Grid fault positioning method, system, equipment and product based on graph neural network

Abstract

The invention discloses a power grid fault positioning method, system, device and product based on a graph neural network, and belongs to the technical field of fault localization, and the method comprises the steps: firstly obtaining an actual power grid line, and generating a graph model; respectively acquiring telemetry data and telecommand data corresponding to each electrical operation device, and correspondingly storing the telemetry data and the telecommand data into each node; training an attention-based space-time diagram convolutional network model and a diagram neural network model, and extracting power grid node features; inputting the power grid node features into a one-dimensional convolutional neural network for node coverage, so as to map the power grid node features to power grid sideline features; and finally, outputting the probability value that each sideline in the sideline set is a fault section, and selecting the corresponding sideline with the maximum probability value as a power grid fault positioning result. The scheme provided by the invention not only has strong robustness for topological change, high-resistance faults and data incomplete scenes, but also remarkably improves the accuracy of power distribution network fault positioning.

Description

Technical Field

[0001] The present invention belongs to the technical field of fault location, and particularly relates to a power grid fault location method, system, device and product based on a graph neural network. Background Art

[0002] In the complex network architecture of the power system, as a key link directly facing users, the safe and stable operation of the distribution network is directly related to the normal order of social production and life. Quickly and accurately locating the fault section is not only an important prerequisite for ensuring power supply reliability, but also a core measure to improve the power grid's emergency response ability and reduce power outage losses. However, traditional fault location methods often face many challenges. On the one hand, these methods mostly rely on manual experience or simple threshold judgment, with limited ability to identify complex and variable fault characteristics, low fault tolerance, and prone to misjudgment or missed judgment under noise interference or signal distortion, resulting in a significant reduction in the accuracy of fault location. On the other hand, traditional methods are inefficient in processing large-scale data, and the fault location process takes a long time, making it difficult to meet the high real-time requirements of modern power systems.

[0003] With the rapid development of artificial intelligence technology, deep learning algorithms have shown great potential in the field of fault diagnosis. However, most deep learning models still have significant limitations when applied to the fault location of distribution networks. Especially in the face of the actual situation where the power grid topology changes frequently, these models often lack sufficient generalization ability. Due to the high dependence of deep learning models on the distribution characteristics of training data, once the power grid structure changes, such as the addition of new lines or equipment transformation, the original model may experience a decline in performance or even failure due to its inability to adapt to new data characteristics.

[0004] Therefore, based on the above deficiencies, how to provide a power grid fault location method, system, device and product based on a graph neural network that can improve the accuracy of distribution network fault location has become an urgent problem to be solved. Summary of the Invention

[0005] The purpose of the present invention is to provide a power grid fault location method, system, device and product based on a graph neural network to solve the problems of serious influence of eddy current or residual magnetism and long scanning time existing in the prior art.

[0006] To achieve the above purpose, the present invention adopts the following technical solutions:

[0007] In the first aspect, the present invention provides a power grid fault location method based on a graph neural network, which includes:

[0008] Obtain the actual power grid lines and generate a graph model based on the actual power grid lines. The graph model includes a node set and an edge set. The node set is used to represent each electrical operating device, and the edge set is used to represent the power line sections;

[0009] Obtain the telemetry data and telecontrol data corresponding to each electrical operating device respectively, and store the telemetry data and telecontrol data corresponding to each electrical operating device into each node of the node set respectively. The telemetry data includes the voltage signal and current time series signal of the corresponding electrical operating device, and the telecontrol data includes the discrete action information of the corresponding electrical operating device;

[0010] Based on the node set, train an attention-based spatio-temporal graph convolutional network model and a graph neural network model, and process the node set according to the attention-based spatio-temporal graph convolutional network model and the graph neural network model to extract power grid node features;

[0011] Input the power grid node features into a one-dimensional convolutional neural network, and use a one-dimensional convolutional kernel for node coverage to map the power grid node features to power grid edge features;

[0012] Reduce the dimension of the power grid edge features, and output the probability value that each edge in the edge set is a fault section, and select the corresponding edge with the largest probability value as the power grid fault location result.

[0013] In a possible design, training an attention-based spatio-temporal graph convolutional network model and a graph neural network model based on the node set includes:

[0014] Obtain the telemetry data of each node in the node set;

[0015] Form a first training set with the telemetry data of each node;

[0016] Based on the first training set, train an attention-based spatio-temporal graph convolutional network model;

[0017] Obtain the telecontrol data of each node in the node set;

[0018] Form a second training set with the telecontrol data of each node;

[0019] Based on the second training set, train a graph neural network model.

[0020] In a possible design, processing the node set according to the attention-based spatio-temporal graph convolutional network model and the graph neural network model to extract power grid node features includes:

[0021] Based on the spatio-temporal graph convolutional network model, project the node set into the temporal attention space. According to the projection of the node set in the temporal attention space, calculate the temporal attention weights of each node in the node set respectively, and obtain the temporal attention matrix based on the temporal attention weights of each node;

[0022] Based on the spatio-temporal graph convolutional network model, project the node set into the spatial attention space. According to the projection of the node set in the spatial attention space, calculate the spatial attention weights of each node in the node set respectively, and obtain the spatial attention matrix based on the spatial attention weights of each node;

[0023] Based on the graph convolution method, process the spatial attention matrix to extract the spatial features of each node in the node set;

[0024] Based on the temporal attention matrix, extract the temporal features of each node in the node set;

[0025] Based on the spatial features and temporal features of each node in the node set, obtain the spatio-temporal fusion features of each node in the node set.

[0026] In one possible design, after obtaining the spatio-temporal fusion features of each node in the node set based on the spatial features and temporal features of each node in the node set, it further includes:

[0027] Based on the graph neural network model, construct a multi-head graph attention mechanism for the node set;

[0028] Use the multi-head graph attention mechanism to calculate the graph attention weights of each node in the node set;

[0029] Process each node in the node set according to the graph attention weights of each node in the node set, so as to calculate the graph attention features through each attention head in the multi-head graph attention mechanism;

[0030] Concatenate the graph attention features calculated by each attention head with the spatio-temporal fusion features of each node in the node set to obtain the grid node features.

[0031] In one possible design, based on the spatio-temporal graph convolutional network model, project the node set into the temporal attention space. According to the projection of the node set in the temporal attention space, calculate the temporal attention weights of each node in the node set respectively, and obtain the temporal attention matrix based on the temporal attention weights of each node, including:

[0032] Use a learnable time linear transformation matrix to linearly process the telemetry data of each node in the node set, so as to project the node set into the time attention space;

[0033] Calculate the time similarity matrix between each node according to the time attention space projection of the node set;

[0034] Add a time bias term matrix to the time similarity matrix between each node, and use an activation function to generate a time attention weight matrix;

[0035] Use a learnable time weight matrix to linearly process the time attention weight matrix, so as to calculate the time attention weights of each node in the node set respectively, and obtain a time attention matrix based on the time attention weights of each node;

[0036] Correspondingly, based on the spatio-temporal graph convolutional network model, project the node set into the spatial attention space, calculate the spatial attention weights of each node in the node set respectively according to the spatial attention space projection of the node set, and obtain a spatial attention matrix based on the spatial attention weights of each node, including:

[0037] Use the time attention matrix to perform weighted processing on each node in the node set to obtain a time weighted feature matrix;

[0038] Use a learnable spatial linear transformation matrix to linearly process each node in the node set, so as to project the node set into the spatial attention space;

[0039] Calculate the spatial similarity matrix between each node according to the spatial attention space projection of the node set;

[0040] Add a spatial bias term matrix to the spatial similarity matrix between each node, and use an activation function to generate a spatial attention weight matrix;

[0041] Use a learnable spatial weight matrix to linearly process the spatial attention weight matrix, so as to calculate the spatial attention weights of each node in the node set respectively, and obtain a spatial attention matrix based on the spatial attention weights of each node.

[0042] In a possible design, based on the graph convolution method, process the spatial attention matrix to extract the spatial features of each node in the node set, including:

[0043] Perform normalization processing on the spatial attention matrix to obtain a normalized spatial attention matrix;

[0044] Using Chebyshev polynomials, perform graph convolution processing on the normalized spatial attention matrix to extract the spatial features of each node in the node set;

[0045] Correspondingly, based on the temporal attention matrix, extracting the temporal features of each node in the node set includes:

[0046] Perform two-dimensional convolution processing on the temporal attention matrix to extract the temporal features of each node in the node set.

[0047] In a possible design, input the power grid node features into a one-dimensional convolutional neural network and use a one-dimensional convolutional kernel to cover the nodes to map the power grid node features to power grid edge features, including:

[0048] Based on the graph model, perform topological arrangement on the power grid node features to form a feature input matrix;

[0049] Perform a convolution operation on the feature input matrix;

[0050] Use the convolutional kernel to slide and cover each adjacent node in the feature input matrix to extract the power grid edge features corresponding to the power grid node features.

[0051] In a second aspect, the present invention provides a power grid fault location system based on a graph neural network, including:

[0052] A data acquisition unit, configured to obtain an actual power grid line and generate a graph model according to the actual power grid line, where the graph model includes a node set and an edge set, the node set is used to represent each electrical operation device, and the edge set is used to represent a power line section;

[0053] A data storage unit, configured to respectively obtain the telemetry data and telecontrol data corresponding to each electrical operation device, and respectively store the telemetry data and telecontrol data corresponding to each electrical operation device into each node of the node set, where the telemetry data includes a voltage signal and a current time series signal of the corresponding electrical operation device, and the telecontrol data includes discrete action information of the corresponding electrical operation device;

[0054] A feature extraction unit, configured to train an attention-based spatio-temporal graph convolution network model and a graph neural network model based on the node set, and process the node set according to the attention-based spatio-temporal graph convolution network model and the graph neural network model to extract power grid node features; and is further configured to input the power grid node features into a one-dimensional convolutional neural network and use a one-dimensional convolutional kernel to cover the nodes to map the power grid node features to power grid edge features;

[0055] The fault location unit is used to reduce the dimension of the grid edge features, output the probability values of each edge in the edge set being a fault section, and select the corresponding edge with the largest probability value as the grid fault location result.

[0056] In a third aspect, the present invention provides an electronic device, including a memory, a processor, and a transceiver that are communicatively connected in sequence. Among them, the memory is used to store computer programs, the transceiver is used to send and receive messages, and the processor is used to read the computer programs and execute the grid fault location method based on the graph neural network as described in the first aspect or any possible design in the first aspect.

[0057] In a fourth aspect, the present invention provides a computer program product containing instructions. When the instructions run on a computer, the computer is made to execute the grid fault location method based on the graph neural network as described in the first aspect or any possible design in the first aspect.

[0058] Beneficial effects: The present invention provides a grid fault location method, system, device, and product based on a graph neural network, belonging to the technical field of fault location. First, the actual grid lines are obtained, and a graph model is generated according to the actual grid lines; then, the telemetry data and telecontrol data corresponding to each electrical operation device are obtained respectively, and the telemetry data and telecontrol data corresponding to each electrical operation device are respectively stored into each node of the node set; based on the node set, a spatio-temporal graph convolutional network model and a graph neural network model based on attention are trained, and according to the spatio-temporal graph convolutional network model and the graph neural network model based on attention, the node set is processed to extract grid node features; then, the grid node features are input into a one-dimensional convolutional neural network, and node coverage is performed using one-dimensional convolutional kernels to map the grid node features to grid edge features; finally, the dimension of the grid edge features is reduced, and the probability values of each edge in the edge set being a fault section are output, and the corresponding edge with the largest probability value is selected as the grid fault location result. The solution provided by the present invention not only has strong robustness to scenarios of topological changes, high-resistance faults, and incomplete data, but also significantly improves the accuracy of distribution network fault location. Description of the Drawings

[0059] Figure 1 It is a schematic flowchart of the steps of the grid fault location method based on the graph neural network provided by the embodiment of the present invention. Detailed Embodiments

[0060] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the present invention in combination with the accompanying drawings and the description of the embodiments or the prior art. Obviously, the following description of the structures of the accompanying drawings is only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can also be obtained based on these drawings. It should be noted here that the description of these embodiments is used to help understand the present invention, but does not constitute a limitation to the present invention.

[0061] It should be understood that although terms such as first and second may be used herein to describe various units, these units should not be limited by these terms. These terms are only used to distinguish one unit from another. For example, the first unit can be called the second unit, and similarly, the second unit can be called the first unit, without departing from the scope of the exemplary embodiments of the present invention.

[0062] It should be understood that for the term "and / or" that may appear in this article, it is only a description of the association relationship of the associated objects, indicating that three relationships can exist. For example, A and / or B can represent: A exists alone, B exists alone, and A and B exist simultaneously; for the term " / and" that may appear in this article, it is a description of another association object relationship, indicating that two relationships can exist. For example, A / and B can represent: A exists alone, and A and B exist alone; in addition, for the character " / " that may appear in this article, generally, it means that the front and rear associated objects are an "or" relationship.

[0063] Embodiment:

[0064] See Figure 1 As shown, for the power grid fault location method based on graph neural network provided in this embodiment, the running steps can but are not limited to the following steps S100 to S500:

[0065] S100. Obtain the actual power grid lines and generate a graph model according to the actual power grid lines. Among them, the graph model includes a node set and an edge set. The node set is used to represent each electrical operating device, and the edge set is used to represent the power line sections.

[0066] S200. Respectively obtain the telemetry data and telecontrol data corresponding to each electrical operating device, and respectively store the telemetry data and telecontrol data corresponding to each electrical operating device into each node of the node set. Among them, the telemetry data includes the voltage signal and current time series signal of the corresponding electrical operating device, and the telecontrol data includes the discrete action information of the corresponding electrical operating device.

[0067] S300. Based on the node set, train an attention-based spatio-temporal graph convolutional network model and a graph neural network model, and process the node set according to the attention-based spatio-temporal graph convolutional network model and the graph neural network model to extract power grid node features;

[0068] Wherein, in a possible implementation manner, in step S300, training an attention-based spatio-temporal graph convolutional network model and a graph neural network model based on the node set includes:

[0069] Obtain the telemetry data of each node in the node set;

[0070] Form a first training set with the telemetry data of each node;

[0071] Train an attention-based spatio-temporal graph convolutional network model based on the first training set;

[0072] Obtain the telecontrol data of each node in the node set;

[0073] Form a second training set with the telecontrol data of each node;

[0074] Train a graph neural network model based on the second training set.

[0075] Wherein, in a possible implementation manner, in step S300, processing the node set according to the attention-based spatio-temporal graph convolutional network model and the graph neural network model to extract power grid node features includes:

[0076] Based on the spatio-temporal graph convolutional network model, project the node set into the temporal attention space, calculate the temporal attention weights of each node in the node set respectively according to the temporal attention space projection of the node set, and obtain a temporal attention matrix based on the temporal attention weights of each node;

[0077] Based on the spatio-temporal graph convolutional network model, project the node set into the spatial attention space, calculate the spatial attention weights of each node in the node set respectively according to the spatial attention space projection of the node set, and obtain a spatial attention matrix based on the spatial attention weights of each node;

[0078] Process the spatial attention matrix based on the graph convolution method to extract the spatial features of each node in the node set;

[0079] Extract the temporal features of each node in the node set based on the temporal attention matrix;

[0080] Based on the spatial features and temporal features of each node in the node set, the spatio-temporal fusion features of each node in the node set are obtained.

[0081] Among them, based on the spatio-temporal graph convolutional network model, the node set is projected into the temporal attention space, and according to the temporal attention space projection of the node set, the temporal attention weights of each node in the node set are calculated respectively, and a temporal attention matrix is obtained based on the temporal attention weights of each node, including:

[0082] Using a learnable temporal linear transformation matrix, the telemetry data of each node in the node set is linearly processed to project the node set into the temporal attention space;

[0083] According to the temporal attention space projection of the node set, a temporal similarity matrix between each node is calculated;

[0084] A temporal bias term matrix is added to the temporal similarity matrix between each node, and an activation function is used to generate a temporal attention weight matrix;

[0085] Using a learnable temporal weight matrix, the temporal attention weight matrix is linearly processed to calculate the temporal attention weights of each node in the node set respectively, and a temporal attention matrix is obtained based on the temporal attention weights of each node;

[0086] Correspondingly, based on the spatio-temporal graph convolutional network model, the node set is projected into the spatial attention space, and according to the spatial attention space projection of the node set, the spatial attention weights of each node in the node set are calculated respectively, and a spatial attention matrix is obtained based on the spatial attention weights of each node, which includes:

[0087] Using the temporal attention matrix, the nodes in the node set are weighted to obtain a temporal weighted feature matrix;

[0088] Using a learnable spatial linear transformation matrix, the nodes in the node set are linearly processed to project the node set into the spatial attention space;

[0089] According to the spatial attention space projection of the node set, a spatial similarity matrix between each node is calculated;

[0090] A spatial bias term matrix is added to the spatial similarity matrix between each node, and an activation function is used to generate a spatial attention weight matrix;

[0091] Using a learnable spatial weight matrix, perform linear processing on the spatial attention weight matrix to calculate the spatial attention weights of each node in the node set respectively, and obtain a spatial attention matrix based on the spatial attention weights of each node.

[0092] Moreover, based on the graph convolution method, process the spatial attention matrix to extract the spatial features of each node in the node set, including:

[0093] Perform normalization processing on the spatial attention matrix to obtain a normalized spatial attention matrix;

[0094] Use Chebyshev polynomials to perform graph convolution processing on the normalized spatial attention matrix to extract the spatial features of each node in the node set;

[0095] Correspondingly, based on the temporal attention matrix, extracting the temporal features of each node in the node set includes:

[0096] Perform two-dimensional convolution processing on the temporal attention matrix to extract the temporal features of each node in the node set.

[0097] Wherein, in a possible implementation manner, in step S300, after obtaining the spatio-temporal fusion features of each node in the node set based on the spatial features and temporal features of each node in the node set, it further includes:

[0098] Based on the graph neural network model, construct a multi-head graph attention mechanism for the node set;

[0099] Use the multi-head graph attention mechanism to calculate the graph attention weights of each node in the node set;

[0100] Process each node in the node set according to the graph attention weights of each node in the node set, so as to calculate graph attention features through each attention head in the multi-head graph attention mechanism;

[0101] Perform feature splicing on the graph attention features calculated by each attention head and the spatio-temporal fusion features of each node in the node set to obtain grid node features.

[0102] S400. Input the grid node features into a one-dimensional convolutional neural network, and use a one-dimensional convolutional kernel for node coverage to map the grid node features to grid edge features;

[0103] Wherein, in step S400, inputting the grid node features into a one-dimensional convolutional neural network and using a one-dimensional convolutional kernel for node coverage to map the grid node features to grid edge features includes:

[0104] Based on the graph model, topologically arrange the grid node features to form a feature input matrix;

[0105] Perform a convolution operation on the feature input matrix;

[0106] Use the convolution kernel to slide and cover each adjacent node in the feature input matrix to extract the grid line features corresponding to the grid node features.

[0107] S500. Reduce the dimension of the grid line features, and output the probability values of each line in the line set being a fault section, and select the corresponding line with the largest probability value as the grid fault location result.

[0108] Through the grid fault location method based on graph neural network in this embodiment, first obtain the actual grid lines and generate a graph model; then respectively obtain the telemetry data and telecontrol data corresponding to each electrical operation device, and store them into each node correspondingly; then train an attention-based spatio-temporal graph convolutional network model and a graph neural network model, and extract the grid node features; then input the grid node features into a one-dimensional convolutional neural network for node coverage to map the grid node features to the grid line features; finally, output the probability values of each line in the line set being a fault section, and select the corresponding line with the largest probability value as the grid fault location result. The solution provided by the present invention not only has strong robustness to scenarios of topological changes, high-resistance faults and incomplete data, but also significantly improves the accuracy of distribution network fault location.

[0109] The second aspect of this embodiment provides a hardware system for implementing the grid fault location method based on graph neural network described in the first aspect of the embodiment, including:

[0110] An acquisition unit, configured to acquire nuclear magnetic scan parameters, where the nuclear magnetic scan parameters include the number of scan repetitions.

[0111] A parameter determination unit, configured to determine, according to the nuclear magnetic scan parameters, the flip time corresponding to each scan dimension and the echo train length corresponding to each flip time in multiple scan dimensions, as well as multiple gradient rotation angle values and the total scan time corresponding to each gradient rotation angle value, where the echo train lengths corresponding to each flip time decrease in sequence.

[0112] A scanning unit configured to generate a gradient magnetic field corresponding to each gradient rotation angle value based on a plurality of gradient rotation angle values, and in each gradient magnetic field, perform pulse excitation on an object to be measured according to the number of scanning repetitions, the flip times corresponding to each scanning dimension, and the echo train lengths corresponding to each flip time, so as to collect nuclear magnetic resonance signals of the object to be measured in each gradient magnetic field after the pulse excitation, wherein the gradient of the gradient magnetic field corresponding to any gradient rotation angle value is constant, and the magnetic field duration is greater than the total scanning time corresponding to the any gradient rotation angle value.

[0113] A nuclear magnetic two-dimensional spectrum generation unit configured to generate a nuclear magnetic two-dimensional spectrum of the object to be measured according to each nuclear magnetic resonance signal.

[0114] For the working process, working details and technical effects of the system provided in this embodiment, reference may be made to the first aspect of the embodiment, which will not be elaborated herein.

[0115] A third aspect of this embodiment provides an electronic device, including: a memory, a processor, and a transceiver that are communicatively connected in sequence, wherein the memory is configured to store a computer program, the transceiver is configured to send and receive messages, and the processor is configured to read the computer program and execute the power grid fault location method based on a graph neural network as described in the first aspect of the embodiment.

[0116] Specifically, the memory may include, but is not limited to, a random access memory (RAM), a read only memory (ROM), a flash memory, a first input first output (FIFO) memory, and / or a first in last out (FILO) memory, etc.; specifically, the processor may include one or more processing cores, such as a 4-core processor or an 8-core processor. The processor may be implemented in at least one hardware form of a digital signal processing (DSP), a field-programmable gate array (FPGA), or a programmable logic array (PLA). At the same time, the processor may also include a main processor and a coprocessor. The main processor is a processor for processing data in the wake state, also known as a central processing unit (CPU); the coprocessor is a low-power processor for processing data in the standby state.

[0117] In some embodiments, the processor may be integrated with a GPU (Graphics Processing Unit), and the GPU is responsible for rendering and drawing the content to be displayed on the display screen. For example, the processor may not be limited to a microprocessor of the STM32F105 series, a reduced instruction set computer (RISC) microprocessor, an X86 architecture processor, or a processor integrated with an embedded neural-network processing unit (NPU); the transceiver may be, but is not limited to, a Wi-Fi wireless transceiver, a Bluetooth wireless transceiver, a General Packet Radio Service (GPRS) wireless transceiver, a ZigBee (low-power local area network protocol based on the IEEE 802.15.4 standard) wireless transceiver, a 3G transceiver, a 4G transceiver, and / or a 5G transceiver, etc. In addition, the device may also include, but is not limited to, a power module, a display screen, and other necessary components.

[0118] For the working process, working details, and technical effects of the electronic device provided in this embodiment, reference may be made to the first aspect of the embodiment, which will not be elaborated here.

[0119] This embodiment provides a computer program product containing instructions. When the instructions run on a computer, the computer is caused to execute the power grid fault location method based on a graph neural network as described in the first aspect of the embodiment, where the computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.

[0120] Finally, it should be noted that the above are only preferred embodiments of the present invention and are not used to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A power grid fault location method based on graph neural network, characterized in that, Including: Obtain the actual power grid line and generate a graph model according to the actual power grid line, where the graph model includes a node set and an edge set, the node set is used to represent each electrical operation device, and the edge set is used to represent the power line section; Obtain the telemetry data and telecontrol data corresponding to each electrical operation device respectively, and store the telemetry data and telecontrol data corresponding to each electrical operation device into each node of the node set respectively, where the telemetry data includes the voltage signal and current time series signal of the corresponding electrical operation device, and the telecontrol data includes the discrete action information of the corresponding electrical operation device; Based on the node set, train a spatio-temporal graph convolutional network model and a graph neural network model based on attention, and process the node set according to the spatio-temporal graph convolutional network model and the graph neural network model based on attention to extract the power grid node features; Input the power grid node features into a one-dimensional convolutional neural network, and use a one-dimensional convolutional kernel for node coverage to map the power grid node features to the power grid edge features; Reduce the dimension of the power grid edge features, and output the probability value that each edge in the edge set is a fault section, and select the corresponding edge with the largest probability value as the power grid fault location result.

2. The power grid fault location method based on a graph neural network according to claim 1, wherein Based on the node set, training a spatio-temporal graph convolutional network model and a graph neural network model based on attention includes: Obtain the telemetry data of each node in the node set; Form a first training set with the telemetry data of each node; Based on the first training set, train a spatio-temporal graph convolutional network model based on attention; Obtain the telecontrol data of each node in the node set; Form a second training set with the telecontrol data of each node; Based on the second training set, train a graph neural network model.

3. The power grid fault location method based on a graph neural network according to claim 1, wherein According to the spatio-temporal graph convolutional network model and the graph neural network model based on attention, processing the node set to extract the power grid node features includes: Based on the spatio-temporal graph convolutional network model, project the node set into the time attention space, and calculate the time attention weights of each node in the node set respectively according to the time attention space projection of the node set, and obtain the time attention matrix based on the time attention weights of each node; Based on the spatio-temporal graph convolutional network model, project the node set into the space attention space, and calculate the space attention weights of each node in the node set respectively according to the space attention space projection of the node set, and obtain the space attention matrix based on the space attention weights of each node; Based on the graph convolution method, process the space attention matrix to extract the space features of each node in the node set; Based on the time attention matrix, extract the time features of each node in the node set; Based on the space features and time features of each node in the node set, obtain the spatio-temporal fusion features of each node in the node set.

4. The power grid fault location method based on a graph neural network according to claim 3, characterized in that After obtaining the spatio-temporal fusion features of each node in the node set based on the space features and time features of each node in the node set, it further includes: Based on the graph neural network model, construct a multi-head graph attention mechanism for the node set; Use the multi-head graph attention mechanism to calculate the graph attention weights of each node in the node set; Process each node in the node set according to the graph attention weights of each node in the node set, so as to calculate the graph attention features through each attention head in the multi-head graph attention mechanism; Perform feature splicing on the graph attention features calculated by each attention head and the spatio-temporal fusion features of each node in the node set to obtain power grid node features.

5. The power grid fault location method based on a graph neural network according to claim 3, wherein, Based on the spatio-temporal graph convolutional network model, project the node set into the temporal attention space. According to the temporal attention space projection of the node set, calculate the temporal attention weights of each node in the node set respectively, and obtain the temporal attention matrix based on the temporal attention weights of each node, including: Use a learnable temporal linear transformation matrix to linearly process the telemetry data of each node in the node set to project the node set into the temporal attention space; Calculate the temporal similarity matrix between each node according to the temporal attention space projection of the node set; Add a temporal bias term matrix to the temporal similarity matrix between each node and use an activation function to generate a temporal attention weight matrix; Use a learnable temporal weight matrix to linearly process the temporal attention weight matrix to calculate the temporal attention weights of each node in the node set respectively, and obtain the temporal attention matrix based on the temporal attention weights of each node; Correspondingly, based on the spatio-temporal graph convolutional network model, project the node set into the spatial attention space. According to the spatial attention space projection of the node set, calculate the spatial attention weights of each node in the node set respectively, and obtain the spatial attention matrix, which includes: Use the temporal attention matrix to perform weighted processing on each node in the node set to obtain a temporally weighted feature matrix; Use a learnable spatial linear transformation matrix to linearly process each node in the node set to project the node set into the spatial attention space; Calculate the spatial similarity matrix between each node according to the spatial attention space projection of the node set; Add a spatial bias term matrix to the spatial similarity matrix between each node and use an activation function to generate a spatial attention weight matrix; Use a learnable spatial weight matrix to linearly process the spatial attention weight matrix to calculate the spatial attention weights of each node in the node set respectively, and obtain the spatial attention matrix based on the spatial attention weights of each node.

6. The power grid fault location method based on a graph neural network according to claim 3, characterized in that, Based on the graph convolution method, process the spatial attention matrix to extract the spatial features of each node in the node set, including: Perform normalization processing on the spatial attention matrix to obtain a normalized spatial attention matrix; Using Chebyshev polynomials, perform graph convolution processing on the normalized spatial attention matrix to extract the spatial features of each node in the node set; Correspondingly, based on the temporal attention matrix, extracting the temporal features of each node in the node set includes: Performing two-dimensional convolution processing on the temporal attention matrix to extract the temporal features of each node in the node set.

7. The power grid fault location method based on a graph neural network according to claim 1, wherein Inputting the power grid node features into a one-dimensional convolutional neural network and using a one-dimensional convolutional kernel to cover the nodes to map the power grid node features to power grid edge features, including: Based on the graph model, performing topological arrangement on the power grid node features to form a feature input matrix; Performing a convolution operation on the feature input matrix; Using the convolutional kernel to slide and cover each adjacent node in the feature input matrix to extract the power grid edge features corresponding to the power grid node features.

8. A power grid fault location system based on a graph neural network, characterized in that, Includes: A data acquisition unit for acquiring the actual power grid lines and generating a graph model according to the actual power grid lines, where the graph model includes a node set and an edge set, the node set is used to represent each electrical operation device, and the edge set is used to represent the power line sections; A data storage unit for respectively acquiring the telemetry data and telecontrol data corresponding to each electrical operation device and storing the telemetry data and telecontrol data corresponding to each electrical operation device into each node of the node set respectively, where the telemetry data includes the voltage signal and current time series signal of the corresponding electrical operation device, and the telecontrol data includes the discrete action information of the corresponding electrical operation device; A feature extraction unit for training a spatio-temporal graph convolutional network model and a graph neural network model based on attention based on the node set, and processing the node set according to the spatio-temporal graph convolutional network model and the graph neural network model based on attention to extract power grid node features; and also for inputting the power grid node features into a one-dimensional convolutional neural network and using a one-dimensional convolutional kernel to cover the nodes to map the power grid node features to power grid edge features; A fault location unit for reducing the dimension of the power grid edge features and outputting the probability value that each edge in the edge set is a fault section, and selecting the corresponding edge with the largest probability value as the power grid fault location result.

9. An electronic device, characterized in that, Includes: A memory, a processor, and a transceiver that are communicatively connected in sequence, where the memory is used to store a computer program, the transceiver is used to send and receive messages, and the processor is used to read the computer program and execute the power grid fault location method based on a graph neural network according to any one of claims 1 to 7.

10. A computer program product comprising instructions, characterized in that, When the instruction runs on a computer, causing the computer to execute the power grid fault location method based on a graph neural network according to any one of claims 1 to 7.

Images