Power grid fault detection method and device, terminal equipment and storage medium

By building a grid diagram model and generating a fault probability matrix, the problem of incomplete grid fault detection in the existing technology is solved, and the efficiency of grid troubleshooting and operational safety are improved.

CN119024096AActive Publication Date: 2024-11-26GUANGDONG POWER GRID CO LTD +1

Patent Information

Application Number
CN202411112931.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-14
Publication Date
2024-11-26
Estimated Expiration
2044-08-14

AI Technical Summary

Technical Problem

The existing power grid fault detection methods are difficult to comprehensively detect possible faults in the power grid, resulting in an extended fault handling time and increasing uncertainty and risks in power grid operation.

Method used

By obtaining the topological structure and power flow information of the power grid, building a graph model, obtaining the power matrix and adjacency matrix, inputting a fault detection model to generate a fault probability matrix, determining the fault probability of each node and edge of the power grid, and improving the comprehensiveness and troubleshooting of fault detection and troubleshooting efficiency.

Benefits of technology

It improves the comprehensiveness and inspection efficiency of power grid fault detection, and enhances the safety and stability of power grid operation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119024096A_ABST
    Figure CN119024096A_ABST
Patent Text Reader

Abstract

The invention discloses a power grid fault detection method and apparatus, a terminal device and a storage medium. The method comprises the steps of obtaining a topological structure and power flow information of a to-be-detected power grid; constructing a to-be-detected power grid graph model according to the topological structure and the power flow information of the to-be-detected power grid; wherein the to-be-detected power grid graph model comprises a plurality of nodes for representing equipment in a to-be-detected power grid and a plurality of edges for representing a connection relationship among the equipment in the to-be-detected power grid; obtaining a power matrix and an adjacent matrix of a to-be-detected power grid according to the to-be-detected power grid graph model; inputting the power matrix and the adjacent matrix of the to-be-detected power grid into a fault detection model, so that the fault detection model outputs a fault probability matrix of the to-be-detected power grid; and determining the fault probability of each node and the fault probability of each edge of the to-be-detected power grid according to the fault probability matrix, so that operation and maintenance personnel perform troubleshooting on the to-be-detected power grid according to the fault probability of each node and the fault probability of each edge of the to-be-detected power grid.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of power grid fault detection, and in particular to a power grid fault detection method, device, terminal equipment and storage medium, device, terminal equipment and storage medium. Background Art

[0002] With the continuous expansion of the scale of power systems and the increasing complexity of power grid structures, the reliability and safe operation of power grids are facing severe challenges. In the existing technologies, fault location usually relies on power grid topology analysis and current flow monitoring methods. These methods usually determine the location of possible faults in the power grid based on the existence of breakpoints, circuit breaks or abnormal current flow in the power grid. However, these existing methods usually determine a branch or node where the fault exists, and then provide it to the operation and maintenance personnel for troubleshooting. This method is difficult to conduct a comprehensive investigation of possible faults in the power grid. When a fault occurs at a fault point that has not been found, it is necessary to repeat the investigation, which leads to a longer fault handling time and increases the uncertainty and risk of power grid operation. Summary of the invention

[0003] The embodiments of the present invention provide a power grid fault detection method, apparatus, terminal equipment and storage medium, which can improve the comprehensiveness and troubleshooting efficiency of power grid fault detection, thereby improving the safety and stability of power grid operation.

[0004] An embodiment of the present invention provides a power grid fault detection method, comprising:

[0005] Obtain the topological structure and power flow information of the power grid to be detected;

[0006] Constructing a graph model of the power grid to be detected according to the topological structure and power flow information of the power grid to be detected; wherein the graph model of the power grid to be detected includes a plurality of nodes representing devices in the power grid to be detected and a plurality of edges representing connection relationships between devices in the power grid to be detected;

[0007] Acquire a power matrix and an adjacency matrix of the power grid to be detected according to the power grid graph model to be detected;

[0008] Inputting the power matrix and the adjacency matrix of the power grid to be detected into the fault detection model so that the fault detection model outputs the fault probability matrix of the power grid to be detected;

[0009] The failure probability of each node and each edge of the power grid to be detected is determined according to the failure probability matrix, so that the operation and maintenance personnel can perform fault troubleshooting on the power grid to be detected according to the failure probability of each node and each edge of the power grid to be detected.

[0010] Furthermore, constructing the fault detection model includes:

[0011] Obtain historical operation data and fault simulation operation data of the power grid to be tested;

[0012] Constructing a training sample set according to the historical operation data and the fault simulation operation data; wherein the training sample set includes a power matrix sample, an adjacency matrix sample and a fault label matrix;

[0013] An initial fault detection model is constructed, and the initial fault detection model is trained with the training sample set, and when a preset convergence condition is reached, a fault detection model is generated.

[0014] Furthermore, the fault detection model includes: a point cloud dynamic graph convolutional network layer, a graph convolutional network layer, a normalization layer and an output layer;

[0015] The step of inputting the power matrix and the adjacency matrix of the power grid to be detected into the fault detection model so that the fault detection model outputs the fault probability matrix of the power grid to be detected comprises:

[0016] Inputting the power matrix and adjacency matrix of the power grid to be detected into the point cloud dynamic graph convolution network layer, so that the point cloud dynamic graph convolution network layer performs edge convolution and dynamic graph update on the power matrix and the adjacency matrix, and extracts the adjacency relationship between each node of the power grid to be detected and the local features of each node;

[0017] Inputting the power matrix and the adjacency matrix of the power grid to be detected into the graph convolutional network layer, so that the graph convolutional network layer determines the weight relationship between the nodes of the power grid to be detected and extracts the global features of the nodes of the power grid to be detected;

[0018] The adjacency relationship between the nodes of the power grid to be detected, the local characteristics of the nodes, the weight relationship between the nodes and the global characteristics of the nodes are transmitted to the normalization layer, so that the normalization layer determines the fault probability matrix of the power grid to be detected according to the adjacency relationship between the nodes of the power grid to be detected, the local characteristics of the nodes, the weight relationship between the nodes and the global characteristics of the nodes, and transmits it to the output layer;

[0019] The output layer outputs the fault probability matrix of the power grid to be detected.

[0020] Furthermore, constructing a training sample set according to the historical operation data and the fault simulation operation data includes:

[0021] According to the historical operation data, the power samples of each node of the power grid to be tested and the adjacency relationship of each node are obtained under the normal operation state;

[0022] According to the fault simulation operation data, power samples of each node of the power grid to be detected, the adjacency relationship of each node and the fault label of each fault position in the power grid to be detected are obtained under the fault state; wherein the fault label is used to identify the fault node and the fault edge;

[0023] Construct power matrix samples, adjacency matrix samples and fault label matrix according to power samples of each node in normal operation, adjacency relations of each node in normal operation, power samples of each node in fault state, adjacency relations of each node in fault state and fault labels of each fault position in the power grid to be detected;

[0024] A training sample set is constructed based on power matrix samples, adjacency matrix samples and fault label matrix.

[0025] Based on the above method embodiment, the present invention provides a corresponding device embodiment;

[0026] An embodiment of the present invention provides a power grid fault detection device, including: a graph model building module, a data acquisition module and a fault detection module;

[0027] The graph model building module is used to obtain the topological structure and power flow information of the power grid to be detected; and to build a graph model of the power grid to be detected according to the topological structure and power flow information of the power grid to be detected; wherein the graph model of the power grid to be detected includes a plurality of nodes representing devices in the power grid to be detected and a plurality of edges representing connection relationships between devices in the power grid to be detected;

[0028] The data acquisition module is used to acquire the power matrix and adjacency matrix of the power grid to be detected according to the power grid graph model to be detected;

[0029] The fault detection module is used to input the power matrix and adjacency matrix of the power grid to be detected into the fault detection model, so that the fault detection model outputs the fault probability matrix of the power grid to be detected; determine the fault probability of each node and the fault probability of each edge of the power grid to be detected according to the fault probability matrix, so that the operation and maintenance personnel can perform fault troubleshooting on the power grid to be detected according to the failure probability of each node and the failure probability of each edge of the power grid to be detected.

[0030] Furthermore, constructing the fault detection model includes:

[0031] Obtain historical operation data and fault simulation operation data of the power grid to be tested;

[0032] Constructing a training sample set according to the historical operation data and the fault simulation operation data; wherein the training sample set includes a power matrix sample, an adjacency matrix sample and a fault label matrix;

[0033] An initial fault detection model is constructed, and the initial fault detection model is trained with the training sample set, and when a preset convergence condition is reached, a fault detection model is generated.

[0034] Furthermore, the fault detection model includes: a point cloud dynamic graph convolutional network layer, a graph convolutional network layer, a normalization layer and an output layer;

[0035] The step of inputting the power matrix and the adjacency matrix of the power grid to be detected into the fault detection model so that the fault detection model outputs the fault probability matrix of the power grid to be detected comprises:

[0036] Inputting the power matrix and adjacency matrix of the power grid to be detected into the point cloud dynamic graph convolution network layer, so that the point cloud dynamic graph convolution network layer performs edge convolution and dynamic graph update on the power matrix and the adjacency matrix, and extracts the adjacency relationship between each node of the power grid to be detected and the local features of each node;

[0037] Inputting the power matrix and the adjacency matrix of the power grid to be detected into the graph convolutional network layer, so that the graph convolutional network layer determines the weight relationship between the nodes of the power grid to be detected and extracts the global features of the nodes of the power grid to be detected;

[0038] The adjacency relationship between the nodes of the power grid to be detected, the local characteristics of the nodes, the weight relationship between the nodes and the global characteristics of the nodes are transmitted to the normalization layer, so that the normalization layer determines the fault probability matrix of the power grid to be detected according to the adjacency relationship between the nodes of the power grid to be detected, the local characteristics of the nodes, the weight relationship between the nodes and the global characteristics of the nodes, and transmits it to the output layer;

[0039] The output layer outputs the fault probability matrix of the power grid to be detected.

[0040] Furthermore, constructing a training sample set according to the historical operation data and the fault simulation operation data includes:

[0041] According to the historical operation data, the power samples of each node of the power grid to be tested and the adjacency relationship of each node are obtained under the normal operation state;

[0042] According to the fault simulation operation data, power samples of each node of the power grid to be detected, the adjacency relationship of each node and the fault label of each fault position in the power grid to be detected are obtained under the fault state; wherein the fault label is used to identify the fault node and the fault edge;

[0043] Construct power matrix samples, adjacency matrix samples and fault label matrix according to power samples of each node in normal operation, adjacency relations of each node in normal operation, power samples of each node in fault state, adjacency relations of each node in fault state and fault labels of each fault position in the power grid to be detected;

[0044] A training sample set is constructed based on power matrix samples, adjacency matrix samples and fault label matrix.

[0045] 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, and when the processor executes the computer program, a power grid fault detection method described in the above-mentioned embodiment of the invention is implemented.

[0046] Another embodiment of the present invention provides a storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the storage medium is located is controlled to execute a power grid fault detection method described in the above-mentioned embodiment of the invention.

[0047] The following beneficial effects are achieved by implementing the present invention:

[0048] The present invention provides a power grid fault detection method, device, terminal equipment and storage medium. The fault detection method constructs a graph model of the power grid to be detected by acquiring the topological structure and power flow information of the power grid to be detected, and then determines the power matrix and adjacency matrix of the power grid to be detected according to the constructed graph model; wherein the graph model includes a plurality of nodes representing devices in the power grid to be detected and a plurality of edges representing the connection relationship between devices in the power grid to be detected; the power matrix and the adjacency matrix of the power grid to be detected are input into the fault detection model so that the fault detection model outputs a fault probability matrix of the power grid to be detected; the fault probability of each node and the fault probability of each edge of the power grid to be detected are determined through the fault probability matrix, and then the fault probability of each node and the fault probability of each edge are provided to power grid operation and maintenance personnel, so that the power grid operation and maintenance personnel perform fault troubleshooting on the power grid according to the fault probability, and determine the fault probability between all nodes and edges in the power grid to be detected through the fault probability matrix, thereby improving the comprehensiveness and troubleshooting efficiency of power grid fault detection, and thus improving the safety and stability of power grid operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a flowchart of a power grid fault detection method provided by an embodiment of the present invention.

[0050] Figure 2 It is a structural diagram of a fault detection model provided by an embodiment of the present invention.

[0051] Figure 3 It is a structural schematic diagram of a power grid fault detection device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0053] like Figure 1 As shown, a power grid fault detection method provided by an embodiment of the present invention includes:

[0054] Step S1: obtaining the topological structure and power flow information of the power grid to be detected;

[0055] Step S2: constructing a graph model of the power grid to be detected according to the topological structure and power flow information of the power grid to be detected; wherein the graph model of the power grid to be detected includes a plurality of nodes representing devices in the power grid to be detected and a plurality of edges representing connection relationships between devices in the power grid to be detected;

[0056] Step S3: acquiring a power matrix and an adjacency matrix of the power grid to be detected according to the power grid graph model to be detected;

[0057] Step S4: inputting the power matrix and the adjacency matrix of the power grid to be detected into the fault detection model, so that the fault detection model outputs the fault probability matrix of the power grid to be detected;

[0058] Step S5: determining the failure probability of each node and each edge of the power grid to be detected according to the failure probability matrix, so that the operation and maintenance personnel can troubleshoot the power grid to be detected according to the failure probability of each node and each edge of the power grid to be detected.

[0059] For steps S1 and S2, the topological structure and power flow information of the power grid to be detected are obtained, and the topological structure and power flow information of the power grid to be detected are encoded into a graph model. In the graph model, the nodes represent the equipment in the power grid to be detected, such as generators, transformers, lines, etc.; the edges represent the connection relationship between the equipment.

[0060] For step S3, according to the constructed graph model of the power grid to be detected, the power matrix and adjacency matrix of the power grid to be detected are obtained based on the node power and edge representation connection information of the graph model; wherein the power matrix contains the active power and reactive power of each node, and each element of the power matrix refers to the active power and reactive power of a node of the power grid to be detected; and the adjacency matrix contains the connection relationship between each node.

[0061] For step S4, the power matrix and the adjacency matrix of the power grid to be detected are input into the fault detection model, so that the fault detection model outputs the fault probability matrix of the power grid to be detected. Exemplarily, the power matrix PQ (including active power P and reactive power Q, with a size of N x 2, N is the total number of nodes of the power grid to be detected) and the adjacency matrix A (with a size of N x N) are input into the fault detection model, so that the fault detection model outputs a fault probability matrix L (size: N x N), and each element in the fault probability matrix represents the probability of a fault occurring between corresponding nodes.

[0062] In a preferred embodiment, constructing the fault detection model includes: acquiring historical operating data and fault simulation operating data of the power grid to be detected; constructing a training sample set based on the historical operating data and fault simulation operating data; wherein the training sample set includes power matrix samples, adjacency matrix samples and fault label matrix; constructing an initial fault detection model, training the initial fault detection model with the training sample set, and generating a fault detection model when a preset convergence condition is reached.

[0063] In another preferred embodiment, the construction of a training sample set based on the historical operation data and the fault simulation operation data includes: obtaining power samples of each node of the power grid to be detected and the adjacency relationship of each node under normal operating conditions according to the historical operation data; obtaining power samples of each node of the power grid to be detected, the adjacency relationship of each node and the fault labels of each fault position in the power grid to be detected under fault conditions according to the fault simulation operation data; wherein the fault labels are used to identify faulty nodes and fault edges; constructing power matrix samples, adjacency matrix samples and fault label matrix according to power samples of each node under normal operating conditions, the adjacency relationship of each node under normal operating conditions, power samples of each node under fault conditions, the adjacency relationship of each node under fault conditions and the fault labels of each fault position in the power grid to be detected; and constructing a training sample set according to the power matrix samples, the adjacency matrix samples and the fault label matrix.

[0064] Specifically, the historical operation data and fault simulation operation data of the power grid to be detected are obtained, and the node power samples of each node in the power grid to be detected under normal operation and the adjacency relationship between each node are obtained based on the historical operation data; the node power samples of each node in the power grid to be detected under fault state, the adjacency relationship between each node and the fault label corresponding to the power grid to be detected under each fault state are obtained based on the fault simulation operation data. It should be noted that the fault label is used to identify the fault node and fault edge in each simulation data. In order to make the fault detection model obtained by training fully cover the possible fault conditions in the power grid, the selected fault simulation data should cover the possible fault conditions in the power grid as much as possible. According to the power samples of each node under normal operation, the adjacency relationship of each node under normal operation, the power samples of each node under fault state, the adjacency relationship of each node under fault state and the fault labels of each fault position in the power grid to be detected, the power matrix samples, adjacency matrix samples and fault label matrix are constructed, and then the training sample set is obtained. An initial fault detection model is constructed, and the selection of the initial fault detection model needs to be a graph neural network to meet the requirements of extracting the structural features of the power grid graph to be detected. Based on the training sample set, the initial fault detection model is trained. In each iterative training process, the power matrix samples and the adjacency matrix samples are used as input, and the predicted fault label matrix is ​​used as output. The loss between the predicted data and the actual data is calculated based on the predicted fault label matrix and the fault label matrix. After the loss is obtained, the network parameters of the initial fault detection model are updated through the back propagation algorithm, and the gradient descent optimization algorithm, such as the Adam algorithm, is used to minimize the loss function. During the training process, the performance of the model is monitored through performance indicators on the validation set, such as prediction accuracy, F1 score, etc. When the performance indicators of the validation set are no longer significantly improved or reach the preset number of training rounds (i.e., the above-mentioned preset convergence conditions), the training of the initial fault detection model is stopped and a fault detection model is generated.

[0065] In a preferred embodiment, the fault detection model includes: a point cloud dynamic graph convolutional network layer, a graph convolutional network layer, a normalization layer and an output layer; the power matrix and the adjacency matrix of the power grid to be detected are input into the fault detection model so that the fault detection model outputs the fault probability matrix of the power grid to be detected, including: inputting the power matrix and the adjacency matrix of the power grid to be detected into the point cloud dynamic graph convolutional network layer so that the point cloud dynamic graph convolutional network layer performs edge convolution and dynamic graph update on the power matrix and the adjacency matrix to extract the adjacency relationship between the nodes of the power grid to be detected and the local features of each node; The power matrix and the adjacency matrix are input into the graph convolutional network layer, so that the graph convolutional network layer determines the weight relationship between the nodes of the power grid to be detected and extracts the global features of the nodes of the power grid to be detected; the adjacency relationship between the nodes of the power grid to be detected, the local features of each node, the weight relationship between the nodes and the global features of each node are transmitted to the normalization layer, so that the normalization layer determines the fault probability matrix of the power grid to be detected according to the adjacency relationship between the nodes of the power grid to be detected, the local features of each node, the weight relationship between the nodes and the global features of each node, and transmits it to the output layer; the output layer outputs the fault probability matrix of the power grid to be detected.

[0066] Specifically, Figure 2As shown, it is a structural schematic diagram of the fault detection model provided by the present invention. The fault detection model mainly includes: a point cloud dynamic graph convolution network layer DGCNN, a graph convolution network layer (GCN, GraphSage and feature-direction MLP parts in the figure), a normalization layer (LayerNorm1() in the figure) and an output layer MLP. After the power matrix and adjacency matrix of the power grid to be detected are input into the fault detection model, the point cloud dynamic graph convolution layer DGCNN first performs edge convolution (EdgeConv) and dynamic graph update on the power matrix and the adjacency matrix to extract the adjacency relationship between each node of the power grid to be detected and the local features of each node. The network layer with DGCNN as the first step of processing can enhance the feature embedding capability of the fault detection model for graph data after processing the power matrix and the adjacency matrix. Edge convolution refers to the convolution operation on the edge of the K nearest neighbor graph of each node to better learn the similarity relationship between nodes; dynamic graph update means that each layer of the network will reconstruct the K nearest neighbor graph according to the output feature graph of the previous layer. Furthermore, the weight relationship between the nodes of the power grid to be detected is determined through the graph convolutional network layer (GCN, GraphSage and feature-direction MLP parts in the figure) and the global features of the nodes of the power grid to be detected are extracted. In this network layer, the first part is composed of GCN and self-attention mechanism. GCN combined with self-attention mechanism can further extract global features in power matrix and adjacency matrix, and can simultaneously consider the different importance between nodes, and determine the weight relationship between nodes according to importance; the second part is composed of GraphSage inductive graph neural network framework. The local features and global features of nodes processed by GCN and self-attention mechanism constitute the node feature matrix. GraphSage generates embedded representation of nodes by sampling and aggregating the features of neighboring nodes. The third part is composed of feature-direction MLP, which further extracts features from the data processed by GraphSage, and transmits all the extracted features, adjacency relations and weight relations to the normalization layer LayerNorm1(), so that the normalization layer LayerNorm1() normalizes the global and local features of the nodes, eliminates the difference in the range of feature values, and accelerates the convergence of the model. Preferably, in order to improve the stability and training efficiency of the model, the fault detection model also includes another normalization layer (i.e., channel direction MLP, ReLU(), Dropout (0.2) and LayerNorm2() in the figure), and further extracts features, ReLU activation function and regularization after the normalization layer LayerNorm1(), and then normalizes the node features again through LayerNorm2(), thereby further improving the stability and training efficiency of the model.Finally, the output layer MLP generates a failure probability matrix containing the failure probability of each node and each edge.

[0067] For step S5, the failure probability of each node and each edge of the power grid to be detected is determined according to the failure probability matrix, and the operation and maintenance personnel are informed so that the operation and maintenance personnel can start comprehensive fault detection of the power grid to be detected based on the failure probability and start with the nodes or edges with high failure probability.

[0068] In order to better illustrate the technical effects that can be achieved by this application, actual examples are given below for illustration.

[0069] For example, assuming that the power grid contains 4 nodes, two quantities are defined for each node, namely active power P and reactive power Q. Then the power matrix PQ will be a 4x2 matrix, each row represents a node, the first column is the active power P, and the second column is the reactive power Q, as shown below:

[0070]

[0071]

[0072] This matrix indicates that node 1 and node 2 are power supply nodes, providing 10MW and 20MW of active power and 5MVar and 10MVar of reactive power respectively; node 3 and node 4 are load nodes, consuming 30MW and 40MW of active power and 15MVar and 20MVar of reactive power respectively.

[0073] Set the adjacency matrix A to [[0,1,0,1],[1,0,1,0],[0,1,0,1],[1,0,1,0]], which means that node 1 is connected to node 2 and node 4, node 2 is connected to node 1 and node 3, node 3 is connected to node 2 and node 4, node 4 is connected to node 1 and node 3, and so on.

[0074] The power matrix PQ and the adjacency matrix A are input into the fault detection model, and the corresponding output is the following fault probability matrix L:

[0075] L=[[0.00,0.70,0.00,0.20],

[0076] [0.70,0.00,0.60,0.00],

[0077] [0.00,0.60,0.00,0.80],

[0078] [0.20,0.00,0.80,0.00]].

[0079] The first row of the fault probability matrix L represents the interpretation as follows: the element value of the first row and the first column is 0.00, indicating that the fault probability of node 1 itself is not considered. The element value of the first row and the second column is 0.70, indicating that the fault probability of the line from node 1 to node 2 is 70%. The element value of the first row and the third column is 0.00, indicating that the fault probability of the line from node 1 to node 3 is 0. The element value of the first row and the fourth column is 0.20, indicating that the fault probability of the line from node 1 to node 4 is 20%. Similarly, the fault probability interpretations corresponding to the second to fourth rows can be derived. That is, the positions to be checked from large to small according to the fault probability are: the line from node 4 to node 3 (failure probability 80%, corresponding to the third row and the fourth column and the fourth row and the third column), the line from node 1 to node 2 (failure probability 70%, corresponding to the first row and the second column and the second row and the first column), the line from node 2 to node 3 (failure probability 60%, corresponding to the second row and the third column and the third row and the second column) and the line from node 1 to node 4 (failure probability 20%, corresponding to the first row and the fourth column and the fourth row and the first column). This data is provided to the operation and maintenance personnel, who conduct investigations in turn, thereby improving the comprehensiveness of troubleshooting.

[0080] Based on the above method embodiment, the present invention provides a corresponding device embodiment.

[0081] like Figure 3 As shown, an embodiment of the present invention provides a power grid fault detection device, including: a graph model building module, a data acquisition module and a fault detection module;

[0082] The graph model building module is used to obtain the topological structure and power flow information of the power grid to be detected; and to build a graph model of the power grid to be detected according to the topological structure and power flow information of the power grid to be detected; wherein the graph model of the power grid to be detected includes a plurality of nodes representing devices in the power grid to be detected and a plurality of edges representing connection relationships between devices in the power grid to be detected;

[0083] The data acquisition module is used to acquire the power matrix and adjacency matrix of the power grid to be detected according to the power grid graph model to be detected;

[0084] The fault detection module is used to input the power matrix and adjacency matrix of the power grid to be detected into the fault detection model, so that the fault detection model outputs the fault probability matrix of the power grid to be detected; determine the fault probability of each node and the fault probability of each edge of the power grid to be detected according to the fault probability matrix, so that the operation and maintenance personnel can perform fault troubleshooting on the power grid to be detected according to the failure probability of each node and the failure probability of each edge of the power grid to be detected.

[0085] In a preferred embodiment, constructing the fault detection model includes:

[0086] Obtain historical operation data and fault simulation operation data of the power grid to be tested;

[0087] Constructing a training sample set according to the historical operation data and the fault simulation operation data; wherein the training sample set includes a power matrix sample, an adjacency matrix sample and a fault label matrix;

[0088] An initial fault detection model is constructed, and the initial fault detection model is trained with the training sample set, and when a preset convergence condition is reached, a fault detection model is generated.

[0089] In a preferred embodiment, the fault detection model includes: a point cloud dynamic graph convolutional network layer, a graph convolutional network layer, a normalization layer and an output layer;

[0090] The step of inputting the power matrix and the adjacency matrix of the power grid to be detected into the fault detection model so that the fault detection model outputs the fault probability matrix of the power grid to be detected comprises:

[0091] Inputting the power matrix and adjacency matrix of the power grid to be detected into the point cloud dynamic graph convolution network layer, so that the point cloud dynamic graph convolution network layer performs edge convolution and dynamic graph update on the power matrix and the adjacency matrix, and extracts the adjacency relationship between each node of the power grid to be detected and the local features of each node;

[0092] Inputting the power matrix and the adjacency matrix of the power grid to be detected into the graph convolutional network layer, so that the graph convolutional network layer determines the weight relationship between the nodes of the power grid to be detected and extracts the global features of the nodes of the power grid to be detected;

[0093] The adjacency relationship between the nodes of the power grid to be detected, the local characteristics of the nodes, the weight relationship between the nodes and the global characteristics of the nodes are transmitted to the normalization layer, so that the normalization layer determines the fault probability matrix of the power grid to be detected according to the adjacency relationship between the nodes of the power grid to be detected, the local characteristics of the nodes, the weight relationship between the nodes and the global characteristics of the nodes, and transmits it to the output layer;

[0094] The output layer outputs the fault probability matrix of the power grid to be detected.

[0095] In a preferred embodiment, the step of constructing a training sample set based on the historical operation data and the fault simulation operation data includes:

[0096] According to the historical operation data, the power samples of each node of the power grid to be tested and the adjacency relationship of each node are obtained under the normal operation state;

[0097] According to the fault simulation operation data, power samples of each node of the power grid to be detected, the adjacency relationship of each node and the fault label of each fault position in the power grid to be detected are obtained under the fault state; wherein the fault label is used to identify the fault node and the fault edge;

[0098] Construct power matrix samples, adjacency matrix samples and fault label matrix according to power samples of each node in normal operation, adjacency relations of each node in normal operation, power samples of each node in fault state, adjacency relations of each node in fault state and fault labels of each fault position in the power grid to be detected;

[0099] A training sample set is constructed based on power matrix samples, adjacency matrix samples and fault label matrix.

[0100] It should be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art may understand and implement it without paying any creative effort.

[0101] Those skilled in the art can clearly understand that for the sake of convenience and brevity, the specific working process of the device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0102] Based on the above method item embodiments, the present invention provides corresponding terminal device item embodiments.

[0103] An 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, wherein when the processor executes the computer program, a power grid fault detection method described in any one of the present invention is implemented.

[0104] The terminal device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0105] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (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 any conventional processor, etc. The processor is the control center of the terminal device, and uses various interfaces and lines to connect various parts of the entire terminal device.

[0106] The memory can be used to store the computer program, and the processor realizes various functions of the terminal device by running or executing the computer program stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required for a function, etc.; the data storage area can store data created according to the use of the mobile phone, etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Med i aCard, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0107] Based on the above method item embodiments, the present invention provides a corresponding storage medium item embodiment.

[0108] An embodiment of the present invention provides a storage medium, wherein the storage medium includes a stored computer program, wherein when the computer program is executed, the device where the storage medium is located is controlled to execute a power grid fault detection method as described in any one of the present inventions.

[0109] The storage medium is a computer-readable storage medium, and the computer program is stored in the computer-readable storage medium. When the computer program is executed by the processor, the steps of each of the above-mentioned method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium.

[0110] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A power grid fault detection method, characterized in that: include: Obtain the topological structure and power flow information of the power grid to be detected; Constructing a graph model of the power grid to be detected according to the topological structure and power flow information of the power grid to be detected; wherein the graph model of the power grid to be detected includes a plurality of nodes representing devices in the power grid to be detected and a plurality of edges representing connection relationships between devices in the power grid to be detected; Acquire a power matrix and an adjacency matrix of the power grid to be detected according to the power grid graph model to be detected; Inputting the power matrix and the adjacency matrix of the power grid to be detected into the fault detection model so that the fault detection model outputs the fault probability matrix of the power grid to be detected; The failure probability of each node and each edge of the power grid to be detected is determined according to the failure probability matrix, so that the operation and maintenance personnel can perform fault troubleshooting on the power grid to be detected according to the failure probability of each node and each edge of the power grid to be detected.

2. A power grid fault detection method according to claim 1, characterized in that: Constructing the fault detection model includes: Obtain historical operation data and fault simulation operation data of the power grid to be tested; Constructing a training sample set according to the historical operation data and the fault simulation operation data; wherein the training sample set includes a power matrix sample, an adjacency matrix sample and a fault label matrix; An initial fault detection model is constructed, and the initial fault detection model is trained with the training sample set, and when a preset convergence condition is reached, a fault detection model is generated.

3. A power grid fault detection method as claimed in claim 2, characterized in that: The fault detection model includes: a point cloud dynamic graph convolutional network layer, a graph convolutional network layer, a normalization layer and an output layer; The step of inputting the power matrix and the adjacency matrix of the power grid to be detected into the fault detection model so that the fault detection model outputs the fault probability matrix of the power grid to be detected comprises: Inputting the power matrix and adjacency matrix of the power grid to be detected into the point cloud dynamic graph convolution network layer, so that the point cloud dynamic graph convolution network layer performs edge convolution and dynamic graph update on the power matrix and the adjacency matrix, and extracts the adjacency relationship between each node of the power grid to be detected and the local features of each node; Inputting the power matrix and the adjacency matrix of the power grid to be detected into the graph convolutional network layer, so that the graph convolutional network layer determines the weight relationship between the nodes of the power grid to be detected and extracts the global features of the nodes of the power grid to be detected; The adjacency relationship between the nodes of the power grid to be detected, the local characteristics of the nodes, the weight relationship between the nodes and the global characteristics of the nodes are transmitted to the normalization layer, so that the normalization layer determines the fault probability matrix of the power grid to be detected according to the adjacency relationship between the nodes of the power grid to be detected, the local characteristics of the nodes, the weight relationship between the nodes and the global characteristics of the nodes, and transmits it to the output layer; The output layer outputs the fault probability matrix of the power grid to be detected.

4. A power grid fault detection method as claimed in claim 3, characterized in that: The step of constructing a training sample set according to the historical operation data and the fault simulation operation data includes: According to the historical operation data, the power samples of each node of the power grid to be tested and the adjacency relationship of each node are obtained under the normal operation state; According to the fault simulation operation data, power samples of each node of the power grid to be detected, the adjacency relationship of each node and the fault label of each fault position in the power grid to be detected are obtained under the fault state; wherein the fault label is used to identify the fault node and the fault edge; Construct power matrix samples, adjacency matrix samples and fault label matrix according to power samples of each node in normal operation, adjacency relations of each node in normal operation, power samples of each node in fault state, adjacency relations of each node in fault state and fault labels of each fault position in the power grid to be detected; A training sample set is constructed based on power matrix samples, adjacency matrix samples and fault label matrix.

5. A power grid fault detection device, characterized in that: include: Graph model building module, data acquisition module and fault detection module; The graph model building module is used to obtain the topological structure and power flow information of the power grid to be detected; Constructing a graph model of the power grid to be detected according to the topological structure and power flow information of the power grid to be detected; wherein the graph model of the power grid to be detected includes a plurality of nodes representing devices in the power grid to be detected and a plurality of edges representing connection relationships between devices in the power grid to be detected; The data acquisition module is used to acquire the power matrix and adjacency matrix of the power grid to be detected according to the power grid graph model to be detected; The fault detection module is used to input the power matrix and adjacency matrix of the power grid to be detected into the fault detection model, so that the fault detection model outputs the fault probability matrix of the power grid to be detected; determine the fault probability of each node and the fault probability of each edge of the power grid to be detected according to the fault probability matrix, so that the operation and maintenance personnel can perform fault troubleshooting on the power grid to be detected according to the failure probability of each node and the failure probability of each edge of the power grid to be detected.

6. A power grid fault detection device as claimed in claim 5, characterized in that: Constructing the fault detection model includes: Obtain historical operation data and fault simulation operation data of the power grid to be tested; Constructing a training sample set according to the historical operation data and the fault simulation operation data; wherein the training sample set includes a power matrix sample, an adjacency matrix sample and a fault label matrix; An initial fault detection model is constructed, and the initial fault detection model is trained with the training sample set, and when a preset convergence condition is reached, a fault detection model is generated.

7. A power grid fault detection device as claimed in claim 6, characterized in that: The fault detection model includes: a point cloud dynamic graph convolutional network layer, a graph convolutional network layer, a normalization layer and an output layer; The step of inputting the power matrix and the adjacency matrix of the power grid to be detected into the fault detection model so that the fault detection model outputs the fault probability matrix of the power grid to be detected comprises: Inputting the power matrix and adjacency matrix of the power grid to be detected into the point cloud dynamic graph convolution network layer, so that the point cloud dynamic graph convolution network layer performs edge convolution and dynamic graph update on the power matrix and the adjacency matrix, and extracts the adjacency relationship between each node of the power grid to be detected and the local features of each node; Inputting the power matrix and the adjacency matrix of the power grid to be detected into the graph convolutional network layer, so that the graph convolutional network layer determines the weight relationship between the nodes of the power grid to be detected and extracts the global features of the nodes of the power grid to be detected; The adjacency relationship between the nodes of the power grid to be detected, the local characteristics of the nodes, the weight relationship between the nodes and the global characteristics of the nodes are transmitted to the normalization layer, so that the normalization layer determines the fault probability matrix of the power grid to be detected according to the adjacency relationship between the nodes of the power grid to be detected, the local characteristics of the nodes, the weight relationship between the nodes and the global characteristics of the nodes, and transmits it to the output layer; The output layer outputs the fault probability matrix of the power grid to be detected.

8. A power grid fault detection device as claimed in claim 7, characterized in that: The step of constructing a training sample set according to the historical operation data and the fault simulation operation data includes: According to the historical operation data, the power samples of each node of the power grid to be tested and the adjacency relationship of each node are obtained under the normal operation state; According to the fault simulation operation data, power samples of each node of the power grid to be detected, the adjacency relationship of each node and the fault label of each fault position in the power grid to be detected are obtained under the fault state; wherein the fault label is used to identify the fault node and the fault edge; Construct power matrix samples, adjacency matrix samples and fault label matrix according to power samples of each node in normal operation, adjacency relations of each node in normal operation, power samples of each node in fault state, adjacency relations of each node in fault state and fault labels of each fault position in the power grid to be detected; A training sample set is constructed based on power matrix samples, adjacency matrix samples and fault label matrix.

9. A terminal device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, a power grid fault detection method as claimed in any one of claims 1 to 4 is implemented.

10. A storage medium, characterized in that: The storage medium includes a stored computer program, wherein when the computer program is executed, the device where the storage medium is located is controlled to execute a power grid fault detection method according to any one of claims 1 to 4.

Citation Information

Patent Citations

  • Power distribution network line fault positioning method and device

    CN111537831A

  • Power information physical system modeling method based on alternating current power flow model

    CN112491034A

  • Grid fault detection method and device based on graph neural network

    CN115733730A

  • Offshore wind plant current collection line fault positioning method adaptive to topological change

    CN118393280A

  • Systems and methods for detecting unauthorized or suspicious financial activity

    US20210334822A1

Cited By

  • Fault detection method and device applied to power distribution network and electronic equipment

    CN119805101A

  • Power distribution network load flow calculation method, device and equipment based on deep learning

    CN120320329A