A power grid fault detection method and device, a terminal device and a storage medium

Abstract

This invention discloses a power grid fault detection method, apparatus, terminal device, and storage medium. The method includes: acquiring the topology and power flow information of the power grid to be detected; constructing a graph model of the power grid to be detected based on the topology and power flow information; wherein the graph model of the power grid to be detected includes several nodes representing devices in the power grid to be detected and several edges representing the connection relationships between devices in the power grid to be detected; acquiring the power matrix and adjacency matrix of the power grid to be detected based on the graph model; inputting the power matrix and adjacency matrix of the power grid to be detected into a fault detection model, so that the fault detection model outputs a fault probability matrix of the power grid to be detected; determining the fault probability of each node and each edge of the power grid to be detected based on the fault probability matrix, so that maintenance personnel can troubleshoot the power grid to be detected based on the fault probabilities of each node and each edge.

Description

Technical Field

[0001] This invention relates to the field of power grid fault detection technology, and in particular to a power grid fault detection method, device, terminal equipment and storage medium. Background Technology

[0002] With the continuous expansion of power system scale and the increasing complexity of power grid structure, the reliability and safe operation of the power grid face severe challenges. In existing technologies, fault location usually relies on power grid topology analysis and current flow monitoring methods. These methods typically determine the location of potential faults in the power grid based on the presence of discontinuities, open circuits, or abnormal current flow. However, these existing methods usually identify a specific branch or node with a fault, which is then provided to maintenance personnel for troubleshooting. This approach is difficult to comprehensively investigate potential faults in the power grid. When a fault occurs at a previously undetected point, repeated investigation is required, leading to prolonged fault handling time and increasing the uncertainty and risk of power grid operation. Summary of the Invention

[0003] This invention provides a power grid fault detection method, device, 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 topology and power flow information of the power grid to be tested;

[0006] A graph model of the power grid to be tested is constructed based on the topology and power flow information of the power grid to be tested; wherein, the graph model of the power grid to be tested includes several nodes representing devices in the power grid to be tested and several edges representing the connection relationships between devices in the power grid to be tested;

[0007] The power matrix and adjacency matrix of the power grid to be detected are obtained based on the power grid graph model to be detected;

[0008] The power matrix and 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.

[0009] The fault probability matrix is ​​used to determine the fault probability of each node and each edge of the power grid to be tested, so that maintenance personnel can troubleshoot the power grid to be tested based on the fault probability of each node and each edge.

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

[0011] Acquire historical operational data and fault simulation operational data of the power grid under test;

[0012] A training sample set is constructed 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;

[0013] An initial fault detection model is constructed, and the initial fault detection model is trained using the training sample set. When a preset convergence condition is met, 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 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, includes:

[0016] The power matrix and adjacency matrix of the power grid to be detected are input 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 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] The power matrix and adjacency matrix of the power grid to be detected are input into the graph convolutional network layer, so that the graph convolutional network layer determines the weight relationship between each node of the power grid to be detected and extracts the global features of each node of the power grid to be detected.

[0018] The adjacency relationships between nodes of the power grid under test, the local features of each node, the weight relationships between each node, and the global features of each node are transmitted to the normalization layer. This enables the normalization layer to determine the fault probability matrix of the power grid under test based on the adjacency relationships between nodes, the local features of each node, the weight relationships between each node, and the global features of each node, and then transmit it to the output layer.

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

[0020] Furthermore, the step of constructing a training sample set based on the historical operating data and the fault simulation operating data includes:

[0021] Based on historical operating data, obtain the power samples of each node in the power grid to be tested and the adjacency relationship of each node under normal operating conditions;

[0022] Based on the fault simulation operation data, the power samples of each node in the power grid to be tested, the adjacency relationship of each node, and the fault labels of each fault location in the power grid to be tested are obtained under the fault state; wherein, the fault labels are used to identify fault nodes and fault edges.

[0023] Based on the power samples of each node under normal operating conditions, the adjacency relationships of each node under normal operating conditions, the power samples of each node under fault conditions, the adjacency relationships of each node under fault conditions, and the fault labels of each fault location in the power grid to be detected, a power matrix sample, an adjacency matrix sample, and a fault label matrix are constructed.

[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 embodiments, the present invention provides corresponding apparatus embodiments;

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

[0027] The graph model construction module is used to obtain the topology and power flow information of the power grid to be tested; and to construct a graph model of the power grid to be tested based on the topology and power flow information of the power grid to be tested; wherein, the graph model of the power grid to be tested includes several nodes representing devices in the power grid to be tested and several edges representing the connection relationships between devices in the power grid to be tested;

[0028] The data acquisition module is used to acquire the power matrix and adjacency matrix of the power grid to be detected based on 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 tested into the fault detection model, so that the fault detection model outputs the fault probability matrix of the power grid to be tested; and to determine the fault probability of each node and each edge of the power grid to be tested based on the fault probability matrix, so that maintenance personnel can troubleshoot the power grid to be tested based on the fault probability of each node and each edge of the power grid to be tested.

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

[0031] Acquire historical operational data and fault simulation operational data of the power grid under test;

[0032] A training sample set is constructed 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;

[0033] An initial fault detection model is constructed, and the initial fault detection model is trained using the training sample set. When a preset convergence condition is met, 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 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, includes:

[0036] The power matrix and adjacency matrix of the power grid to be detected are input 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 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] The power matrix and adjacency matrix of the power grid to be detected are input into the graph convolutional network layer, so that the graph convolutional network layer determines the weight relationship between each node of the power grid to be detected and extracts the global features of each node of the power grid to be detected.

[0038] The adjacency relationships between nodes of the power grid under test, the local features of each node, the weight relationships between each node, and the global features of each node are transmitted to the normalization layer. This enables the normalization layer to determine the fault probability matrix of the power grid under test based on the adjacency relationships between nodes, the local features of each node, the weight relationships between each node, and the global features of each node, and then transmit it to the output layer.

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

[0040] Furthermore, the step of constructing a training sample set based on the historical operating data and the fault simulation operating data includes:

[0041] Based on historical operating data, obtain the power samples of each node in the power grid to be tested and the adjacency relationship of each node under normal operating conditions;

[0042] Based on the fault simulation operation data, the power samples of each node in the power grid to be tested, the adjacency relationship of each node, and the fault labels of each fault location in the power grid to be tested are obtained under the fault state; wherein, the fault labels are used to identify fault nodes and fault edges.

[0043] Based on the power samples of each node under normal operating conditions, the adjacency relationships of each node under normal operating conditions, the power samples of each node under fault conditions, the adjacency relationships of each node under fault conditions, and the fault labels of each fault location in the power grid to be detected, a power matrix sample, an adjacency matrix sample, and a fault label matrix are constructed.

[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. When the processor executes the computer program, it implements a power grid fault detection method as described in the above-described embodiment of the invention.

[0046] Another embodiment of the present invention provides a storage medium including a stored computer program, wherein, when the computer program is executed, it controls the device where the storage medium is located to execute the power grid fault detection method described in the above-described embodiment of the invention.

[0047] The following benefits can be obtained by implementing the present invention:

[0048] This invention provides a power grid fault detection method, apparatus, terminal equipment, and storage medium. The fault detection method constructs a graph model of the power grid by acquiring its topology and power flow information, and then determines the power matrix and adjacency matrix of the power grid based on the constructed graph model. The graph model includes nodes representing devices in the power grid and edges representing the connections between devices. The power matrix and adjacency matrix of the power grid are input into the fault detection model, which outputs a fault probability matrix. The fault probability matrix determines the fault probability of each node and each edge in the power grid, and provides these probabilities to power grid maintenance personnel. This allows maintenance personnel to troubleshoot power grid faults based on the fault probabilities. By determining the fault probabilities between all nodes and edges in the power grid, the comprehensiveness and efficiency of power grid fault detection are improved, thereby enhancing the safety and stability of power grid operation. Attached Figure Description

[0049] Figure 1 This is a flowchart illustrating a power grid fault detection method according to an embodiment of the present invention.

[0050] Figure 2 This is a schematic diagram of the structure of a fault detection model provided in an embodiment of the present invention.

[0051] Figure 3 This is a schematic diagram of the structure of a power grid fault detection device provided in an embodiment of the present invention. Detailed Implementation

[0052] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0053] like Figure 1 The image shows a power grid fault detection method provided by an embodiment of the present invention, comprising:

[0054] Step S1: Obtain the topology and power flow information of the power grid to be tested;

[0055] Step S2: Construct a graph model of the power grid to be tested based on the topology and power flow information of the power grid to be tested; wherein, the graph model of the power grid to be tested includes several nodes representing devices in the power grid to be tested and several edges representing the connection relationships between devices in the power grid to be tested;

[0056] Step S3: Obtain the power matrix and adjacency matrix of the power grid to be detected based on the power grid graph model to be detected;

[0057] Step S4: 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;

[0058] Step S5: Determine the fault probability of each node and each edge of the power grid to be tested based on the fault probability matrix, so that maintenance personnel can troubleshoot the power grid to be tested based on the fault probability of each node and each edge of the power grid to be tested.

[0059] For steps S1 and S2, the topology and power flow information of the power grid to be tested are obtained. The topology and power flow information of the power grid to be tested are encoded into a graph model. In the graph model, nodes represent devices in the power grid to be tested, such as generators, transformers, and lines; edges represent the connection relationships between devices.

[0060] For step S3, based on the constructed graph model of the power grid to be tested, the power matrix and adjacency matrix of the power grid to be tested 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 in the power grid to be tested; the adjacency matrix contains the connection relationship between each node.

[0061] For step S4, the power matrix and 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. For example, the power matrix PQ (including active power P and reactive power Q, with a size of N x 2, where N is the total number of nodes in 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 the fault probability matrix L (with a size of N x N), where each element in the fault probability matrix represents the probability of a fault occurring between the 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 step of constructing a training sample set based on the historical operating data and fault simulation operating data includes: obtaining power samples and adjacency relationships of each node in the power grid to be detected under normal operating conditions based on the historical operating data; obtaining power samples, adjacency relationships, and fault labels of each fault location in the power grid to be detected under fault conditions based on the fault simulation operating data; wherein the fault labels are used to identify fault nodes and fault edges; constructing a power matrix sample, an adjacency matrix sample, and a fault label matrix based on the power samples, adjacency relationships, and fault label matrix of each node under normal operating conditions, the power samples, adjacency relationships, and fault labels of each fault location in the power grid to be detected; and constructing a training sample set based on the power matrix sample, adjacency matrix sample, and fault label matrix.

[0064] Specifically, historical operating data and fault simulation operating data of the power grid to be tested are acquired. Based on the historical operating data, node power samples and adjacency relationships between nodes in the power grid under normal operating conditions are obtained. Based on the fault simulation operating data, node power samples, adjacency relationships between nodes, and fault labels corresponding to each fault state in the power grid to be tested are obtained. It should be noted that fault labels are used to identify fault nodes and fault edges in each simulation data. To ensure that the trained fault detection model can comprehensively cover possible fault situations in the power grid, the selected fault simulation data should cover as many possible fault situations in the power grid as possible. Based on the power samples, adjacency relationships, and fault labels of each node under normal operating conditions, fault states, and fault locations in the power grid to be tested, power matrix samples, adjacency matrix samples, and fault label matrices are constructed to obtain the training sample set. An initial fault detection model is constructed. This initial fault detection model should be a graph neural network to meet the requirement of extracting the graph structure features of the power grid to be tested. Based on this training sample set, the initial fault detection model is trained. In each training iteration, the power matrix samples and adjacency matrix samples are used as inputs, and the predicted fault label matrix is ​​used as the 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 obtaining the loss, the network parameters of the initial fault detection model are updated using the backpropagation algorithm, and the loss function is minimized using a gradient descent optimization algorithm, such as the Adam algorithm. During training, the model performance is monitored using performance metrics on the validation set, such as prediction accuracy and F1 score. When the performance metrics on the validation set no longer improve significantly or reach the preset number of training epochs (i.e., the preset convergence condition mentioned above), the training of the initial fault detection model is stopped, and a new 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 step of inputting 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 a fault probability matrix of the power grid to be detected, includes: inputting the power matrix and 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 adjacency matrix, extracting the adjacency relationships between nodes of the power grid to be detected and the local features of each node; ... The power matrix and adjacency matrix are input into a graph convolutional network layer, enabling the graph convolutional network layer to determine the weight relationships between nodes in the power grid to be detected and extract the global features of each node in the power grid to be detected. The adjacency relationships between nodes in the power grid to be detected, the local features of each node, the weight relationships between each node, and the global features of each node are then transmitted to a normalization layer. The normalization layer determines the fault probability matrix of the power grid to be detected based on the adjacency relationships between nodes in the power grid to be detected, the local features of each node, the weight relationships between each node, 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, such as Figure 2The diagram shows the structure of the fault detection model provided by this invention. The fault detection model mainly includes: a point cloud dynamic graph convolutional network layer (DGCNN), a graph convolutional network layer (GCN, GraphSage, and feature-direction MLP in the diagram), a normalization layer (LayerNorm1() in the diagram), and an output layer (MLP). After inputting the power matrix and adjacency matrix of the power grid to be detected into the fault detection model, the point cloud dynamic graph convolutional layer (DGCNN) first performs edge convolution (EdgeConv) and dynamic graph update on the power matrix and adjacency matrix to extract the adjacency relationships between nodes in the power grid to be detected and the local features of each node. Using DGCNN as the first processing layer improves the feature embedding capability of the fault detection model for graph data after processing the power matrix and adjacency matrix. Edge convolution refers to performing convolution operations on the edges of the K nearest neighbor graph of each node to better learn the similarity relationships between nodes; dynamic graph update means that each layer of the network reconstructs the K nearest neighbor graph based on the output feature map of the previous layer. Furthermore, a graph convolutional network layer (GCN, GraphSage, and feature-direction MLP in the diagram) is used to determine the weight relationships between nodes in the power grid to be detected and to extract the global features of each node. In this network layer, the first part consists of GCN and a self-attention mechanism. By combining GCN with the self-attention mechanism, global features in the power matrix and adjacency matrix can be further extracted, and the different importance of nodes can be considered simultaneously, determining the weight relationships between nodes based on their importance. The second part consists of the GraphSage inductive graph neural network framework. The local and global features of the nodes processed by GCN and the self-attention mechanism constitute the node feature matrix. GraphSage generates the embedded representation of the nodes by sampling and aggregating the features of neighboring nodes. The third part consists of a feature-direction MLP, which further extracts features from the data processed by GraphSage and transmits all extracted features, adjacency relationships, and weight relationships to the normalization layer LayerNorm1(). This allows LayerNorm1() to normalize the global and local features of nodes, eliminating differences in feature value ranges and accelerating model convergence. Preferably, to improve model stability and training efficiency, the fault detection model also includes another normalization layer (i.e., the connection MLP, ReLU(), Dropout(0.2), and LayerNorm2() parts in the figure). After further feature extraction, ReLU activation function, and regularization processing after the normalization layer LayerNorm1(), LayerNorm2() normalizes the node features again, thereby further improving model stability and training efficiency.Finally, the output layer MLP generates a failure probability matrix containing the failure probability of each node and each edge.

[0067] For step S5, determine the fault probability of each node and each edge of the power grid to be tested based on the fault probability matrix, and inform the operation and maintenance personnel so that they can start a comprehensive fault investigation of the power grid to be tested based on the fault probability of the node or edge with the high fault probability.

[0068] To better illustrate the technical effects achievable by this application, the following practical examples are provided.

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

[0070]

[0071]

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

[0073] Let the adjacency matrix A be [[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] Inputting the power matrix PQ and the adjacency matrix A into the fault detection model will output 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 can be interpreted as follows: A value of 0.00 in the first row and first column indicates that node 1 itself has no fault probability. A value of 0.70 in the first row and second column indicates that the fault probability of the line from node 1 to node 2 is 70%. A value of 0.00 in the first row and third column indicates that the fault probability of the line from node 1 to node 3 is 0%. A value of 0.20 in the first row and fourth column indicates that the fault probability of the line from node 1 to node 4 is 20%. Similarly, the fault probability interpretations for the second to fourth rows can be derived. That is, the lines to be checked, in descending order of fault probability, are: the line from node 4 to node 3 (fault probability 80%, corresponding to the third row and fourth column and the fourth row and third column), the line from node 1 to node 2 (fault probability 70%, corresponding to the first row and second row and first column), the line from node 2 to node 3 (fault probability 60%, corresponding to the second row and third column and the third row and second column), and the line from node 1 to node 4 (fault probability 20%, corresponding to the first row and fourth column and the fourth row and first column). This data is provided to operations and maintenance personnel, who then conduct a thorough investigation, thereby improving the comprehensiveness of troubleshooting.

[0080] Based on the above method embodiments, the present invention provides corresponding apparatus embodiments.

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

[0082] The graph model construction module is used to obtain the topology and power flow information of the power grid to be tested; and to construct a graph model of the power grid to be tested based on the topology and power flow information of the power grid to be tested; wherein, the graph model of the power grid to be tested includes several nodes representing devices in the power grid to be tested and several edges representing the connection relationships between devices in the power grid to be tested;

[0083] The data acquisition module is used to acquire the power matrix and adjacency matrix of the power grid to be detected based on 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 tested into the fault detection model, so that the fault detection model outputs the fault probability matrix of the power grid to be tested; and to determine the fault probability of each node and each edge of the power grid to be tested based on the fault probability matrix, so that maintenance personnel can troubleshoot the power grid to be tested based on the fault probability of each node and each edge of the power grid to be tested.

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

[0086] Acquire historical operational data and fault simulation operational data of the power grid under test;

[0087] A training sample set is constructed 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;

[0088] An initial fault detection model is constructed, and the initial fault detection model is trained using the training sample set. When a preset convergence condition is met, 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 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, includes:

[0091] The power matrix and adjacency matrix of the power grid to be detected are input 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 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] The power matrix and adjacency matrix of the power grid to be detected are input into the graph convolutional network layer, so that the graph convolutional network layer determines the weight relationship between each node of the power grid to be detected and extracts the global features of each node of the power grid to be detected.

[0093] The adjacency relationships between nodes of the power grid under test, the local features of each node, the weight relationships between each node, and the global features of each node are transmitted to the normalization layer. This enables the normalization layer to determine the fault probability matrix of the power grid under test based on the adjacency relationships between nodes, the local features of each node, the weight relationships between each node, and the global features of each node, and then transmit it to the output layer.

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

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

[0096] Based on historical operating data, obtain the power samples of each node in the power grid to be tested and the adjacency relationship of each node under normal operating conditions;

[0097] Based on the fault simulation operation data, the power samples of each node in the power grid to be tested, the adjacency relationship of each node, and the fault labels of each fault location in the power grid to be tested are obtained under the fault state; wherein, the fault labels are used to identify fault nodes and fault edges.

[0098] Based on the power samples of each node under normal operating conditions, the adjacency relationships of each node under normal operating conditions, the power samples of each node under fault conditions, the adjacency relationships of each node under fault conditions, and the fault labels of each fault location in the power grid to be detected, a power matrix sample, an adjacency matrix sample, and a fault label matrix are constructed.

[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 illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

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

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

[0103] One embodiment of the present invention provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements a power grid fault detection method according to any one of the present invention.

[0104] The terminal device can be a desktop computer, laptop, handheld computer, or cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0105] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the terminal device, connecting all parts of the terminal device via various interfaces and lines.

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

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

[0108] One embodiment of the present invention provides a storage medium including a stored computer program, wherein, when the computer program is executed, it controls the device where the storage medium is located to execute a power grid fault detection method according to any one of the present invention.

[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 a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0110] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for detecting power grid faults, characterized in that, include: Obtain the topology and power flow information of the power grid to be tested; A graph model of the power grid to be tested is constructed based on the topology and power flow information of the power grid to be tested; wherein, the graph model of the power grid to be tested includes several nodes representing devices in the power grid to be tested and several edges representing the connection relationships between devices in the power grid to be tested; The power matrix and adjacency matrix of the power grid to be detected are obtained based on the power grid graph model to be detected; The power matrix and 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. The fault probability matrix is ​​used to determine the fault probability of each node and each edge of the power grid to be tested, so that maintenance personnel can troubleshoot the power grid to be tested based on the fault probability of each node and each edge of the power grid to be tested. 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 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, includes: The power matrix and adjacency matrix of the power grid to be detected are input 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 adjacency matrix, and extracts the adjacency relationship between each node of the power grid to be detected and the local features of each node. The power matrix and adjacency matrix of the power grid to be detected are input into the graph convolutional network layer, so that the graph convolutional network layer determines the weight relationship between each node of the power grid to be detected and extracts the global features of each node of the power grid to be detected. The adjacency relationships between nodes of the power grid under test, the local features of each node, the weight relationships between each node, and the global features of each node are transmitted to the normalization layer. This enables the normalization layer to determine the fault probability matrix of the power grid under test based on the adjacency relationships between nodes, the local features of each node, the weight relationships between each node, and the global features of each node, and then transmit it to the output layer. The output layer outputs the fault probability matrix of the power grid to be tested.

2. The power grid fault detection method as described in claim 1, characterized in that, Constructing the fault detection model includes: Acquire historical operational data and fault simulation operational data of the power grid under test; A training sample set is constructed 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; An initial fault detection model is constructed, and the initial fault detection model is trained using the training sample set. When a preset convergence condition is met, a fault detection model is generated.

3. The power grid fault detection method as described in claim 2, characterized in that, The step of constructing a training sample set based on the historical operating data and fault simulation operating data includes: Based on historical operating data, obtain the power samples of each node in the power grid to be tested and the adjacency relationship of each node under normal operating conditions; Based on the fault simulation operation data, the power samples of each node in the power grid to be tested, the adjacency relationship of each node, and the fault labels of each fault location in the power grid to be tested are obtained under the fault state; wherein, the fault labels are used to identify fault nodes and fault edges. Based on the power samples of each node under normal operating conditions, the adjacency relationships of each node under normal operating conditions, the power samples of each node under fault conditions, the adjacency relationships of each node under fault conditions, and the fault labels of each fault location in the power grid to be detected, a power matrix sample, an adjacency matrix sample, and a fault label matrix are constructed. A training sample set is constructed based on power matrix samples, adjacency matrix samples, and fault label matrix.

4. A power grid fault detection device, characterized in that, include: The module includes a graphical model building module, a data acquisition module, and a fault detection module. The graph model construction module is used to obtain the topology and power flow information of the power grid to be detected; A graph model of the power grid to be tested is constructed based on the topology and power flow information of the power grid to be tested; wherein, the graph model of the power grid to be tested includes several nodes representing devices in the power grid to be tested and several edges representing the connection relationships between devices in the power grid to be tested; The data acquisition module is used to acquire the power matrix and adjacency matrix of the power grid to be detected based on 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 a fault probability matrix of the power grid to be detected; and to determine the fault probability of each node and each edge of the power grid to be detected based on the fault probability matrix, so that maintenance personnel can perform fault investigation on the power grid to be detected based on the fault probability of each node and each edge of the power grid to be detected; 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 adjacency matrix of the power grid to be detected into the fault detection model, so that the fault detection model outputs a fault probability matrix of the power grid to be detected, includes: inputting the power matrix and 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 outputs a fault probability matrix of the power grid to be detected ... The dynamic graph convolutional network layer performs edge convolution and dynamic graph update on the power matrix and adjacency matrix to extract the adjacency relationships between nodes of the power grid to be detected and the local features of each node. The power matrix and adjacency matrix of the power grid to be detected are input into the graph convolutional network layer so that the graph convolutional network layer determines the weight relationship between nodes of the power grid to be detected and extracts the global features of each node. The adjacency relationships between nodes of the power grid to be detected, the local features of each node, the weight relationship between each node, 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 based on the adjacency relationships between nodes of the power grid to be detected, the local features of each node, the weight relationship between each node, 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.

5. The power grid fault detection device as described in claim 4, characterized in that, Constructing the fault detection model includes: Acquire historical operational data and fault simulation operational data of the power grid under test; A training sample set is constructed 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; An initial fault detection model is constructed, and the initial fault detection model is trained using the training sample set. When a preset convergence condition is met, a fault detection model is generated.

6. The power grid fault detection device as described in claim 5, characterized in that, The step of constructing a training sample set based on the historical operating data and fault simulation operating data includes: Based on historical operating data, obtain the power samples of each node in the power grid to be tested and the adjacency relationship of each node under normal operating conditions; Based on the fault simulation operation data, the power samples of each node in the power grid to be tested, the adjacency relationship of each node, and the fault labels of each fault location in the power grid to be tested are obtained under the fault state; wherein, the fault labels are used to identify fault nodes and fault edges. Based on the power samples of each node under normal operating conditions, the adjacency relationships of each node under normal operating conditions, the power samples of each node under fault conditions, the adjacency relationships of each node under fault conditions, and the fault labels of each fault location in the power grid to be detected, a power matrix sample, an adjacency matrix sample, and a fault label matrix are constructed. A training sample set is constructed based on power matrix samples, adjacency matrix samples, and fault label matrix.

7. A terminal device, characterized in that, The device includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements a power grid fault detection method as described in any one of claims 1 to 3.

8. A storage medium, characterized in that, The storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device where the storage medium is located to perform a power grid fault detection method as described in any one of claims 1 to 3.

Images