A method, device, equipment, medium, and product for fault location in distribution networks based on target spatiotemporal graph neural networks.

Abstract

This invention discloses a method, apparatus, device, medium, and product for fault location in distribution networks based on a target spatiotemporal graph neural network. The method includes: acquiring the spatial connection state matrix of each node in the distribution network and voltage data of each node within a preset time period; reconstructing the voltage data of each node in the distribution network within the preset time period to obtain reconstructed voltage data of each node; determining a similarity coefficient matrix between nodes based on the reconstructed voltage data of each node; inputting the spatial connection state matrix of each node, the reconstructed voltage data of each node, and the similarity coefficient matrix between nodes into a target spatiotemporal graph neural network to obtain the fault state of each node; and determining the fault section in the distribution network based on the fault state of each node. Through the technical solution of this invention, the accuracy and efficiency of fault location in distribution networks can be improved.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of communication technology, and in particular to a method, apparatus, equipment, medium and product for fault location in power distribution networks based on a target spatiotemporal graph neural network. Background Technology

[0002] The reliability of power supply in a distribution network is closely related to the power quality of users. Accurately and quickly locating faults is of great significance for ensuring power supply reliability and the safe operation of the distribution network.

[0003] Existing fault location methods for distribution networks are mainly divided into traditional methods and optimized algorithms. Traditional fault location methods include impedance methods, traveling wave methods, and matrix methods. Although traditional fault location methods are simple in principle and fast in calculation, they have poor applicability and low fault tolerance in distribution networks after the integration of new energy sources. Optimized algorithms exhibit redundant iterations near the end of the calculation, thus fault location based on optimized algorithms suffers from low accuracy and low solution efficiency. Summary of the Invention

[0004] This invention provides a method, device, equipment, medium, and product for fault location in power distribution networks based on a target spatiotemporal graph neural network, in order to improve the accuracy and efficiency of fault location in power distribution networks.

[0005] According to one aspect of the present invention, a method for fault location in a distribution network based on a target spatiotemporal graph neural network is provided, comprising:

[0006] Obtain the spatial connection state matrix of each node in the distribution network and the voltage data of each node within a preset time period;

[0007] The voltage data of each node in the power distribution network within a preset time period are reconstructed to obtain the reconstructed voltage data of each node.

[0008] Based on the reconstructed voltage data of each node, determine the similarity coefficient matrix between nodes;

[0009] The spatial connection state matrix of each node in the power distribution network, the reconstructed voltage data of each node, and the similarity coefficient matrix between nodes are input into the target spatiotemporal graph neural network to obtain the fault state of each node.

[0010] Based on the fault status of each node in the distribution network, the fault sections in the distribution network are determined.

[0011] According to another aspect of the present invention, a distribution network fault location device based on a target spatiotemporal graph neural network is provided, the distribution network fault location device based on the target spatiotemporal graph neural network comprising:

[0012] The voltage data acquisition module is used to acquire the spatial connection status matrix of each node in the distribution network and the voltage data of each node within a preset time period.

[0013] The reconstructed voltage data determination module is used to reconstruct the voltage data of each node in the distribution network within a preset time period to obtain the reconstructed voltage data of each node.

[0014] The similarity coefficient matrix determination module is used to determine the similarity coefficient matrix between nodes based on the reconstructed voltage data of each node.

[0015] The fault state determination module of the node is used to input the spatial connection state matrix of each node in the distribution network, the reconstructed voltage data of each node, and the similarity coefficient matrix between nodes into the target spatiotemporal graph neural network to obtain the fault state of each node.

[0016] The fault section determination module is used to determine the fault section in the distribution network based on the fault status of each node in the distribution network.

[0017] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0018] At least one processor; and

[0019] A memory communicatively connected to the at least one processor; wherein,

[0020] The memory stores a computer program that can be executed by the at least one processor, which enables the at least one processor to execute the distribution network fault location method based on the target spatiotemporal graph neural network according to any embodiment of the present invention.

[0021] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions, the computer instructions being configured to cause a processor to execute and implement the distribution network fault location method based on a target spatiotemporal graph neural network as described in any embodiment of the present invention.

[0022] According to another aspect of the present invention, a computer program product is provided, which, when executed by a processor, implements the distribution network fault location method based on a target spatiotemporal graph neural network as described in any of the embodiments of the present invention.

[0023] This invention first reconstructs the voltage data of each node in the distribution network within a preset time period to obtain reconstructed voltage data for each node. The reconstructed voltage data allows for the extraction of key features related to distribution network faults. Then, the spatial connection state matrix of each node, the reconstructed voltage data of each node, and the similarity coefficient matrix between nodes are input into a target spatiotemporal graph neural network to obtain the fault state of each node. Based on the fault state of each node in the distribution network, the fault section in the distribution network is determined. This method maximizes the use of spatiotemporal information through the target spatiotemporal graph neural network to extract fault location features, solving the problems of poor applicability, low fault tolerance, low accuracy, and long location time of current fault location methods in active distribution networks. This leads to more accurate and faster fault section location. Fast and accurate fault location technology can effectively shorten the fault investigation time of the distribution network and reduce power outage duration, which is of great significance for the safe and reliable operation of the distribution system.

[0024] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0025] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 This is a flowchart of a power distribution network fault location method based on a target spatiotemporal graph neural network in an embodiment of the present invention;

[0027] Figure 2 This is a schematic diagram of the structure of a temporal convolutional layer in an embodiment of the present invention;

[0028] Figure 3 This is a schematic diagram of the structure of a power distribution network fault location device based on a target spatiotemporal graph neural network in an embodiment of the present invention;

[0029] Figure 4 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0030] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. 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 should fall within the scope of protection of the present invention.

[0031] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0032] It is understood that before using the technical solutions disclosed in the various embodiments of this disclosure, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in this disclosure in an appropriate manner in accordance with relevant laws and regulations, and user authorization should be obtained.

[0033] Example 1

[0034] Figure 1 This is a flowchart illustrating a distribution network fault location method based on a target spatiotemporal graph neural network, provided in an embodiment of the present invention. This embodiment is applicable to distribution network fault location. The method can be executed by the distribution network fault location device based on the target spatiotemporal graph neural network in this embodiment. This device can be implemented in software and / or hardware, such as... Figure 1 As shown, the method specifically includes the following steps:

[0035] S110: Obtain the spatial connection status matrix of each node in the distribution network and the voltage data of each node within a preset time period.

[0036] In this embodiment, the nodes in the power distribution network include: substations, switching stations, monitoring nodes of power distribution stations, generators, loads, and other control elements.

[0037] In this embodiment, the size of the spatial connection state matrix of each node in the distribution network is N×N, where N is the total number of nodes in the distribution network. The spatial connection state matrix of each node in the distribution network is used to characterize the control connection state between each node in the distribution network. For example, when node i and node j are connected, the corresponding element (A) of the matrix... i A j If the value is 1, then the value is 0; otherwise, the value is 0.

[0038] In this embodiment, the preset time period can be a pre-set time period, and the length of the preset time period is not limited in this embodiment of the invention.

[0039] In this embodiment, the voltage data of each node in the distribution network within a preset time period can be obtained by acquiring the voltage data of each node in the distribution network at each moment within the preset time period.

[0040] S120, reconstruct the voltage data of each node in the distribution network within a preset time period to obtain the reconstructed voltage data of each node.

[0041] In this embodiment, the reconstructed voltage data of the node includes: residual terms and at least one first-order intrinsic mode function.

[0042] In this embodiment, the voltage data of each node in the distribution network within a preset time period is reconstructed to obtain the reconstructed voltage data of each node. This can be achieved by: decomposing the voltage data of the node within the preset time period to obtain multi-order intrinsic modulus functions and residual terms; filtering the intrinsic modulus functions based on the correlation coefficients between the intrinsic modulus functions and the voltage data of the node within the preset time period to obtain target intrinsic modulus functions; and reconstructing the voltage data based on the residual terms and target intrinsic modulus functions to obtain the reconstructed voltage data of the node.

[0043] Optionally, the voltage data of each node in the distribution network within a preset time period is reconstructed to obtain the reconstructed voltage data of each node, including:

[0044] The voltage data of each node within a preset time period are reconstructed to obtain the reconstructed voltage data of each node.

[0045] In this embodiment, the reconstruction operation includes: decomposing the voltage data of the node within a preset time period to obtain multi-order intrinsic modulus functions and residual terms; obtaining the correlation coefficients between each order intrinsic modulus function and the voltage data of the node within the preset time period; filtering each order intrinsic modulus function based on the correlation coefficients between each order intrinsic modulus function and the voltage data of the node within the preset time period to obtain a target intrinsic modulus function; and reconstructing the voltage data based on the residual terms and the target intrinsic modulus function to obtain the reconstructed voltage data of the node.

[0046] In this embodiment, the method for decomposing the voltage data of a node within a preset time period to obtain multi-order intrinsic mode functions and residual terms can be as follows: The voltage data is decomposed into multi-order intrinsic mode functions using an improved adaptive noise complete set empirical mode decomposition to reflect the local characteristics of the voltage data in each frequency range. The decomposed voltage data can be represented as the superposition of multi-order intrinsic mode functions and residual terms, for example:

[0047]

[0048] Where U(t) represents the voltage data of the node within a preset time period. For the first The intrinsic modulus function is given by R(t), where R(t) is the residual term.

[0049] In this embodiment, the correlation coefficient between intrinsic modulus functions of each order and the voltage data of nodes within a preset time period can be obtained as follows: obtain the covariance operator between intrinsic modulus functions of each order and the voltage data of nodes within a preset time period; obtain the standard deviation of intrinsic modulus functions of each order and the standard deviation of voltage data of nodes within a preset time period; and determine the correlation coefficient between intrinsic modulus functions of each order and voltage data of nodes within a preset time period based on the covariance operator, the standard deviation of intrinsic modulus functions of each order, and the standard deviation of voltage data of nodes within a preset time period.

[0050] In this embodiment, the intrinsic mode functions of each order are screened based on the correlation coefficient between the intrinsic mode functions of each order and the voltage data of the nodes within a preset time period to obtain the target intrinsic mode function. This can be done by: determining the intrinsic mode functions with correlation coefficients greater than or equal to the correlation coefficient threshold as the target intrinsic mode functions; or, determining the intrinsic mode function with the largest correlation coefficient as the target intrinsic mode function; or, if all correlation coefficients are less than the correlation coefficient threshold, then determining the intrinsic mode function with the largest correlation coefficient as the target intrinsic mode function.

[0051] In this embodiment, the correlation coefficient threshold can be a pre-set correlation coefficient, for example, the correlation coefficient threshold can be 0.2.

[0052] In this embodiment, the reconstructed voltage data of a node is obtained by reconstructing the voltage data based on the residual term and the target intrinsic mode function. The formula for reconstructing the voltage data is as follows:

[0053] U new (t)=∑E new (t)+R(t);

[0054] In this embodiment, U new (t) represents the reconstructed voltage data, E newR(t) represents all target intrinsic mode functions with correlation coefficients greater than the correlation coefficient threshold, and R(t) is the residual term. It should be noted that if there are multiple target intrinsic mode functions with correlation coefficients greater than the correlation coefficient threshold, the sum of the multiple target intrinsic mode functions is added to the residual term to obtain the reconstructed voltage data.

[0055] The technical solution provided in this embodiment can better mine the amplitude characteristics of fault voltage data by reconstructing voltage data. Compared with the original voltage data, the reconstructed voltage data has more obvious fault information.

[0056] Optionally, obtain the correlation coefficients between the intrinsic mode functions of each order and the voltage data of the nodes within a preset time period, including:

[0057] Obtain the covariance operator of the intrinsic mode functions of each order and the voltage data of the nodes within a preset time period.

[0058] In this embodiment, the method for obtaining the covariance operator of each order intrinsic mode function and the voltage data of the node within a preset time period can be as follows: the covariance operator of each order intrinsic mode function and the voltage data of the node within a preset time period is calculated through the covariance operator function.

[0059] Obtain the standard deviation of the intrinsic modulus functions of each order and the standard deviation of the voltage data of the nodes within a preset time period.

[0060] Based on the covariance operator, the standard deviation of each order of intrinsic modulus function, and the standard deviation of the voltage data of the node within a preset time period, the correlation coefficient between each order of intrinsic modulus function and the voltage data of the node within a preset time period is determined.

[0061] In this embodiment, the correlation coefficient between the intrinsic mode functions of each order and the voltage data of the node within the preset time period is determined based on the covariance operator, the standard deviation of each order intrinsic mode function, and the standard deviation of the voltage data of the node within the preset time period. This can be achieved by: determining the correlation coefficient between the intrinsic mode functions of each order and the voltage data of the node within the preset time period based on the covariance operator, the standard deviation of each order intrinsic mode function, the standard deviation of the voltage data of the node within the preset time period, and a first formula. The first formula is as follows:

[0062]

[0063] In this embodiment, For the first The correlation coefficient between the intrinsic eigenmode function and the voltage data of the nodes within a preset time period, where cov(.) is the covariance operator function. and σ U They represent The standard deviation of U(t).

[0064] In this embodiment, the correlation coefficient can be the Pearson correlation coefficient. The Pearson correlation coefficient is used to measure the linear relationship between two variables.

[0065] The technical solution provided in this embodiment uses the Pearson correlation coefficient to select the intrinsic mode function that is strongly correlated with the original voltage data for voltage data reconstruction, thereby preserving the main fault characteristics of the voltage data.

[0066] S130, Based on the reconstructed voltage data of each node, determine the similarity coefficient matrix between nodes.

[0067] In this embodiment, the similarity coefficient matrix between nodes can be determined based on the reconstructed voltage data of each node as follows: the mean value of each reconstructed voltage of each node is determined based on the multiple reconstructed voltages of each node at each time point; and the similarity coefficient matrix between nodes is determined based on the multiple reconstructed voltages of each node at each time point and the mean value of each reconstructed voltage of each node.

[0068] Optionally, the reconstructed voltage data of the node includes: multiple reconstructed voltages of each node at each time point;

[0069] Based on the reconstructed voltage data of each node, a similarity coefficient matrix between nodes is determined, including:

[0070] Based on the multiple reconstruction voltages of each node at each time point, determine the average value of each reconstruction voltage of each node;

[0071] The similarity coefficient matrix between nodes is determined based on the multiple reconstruction voltages of each node at each time point and the mean of each reconstruction voltage of each node.

[0072] In this embodiment, the similarity coefficient matrix between nodes can be determined by the following method based on the multiple reconstruction voltages of each node at each time point and the mean of the multiple reconstruction voltages of each node: determine the product of the deviations and the product of squared deviations of each node based on the multiple reconstruction voltages of each node at each time point and the mean of the multiple reconstruction voltages of each node; determine the similarity coefficient matrix between nodes based on the product of the deviations and the product of squared deviations of each node.

[0073] In this embodiment, the similarity coefficient matrix between nodes can be determined based on the multiple reconstruction voltages of each node at each time point and the mean of the multiple reconstruction voltages of each node. This can be achieved by: determining the similarity coefficient matrix between nodes based on the multiple reconstruction voltages of each node at each time point, the mean of the multiple reconstruction voltages of each node, and a second formula. The second formula is as follows:

[0074]

[0075] In this embodiment, eij X is the similarity coefficient between node i and node j; i (p,q) represents matrix X n The reconstructed voltage data of the p-th phase (p=1,2,3 corresponding to phases A, B, and C respectively) of node i at the q-th time point (q=1,…,T, where T is a preset time period) is obtained. It is X i The mean matrix, X j (p,q) represents matrix X n The reconstructed voltage data of the p-th phase (p=1,2,3 corresponding to phases A, B, and C respectively) of node j at the q-th time point (q=1,…,T, where T is a preset time period) is obtained. It is X j The mean matrix, matrix X n =[U a U b U c ], where n is the node number, U a To determine the A-phase reconfiguration voltage of each node within a preset time period, U b To determine the B-phase reconfiguration voltage of each node within a preset time period, U c This refers to the C-phase reconfiguration voltage of each node within a preset time period.

[0076] S140, input the spatial connection state matrix of each node in the distribution network, the reconstructed voltage data of each node, and the similarity coefficient matrix between nodes into the target spatiotemporal graph neural network to obtain the fault state of each node.

[0077] In this embodiment, the method of inputting the spatial connection state matrix of each node in the distribution network, the reconstructed voltage data of each node, and the similarity coefficient matrix between nodes into the target spatiotemporal graph neural network to obtain the fault state of each node can be as follows: construct graph data based on the spatial connection state matrix of each node in the distribution network, the reconstructed voltage data of each node, and the similarity coefficient matrix between nodes; input the graph data into the target spatiotemporal graph neural network to obtain the fault state of each node.

[0078] In this embodiment, generators, loads, and other components of the distribution network are used as nodes, and the connecting lines between them are used as edges to form the topology of the distribution network. Based on the topology of the distribution network and the reconstructed voltage data of the nodes, graph data G{X, E, A} is constructed, where X is a matrix composed of the reconstructed voltage data of each node; E is a matrix composed of the similarity coefficients between nodes; and A is a matrix composed of the spatial connection states of each node.

[0079] In this embodiment, the target spatiotemporal graph neural network is obtained by iteratively training the neural network to be trained using the target sample set.

[0080] In this embodiment, the target spatiotemporal graph neural network can be an improved spatiotemporal graph neural network, which includes, from input to output, an attention module, a spatiotemporal convolution module, and an output module.

[0081] In this embodiment, graph data is constructed by combining the spatial connection state matrix of each node in the distribution network, the reconstructed voltage data of each node, and the similarity coefficient matrix between nodes. This maximizes the representation of spatiotemporal information in the data during construction, which is beneficial to improving the ability of the target spatiotemporal graph neural network to cope with the variable and complex topology of the active distribution network.

[0082] Optionally, the target spatiotemporal graph neural network includes, from input to output, an attention module, a spatiotemporal convolution module, and an output module.

[0083] The spatial connection state matrix of each node in the distribution network, the reconstructed voltage data of each node, and the similarity coefficient matrix between nodes are input into the target spatiotemporal graph neural network to obtain the fault state of each node, including:

[0084] The reconstructed voltage data of each node is input into the attention module to obtain the initial feature matrix of each node.

[0085] In this embodiment, the attention module includes, from input to output, a single-layer feedforward neural subnetwork and a nonlinear activation layer.

[0086] In this embodiment, the reconstructed voltage data of each node is input into the attention module to obtain the initial feature matrix of each node, including: inputting the reconstructed voltage data of each node into the single-layer feedforward neural subnetwork to obtain the attention coefficients between each node and all neighboring nodes; inputting the attention coefficients between each node and all neighboring nodes and the reconstructed voltage data of each node into the nonlinear activation layer to obtain the initial feature matrix of each node.

[0087] The initial feature matrix of each node, the similarity coefficient matrix between the nodes, and the spatial connection state matrix of each node are input into the spatiotemporal convolution module to obtain the voltage feature matrix of each node and the temporal feature matrix of the similarity coefficient between the nodes.

[0088] In this embodiment, the spatiotemporal convolution module includes, from input to output, a temporal convolutional layer and a spatial convolutional layer. It should be noted that the method for inputting the initial feature matrix of each node, the similarity coefficient matrix between nodes, and the spatial connection state matrix of each node into the spatiotemporal convolution module to obtain the voltage feature matrix of each node and the temporal feature matrix of the similarity coefficients between nodes can be as follows: inputting the initial feature matrix of each node and the similarity coefficient matrix between nodes into the temporal convolutional layer to obtain the temporal feature matrix of each node and the temporal feature matrix of the similarity coefficients between nodes; and inputting the temporal feature matrix of each node and the spatial connection state matrix of each node into the spatial convolutional layer to obtain the voltage feature matrix of each node.

[0089] The voltage feature matrix of each node and the time-domain feature matrix of the similarity coefficient between nodes are input into the output module to obtain the fault status of each node.

[0090] In this embodiment, the output module includes, from input to output, a fully connected layer and an activation layer. The method for obtaining the fault state of each node by inputting the voltage feature matrix of each node and the time-domain feature matrix of the similarity coefficients between nodes into the output module can be as follows: input the voltage feature matrix of each node and the time-domain feature matrix of the similarity coefficients between nodes into the fully connected layer to obtain the target feature matrix; input the target feature matrix into the activation layer to obtain the node fault index of each node; and determine the fault state of each node based on the node fault index.

[0091] In this embodiment, the output expression of node i is:

[0092]

[0093] In this embodiment, δ Re (·) represents the activation function, such as the Rectified Linear Unit (ReLU) activation function, δ S (·) represents the sigmoid function, where ω1 and ω2 are coefficient matrices in the fully connected layer. Let i be the predicted fault state of node i. This indicates that there is a fault in the branch where node i is located. This indicates that the branch containing node i is in normal working condition.

[0094] Optionally, the attention module includes, from input to output, a single-layer feedforward neural subnetwork and a nonlinear activation layer.

[0095] The reconstructed voltage data of each node is input into the attention module to obtain the initial feature matrix of each node, including:

[0096] The reconstructed voltage data of each node is input into the single-layer feedforward neural subnetwork to obtain the attention coefficients between each node and all its neighboring nodes.

[0097] In this embodiment, a single-layer feedforward neural subnetwork is used to perform a nonlinear transformation on the input features, mapping them to a new feature space. This enables the target spatiotemporal graph neural network to capture more complex feature relationships and enhance the feature representation capability. A single-layer feedforward neural subnetwork typically includes two fully connected layers and one nonlinear activation layer, with the nonlinear activation layer positioned between the two fully connected layers.

[0098] In this embodiment, different weights are assigned to the inputs in the attention module to distinguish the importance of different elements, thereby extracting more crucial information and achieving better results. In this embodiment, a single-layer feedforward neural subnetwork is selected to calculate the attention coefficients between node i and all its neighboring nodes j.

[0099] In this embodiment, the method for inputting the reconstructed voltage data of each node into the single-layer feedforward neural subnetwork to obtain the attention coefficients between each node and all its neighboring nodes can be as follows: input the reconstructed voltage data of each node into the single-layer feedforward neural subnetwork to obtain the attention coefficients between nodes; calculate the attention coefficients between each node and all its neighboring nodes based on the attention coefficients between nodes and the third formula, whereby the third formula is:

[0100] a ij =∑exp(y i,j );

[0101] In this embodiment, a ij Let be the attention coefficient between node i and all its neighboring nodes j;

[0102]

[0103] In this embodiment, a ij ' is the attention coefficient between the normalized node i and all its neighboring nodes j, y i,j Let ∑exp(y be the attention coefficient between node i and node j) i,j ) represents the attention coefficient between node i and all its neighboring nodes j.

[0104] The attention coefficients between each node and all its neighboring nodes, and the reconstructed voltage data of each node, are input into the nonlinear activation layer to obtain the initial feature matrix of each node.

[0105] In this embodiment, the initial feature matrix of each node is obtained by inputting the attention coefficients between each node and all its neighboring nodes and the reconstructed voltage data of each node into the nonlinear activation layer.

[0106] In this embodiment, the target spatiotemporal graph neural network includes an attention module, a spatiotemporal convolution module, and an output module. These modules are connected via a ReLU activation function. The attention module assigns higher weights to more important nodes, resulting in better learning performance. By assigning different weights to the inputs, the attention module distinguishes the importance of different elements, thereby extracting more crucial information. The attention module in the target spatiotemporal graph neural network can better integrate the correlations between features into the network.

[0107] Optionally, the spatiotemporal convolution module includes, from input to output, a temporal convolutional layer and a spatial convolutional layer.

[0108] The initial feature matrix of each node, the similarity coefficient matrix between the nodes, and the spatial connection state matrix of each node are input into the spatiotemporal convolution module to obtain the voltage feature matrix of each node and the temporal feature matrix of the similarity coefficients between the nodes, including:

[0109] The initial feature matrix of each node and the similarity coefficient matrix between the nodes are input into the temporal convolutional layer to obtain the temporal feature matrix of each node and the temporal feature matrix of the similarity coefficient between the nodes.

[0110] In this embodiment, the structure of the temporal convolutional layer is as follows: Figure 2 As shown, the temporal convolutional layer filters input features by adding gated linear units to the convolutional neural network and processes the initial feature matrix of each node and the similarity coefficient matrix between nodes in parallel.

[0111] The convolution process is shown in the following equation:

[0112] X l+1 =η(X) l );

[0113]

[0114] In this embodiment, X l E is the initial feature matrix of each node in the l-th temporal convolutional layer. l The similarity coefficient matrix of each node in the l-th temporal convolutional layer is given by input; η(·) and These are two one-dimensional convolution kernel functions, where η(·) is used to extract the temporal features of the initial feature matrix of each node. Used to extract the temporal features of the similarity coefficient matrix of each node; δ S (·) is the sigmoid activation function. This indicates that corresponding elements are multiplied.

[0115] It should be noted that when a line fault occurs, voltage change is the primary fault characteristic. In contrast, the similarity coefficient between nodes is not a direct manifestation of the fault and can be considered a secondary fault characteristic.

[0116] The temporal feature matrix and spatial connection state matrix of each node are input into the spatial convolutional layer to obtain the voltage feature matrix of each node.

[0117] In this embodiment, the voltage feature matrix of each node is a feature matrix that incorporates spatial information.

[0118] In this embodiment, the spatial convolutional layer enhances fault information based on the spatial topological information of the graph data, enabling the target spatiotemporal graph neural network to exhibit high generalization ability under different topological scenarios. The spatial convolutional layer combines the spatial information in the spatial connection state matrix of each node in the graph data and extracts spatial features through spectral graph convolution based on the principle of graph Fourier transform.

[0119] Optionally, the training process of the target spatiotemporal graph neural network includes:

[0120] Obtain the target sample set.

[0121] In this embodiment, the target sample set includes: training samples and the historical fault states of each node corresponding to the training samples. The training samples include: historical voltage data of each node in the distribution network and historical spatial connection state matrix of each node in the distribution network.

[0122] In this embodiment, the historical voltage data of each node in the distribution network can be historical voltage data acquired in advance over a certain period. The historical spatial connectivity state matrix of each node in the distribution network can be determined based on the historical topology of the distribution network. It should be noted that the acquisition time of the historical voltage data and the historical spatial connectivity state matrix of each node in the distribution network is the same. That is, if the acquisition time of the historical voltage data of each node in the distribution network is June 2nd of the year before last, then the acquisition time of the historical spatial connectivity state matrix of each node in the distribution network should also be June 2nd of the year before last. This is only an example and does not impose any restrictions on the specific acquisition time.

[0123] Based on the historical voltage data of each node in the power distribution network, the reconstructed voltage data samples of each node are obtained.

[0124] In this embodiment, the method for reconstructing the historical voltage data of each node can be the same as the method for reconstructing the voltage data of each node, and will not be described again here.

[0125] Based on the reconstructed voltage data samples of each node, a similarity coefficient matrix sample between nodes is determined.

[0126] In this embodiment, the method of determining the similarity coefficient matrix sample between nodes based on the reconstructed voltage data sample of each node is the same as the method of determining the similarity coefficient matrix between nodes based on the reconstructed voltage data of each node, and will not be described again here.

[0127] The similarity coefficient matrix samples between the nodes, the historical spatial connection status of each node in the distribution network, and the reconstructed voltage data samples of each node are input into the neural network to be trained to obtain the predicted fault status of each node.

[0128] In this embodiment, the predicted fault state of each node can be the predicted node fault index of each node.

[0129] In this embodiment, the neural network to be trained, from input to output, includes: an attention module to be trained, a spatiotemporal convolution module to be trained, and an output module to be trained. The attention module to be trained, from input to output, includes: a single-layer feedforward neural subnetwork to be trained and a nonlinear activation layer to be trained; the spatiotemporal convolution module to be trained, from input to output, includes: a temporal convolutional layer to be trained and a spatial convolutional layer to be trained.

[0130] The parameters of the neural network to be trained are trained based on the differences between the predicted fault states of each node and the historical fault states of each node corresponding to the training samples.

[0131] In this embodiment, the method for training the parameters of the neural network to be trained based on the difference between the predicted fault state of each node and the historical fault state of each node corresponding to the training sample can be as follows: The parameters of the neural network to be trained are trained based on the predicted fault state of each node, the historical fault state of each node corresponding to the training sample, and a loss function; the loss function is:

[0132]

[0133] In this embodiment, Y is the actual fault state matrix of the node. Let y be the predicted fault state matrix of the node. i This represents the actual fault state of node i. Let N be the predicted fault state of node i, and N be the total number of nodes.

[0134] S150, based on the fault status of each node in the distribution network, determines the fault section in the distribution network.

[0135] In this embodiment, the method for determining the fault section in the distribution network based on the fault status of each node in the distribution network can be as follows: determine the fault section in the distribution network based on the location information of the nodes in the fault state in the distribution network.

[0136] In a specific example, the three-phase voltage signals of nodes in historical fault data of a distribution network are acquired. An improved Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (ICEEMDAN) is used to decompose the voltage signals of each node, and then the voltage signals are reconstructed based on the correlation between the decomposed signals and the original signals. The similarity coefficient matrix between nodes is calculated, and the spatial connectivity state matrix of each node in the distribution network is generated according to the distribution network topology. Combined with the reconstructed voltage signals of each node, graph data is constructed. All fault samples in the graph data are divided into training and testing sets according to a certain ratio. A neural network to be trained is built, and model parameters such as the input / output dimensions, learning rate, and number of iterations for each layer are set according to actual needs. The fault localization problem is transformed into a multi-label classification problem, and a multi-label classification loss function is used for model training. The neural network to be trained is first evaluated on the testing set, and the neural network with the highest accuracy is selected as the target spatiotemporal graph neural network for application. In practical fault identification applications, the spatial connection state matrix of each node in the distribution network and the voltage data of each node within a preset time period are obtained. After data processing, the data is input into the target spatiotemporal neural network, and the fault section is determined based on the output node fault index.

[0137] In another specific example, the fault location problem can be modeled as a multi-label classification problem about nodes, that is, establishing a mapping relationship to classify the graph data according to the node number where the faulty branch is located, as shown in the following formula:

[0138] Y = f(X, E, A);

[0139] In this embodiment, Y = [y1, y2, y3, ..., y n ] represents the fault status of each node. If a fault occurs, y n If the value is 1, then it is 0 otherwise. The fault section can be determined based on the node status information. n is the node number, X is a matrix composed of the reconstructed voltage data of each node, E is a matrix composed of the similarity coefficients between nodes, and A is a matrix composed of the spatial connection status of each node.

[0140] The technical solution provided by the embodiments of the present invention has high generalization in different topology scenarios; at the same time, it can maximize the spatiotemporal data information characteristics of the power grid, and improve the efficiency and accuracy of fault location under different fault conditions and noise interference environments.

[0141] The technical solution of this embodiment first reconstructs the voltage data of each node in the distribution network within a preset time period to obtain the reconstructed voltage data of each node. The reconstructed voltage data allows for the extraction of key features related to distribution network faults. Then, the spatial connection state matrix of each node, the reconstructed voltage data of each node, and the similarity coefficient matrix between nodes are input into a target spatiotemporal graph neural network to obtain the fault state of each node. Based on the fault state of each node in the distribution network, the fault section in the distribution network is determined. This approach maximizes the use of spatiotemporal information through the target spatiotemporal graph neural network to extract fault location features, solving the problems of poor applicability, low fault tolerance, low accuracy, and long location time of current fault location methods in active distribution networks. This leads to more accurate and faster fault section location. Fast and accurate fault location technology can effectively shorten the fault investigation time of the distribution network and reduce power outage duration, which is of great significance for the safe and reliable operation of the distribution system.

[0142] Example 2

[0143] Figure 3 This is a schematic diagram of a distribution network fault location device based on a target spatiotemporal graph neural network, provided as an embodiment of the present invention. This embodiment is applicable to distribution network fault location. The device can be implemented using software and / or hardware, and can be integrated into any device that provides distribution network fault location functionality, such as… Figure 3 As shown, the distribution network fault location device based on the target spatiotemporal graph neural network specifically includes: a voltage data acquisition module 310, a reconstructed voltage data determination module 320, a similarity coefficient matrix determination module 330, a node fault state determination module 340, and a fault section determination module 350.

[0144] Among them, the voltage data acquisition module is used to acquire the spatial connection status matrix of each node in the distribution network and the voltage data of each node within a preset time period.

[0145] The reconstructed voltage data determination module is used to reconstruct the voltage data of each node in the distribution network within a preset time period to obtain the reconstructed voltage data of each node.

[0146] The similarity coefficient matrix determination module is used to determine the similarity coefficient matrix between nodes based on the reconstructed voltage data of each node.

[0147] The fault state determination module of the node is used to input the spatial connection state matrix of each node in the distribution network, the reconstructed voltage data of each node, and the similarity coefficient matrix between nodes into the target spatiotemporal graph neural network to obtain the fault state of each node.

[0148] The fault section determination module is used to determine the fault section in the distribution network based on the fault status of each node in the distribution network.

[0149] The above-described products can perform the methods provided in any embodiment of the present invention, and have the corresponding functional modules and beneficial effects for performing the methods.

[0150] Example 3

[0151] Figure 4 A schematic diagram of an electronic device 10 that can be used to implement embodiments of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0152] like Figure 4 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 may also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0153] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0154] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as the distribution network fault location method based on a target spatiotemporal graph neural network.

[0155] In some embodiments, the distribution network fault location method based on a target spatiotemporal graph neural network can be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the distribution network fault location method based on a target spatiotemporal graph neural network described above can be performed. Alternatively, in other embodiments, processor 11 can be configured to execute the distribution network fault location method based on a target spatiotemporal graph neural network by any other suitable means (e.g., by means of firmware).

[0156] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0157] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0158] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0159] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0160] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0161] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0162] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0163] This invention also provides a computer program product, including a computer program that, when executed by a processor, implements the distribution network fault location method based on a target spatiotemporal graph neural network according to any embodiment of the invention.

[0164] In implementing the computer program product, computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof. Programming languages ​​include object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0165] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for fault location in a distribution network based on a target spatiotemporal graph neural network, characterized in that, include: Obtain the spatial connection state matrix of each node in the distribution network and the voltage data of each node within a preset time period; The voltage data of each node in the power distribution network within a preset time period are reconstructed to obtain the reconstructed voltage data of each node. Based on the reconstructed voltage data of each node, determine the similarity coefficient matrix between nodes; The spatial connection state matrix of each node in the power distribution network, the reconstructed voltage data of each node, and the similarity coefficient matrix between nodes are input into the target spatiotemporal graph neural network to obtain the fault state of each node. Based on the fault status of each node in the distribution network, determine the fault section in the distribution network; The acquisition of voltage data of each node in the distribution network within a preset time period includes: Obtain voltage data of each node in the distribution network at each moment within a preset time period.

2. The method according to claim 1, characterized in that, The voltage data of each node in the distribution network within a preset time period are reconstructed to obtain the reconstructed voltage data of each node, including: The voltage data of each node within a preset time period are reconstructed to obtain the reconstructed voltage data of each node. The reconstruction operation includes: decomposing the voltage data of the nodes within a preset time period to obtain multi-order intrinsic mode functions and residual terms; Obtain the correlation coefficients between the intrinsic mode functions of each order and the voltage data of the nodes within a preset time period; Based on the correlation coefficient between the intrinsic mode functions of each order and the voltage data of the nodes within a preset time period, the intrinsic mode functions of each order are screened to obtain the target intrinsic mode functions; The reconstructed voltage data of the nodes is obtained by reconstructing the voltage data based on the residual terms and the target intrinsic mode function.

3. The method according to claim 2, characterized in that, Obtain the correlation coefficients between the intrinsic mode functions of each order and the voltage data of the nodes within a preset time period, including: Obtain the covariance operator of intrinsic mode functions of each order and node voltage data within a preset time period; Obtain the standard deviation of the intrinsic mode functions of each order and the standard deviation of the voltage data of the nodes within a preset time period; Based on the covariance operator, the standard deviation of each order of intrinsic modulus function, and the standard deviation of the voltage data of the node within a preset time period, the correlation coefficient between each order of intrinsic modulus function and the voltage data of the node within a preset time period is determined.

4. The method according to claim 1, characterized in that, The target spatiotemporal graph neural network includes, from input to output, an attention module, a spatiotemporal convolution module, and an output module. The spatial connection state matrix of each node in the distribution network, the reconstructed voltage data of each node, and the similarity coefficient matrix between nodes are input into the target spatiotemporal graph neural network to obtain the fault state of each node, including: The reconstructed voltage data of each node is input into the attention module to obtain the initial feature matrix of each node; The initial feature matrix of each node, the similarity coefficient matrix between the nodes, and the spatial connection state matrix of each node are input into the spatiotemporal convolution module to obtain the voltage feature matrix of each node and the temporal feature matrix of the similarity coefficient between the nodes. The voltage feature matrix of each node and the time-domain feature matrix of the similarity coefficient between nodes are input into the output module to obtain the fault status of each node.

5. The method according to claim 4, characterized in that, The attention module, from input to output, comprises: a single-layer feedforward neural subnetwork and a nonlinear activation layer. The reconstructed voltage data of each node is input into the attention module to obtain the initial feature matrix of each node, including: The reconstructed voltage data of each node is input into the single-layer feedforward neural subnetwork to obtain the attention coefficients between each node and all its neighboring nodes. The attention coefficients between each node and all its neighboring nodes, and the reconstructed voltage data of each node, are input into the nonlinear activation layer to obtain the initial feature matrix of each node.

6. The method according to claim 4, characterized in that, The spatiotemporal convolution module includes, from input to output, a temporal convolutional layer and a spatial convolutional layer. The initial feature matrix of each node, the similarity coefficient matrix between the nodes, and the spatial connection state matrix of each node are input into the spatiotemporal convolution module to obtain the voltage feature matrix of each node and the temporal feature matrix of the similarity coefficients between the nodes, including: The initial feature matrix of each node and the similarity coefficient matrix between the nodes are input into the temporal convolutional layer to obtain the temporal feature matrix of each node and the temporal feature matrix of the similarity coefficient between the nodes. The temporal feature matrix and spatial connection state matrix of each node are input into the spatial convolutional layer to obtain the voltage feature matrix of each node.

7. The method according to claim 1, characterized in that, The training process of the target spatiotemporal graph neural network includes: Obtain a target sample set, wherein the target sample set includes: training samples and the historical fault states of each node corresponding to the training samples, and the training samples include: historical voltage data of each node in the distribution network and historical spatial connection state matrix of each node in the distribution network; Based on the historical voltage data of each node in the distribution network, the reconstructed voltage data samples of each node are obtained. Based on the reconstructed voltage data samples of each node, determine the similarity coefficient matrix samples between nodes; The similarity coefficient matrix samples between the nodes, the historical spatial connection status of each node in the distribution network, and the reconstructed voltage data samples of each node are input into the neural network to be trained to obtain the predicted fault status of each node. The parameters of the neural network to be trained are trained based on the differences between the predicted fault states of each node and the historical fault states of each node corresponding to the training samples.

8. The method according to claim 1, characterized in that, The reconstructed voltage data of the node includes: multiple reconstructed voltages of each node at each time point; Based on the reconstructed voltage data of each node, a similarity coefficient matrix between nodes is determined, including: Based on the multiple reconstruction voltages of each node at each time point, determine the average value of each reconstruction voltage of each node; The similarity coefficient matrix between nodes is determined based on the multiple reconstruction voltages of each node at each time point and the mean of each reconstruction voltage of each node.

9. A distribution network fault location device based on a target spatiotemporal graph neural network, characterized in that, include: The voltage data acquisition module is used to acquire the spatial connection status matrix of each node in the distribution network and the voltage data of each node within a preset time period. The reconstructed voltage data determination module is used to reconstruct the voltage data of each node in the distribution network within a preset time period to obtain the reconstructed voltage data of each node. The similarity coefficient matrix determination module is used to determine the similarity coefficient matrix between nodes based on the reconstructed voltage data of each node. The fault state determination module of the node is used to input the spatial connection state matrix of each node in the distribution network, the reconstructed voltage data of each node, and the similarity coefficient matrix between nodes into the target spatiotemporal graph neural network to obtain the fault state of each node. The fault section determination module is used to determine the fault section in the distribution network based on the fault status of each node in the distribution network. The acquisition of voltage data of each node in the distribution network within a preset time period includes: Obtain voltage data of each node in the distribution network at each moment within a preset time period.

10. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the power distribution network fault location method based on the target spatiotemporal graph neural network as described in any one of claims 1-8.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute the power distribution network fault location method based on a target spatiotemporal graph neural network as described in any one of claims 1-8.

12. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the distribution network fault location method based on a target spatiotemporal graph neural network according to any one of claims 1-8.

Images