Power distribution network voltage flow state estimation method and system based on GAT

Abstract

The invention relates to a GAT-based power distribution network voltage power flow state estimation method and system, and the method comprises the following steps: obtaining the historical measurement state quantity of a target power distribution network, obtaining the power flow data of multiple time sections through simulation, and constructing a sample set by taking the power flow data of multiple time sections as a real label; constructing a GAT graph attention network, setting a node number and a connection relation of the GAT graph attention network based on the topological information of the target power distribution network, and inputting node feature information; inputting the sample set into the constructed GAT map attention network, outputting a power flow data predicted value of a node through the GAT map attention network, calculating loss through the power flow data predicted value of the corresponding node and a real label of the power flow data, updating model parameters of the GAT map attention network through the calculated loss, and obtaining the power flow data of the corresponding node. Obtaining a trained power distribution network state estimation model; and estimating the tidal current state of the target power distribution network through the trained power distribution network state estimation model.

Description

Technical Field

[0001] The invention relates to a method and system for estimating the voltage flow state of a distribution network based on GAT, and belongs to the technical field of power system state estimation. Background Art

[0002] In the first half of 2024, the country's new installed capacity of renewable energy power generation was 134 million kilowatts, a year-on-year increase of 24%, accounting for 88% of the country's new installed power capacity. By the end of June 2024, the country's installed capacity of renewable energy power generation reached 1.653 billion kilowatts, a year-on-year increase of 25%, accounting for about 53.8% of my country's total installed power generation capacity. With the vigorous development of renewable energy power generation, the scale of renewable energy power generation has also expanded, resulting in more complex power system trends. The main direction of the flow of traditional power systems is from power plants to loads, but when renewable energy power generation is connected to the grid in large quantities, especially when distributed renewable energy power generation is connected to the grid in large quantities, the flow of power systems has also appeared in various complex phenomena such as reverse flow. At present, the volatility of renewable energy power generation, the randomness of system load fluctuations, the complexity of power system power quality problems caused by the large-scale access of power electronic devices, and the uncertainty of system measurement equipment sometimes caused by errors, etc., all of these problems have caused more complex power system power quality problems, and also made the operation of the power system more complicated, increasing the uncertainty of the grid operation state, causing troubles for power system dispatching, and also bringing a series of challenges to the safe and stable operation of the power system.

[0003] In summary, estimating the state of the power system can provide a clear estimate of the power quality problems caused by the current expansion of the scale of new energy access and the sharp increase in the number of power electronic equipment access. It can promote the organic integration of new energy power generation and the power grid, help new energy power generation to better play a positive role in the power grid, provide a certain reference for the power grid dispatching system, and help power grid companies grasp the overall state of the power system. Estimating the state of the power system can also help reduce the production losses of enterprises caused by power system state problems, and can help the power grid better monitor the overall situation, which is of great significance to the dispatching process of the power system and the management and control of the power grid.

[0004] The goal of power system state estimation is to obtain the voltage, phase angle, active power, reactive power and other states of each node in the current system, so as to provide support for subsequent power system stability assessment or power system dispatch. Currently, there are two main methods used for estimation: one is the traditional physical model-based method, and the other is the data-driven method.

[0005] At present, the state estimation method based on weighted least square (WLS) has been widely used in the physical model-based methods. The WLS estimation result has the characteristics of minimum variance and unbiasedness. It is a very excellent state estimation algorithm in the physical model-driven method when only Gaussian white noise is contained. However, in the actual power grid operation, the "raw data" collected by the measurement system contains non-Gaussian noise and bad data, which makes the WLS estimation convergence poor and the estimation result unusable. There is also a multi-region distributed state estimation method for power systems, which can effectively improve the computational efficiency and robustness of centralized state estimation systems, and is suitable for state estimation of power grids with more complex structures and larger measurement data volumes. There is also an existing measurement linear equivalent transformation method that is improved by state prediction and pseudo-measurement recursive transformation, shortening the state update cycle. There is also a method that converts the weighted least absolute value (WLAV) estimation into a nonlinear transformation and sub-programming problem, which improves the speed of WLAV solution, but the bilinear algorithm will reduce the redundancy of state estimation and affect the estimation accuracy.

[0006] Most of the data-driven methods use neural network methods, including data-driven robust state estimation methods based on offline learning and online matching, which improves offline learning efficiency. There are also methods that use deep neural networks and apply correlation analysis to extract strongly correlated measurements to achieve distributed estimation, while improving the robustness and efficiency of the model. There is also a physics-guided deep learning method that learns time correlation through deep neural networks and models machine learning based on physics. Compared with pure data-driven methods, this method has better interpretability.

[0007] However, the topology of power grids is becoming increasingly complex and the access components are becoming more diversified. The above methods do not consider the impact of the topological structure information of each node and its lines in the power grid on the accuracy of the method. Therefore, this type of method cannot adapt to the topological changes that occur during the actual operation of the power system. Summary of the invention

[0008] In order to solve the above problems existing in the prior art, the present invention proposes a method and system for estimating the voltage flow state of a distribution network based on GAT.

[0009] The technical solution of the present invention is as follows:

[0010] On the one hand, the present invention provides a method for estimating the voltage flow state of a distribution network based on GAT, comprising the following steps:

[0011] Obtain the historical measurement state of the target distribution network, obtain the power flow data of multiple time sections through simulation, and construct a sample set using the power flow data of multiple time sections as the true label;

[0012] Construct a GAT graph attention network, set the number of nodes and connection relationships of the GAT graph attention network based on the topology information of the target distribution network, and input node feature information;

[0013] The sample set is input into the constructed GAT graph attention network, and the power flow data prediction value of the node is output through the GAT graph attention network. The loss is calculated by the power flow data prediction value of the corresponding node and the true label of the power flow data. The model parameters of the GAT graph attention network are updated by the calculated loss to obtain the trained distribution network state estimation model.

[0014] The power flow state of the target distribution network is estimated through the trained distribution network state estimation model.

[0015] As a preferred implementation, in the step of constructing a sample set using the flow data of multiple time sections as true labels, it also includes:

[0016] Based on the real multi-time section power flow data obtained by simulation, data expansion operation is performed, and a sample set is constructed with the expanded power flow data.

[0017] As a preferred implementation, the node feature information includes:

[0018] Node voltage amplitude, branch phase angle difference, node injected active power and node injected reactive power.

[0019] As a preferred implementation, the step of constructing the GAT graph attention network includes:

[0020] The graph data is constructed based on the known topological relationship of the power nodes in the target distribution network. The graph includes N power nodes. For the target power node, the attention mechanism is used to calculate the attention coefficient between it and its neighboring nodes, as follows:

[0021]

[0022] Among them, C ij is the attention coefficient, f(·,·) represents the attention calculation function, σ1 is the LeakyReLU function, represents the first parameter matrix to be trained, W represents the second parameter matrix to be trained, T represents the transpose, represents the feature embedding vector of the target power node i, represents the feature embedding vector of the jth neighbor node of the target power node i, || represents the series operation;

[0023] The calculated attention coefficients of all neighbor nodes are normalized to obtain the final weight distribution coefficients, as follows:

[0024]

[0025] Among them, N i is the set of neighbor nodes of target power node i, Represents the feature embedding vector of the kth neighbor node of the target power node i;

[0026] The output vector of the target power node is obtained according to the final weight allocation coefficient, as follows:

[0027]

[0028] in, is the output vector of the target power node i, and σ2 is the ELU activation function.

[0029] On the other hand, the present invention also provides a distribution network voltage flow state estimation system based on GAT, comprising:

[0030] The data set construction module is used to obtain the historical measurement state of the target distribution network, obtain the power flow data of multiple time sections through simulation, and construct the sample set with the power flow data of multiple time sections as the real label;

[0031] The GAT model building module is used to build a GAT graph attention network, set the number of nodes and connection relationships of the GAT graph attention network based on the topological information of the target distribution network, and input node feature information;

[0032] The model training module is used to input the sample set into the constructed GAT graph attention network, output the power flow data prediction value of the node through the GAT graph attention network, calculate the loss through the power flow data prediction value of the corresponding node and the true label of the power flow data, and update the model parameters of the GAT graph attention network through the calculated loss to obtain the trained distribution network state estimation model;

[0033] The power flow state estimation module is used to estimate the power flow state of the target distribution network through the trained distribution network state estimation model.

[0034] As a preferred implementation, in the step of constructing a sample set using the flow data of multiple time sections as true labels, it also includes:

[0035] Based on the real multi-time section power flow data obtained by simulation, data expansion operation is performed, and a sample set is constructed with the expanded power flow data.

[0036] As a preferred implementation, the node feature information includes:

[0037] Node voltage amplitude, branch phase angle difference, node injected active power and node injected reactive power.

[0038] As a preferred implementation, the step of constructing the GAT graph attention network includes:

[0039] The graph data is constructed based on the known topological relationship of the power nodes in the target distribution network. The graph includes N power nodes. For the target power node, the attention mechanism is used to calculate the attention coefficient between it and its neighboring nodes, as follows:

[0040]

[0041] Among them, C ij is the attention coefficient, f(·,·) represents the attention calculation function, σ1 is the LeakyReLU function, represents the first parameter matrix to be trained, W represents the second parameter matrix to be trained, T represents the transpose, represents the feature embedding vector of the target power node i, represents the feature embedding vector of the jth neighbor node of the target power node i, || represents the series operation;

[0042] The calculated attention coefficients of all neighbor nodes are normalized to obtain the final weight distribution coefficients, as follows:

[0043]

[0044] Among them, N i is the set of neighbor nodes of target power node i, Represents the feature embedding vector of the kth neighbor node of the target power node i;

[0045] The output vector of the target power node is obtained according to the final weight allocation coefficient, as follows:

[0046]

[0047] in, is the output vector of the target power node i, and σ2 is the ELU activation function.

[0048] On the other hand, the present invention also proposes an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for estimating the voltage flow state of a distribution network based on GAT as described in any embodiment of the present invention is implemented.

[0049] On the other hand, the present invention further proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for estimating the voltage flow state of a distribution network based on GAT as described in any embodiment of the present invention.

[0050] Additional aspects and advantages of the present invention will be set forth in the following description, and some of them will be apparent from the description, or may be understood by practicing the present invention. In addition, the various aspects and advantages of the present invention may be realized and obtained by the method steps and combinations particularly pointed out in the appended claims. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 The figure is a schematic diagram of the method flow of the first embodiment of the present invention. DETAILED DESCRIPTION

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

[0053] It should be understood that the step numbers used in this document are only for convenience of description and are not intended to limit the order in which the steps are executed.

[0054] It should be understood that the terms used in the present specification are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0055] The terms “include” and “comprising” indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0056] The term "and / or" means and includes any and all possible combinations of one or more of the associated listed items.

[0057] Embodiment 1:

[0058] See also Figure 1 ,This embodiment provides a method for estimating the voltage flow state of a distribution network based on GAT, which is applied in a distribution network, where the connection topology structure is clear, the line parameters are clear, and most nodes in the distribution network need to have voltage measurement and power measurement.

[0059] The method comprises the following steps:

[0060] S100, obtaining historical measurement state quantities of the target distribution network, and obtaining flow data of multiple time sections as real data through simulation, normalizing the flow data of multiple time sections and making a sample set in the form of graph data according to the obtained distribution network topology information; and dividing the sample set into a test set and a training set.

[0061] In one embodiment, step S100 further includes: performing a data expansion operation based on the real multi-time section power flow data obtained by simulation, and constructing a sample set with the power flow data after data expansion. The data expansion operation can be performed by adding Gaussian white noise and bad data to the real multi-time section power flow data, and preventing overfitting and improving model robustness by data expansion.

[0062] S200. Construct a GAT graph attention network, set the number of nodes and connection relationships of the GAT graph attention network based on the topology information of the target distribution network, and input node feature information; wherein the node feature information is set as a four-dimensional vector of node voltage amplitude, branch phase angle difference, node injected active power, and node injected reactive power, and input the node feature information at the measurement node.

[0063] The Graph Attention Network (GAT) applies the attention mechanism to graph-structured data, and uses a variable attention mechanism on the weights of the edges between nodes instead of the unchanged weight assignment form, so it can be adaptable to changes in the topological structure. GAT integrates the graph structure and node features in the model graph. Through the attention mechanism, the node can pay attention to the characteristics of the neighborhood and learn different attention weights, thereby automatically determining the importance and relevance between the current node and its neighborhood, and then integrating the feature information of the neighboring nodes to generate the embedded representation of the target node.

[0064] In one embodiment, the main task of the GAT graph attention network constructed in step S200 is to train the attention coefficients between power nodes and aggregate the feature information of neighboring nodes. In order to model and analyze the power nodes in the distribution network, a graph data is constructed according to the known topological relationship of the power nodes in the target distribution network. The graph includes N power nodes, where the node feature information vector of power node i is h i ; For the target power node, the attention mechanism is used to calculate the attention coefficient C between it and its neighboring nodes ij , that is, the degree of association of neighbor node j to target power node i, as follows:

[0065]

[0066] Among them, C ij is the attention coefficient, f(·,·) represents the attention calculation function, σ1 is the LeakyReLU function, represents the first parameter matrix to be trained, W represents the second parameter matrix to be trained, T represents the transpose, represents the feature embedding vector of the target power node i, represents the feature embedding vector of the jth neighbor node of the target power node i, || represents the series operation;

[0067] In order to better distribute the weights, it is also necessary to use the softmax function to normalize the relevance of all neighbor nodes to obtain the final weight distribution coefficient, as follows:

[0068]

[0069] Among them, N i is the set of neighbor nodes of target power node i, Represents the feature embedding vector of the kth neighbor node of the target power node i;

[0070] Then, the calculated normalized final weight distribution coefficient is used to aggregate the feature information of the neighboring nodes to obtain the output vector of the target power node, as follows:

[0071]

[0072] in, is the output vector of the target power node i, and σ2 is the ELU activation function.

[0073] In one embodiment, a multi-head attention mechanism is introduced to improve the learning ability and robustness of the model, that is, multiple sets of independent attention mechanisms are used, and then the output results are spliced. Different methods are used at different stages of the model. In the middle layer, multiple sets of target node embedding representations generated by the multi-head attention mechanism are directly merged:

[0074]

[0075] Where R is the number of heads of multi-head attention, r = 1, 2...R. For the last output layer, the mean merging of multiple groups of target node embeddings generated by the multi-head attention mechanism is adopted:

[0076]

[0077] S300, then start the training process, during which the attention mechanism will automatically assign weights to simulate the influence of line impedance; input the training set into the constructed GAT graph attention network, output the predicted value of the flow data of the node through the GAT graph attention network, and set the model output to the voltage amplitude and active and reactive power of the destination node; calculate the loss through the predicted value of the flow data of the corresponding node and the real label of the flow data, and update the model parameters of the GAT graph attention network by the back propagation algorithm through the calculated loss. Finally, test the GAT graph attention network through the test set, and obtain the trained distribution network state estimation model after meeting the requirements.

[0078] S400: Estimating the power flow state of the target distribution network by using the trained distribution network state estimation model.

[0079] Based on the above implementation, the GAT-based distribution network voltage flow state estimation method provided in this embodiment can effectively evaluate the voltage status of the destination node in the distribution network and the flow status in the distribution network. In the case of line failure in the distribution network or temporary line cutting due to line maintenance, the distribution network flow will also change to a certain extent, which may also change the voltage status at each node. Rapid and timely reasonable evaluation of the flow in the distribution network and the voltage of the node of interest can provide support for the safe operation, operation and maintenance control, and quality and efficiency improvement of the power grid.

[0080] Embodiment 2:

[0081] This embodiment provides a distribution network voltage flow state estimation system based on GAT, including:

[0082] A data set construction module is used to obtain the historical measurement state of the target distribution network, obtain the power flow data of multiple time sections through simulation, and construct a sample set with the power flow data of multiple time sections as the real label; this module is used to implement the function of step S100 in the first embodiment, and will not be repeated here;

[0083] A GAT model construction module is used to construct a GAT graph attention network, set the number of nodes and connection relationships of the GAT graph attention network based on the topology information of the target distribution network, and input node feature information; this module is used to implement the function of step S200 in the first embodiment, which will not be repeated here;

[0084] A model training module is used to input the sample set into the constructed GAT graph attention network, output the power flow data prediction value of the node through the GAT graph attention network, calculate the loss through the power flow data prediction value of the corresponding node and the real label of the power flow data, and update the model parameters of the GAT graph attention network through the calculated loss to obtain a trained distribution network state estimation model; this module is used to implement the function of step S300 in the first embodiment, which will not be repeated here;

[0085] The power flow state estimation module is used to estimate the power flow state of the target distribution network through the trained distribution network state estimation model; this module is used to implement the function of step S400 in the first embodiment, which will not be described in detail here.

[0086] As a preferred implementation of this embodiment, in the step of constructing a sample set using the power flow data of multiple time sections as true labels, it also includes:

[0087] Based on the real multi-time section power flow data obtained by simulation, data expansion operation is performed, and a sample set is constructed with the expanded power flow data.

[0088] As a preferred implementation of this embodiment, the node feature information includes:

[0089] Node voltage amplitude, branch phase angle difference, node injected active power and node injected reactive power.

[0090] As a preferred implementation of this embodiment, the step of constructing the GAT graph attention network includes:

[0091] The graph data is constructed based on the known topological relationship of the power nodes in the target distribution network. The graph includes N power nodes. For the target power node, the attention mechanism is used to calculate the attention coefficient between it and its neighboring nodes, as follows:

[0092]

[0093] Among them, C ij is the attention coefficient, f(·,·) represents the attention calculation function, σ1 is the LeakyReLU function, represents the first parameter matrix to be trained, W represents the second parameter matrix to be trained, T represents the transpose, represents the feature embedding vector of the target power node i, represents the feature embedding vector of the jth neighbor node of the target power node i, || represents the series operation;

[0094] The calculated attention coefficients of all neighbor nodes are normalized to obtain the final weight distribution coefficients, as follows:

[0095]

[0096] Among them, N i is the set of neighbor nodes of target power node i, Represents the feature embedding vector of the kth neighbor node of the target power node i;

[0097] The output vector of the target power node is obtained according to the final weight allocation coefficient, as follows:

[0098]

[0099] in, is the output vector of the target power node i, and σ2 is the ELU activation function.

[0100] Embodiment three:

[0101] This embodiment proposes an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the method for estimating the voltage flow state of a distribution network based on GAT as described in any embodiment of the present invention is implemented.

[0102] Embodiment 4:

[0103] This embodiment provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the method for estimating the voltage flow state of a distribution network based on GAT as described in any embodiment of the present invention is implemented.

[0104] In the embodiments of the present application, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent the existence of A alone, the existence of A and B at the same time, and the existence of B alone. Among them, A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b and c can be represented by: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, c can be single or multiple.

[0105] Those of ordinary skill in the art will appreciate that the various units and algorithm steps described in the embodiments disclosed herein can be implemented in a combination of electronic hardware, computer software, and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0106] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0107] In several embodiments provided in the present application, if any function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory; hereinafter referred to as: ROM), random access memory (Random Access Memory; hereinafter referred to as: RAM), disk or optical disk, and other media that can store program codes.

[0108] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for estimating voltage flow state of distribution network based on GAT, characterized in that: The following steps are involved: Obtain the historical measurement state of the target distribution network, obtain the power flow data of multiple time sections through simulation, and construct a sample set using the power flow data of multiple time sections as the true label; Construct a GAT graph attention network, set the number of nodes and connection relationships of the GAT graph attention network based on the topology information of the target distribution network, and input node feature information; The sample set is input into the constructed GAT graph attention network, and the power flow data prediction value of the node is output through the GAT graph attention network. The loss is calculated by the power flow data prediction value of the corresponding node and the true label of the power flow data. The model parameters of the GAT graph attention network are updated by the calculated loss to obtain the trained distribution network state estimation model. The power flow state of the target distribution network is estimated through the trained distribution network state estimation model.

2. A method for estimating voltage flow state of a distribution network based on GAT according to claim 1, characterized in that: The step of constructing a sample set using the multi-time section flow data as the true label also includes: Based on the real multi-time section power flow data obtained by simulation, data expansion operation is performed, and a sample set is constructed with the expanded power flow data.

3. The method for estimating the voltage flow state of a distribution network based on GAT according to claim 1, characterized in that: The node feature information includes: Node voltage amplitude, branch phase angle difference, node injected active power and node injected reactive power.

4. The method for estimating the voltage flow state of a distribution network based on GAT according to claim 1, characterized in that: The steps of constructing the GAT graph attention network include: The graph data is constructed based on the known topological relationship of the power nodes in the target distribution network. The graph includes N power nodes. For the target power node, the attention mechanism is used to calculate the attention coefficient between it and its neighboring nodes, as follows: Among them, C ij is the attention coefficient, f(·,·) represents the attention calculation function, σ1 is the LeakyReLU function, represents the first parameter matrix to be trained, W represents the second parameter matrix to be trained, T represents the transpose, represents the feature embedding vector of the target power node i, represents the feature embedding vector of the jth neighbor node of the target power node i, || represents the series operation; The calculated attention coefficients of all neighbor nodes are normalized to obtain the final weight distribution coefficients, as follows: Among them, N i is the set of neighbor nodes of target power node i, Represents the feature embedding vector of the kth neighbor node of the target power node i; The output vector of the target power node is obtained according to the final weight allocation coefficient, as follows: in, is the output vector of the target power node i, and σ2 is the ELU activation function.

5. A distribution network voltage flow state estimation system based on GAT, characterized in that: include: The data set construction module is used to obtain the historical measurement state of the target distribution network, obtain the power flow data of multiple time sections through simulation, and construct the sample set with the power flow data of multiple time sections as the real label; The GAT model building module is used to build a GAT graph attention network, set the number of nodes and connection relationships of the GAT graph attention network based on the topological information of the target distribution network, and input node feature information; The model training module is used to input the sample set into the constructed GAT graph attention network, output the power flow data prediction value of the node through the GAT graph attention network, calculate the loss through the power flow data prediction value of the corresponding node and the true label of the power flow data, and update the model parameters of the GAT graph attention network through the calculated loss to obtain the trained distribution network state estimation model; The power flow state estimation module is used to estimate the power flow state of the target distribution network through the trained distribution network state estimation model.

6. A method for estimating voltage flow state of distribution network based on GAT according to claim 5, characterized in that: The step of constructing a sample set using the multi-time section flow data as the true label also includes: Based on the real multi-time section power flow data obtained by simulation, data expansion operation is performed, and a sample set is constructed with the expanded power flow data.

7. A distribution network voltage flow state estimation system based on GAT according to claim 5, characterized in that: The node feature information includes: Node voltage amplitude, branch phase angle difference, node injected active power and node injected reactive power.

8. A distribution network voltage flow state estimation system based on GAT according to claim 5, characterized in that: The steps of constructing the GAT graph attention network include: The graph data is constructed based on the known topological relationship of the power nodes in the target distribution network. The graph includes N power nodes. For the target power node, the attention mechanism is used to calculate the attention coefficient between it and its neighboring nodes, as follows: Among them, C ij is the attention coefficient, f(·,·) represents the attention calculation function, σ1 is the LeakyReLU function, represents the first parameter matrix to be trained, W represents the second parameter matrix to be trained, T represents the transpose, represents the feature embedding vector of the target power node i, represents the feature embedding vector of the jth neighbor node of the target power node i, || represents the series operation; The calculated attention coefficients of all neighbor nodes are normalized to obtain the final weight distribution coefficients, as follows: Among them, N i is the set of neighbor nodes of target power node i, Represents the feature embedding vector of the kth neighbor node of the target power node i; The output vector of the target power node is obtained according to the final weight allocation coefficient, as follows: in, is the output vector of the target power node i, and σ2 is the ELU activation function.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the GAT-based distribution network voltage flow state estimation method is implemented as described in any one of claims 1 to 4.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method for estimating the voltage flow state of a distribution network based on GAT as described in any one of claims 1 to 4 is implemented.

Images