct an initial directed graph representing the operating states of various devices in the power grid according to the current power data and historical fault data; among them, the power data includes: power grid devices, operating parameters, and topological structures; The directed graph optimization module is used to acquire the historical power data in the area to be measured, adjust the weights of the directed edges in the initial directed graph according to the historical power data and the historical fault data, obtain the optimized directed graph, and the conditional probability distribution of each node in the optimized directed graph; among them, the conditional probability distribution represents the conditional probability that when the parent node is in different node states, its child node is in different node states under different node states of the parent node; The abnormal event sequence generation module is used to perform multiple Monte Carlo random samplings on the optimized directed graph. Each time a Monte Carlo random sampling is performed, an abnormal event sequence is generated according to the conditional probability distribution of each node, and then several abnormal event sequences are obtained; The target abnormal event sequence determination module is used to determine the actual node state of each node in the optimized directed graph according to the operating parameters in the current power data, and for each abnormal event sequence, match the node state features of the actual node state with the event state features of the abnormal events in the abnormal event sequence to determine the target abnormal event sequence; The power grid abnormal recognition module is used to identify the abnormal events occurring in the power grid according to the target abnormal event sequence.

8. The power grid abnormal event recognition device based on a directed graph according to claim 7, characterized in that, The data acquisition module includes: A data extraction unit and an initial directed graph construction unit; The data extraction unit is used to extract the entities, relationships, and attributes of the power data from the current power data and historical fault data according to the knowledge extraction technology; The initial directed graph construction unit is configured to construct the initial directed graph by using the entity as a node, the relationship as a directed edge, and the attribute as the node feature of the node.

9. A terminal device, characterized in that, It 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 method for identifying abnormal grid events based on a directed graph according to any one of claims 1 to 6.

10. A storage medium, characterized in that, The storage medium includes a stored computer program. When the computer program runs, it controls the device where the storage medium is located to execute a method for identifying abnormal grid events based on a directed graph according to any one of claims 1 to 6.

Images