Power distribution network key node identification method and device, terminal and medium

Abstract

The invention discloses a power distribution network key node identification method and device, a terminal and a medium, and relates to the technical field of power distribution networks. According to the scheme provided by the invention, based on node topological data and historical operation data of a power distribution network, different node relation graph data are constructed according to the node topological data; the method comprises the following steps: firstly, obtaining a time sequence aggregation model, then obtaining a spatial-temporal feature vector through a time sequence aggregation model by utilizing spatial features obtained based on node relation graph data and historical operation data representing time sequence features, then selecting any target node, shielding the spatial-temporal feature vectors of the target node and related neighborhood nodes, and carrying out simulation operation through a multi-task prediction model to obtain a multi-task prediction model; according to the method, the loss errors of the target node before and after shielding are respectively obtained, the influence degree of the fault of the target node on the whole power distribution network is measured according to the difference of the two groups of loss errors, then the key fragile node in the power distribution network is identified, and the technical problem of low accuracy of key node identification in the prior art is solved.

Description

Technical Field

[0001] The present application relates to the technical field of distribution networks, and particularly to a method, device, terminal and medium for identifying key nodes in a distribution network. Background Art

[0002] With the expansion of the scale of the power system and the frequent occurrence of extreme weather events, the resilience assessment of the power system and the identification of key nodes have become important issues for ensuring power supply safety. Traditional methods mainly rely on physical model simulation and complex network theory, but these methods are difficult to cope with complex spatio-temporal coupling dynamic scenarios. The mainstream methods in the field of key node identification include: complex network theory, machine learning models, interpretability methods, etc. In recent years, technologies based on graph neural networks (GNNs) and spatio-temporal modeling have been gradually applied to power system analysis, but existing methods have problems such as single dimension, poor multi-task collaboration, and insufficient interpretability in key node identification, resulting in the technical problem of low accuracy in key node identification in practical applications. Summary of the Invention

[0003] The present application provides a method, device, terminal and medium for identifying key nodes in a distribution network, which is used to solve the technical problem of low accuracy in existing key node identification.

[0004] To solve the above technical problem, in the first aspect of the present application, a method for identifying key nodes in a distribution network is provided, including:

[0005] Obtain the node topology data and historical operation data of the distribution network;

[0006] According to the node topology data, construct different node relationship graph data through multiple preset graph data construction conditions;

[0007] Perform fusion processing on each of the node relationship graph data to obtain the spatial features of the distribution network, and then according to the historical operation data and the spatial features, obtain the spatio-temporal feature vector of the distribution network through a preset time series aggregation model;

[0008] According to the spatio-temporal feature vectors corresponding to each node in the distribution network, perform simulation operations in combination with a multi-task prediction model to obtain a multi-task loss error reference value;

[0009] Determine any target node from the distribution network, combine the neighborhood nodes corresponding to the target node, mask the spatio-temporal feature vectors of the target node and the neighborhood nodes, and then perform simulation operations through the multi-task prediction model to obtain a multi-task loss error verification value;

[0010] According to the error difference between the multi-task loss error reference value and the multi-task loss error verification value, determine the key node identification result of the target node.

[0011] Preferably, the node relationship graph data includes: topological adjacency graph data, geographical distance graph data, and node type similarity graph data.

[0012] Preferably, the specific process of fusing each of the node relationship graph data to obtain the spatial characteristics of the distribution network includes:

[0013] Extract the spatial embedding features from each node relationship graph data through a graph neural network, and then fuse the spatial embedding features of each node relationship graph data to obtain the spatial characteristics of the distribution network.

[0014] Preferably, the specific process of performing simulation operations based on the spatio-temporal feature vectors corresponding to each node in the distribution network and combining a multi-task prediction model to obtain a multi-task loss error reference value includes:

[0015] Perform simulation operations based on the spatio-temporal feature vectors corresponding to each node in the distribution network and combine a multi-task prediction model to obtain multiple single-task prediction data;

[0016] Obtain the multi-task loss error reference value based on the sum of the loss errors between each single-task prediction data and the corresponding single-task true data.

[0017] Preferably, the single-task prediction data includes at least two of load gap prediction data, fault probability prediction data, restoration duration prediction data, transmission flow prediction data, and charging demand prediction data.

[0018] Preferably, the specific process of determining any target node from the distribution network, combining the neighborhood nodes corresponding to the target node, masking the spatio-temporal feature vectors of the target node and the neighborhood nodes, and then performing simulation operations through a multi-task prediction model to obtain a multi-task loss error verification value includes:

[0019] Determine any target node from the distribution network, combine the neighborhood nodes corresponding to the target node, and mask the spatio-temporal feature vectors of the target node and the neighborhood nodes according to the Extra masking logic and / or Intra masking logic.

[0020] Based on the spatio-temporal feature vectors of the distribution network after node masking processing, perform simulation operations through a multi-task prediction model to obtain a first multi-task loss error verification value and / or a second multi-task loss error verification value, where the first multi-task loss error verification value is the multi-task loss error verification value obtained based on the spatio-temporal feature vectors of the distribution network processed according to the Extra masking logic, and the second multi-task loss error verification value is the multi-task loss error verification value obtained based on the spatio-temporal feature vectors of the distribution network processed according to the Intra masking logic.

[0021] Preferably, determining the key node recognition result of the target node according to the error difference between the multi-task loss error reference value and the multi-task loss error verification value specifically includes:

[0022] According to the error difference between the multi-task loss error reference value and the multi-task loss error verification value, by comparing the error differences corresponding to each target node, determine the key node recognition result of the target node.

[0023] Meanwhile, a second aspect of the present application provides a device for identifying key nodes in a distribution network, including:

[0024] A distribution network data acquisition unit for acquiring node topology data and historical operation data of the distribution network;

[0025] A graph data construction unit for constructing different node relationship graph data according to the node topology data through multiple preset graph data construction conditions;

[0026] A spatio-temporal feature extraction unit for fusing each of the node relationship graph data to obtain the spatial features of the distribution network, and then according to the historical operation data and the spatial features, through a preset time series aggregation model, obtaining the spatio-temporal feature vector of the distribution network;

[0027] A first error calculation unit for performing a simulation operation according to the spatio-temporal feature vectors corresponding to each node in the distribution network in combination with a multi-task prediction model to obtain a multi-task loss error reference value;

[0028] A second error calculation unit for determining any target node from the distribution network, combining the neighborhood nodes corresponding to the target node, masking the spatio-temporal feature vectors of the target node and the neighborhood nodes, and then performing a simulation operation through a multi-task prediction model to obtain a multi-task loss error verification value;

[0029] A key node recognition unit for determining the key node recognition result of the target node according to the error difference between the multi-task loss error reference value and the multi-task loss error verification value.

[0030] A third aspect of the present application provides a terminal for identifying key nodes in a distribution network, including: a memory and a processor;

[0031] The memory is used for storing program codes, and the program codes are used to implement a method for identifying key nodes in a distribution network provided in the first aspect of the present application;

[0032] The processor is used for reading and executing the program codes.

[0033] A fourth aspect of the present application provides a computer-readable storage medium, in which program code is stored, and the program code is used to be read and executed by a processor to implement a method for identifying key nodes of a distribution network as provided in the first aspect of the present application.

[0034] As can be seen from the above technical solutions, the present application has the following advantages:

[0035] The solution provided by the present application is based on the node topology data and historical operation data of the distribution network. First, according to the node topology data, different node relationship graph data are constructed, and then through the fusion of each node relationship graph data, the spatial characteristics of the distribution network are obtained. Then, using the spatial characteristics and the historical operation data representing the temporal characteristics, through a preset temporal aggregation model for calculation, a spatio-temporal feature vector is obtained. Then, an arbitrary target node is determined from the distribution network, and the spatio-temporal feature vectors of the target node and related neighborhood nodes are masked. Through a multi-task prediction model for simulation calculation, the loss errors before and after masking the target node are obtained respectively. According to the difference between the two sets of loss errors, the impact degree of the failure of the target node on the overall distribution network is measured, and then the key nodes in the distribution network are identified. This solution can comprehensively capture the spatio-temporal dynamic characteristics of the power system, improve the prediction accuracy of node resilience, and solve the technical problem of low accuracy in identifying existing key nodes through multi-dimensional modeling, task collaborative training, efficient calculation, and transparent decision support. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0037] Figure 1 It is a schematic flowchart of an embodiment of a method for identifying key nodes of a distribution network provided by the present application.

[0038] Figure 2 It is a schematic diagram of the multi-graph structure and its fusion in the spatio-temporal model in a method for identifying key nodes of a distribution network provided by the present application.

[0039] Figure 3 It is a schematic diagram of Extra and Intra masking in interpretable analysis.

[0040] Figure 4 It is a schematic structural diagram of an embodiment of a device for identifying key nodes of a distribution network provided by the present application.

[0041] Figure 5Schematic structural diagram of an embodiment of a key node identification terminal for this application. Detailed implementation manners

[0042] This application embodiment provides a method, device, terminal and medium for identifying key nodes in a distribution network, which are used to solve the technical problem of low accuracy in identifying existing key nodes.

[0043] To make the invention purpose, features and advantages of this application more obvious and understandable, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of this application. Obviously, the embodiments described below are only a part of the embodiments of this application, rather than all the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the protection scope of this application.

[0044] First, a detailed description of an embodiment of a method for identifying key nodes in a distribution network provided by this application is as follows:

[0045] Please refer to Figure 1 and Figure 2 , a method for identifying key nodes in a distribution network provided by an embodiment of this application includes:

[0046] Step 101, obtain node topology data and historical operation data of the distribution network;

[0047] It should be noted that, first, according to the structure of the distribution network to be analyzed, the node topology data and historical operation data of the distribution network system are obtained.

[0048] Step 102, according to the node topology data, construct different node relationship graph data through multiple preset graph data construction conditions;

[0049] It should be noted that in step 102, according to the node topology data obtained in the previous step, different node relationship graph data are constructed. For example, assume that there are M nodes in the power system (such as substations, switch stations, distribution areas, important load centers, etc.). This embodiment presets multiple graph data construction conditions from different perspectives for constructing K pieces of node relationship graph data , each graph is composed of a node set and a corresponding adjacency matrix constitute, and the adjacency matrix of each graph data represents different association relationships between each node of the distribution network.

[0050] More specifically, the node relationship graph data mentioned in this embodiment mainly includes: topological adjacency graph data, geographical distance graph data and node type similarity graph data;

[0051] Among them, the adjacency matrix of the topological adjacency graph is specifically used to reflect the topological connection relationship between each node. For example, if node i and node j are directly connected in a network topology, then ; otherwise it is 0;

[0052] The adjacency matrix of the geographical distance graph is specifically used to reflect the geographical distance between each node. For example, according to the geographical distance perform normalization to obtain , and methods such as or can be adopted;

[0053] The adjacency matrix of the node type similarity graph is used to reflect the functional attribute similarity or node type similarity between each node. If node i and j belong to the same type of node or nodes with similar functions, then , otherwise it is 0.

[0054] In addition to the above three main graph data, more graphs can be expanded according to the needs of the actual scenario, such as the "historical fault correlation graph", the "meteorological zoning graph", etc., so as to reflect the node associations in more dimensions, which will not be elaborated here.

[0055] Step 103: Perform fusion processing on the data of each node relationship graph to obtain the spatial characteristics of the distribution network, and then according to the historical operation data and spatial characteristics, through a preset time series aggregation model, obtain the spatio-temporal characteristic vector of the distribution network;

[0056] It should be noted that when performing graph fusion, the GraphSAGE module is mainly used to extract the spatial feature representation of each node under each graph through a graph neural network. For each graph, GraphSAGE calculates the embedding representation of each node by performing an aggregation operation on the neighborhood of each node. Assuming that at a certain layer, the representation of node i is , then the update process of this layer is:

[0057]

[0058] In the formula, is the learnable parameter matrix, is the embedding representation of node i at the l-th layer, is the non-linear activation function (such as ReLU), is the neighbor set of node i in a certain graph, is the aggregation function, such as mean, max-pool or LSTM aggregation, etc.

[0059] The input information of the aggregation function is the feature matrix and the adjacency matrix , in the AGG function, the features of the neighbor nodes of node i are combined by an aggregation method. For node i, it aggregates the features of its neighbor nodes . In the first layer of the GraphSAGE module, the feature representation of the node is extracted from the node feature matrix X_(i,t). In other words, the initial feature representation is directly obtained from the input matrix.

[0060] If there are K graphs, several layers of GraphSAGE operations can be performed on each graph respectively to obtain the representation of node i under each graph , and then they are concatenated as:

[0061]

[0062] Among them, square brackets and "||" are used to represent the feature concatenation operation.

[0063] Step 104: According to the spatio-temporal feature vectors corresponding to each node in the distribution network, perform simulation operations in combination with the multi-task prediction model to obtain the multi-task loss error benchmark value;

[0064] It should be noted that then multi-task output and loss processing are carried out. Suppose there are N tasks, and the prediction target of each task n is denoted as . In the last layer of the network, a small network is designed for each task:

[0065]

[0066] Among them, represents the predicted value of the nth task, represents the exclusive output structure of task 𝑛, is the learnable parameter of this network. Let the true label be , the multi-task loss (using mean square error MSE) can be defined, and the expression is as follows:

[0067]

[0068] Among them, represents all the parameters of the shared layer (spatial + temporal module); represents the parameters of the nth task head; represents the set of time indices covered by the training set. During backpropagation, the gradients of multiple tasks will jointly update the shared layer and their respective exclusive layers, and thus multiple resilience-related indicators can be predicted simultaneously.

[0069] Next, this model will output the predicted value of each task and an error value , the specific type of error value depends on the input information of the model. If the input is the information without masking, the initial error will be obtained. , that is, the multi-task loss error benchmark value mentioned in this application, which represents the total error of multi-task prediction when nodes are not masked. Assuming taking three tasks as an example (A: load gap, B: restoration duration, C: fault scope), the expression is as follows:

[0070]

[0071] Among them, the load gap: represents the gap between the electricity demand and the supply of node i at time t in an extreme scenario;

[0072] The restoration duration: represents the time required from power outage to normal power supply restoration when a node or line fails;

[0073] The fault affected scope: represents the potential impact degree or the depth of failure propagation of a node fault on surrounding nodes or the entire system.

[0074] Step 105: Determine any target node from the distribution network, combine the neighborhood nodes corresponding to the target node, mask the spatio-temporal feature vectors of the target node and the neighborhood nodes, and then perform simulation operations through the multi-task prediction model to obtain the multi-task loss error verification value;

[0075] Step 106: Determine the key node identification result of the target node according to the error difference between the multi-task loss error benchmark value and the multi-task loss error verification value.

[0076] It should be noted that in Step 105 and Step 106, to identify the key nodes and their importance to the whole network or local prediction, the logical framework of "Mask-Compute-Analyze" (mask-compute-analyze) for interpretability analysis is adopted. After training, to further explore the importance of a certain node (or neighborhood) in "resilience prediction", the following three steps are adopted:

[0077] Mask: Select the target node and its k-hop neighborhood, set the input data of these nodes to zero, and simulate a serious failure or data loss of this node;

[0078] Compute: Re-infer the model based on the masked input, obtain the new multi-task prediction result, and then calculate the error;

[0079] Specifically, there are two masking strategies for node , which are divided into Extra masking (masking the node itself + neighborhood) and Intra masking (only masking the neighborhood, excluding ), and the schematic diagram is asFigure 3 As shown, among which, Figure 3 Part (a) is a schematic diagram of the shielding effect of node and its neighborhood under the Extra shielding strategy, Figure 3 Part (b) is a schematic diagram of the shielding effect of node and its neighborhood under the Intra shielding strategy.

[0080] Among which, the Extra shielding logic mainly analyzes the impact on the entire network. Define the k-hop neighborhood (including itself) of node as . According to the small-world network theory, the influence range of a node is mainly concentrated within its 3-hop neighborhood. The contribution degree of nodes beyond this range to the target node decays exponentially. Assume k = 3, then the error output by the multi-task spatio-temporal learning model is:

[0081]

[0082] Among which, is the total error after the target node and its 3-hop neighborhood, that is, the first multi-task loss error verification value mentioned in this application, is the set of 3-hop neighborhood nodes of node .

[0083] Intra shielding mainly analyzes the impact on the node itself. This analysis only shields the neighborhood of node but does not shield itself. At this time, the error output by the multi-task spatio-temporal learning model is:

[0084]

[0085] Among which, is the total error after the target node and its 3-hop neighborhood, that is, the second multi-task loss error verification value mentioned in this application, represents only shielding the 3-hop neighborhood of (retaining itself).

[0086] It can be understood that in actual node analysis applications, any one of the node shielding strategies can be selected according to actual needs, or both shielding strategies can be used. For the convenience of description, this embodiment only takes the case of using both strategies as an example for description.

[0087] Analyze: Compare the error changes before and after setting to zero to measure the marginal contribution or influence of this node / neighborhood on the prediction result.

[0088] It should be noted that when performing analyze analysis, the way to compare the error change is to subtract the initial error from the error after masking. Specifically, they are respectively:

[0089]

[0090]

[0091] Among them, is the marginal contribution of node to the overall network prediction error. The larger the positive value, the more significant the impact of the failure of this node on the overall resilience of the system; is the dependence degree of node on its own prediction error. The larger the positive value, the more the resilience of this node depends on the surrounding area.

[0092] Based on the Extra masking analysis, select several nodes with the largest values (such as Top10, which can be selected in combination with the actual node scale). The failure of these nodes will cause a significant increase in the total multi-task loss, and emergency power supplies should be preferentially configured;

[0093] Based on the Intra masking analysis, if ( is a preset threshold, which can be selected in combination with the actual situation), it indicates that this node is vulnerable to the influence of neighborhood failures and the local topological connection needs to be strengthened;

[0094] Finally, based on the comprehensive multi-task analysis result information, by comparing the and calculated for each node, if the node is larger in the and calculated for multiple tasks, it indicates that the probability of this node being a system-level critical vulnerability point is greater.

[0095] Through the solution provided by this application, the most critical nodes in the distribution network can be found more accurately, that is, the nodes that will have a major impact on the entire power grid once a problem occurs can be identified, and the resilience of the power grid can be evaluated, that is, the recovery efficiency of the power grid when encountering faults can be evaluated, as well as the places where problems are most likely to occur, which can help the power operation and maintenance department quickly identify critical nodes and give priority to emergency repairs during disasters, reducing the power outage time and scope; guiding reinforcement and expansion in power grid planning to improve the overall resilience; and monitoring the status of critical nodes in real time during daily operation and maintenance, giving early warnings of potential risks, and avoiding large-scale power outages.

[0096] The above is a detailed description of an embodiment of a method for identifying critical nodes in a distribution network provided by this application. The following is a detailed description of a device for identifying critical nodes in a distribution network provided by this application, specifically as follows:

[0097] Please refer toFigure 4 , a key node identification device for a distribution network provided by an embodiment of the present application includes:

[0098] A distribution network data acquisition unit 201 for acquiring node topology data and historical operation data of the distribution network;

[0099] A graph data construction unit 202 for constructing different node relationship graph data according to the node topology data through multiple preset graph data construction conditions;

[0100] A spatio-temporal feature extraction unit 203 for fusing and processing each node relationship graph data to obtain the spatial features of the distribution network, and then obtaining the spatio-temporal feature vector of the distribution network through a preset time series aggregation model according to the historical operation data and the spatial features;

[0101] A first error calculation unit 204 for performing simulation operations according to the spatio-temporal feature vectors corresponding to each node in the distribution network in combination with a multi-task prediction model to obtain a multi-task loss error reference value;

[0102] A second error calculation unit 205 for determining any target node from the distribution network, combining the neighborhood nodes corresponding to the target node, masking the spatio-temporal feature vectors of the target node and the neighborhood nodes, and then performing simulation operations through a multi-task prediction model to obtain a multi-task loss error verification value;

[0103] A key node identification unit 206 for determining the key node identification result of the target node according to the error difference between the multi-task loss error reference value and the multi-task loss error verification value.

[0104] In addition, the present application also correspondingly provides a detailed description of embodiments of a key node identification terminal for a distribution network and a computer-readable storage medium, as follows:

[0105] As Figure 5 shown, an embodiment of a key node identification terminal for a distribution network provided by the present application, the implementation types of the terminal include but are not limited to: personal computers, industrial computers, servers, and embedded intelligent devices. The main components of the terminal include: a memory 33 and a processor 31, and the memory 33 and the processor 31 can be connected through a communication bus 34;

[0106] The memory 33 is used to store program codes, and the program codes are used to implement a key node identification method for a distribution network provided by the above embodiment;

[0107] The processor 31 is used to read and execute the program codes.

[0108] A computer-readable storage medium provided by the present application stores program codes, which are used to be read and executed by a processor to implement a method for identifying key nodes in a distribution network as provided in the above embodiments.

[0109] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described terminals, devices, and units can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0110] In several embodiments provided by the present application, it should be understood that the disclosed terminals, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed mutual coupling or direct coupling or communication connection can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be in an electrical, mechanical, or other form.

[0111] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects and do not necessarily need to describe a specific order or sequence. It should be understood that such used data can be interchanged under appropriate circumstances so that the embodiments of the present application described herein, for example, can be implemented in an order other than those illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units does not necessarily need to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products, or devices.

[0112] It should be understood that in this application, "at least one (item)" means one or more, and "a plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that there can be three relationships. For example, "A and / or B" can represent: only A exists, only B exists, and both A and B exist at the same time. Among them, A and B can be singular or plural. The character " / " generally represents an "or" relationship between the associated objects before and after. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single items (items) or plural items (items). For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0113] The unit described as a separate component may or may not be physically separated. The component shown as a unit may or may not be a physical unit, that is, it may be located in one place, or it may be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0114] In addition, each functional unit in various embodiments of the present invention can be integrated into a processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of a software functional unit.

[0115] If the above-mentioned integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical disks, etc., which can store program codes.

[0116] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the various embodiments of the present application.

Claims

1. A method for identifying key nodes in a distribution network, characterized in that Including: Obtain the node topology data and historical operation data of the distribution network; According to the node topology data, construct different node relationship graph data through multiple preset graph data construction conditions; Fuse and process each of the node relationship graph data to obtain the spatial characteristics of the distribution network, and then according to the historical operation data and the spatial characteristics, obtain the spatio-temporal feature vector of the distribution network through a preset time-series aggregation model; According to the spatio-temporal feature vectors corresponding to each node in the distribution network, perform simulation operations in combination with a multi-task prediction model to obtain a multi-task loss error reference value; Determine any target node from the distribution network, combine the neighborhood nodes corresponding to the target node, mask the spatio-temporal feature vectors of the target node and the neighborhood nodes, and then perform simulation operations through a multi-task prediction model to obtain a multi-task loss error verification value; Determine the key node identification result of the target node according to the error difference between the multi-task loss error reference value and the multi-task loss error verification value.

2. The method for identifying key nodes of a distribution network according to claim 1, wherein The node relationship graph data includes: topological adjacency graph data, geographical distance graph data, and node type similarity graph data.

3. A method for identifying key nodes of a distribution network according to claim 1, characterized in that The fusing and processing each of the node relationship graph data to obtain the spatial characteristics of the distribution network specifically includes: Extract the spatial embedding features in each node relationship graph data through a graph neural network, and then fuse and process the spatial embedding features of each of the node relationship graph data to obtain the spatial characteristics of the distribution network.

4. A method for identifying key nodes in a distribution network according to claim 1, characterized in that, The performing simulation operations in combination with a multi-task prediction model according to the spatio-temporal feature vectors corresponding to each node in the distribution network to obtain a multi-task loss error reference value specifically includes: Perform simulation operations in combination with a multi-task prediction model according to the spatio-temporal feature vectors corresponding to each node in the distribution network to obtain multiple single-task prediction data; Obtain the multi-task loss error reference value according to the sum of the loss errors of each single-task prediction data and the corresponding single-task real data.

5. A method for identifying key nodes in a distribution network according to claim 4, characterized in that The single-task prediction data includes at least two of load gap prediction data, fault probability prediction data, restoration duration prediction data, transmission flow prediction data, and charging demand prediction data.

6. The key node identification method for a distribution network according to claim 4, characterized in that The determining any target node from the distribution network, combining the neighborhood nodes corresponding to the target node, masking the spatio-temporal feature vectors of the target node and the neighborhood nodes, and then performing simulation operations through a multi-task prediction model to obtain a multi-task loss error verification value specifically includes: Determine any target node from the distribution network, combine the neighborhood nodes corresponding to the target node, and mask the spatio-temporal feature vectors of the target node and the neighborhood nodes according to the Extra masking logic and / or Intra masking logic, Based on the spatio-temporal feature vector of the distribution network after node masking processing, through a multi-task prediction model for simulation operations, a first multi-task loss error verification value and / or a second multi-task loss error verification value are obtained, where the first multi-task loss error verification value is a multi-task loss error verification value obtained based on the spatio-temporal feature vector of the distribution network processed according to the Extra masking logic, and the second multi-task loss error verification value is a multi-task loss error verification value obtained based on the spatio-temporal feature vector of the distribution network processed according to the Intra masking logic.

7. A method for identifying key nodes in a distribution network according to claim 6, characterized in that, The determining of the key node recognition result of the target node according to the error difference between the multi-task loss error reference value and the multi-task loss error verification value specifically includes: According to the error difference between the multi-task loss error reference value and the multi-task loss error verification value, by comparing the error differences corresponding to each target node, the key node recognition result of the target node is determined.

8. A key node identification device for a distribution network, characterized in that, Includes: A distribution network data acquisition unit for acquiring the node topology data and historical operation data of the distribution network; A graph data construction unit for constructing different node relationship graph data according to the node topology data through a plurality of preset graph data construction conditions; A spatio-temporal feature extraction unit for fusing each of the node relationship graph data to obtain the spatial feature of the distribution network, and then according to the historical operation data and the spatial feature, through a preset time series aggregation model, obtaining the spatio-temporal feature vector of the distribution network; A first error calculation unit for performing simulation operations in combination with a multi-task prediction model according to the spatio-temporal feature vectors corresponding to each node in the distribution network to obtain a multi-task loss error reference value; A second error calculation unit for determining any target node from the distribution network, combining the neighborhood nodes corresponding to the target node, masking the spatio-temporal feature vectors of the target node and the neighborhood nodes, and then performing simulation operations through a multi-task prediction model to obtain a multi-task loss error verification value; A key node recognition unit for determining the key node recognition result of the target node according to the error difference between the multi-task loss error reference value and the multi-task loss error verification value.

9. A key node identification terminal for a distribution network, characterized in that, Includes: A memory and a processor; The memory is used to store program codes, and the program codes are used to implement a method for identifying key nodes in a distribution network according to any one of claims 1 to 7; The processor is used to read and execute the program codes.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores program codes, and the program codes are used to be read and executed by the processor to implement a method for identifying key nodes in a distribution network according to any one of claims 1 to 7.

Images