Power distribution network single-phase disconnection fault judgment method and device, terminal equipment and computer readable storage medium

Abstract

The invention discloses a power distribution network single-phase disconnection fault judgment method and device, terminal equipment and a computer readable storage medium, and the method comprises the steps: obtaining a to-be-evaluated power distribution network topology structure diagram, and determining a first node feature and a first adjacent matrix of each node in the diagram; the first node feature comprises a three-phase circuit effective value on a power distribution network line; and inputting the first node feature of each node and the first adjacent matrix into a trained single-phase broken line fault positioning model, so that the model performs forward propagation for several times according to the first node feature of each node and the first adjacent matrix, and the model aggregates node features of neighbor nodes of each node according to a connection relationship of the nodes in the first adjacent matrix, and outputs a category label of each node according to the aggregated node features of each node in each forward propagation. By implementing the method, the time and resource cost of retraining can be reduced, and the mobility of the model is improved.

Description

Technical Field

[0001] The present invention relates to the field of power system fault location, and in particular to a method, device, terminal equipment and computer-readable storage medium for distinguishing single-phase disconnection faults in a distribution network. Background Art

[0002] The single-phase disconnection fault identification of the distribution network in the power system plays a very important role in the safe operation of the distribution network. In order to solve the problem of disconnection faults in the distribution network under extreme weather conditions, it is necessary to establish a safe and reliable distribution network status detection and fault location system, analyze the working status of the distribution network equipment and the power outage of the user, handle the fault in time and take preventive measures. The existing disconnection fault location methods include the traditional fault location method based on electrical quantity threshold and the fault location method based on artificial intelligence. There are certain limitations. For example, the traditional manual location method is easily affected by factors such as operation mode, grounding resistance, load size, environment, etc., and the fault location accuracy is low; and the graph deep learning in the current new generation of artificial intelligence technology, such as GCN, belongs to the direct push graph neural network. During the training process, all the nodes of the whole graph need to be input at one time. If the topology map changes, it needs to be retrained, which is not suitable for fault location tasks with frequent topology changes. Therefore, how to make the new fault location method meet the requirements of positioning accuracy and have portability is a problem that needs to be studied. Summary of the invention

[0003] The embodiments of the present invention provide a method, an apparatus, a terminal device and a computer-readable storage medium for distinguishing a single-phase line break fault in a distribution network, which can reduce the time and resource cost of retraining and improve the portability of the model.

[0004] An embodiment of the present invention provides a method for determining a single-phase disconnection fault in a power distribution network, comprising:

[0005] Obtain a topological structure diagram of a distribution network to be evaluated;

[0006] According to the distribution network topology diagram to be evaluated, determining the first node feature and the first adjacency matrix of each node in the distribution network topology diagram to be evaluated; the first node feature includes: the effective value of the three-phase circuit on the distribution network line;

[0007] The first node feature and the first adjacency matrix of each node are input into the trained single-phase line break fault location model, so that the single-phase line break fault location model performs several forward propagations according to the first node feature and the first adjacency matrix of each node, and outputs the category label of each node; wherein, in each forward propagation process, the single-phase line break fault location model aggregates the node features of the neighboring nodes of each node according to the connection relationship of the nodes in the first adjacency matrix, and obtains the aggregated node features of each node; and outputs the category label of each node according to the aggregated node features of each node in each forward propagation.

[0008] Furthermore, the single-phase line break fault location model aggregates the node features of the neighboring nodes of each node according to the connection relationship of the nodes in the first adjacency matrix to obtain the aggregated node features of each node, including:

[0009] The single-phase line break fault location model performs neighbor sampling on each node in the distribution network topology diagram to be evaluated according to the connection relationship of the nodes in the first adjacency matrix and the sampling number preset in the single-phase line break fault location model to determine the neighbor nodes of each node;

[0010] The node features of the neighboring nodes of each node are aggregated through the mean aggregation function to obtain the aggregated node features of each node.

[0011] Furthermore, according to the aggregate node features of each node in each forward propagation, the category label of each node is output, including:

[0012] According to the aggregate node features of each node in each forward propagation, the feature value of each node is calculated by the following formula:

[0013]

[0014] The feature value of each node is input into the fully connected layer in the single-phase line break fault location model, so that the fully connected layer maps the feature value of each node to the category space, and obtains the probability distribution of each node in each category;

[0015] The probability distribution of each node in each category is passed through the softmax function to obtain the category label of each node;

[0016] in, Represents node v i The eigenvalue of the kth forward propagation, σ(·) represents the activation function, W k represents the weight matrix of the kth forward propagation, Represents node v i Aggregated node features of the kth forward pass.

[0017] Furthermore, the single-phase line break fault location model is obtained by:

[0018] Obtain several distribution network topology diagrams to be trained and the category label of each corresponding node;

[0019] According to the distribution network topology diagram to be trained, determining a second node feature and a second adjacency matrix of each node in the distribution network topology diagram to be trained;

[0020] Build an inductive graph neural network model and set the initial parameters of the inductive graph neural network model;

[0021] The second node feature and the second adjacency matrix of each node in each distribution network topology diagram to be trained and the corresponding category label of each node are respectively input into the inductive graph neural network model for iterative training until the loss function converges to obtain a trained single-phase line break fault location model.

[0022] Furthermore, the initial parameters of the inductive graph neural network model include: number of layers, number of hidden channels, batch normalization, dropout rate, and learning rate.

[0023] Based on the above method embodiment, the present invention provides a corresponding device embodiment, including: a topology map acquisition module, a feature determination module and a fault identification module;

[0024] A topology diagram acquisition module is used to acquire a topology diagram of a distribution network to be evaluated;

[0025] The feature determination module is used to determine the first node feature and the first adjacency matrix of each node in the distribution network topology diagram to be evaluated according to the distribution network topology diagram to be evaluated; the first node feature includes: the effective value of the three-phase circuit on the distribution network line;

[0026] A fault identification module is used to input the first node feature and the first adjacency matrix of each node into the trained single-phase line break fault location model, so that the single-phase line break fault location model performs several forward propagations according to the first node feature and the first adjacency matrix of each node, and outputs the category label of each node; wherein, in each forward propagation process, the single-phase line break fault location model aggregates the node features of the neighboring nodes of each node according to the connection relationship of the nodes in the first adjacency matrix, and obtains the aggregated node features of each node; and outputs the category label of each node according to the aggregated node features of each node in each forward propagation.

[0027] Furthermore, the fault identification module includes: a sampling submodule and an aggregation submodule;

[0028] A sampling submodule is used for the single-phase line break fault location model to perform neighbor sampling on each node in the distribution network topology diagram to be evaluated according to the connection relationship of the nodes in the first adjacency matrix and the sampling number preset in the single-phase line break fault location model, and determine the neighbor nodes of each node;

[0029] The aggregation submodule is used to aggregate the node features of the neighboring nodes of each node through the mean aggregation function to obtain the aggregated node features of each node.

[0030] Furthermore, the fault identification module includes: a feature value calculation submodule, a space mapping submodule and a classification submodule;

[0031] The eigenvalue calculation submodule is used to calculate the eigenvalue of each node according to the aggregate node features of each node in each forward propagation by the following formula:

[0032]

[0033] in, Represents node v i The eigenvalue of the kth forward propagation, σ(·) represents the activation function, W k represents the weight matrix of the kth forward propagation, Represents node v i Aggregate node features of the kth forward propagation;

[0034] A space mapping submodule is used to input the feature value of each node into the fully connected layer in the single-phase line break fault location model, so that the fully connected layer maps the feature value of each node to the category space to obtain the probability distribution of each node in each category;

[0035] The classification submodule is used to obtain the category label of each node through the softmax function for the probability distribution of each node in each category.

[0036] Based on the above-mentioned method item embodiment, the present invention provides a corresponding terminal device item embodiment, including: a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the steps of the distribution network single-phase line break fault judgment method as described in the present invention are implemented.

[0037] Based on the above-mentioned method item embodiment, the present invention provides a corresponding computer-readable storage medium item embodiment, including: a stored computer program, when the computer program is running, controlling the device where the computer-readable storage medium is located to execute the steps of the distribution network single-phase line break fault judgment method as described in the present invention.

[0038] Compared with the prior art, the beneficial effects of the embodiment of this solution are:

[0039] The present invention obtains a distribution network topology diagram to be evaluated, and then determines the first node feature and the first adjacency matrix of each node in the distribution network topology diagram to be evaluated according to the distribution network topology diagram to be evaluated, wherein the first node feature includes the effective value of the three-phase circuit on the distribution network line, and the effective value of the three-phase circuit on the distribution network line is used as the node feature, and the single-phase line break fault can be judged and located by monitoring the effective value of the current of the node and the adjacent node. Next, the first node feature and the first adjacency matrix of each node are input into the trained single-phase line break fault location model, so that the single-phase line break fault location model performs several forward propagations according to the first node feature and the first adjacency matrix of each node, and outputs the category label of each node; wherein, in each forward propagation process, the single-phase line break fault location model aggregates the node features of the adjacent nodes of each node according to the connection relationship of the nodes in the first adjacency matrix, and obtains the aggregated node features of each node; finally, according to the aggregated node features of each node obtained in the last forward propagation, the category label of each node is output. The process of aggregating node features is based on the connection relationship between nodes. In this process, the model determines the adjacent nodes of each node through the adjacency matrix and aggregates them according to the characteristics of the adjacent nodes. In other words, the aggregation method of the model is independent of the specific topology. Therefore, when the topology changes, you only need to update the adjacency matrix without retraining the model.

[0040] In summary, the present invention helps the model to better understand the contextual information of the node by aggregating node features according to the adjacency matrix in the model, and locates the fault based on this information. In the fault location task where the topology graph changes frequently, the time and resource cost of retraining is greatly reduced, and the flexibility and portability of the model are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a flow chart of a method for determining a single-phase disconnection fault in a distribution network provided by an embodiment of the present invention;

[0042] Figure 2 It is a structural schematic diagram of a single-phase disconnection fault location model provided by an embodiment of the present invention;

[0043] Figure 3 It is a flow chart of a single-phase disconnection fault location model training process provided by an embodiment of the present invention;

[0044] Figure 4 is a Loss curve diagram of the GCN, GAT and GraphSAGE models provided in one embodiment of the present invention;

[0045] Figure 5It is a node positioning accuracy curve diagram of the GCN, GAT and GraphSAGE models provided in one embodiment of the present invention;

[0046] Figure 6 It is a structural schematic diagram of a single-phase line break fault determination device for a distribution network provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0048] In the description of the present invention, it should be understood that the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features.

[0049] like Figure 1 As shown, an embodiment of the present invention provides a method for determining a single-phase disconnection fault in a power distribution network, the method comprising at least the following steps:

[0050] Step S1: obtaining a topological structure diagram of a distribution network to be evaluated;

[0051] For step S1, a distribution network topology diagram to be evaluated is obtained from a distribution network management system. The distribution network topology diagram is a diagram describing the connection and relationship between components of the distribution network, and it shows the connection mode and topology structure between various nodes in the distribution network.

[0052] Step S2: According to the distribution network topology diagram to be evaluated, determine the first node feature and the first adjacency matrix of each node in the distribution network topology diagram to be evaluated; the first node feature includes: the effective value of the three-phase circuit on the distribution network line;

[0053] For step S2, for the distribution network topology diagram to be evaluated, it is first necessary to determine the position of each node and the effective value of the three-phase circuit of their corresponding circuits, and the dimension is three-dimensional.

[0054] Next, according to the distribution network topology diagram to be evaluated, the first adjacency matrix is ​​constructed to represent the connection relationship between nodes. This matrix is ​​a matrix representation of the node connection relationship of the distribution network topology diagram. By analyzing each node one by one and marking the connection relationship between nodes in the matrix according to the connection mode in the topology diagram, if there is a connection between two nodes, the corresponding position in the matrix is ​​marked as 1; if there is no connection between the two nodes, it is marked as 0.

[0055] It should be noted that when a single-phase line break occurs, in the case of an ungrounded fault on both sides, the upstream and downstream currents of the fault phase are both 0, and the currents of the two non-fault phases are equal in magnitude and opposite in direction, one of which is negative, but regardless of the direction, they are all positive. When a single-phase line break is accompanied by grounding on the power supply side or grounding on the load side, there will be a grounding current in the line, but it is still lower than the normal value and close to 0, while the current directions of the two normal phases are the same and both are positive. The three-phase line current is no longer symmetrical during a single-phase line break fault. Compared with a short-circuit fault, during a single-phase grounding short-circuit fault, the short-circuit current of the fault phase is very large, which means that the current flows abnormally through the fault phase, while the currents of the other two phases are 0, indicating that the current on the fault phase cannot pass through the other two phases. Therefore, it is not possible to judge whether it is a single-phase line break fault or a short-circuit fault based solely on the current value of 0, and the present invention uses an inductive graph neural network model to learn and judge from the asymmetric characteristics of a single-phase line break fault.

[0056] Step S3: input the first node feature and the first adjacency matrix of each node into the trained single-phase line break fault location model, so that the single-phase line break fault location model performs several forward propagations according to the first node feature and the first adjacency matrix of each node, and outputs the category label of each node; wherein, in each forward propagation process, the single-phase line break fault location model aggregates the node features of the neighboring nodes of each node according to the connection relationship of the nodes in the first adjacency matrix, and obtains the aggregated node features of each node; outputs the category label of each node according to the aggregated node features of each node in each forward propagation.

[0057] In a preferred embodiment, the single-phase line break fault location model aggregates node features of neighboring nodes of each node according to the connection relationship of nodes in the first adjacency matrix to obtain aggregated node features of each node, including:

[0058] The single-phase line break fault location model performs neighbor sampling on each node in the distribution network topology diagram to be evaluated according to the connection relationship of the nodes in the first adjacency matrix and the sampling number preset in the single-phase line break fault location model to determine the neighbor nodes of each node;

[0059] The node features of the neighboring nodes of each node are aggregated through the mean aggregation function to obtain the aggregated node features of each node.

[0060] In a preferred embodiment, the category label of each node is output according to the aggregate node features of each node in each forward propagation, including:

[0061] According to the aggregate node features of each node in each forward propagation, the feature value of each node is calculated by the following formula:

[0062]

[0063] The feature value of each node is input into the fully connected layer in the single-phase line break fault location model, so that the fully connected layer maps the feature value of each node to the category space, and obtains the probability distribution of each node in each category;

[0064] The probability distribution of each node in each category is passed through the softmax function to obtain the category label of each node;

[0065] in, Represents node v i The eigenvalue of the kth forward propagation, σ(·) represents the activation function, W k represents the weight matrix of the kth forward propagation, Represents node v i Aggregated node features of the kth forward pass.

[0066] For step S3, Figure 2 As shown in the figure, first, the first node feature and the first adjacency matrix of each node are input into the single-phase line break fault location model. The model includes 4 SAGEConv graph convolution layers, and the input data is forward propagated once after passing through a SAGEConv graph convolution layer. In each forward propagation process, the model is divided into a sampling stage and an aggregation stage.

[0067] In the sampling phase, the model selects the neighbor nodes according to the connection relationship of the nodes in the first adjacency matrix and the preset number of neighbor node samplings in each layer. L , determine the neighbor nodes of each node layer by layer, and then, in the aggregation stage, according to the connection relationship of the nodes in the first adjacency matrix, aggregate the node features of the neighbor nodes of each node through the following formula to obtain the aggregated node features of each node:

[0068]

[0069] in, Represents node v i Aggregate node features of the kth forward propagation, AGGREGATE k (·) represents the aggregation function of the kth forward propagation, Represents the features of node v in the kth forward propagation. This aggregation process can help the model understand the relationships and interactions between nodes.

[0070] According to the aggregate node features of each node in each forward propagation, the feature value of each node is calculated by the following feature propagation formula:

[0071]

[0072] in, Represents node v i The eigenvalue of the kth forward propagation, σ(·) represents the activation function, W k represents the weight matrix of the kth forward propagation, Represents node v i Aggregate node features of the kth forward propagation. The feature propagation formula uses the activation function and weight matrix to convert the aggregate node features of each node into feature values.

[0073] Then, the feature value of each node is input into the fully connected layer in the single-phase disconnection fault location model, and the feature value of each node is mapped to the category space through the fully connected layer. In this way, the probability distribution of each node in each category can be obtained, and the fault location task is converted into a node classification problem. Finally, the estimated value of each node category in the graph is output through the softmax function, indicating whether each node has a single-phase disconnection fault. 0 means that the node has no fault, and 1 means that the node has a single-phase disconnection fault.

[0074] In addition to outputting the probability value of the fault, the model can also output the location information of the nodes classified as 1. By analyzing the output results, it is possible to determine which nodes have single-phase disconnection faults. The location information of these nodes can be used to locate the specific location of the fault, helping operation and maintenance personnel to quickly locate and repair the fault.

[0075] The training process of the single-phase line break fault location model is described in detail below:

[0076] like Figure 3 As shown in FIG. 1 , the training process of the single-phase line break fault location model includes the following steps:

[0077] Step S301: obtaining a number of distribution network topology diagrams to be trained and the category label of each corresponding node;

[0078] For step S301, a training set is obtained. The training set consists of multiple distribution network topology diagrams. Each distribution network topology diagram includes a node feature x, an adjacency matrix edge_index, and an output node label y. The node feature x is the effective value of the three-phase circuit after a fault on the node in the diagram, i.e., the line in the distribution network, and the dimension is three-dimensional. The output y is the label of all nodes in the diagram. In this embodiment, there are two cases, y=0 indicates that there is no fault, and y=1 indicates that a line break fault occurs on the line.

[0079] In Simulink, an IEEE123 distribution network system similar to the actual operation was built. The distribution network contains 5 bus tie breakers and 16 distributed power sources IIDG. In addition, according to the structure of the distribution network, six operation modes were built. In the first five operation modes, one bus tie breaker was disconnected in turn. In the sixth operation mode, the topology diagram changed significantly, that is, on the basis of the fifth operation mode, the <160,67> line was additionally disconnected, resulting in a reduction of 16 lines and 14 bus nodes in the topology diagram, and the topology diagram changed significantly. After setting the fault parameters, simulation was performed. In each round of simulation, a line was randomly selected to have a single-phase line break fault, and the line break type, line break location, line break time, transition resistance size and operation mode were randomly set. The effective value of the three-phase current on each line after the fault was obtained through simulation. 5 / 6 of the data sets formed by the first four operation modes were extracted as training sets, and the remaining data sets formed with the fifth and sixth operation modes were used as test sets one, two and three. In this embodiment, the training set contains 7080 images in total, and the three test sets contain 1416, 2124 and 1836 images respectively.

[0080] Step S302: determining a second node feature and a second adjacency matrix of each node in the distribution network topology diagram to be trained according to the distribution network topology diagram to be trained;

[0081] For step S302, for the distribution network topology diagram to be trained, the effective value of the three-phase circuit of each node is used as the second node feature, and a second adjacency matrix is ​​constructed to represent the connection relationship between the nodes. The second adjacency matrix is ​​a matrix representation of the node connection relationship of the distribution network topology diagram.

[0082] Step S303: construct an inductive graph neural network model and set initial parameters of the inductive graph neural network model;

[0083] In a preferred embodiment, the initial parameters of the inductive graph neural network model include: the number of layers, the number of hidden channels, batch normalization, the dropout rate and the learning rate.

[0084] For step S303, an inductive graph neural network model is constructed, and the parameters of the model are set as follows: layer (number of layers) = 4, hidden_channels (number of hidden channels) = 64, batchnorm (batch normalization) = True, dropout (dropout rate) = 0.0, learning_rate (learning rate) = 0.01.

[0085] Step S304: The second node feature and the second adjacency matrix of each node in each distribution network topology diagram to be trained and the corresponding category label of each node are respectively input into the inductive graph neural network model for iterative training until the loss function converges to obtain a trained single-phase line break fault location model.

[0086] For step S304, for each distribution network topology diagram to be trained, the second node feature and the second adjacency matrix of each node in each distribution network topology diagram to be trained and the corresponding category label of each node are respectively input into the inductive graph neural network model for iterative training. In each training epoch, the loss function value of the model can be calculated, and according to the change of the loss function value, an image is drawn to observe the training effect of the model.

[0087] In this process, nll_loss in PyTorch can be used as the loss function of the model. nll_loss is a negative log-likelihood loss function that is used to calculate the deviation between the predicted value and the target value. Its parameters include input data, target data, weight, reduction method, and ignore index. In this embodiment, the predicted probability obtained by log_softmax is used as input, and the target data is the real category label. The nll_loss formula is defined as follows:

[0088] loss(x,y)=L={l1,...,l n} T ;

[0089] Among them, l n =-w yn x n,yn , w yn is the weight of each category, all 1 by default, x n,yn Indicates the probability of the target category, that is, the x of the corresponding category is negative. In this embodiment, after about 60 epochs of testing, the model will reach convergence and a trained single-phase disconnection fault location model will be obtained.

[0090] Next, the trained single-phase line break fault location model is tested by the test set, and the GCN, GAT and GraphSAGE models are trained to converge (GraphSAGE is the single-phase line break fault location model trained by the present invention), and the Loss curve and the node location accuracy curve are plotted, as shown in Fig. Figure 4 and Figure 5 shown.

[0091] Test sets 1, 2 and 3 are tested respectively to compare their node positioning accuracy. The results are shown in Table 1:

[0092] Table 1. Node classification accuracy of GCN, GAT and GraphSAGE models in three test sets

[0093]

[0094] By comparison, it is found that GraphSAGE has a higher node location accuracy than GCN and GAT, especially in test set 3 where the topology changes greatly. Therefore, the GraphSAGE model is transferable and can better adapt to the single-phase line break fault location task in distribution networks where the topology changes frequently.

[0095] It should be noted that the present invention can also determine a two-phase line break fault or a three-phase line break fault. The current characteristics of these two line break faults are different from those of a single-phase line break fault. By allowing the model to learn the characteristics of the nodes, the location and type of the line break can be determined.

[0096] like Figure 6 As shown, based on the above method embodiment, a corresponding device embodiment is provided;

[0097] An embodiment of the present invention provides a single-phase disconnection fault identification device for a power distribution network, comprising: a topology diagram acquisition module, a feature determination module, and a fault identification module;

[0098] A topology diagram acquisition module is used to acquire a topology diagram of a distribution network to be evaluated;

[0099] The feature determination module is used to determine the first node feature and the first adjacency matrix of each node in the distribution network topology diagram to be evaluated according to the distribution network topology diagram to be evaluated; the first node feature includes: the effective value of the three-phase circuit on the distribution network line;

[0100] A fault identification module is used to input the first node feature and the first adjacency matrix of each node into the trained single-phase line break fault location model, so that the single-phase line break fault location model performs several forward propagations according to the first node feature and the first adjacency matrix of each node, and outputs the category label of each node; wherein, in each forward propagation process, the single-phase line break fault location model aggregates the node features of the neighboring nodes of each node according to the connection relationship of the nodes in the first adjacency matrix, and obtains the aggregated node features of each node; and outputs the category label of each node according to the aggregated node features of each node in each forward propagation.

[0101] In a preferred embodiment, the fault identification module includes: a sampling submodule and an aggregation submodule;

[0102] A sampling submodule is used for the single-phase line break fault location model to perform neighbor sampling on each node in the distribution network topology diagram to be evaluated according to the connection relationship of the nodes in the first adjacency matrix and the sampling number preset in the single-phase line break fault location model, and determine the neighbor nodes of each node;

[0103] The aggregation submodule is used to aggregate the node features of the neighboring nodes of each node through the mean aggregation function to obtain the aggregated node features of each node.

[0104] In a preferred embodiment, the fault identification module includes: a feature value calculation submodule, a space mapping submodule and a classification submodule;

[0105] The eigenvalue calculation submodule is used to calculate the eigenvalue of each node according to the aggregate node features of each node in each forward propagation by the following formula:

[0106]

[0107] in, Represents node v i The eigenvalue of the kth forward propagation, σ(·) represents the activation function, W k represents the weight matrix of the kth forward propagation, Represents node v i Aggregate node features of the kth forward propagation;

[0108] A space mapping submodule is used to input the feature value of each node into the fully connected layer in the single-phase line break fault location model, so that the fully connected layer maps the feature value of each node to the category space to obtain the probability distribution of each node in each category;

[0109] The classification submodule is used to obtain the category label of each node through the softmax function for the probability distribution of each node in each category.

[0110] It can be understood that the above-mentioned device item embodiments correspond to the method item embodiments of the present invention, and can implement the distribution network single-phase line break fault judgment method provided by any one of the above-mentioned method item embodiments of the present invention.

[0111] It should be noted that the device embodiments described above are merely illustrative, and some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. Those of ordinary skill in the art may understand and implement the present invention without creative work.

[0112] Based on the above-mentioned embodiment of the distribution network single-phase line break fault judgment method, another embodiment of the present invention provides a terminal device, which includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the distribution network single-phase line break fault judgment method of any embodiment of the present invention is implemented.

[0113] Exemplarily, in this embodiment, the computer program may be divided into one or more modules, which are stored in the memory and executed by the processor to implement the present invention. The one or more module elements may be a series of computer program instruction segments capable of implementing specific functions, which are used to describe the execution process of the computer program in the terminal device.

[0114] The terminal device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0115] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the terminal device, and uses various interfaces and lines to connect various parts of the entire terminal device.

[0116] Based on the above method embodiment, another embodiment is provided: another embodiment of the present invention provides a computer-readable storage medium, including a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the distribution network single-phase line break fault judgment method described in any one of the above method embodiments of the present invention.

[0117] Wherein, the module / unit integrated in the single-phase disconnection fault determination device / terminal device of the distribution network, if implemented in the form of a software functional unit and sold or used as an independent product, can be stored in a computer-readable storage medium. Based on such an understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium, etc.

[0118] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for distinguishing single-phase disconnection fault in a distribution network, characterized in that: include: Obtain a topological structure diagram of a distribution network to be evaluated; According to the distribution network topology diagram to be evaluated, determining a first node feature and a first adjacency matrix of each node in the distribution network topology diagram to be evaluated; The first node feature includes: the effective value of the three-phase circuit on the distribution network line; The first node feature and the first adjacency matrix of each node are input into the trained single-phase line break fault location model, so that the single-phase line break fault location model performs several forward propagations according to the first node feature and the first adjacency matrix of each node, and outputs the category label of each node; wherein, in each forward propagation process, the single-phase line break fault location model aggregates the node features of the neighboring nodes of each node according to the connection relationship of the nodes in the first adjacency matrix to obtain the aggregated node features of each node; and outputs the category label of each node according to the aggregated node features of each node in each forward propagation.

2. The method for determining a single-phase disconnection fault in a power distribution network according to claim 1, characterized in that: The single-phase line break fault location model aggregates node features of neighboring nodes of each node according to the connection relationship of nodes in the first adjacency matrix to obtain aggregated node features of each node, including: The single-phase line break fault location model performs neighbor sampling on each node in the distribution network topology diagram to be evaluated according to the connection relationship of the nodes in the first adjacency matrix and the sampling number preset in the single-phase line break fault location model to determine the neighbor nodes of each node; The node features of the neighboring nodes of each node are aggregated through the mean aggregation function to obtain the aggregated node features of each node.

3. The method for determining a single-phase disconnection fault in a power distribution network according to claim 2, characterized in that: According to the aggregate node features of each node in each forward propagation, the category label of each node is output, including: According to the aggregate node features of each node in each forward propagation, the feature value of each node is calculated by the following formula: Inputting the feature value of each node into the fully connected layer in the single-phase line break fault location model, so that the fully connected layer maps the feature value of each node to the category space to obtain the probability distribution of each node in each category; The probability distribution of each node in each category is passed through the softmax function to obtain the category label of each node; in, Represents node v i The eigenvalue of the kth forward propagation, σ(·) represents the activation function, W k represents the weight matrix of the kth forward propagation, Represents node v i Aggregated node features for the kth forward pass.

4. The method for determining a single-phase disconnection fault in a power distribution network according to claim 3, characterized in that: The single-phase line break fault location model is obtained by: Obtain several distribution network topology diagrams to be trained and the category label of each corresponding node; According to the distribution network topology diagram to be trained, determining a second node feature and a second adjacency matrix of each node in the distribution network topology diagram to be trained; Constructing an inductive graph neural network model and setting initial parameters of the inductive graph neural network model; The second node feature and the second adjacency matrix of each node in each distribution network topology diagram to be trained and the corresponding category label of each node are respectively input into the inductive graph neural network model for iterative training until the loss function converges to obtain a trained single-phase line break fault location model.

5. The method for determining a single-phase disconnection fault in a power distribution network according to claim 4, characterized in that: The initial parameters of the inductive graph neural network model include: number of layers, number of hidden channels, batch normalization, dropout rate and learning rate.

6. A single-phase disconnection fault identification device for a distribution network, characterized in that: include: Topological structure map acquisition module, feature determination module and fault identification module; The topology diagram acquisition module is used to acquire the topology diagram of the distribution network to be evaluated; The feature determination module is used to determine a first node feature and a first adjacency matrix of each node in the distribution network topology diagram to be evaluated according to the distribution network topology diagram to be evaluated; The first node feature includes: the effective value of the three-phase circuit on the distribution network line; The fault identification module is used to input the first node feature and the first adjacency matrix of each node into the trained single-phase line break fault location model, so that the single-phase line break fault location model performs several forward propagations according to the first node feature and the first adjacency matrix of each node, and outputs the category label of each node; wherein, in each forward propagation process, the single-phase line break fault location model aggregates the node features of the neighboring nodes of each node according to the connection relationship of the nodes in the first adjacency matrix to obtain the aggregated node features of each node; and outputs the category label of each node according to the aggregated node features of each node in each forward propagation.

7. The single-phase disconnection fault identification device of the power distribution network according to claim 6, characterized in that: The fault identification module includes: a sampling submodule and an aggregation submodule; The sampling submodule is used for the single-phase line break fault location model to perform neighbor sampling on each node in the distribution network topology diagram to be evaluated according to the connection relationship of the nodes in the first adjacency matrix and the sampling number preset in the single-phase line break fault location model, so as to determine the neighbor nodes of each node; The aggregation submodule is used to aggregate the node features of the neighboring nodes of each node through the mean aggregation function to obtain the aggregated node features of each node.

8. The single-phase disconnection fault identification device for power distribution network according to claim 7, characterized in that: The fault identification module includes: a feature value calculation submodule, a space mapping submodule and a classification submodule; The eigenvalue calculation submodule is used to calculate the eigenvalue of each node according to the aggregate node features of each node in each forward propagation by the following formula: in, Represents node v i The eigenvalue of the kth forward propagation, σ(·) represents the activation function, W k represents the weight matrix of the kth forward propagation, Represents node v i Aggregate node features of the kth forward propagation; The spatial mapping submodule is used to input the feature value of each node into the fully connected layer in the single-phase disconnection fault location model, so that the fully connected layer maps the feature value of each node to the category space to obtain the probability distribution of each node in each category; The classification submodule is used to obtain the category label of each node by applying the softmax function to the probability distribution of each node in each category.

9. A terminal device, characterized in that: include: A processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the method for determining a single-phase line break fault in a distribution network as described in any one of claims 1 to 5 is implemented.

10. A computer-readable storage medium, characterized in that: include: A stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the method for determining a single-phase line break fault in a distribution network as described in any one of claims 1 to 5.

Images