Transformer area distribution network data sensing method and system based on space-time diagram neural network

Abstract

The invention discloses a transformer area distribution network data sensing method and system based on a space-time diagram neural network, and relates to the field of transformer area distribution networks, and the method comprises the steps: constructing an undirected graph according to an electrical network of a transformer area distribution network, generating an adjacent matrix, and obtaining a Laplacian matrix through normalization processing; constructing a time sequence diagram neural network model based on a Laplacian matrix; the method comprises the following steps: recording input characteristics and state estimation data of transformer area distribution network nodes, constructing label training data, and storing the label training data into a historical database for model training; training a time sequence diagram neural network model based on historical data, and performing distribution network state estimation by using the trained model; the system monitors the state of the power grid in real time through the cloud server and performs supplementary training on the model; the space-time diagram neural network provided by the invention combines the advantages of a diagram neural network and a sequence model, considers the topological structure of the power grid and the association between the nodes, and improves the accuracy of node sensing data.

Description

Technical Field

[0001] The present invention relates to the field of substation distribution networks, and in particular to a substation distribution network data perception method and system based on spatiotemporal graph neural network. Background Art

[0002] In the operation control of the new energy consumption of the distribution network in the substation, obtaining accurate voltage and power information of each node in the system is the basis of the operation control algorithm of the new energy consumption in the substation. Due to the influence of factors such as communication noise, environmental noise, parameter errors, and the lack of measurement devices at some points, the data directly measured often cannot fully meet the control requirements.

[0003] Distribution network data perception can calculate the current operating status data of the system based on various measurement information of the system, sampling state estimation and other methods. State estimation plays a vital role in the operation and control of power system operation and new energy consumption. The goal of state estimation in the distribution network is to estimate the state of the distribution system with high precision, completeness and reliability by using appropriate algorithms based on the measurement information of the distribution system and supplemented by pseudo-measurement data such as load forecasting. The characteristics of distribution network state estimation are that the number of input measurement data is small relative to the entire system and the data error is large. In the past, machine learning methods based on big data and neural networks usually only considered the numerical characteristics of a large amount of data, and trained the neural network model through a large amount of historical data. Such methods did not take into account the topological structure of the power grid, and did not build information such as the connection relationship between the calculation point and other power grid nodes and the operating status of the nodes associated with the calculation point into the neural network model. Summary of the invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by the present invention is: how to improve the perception and analysis capabilities of distribution network data in the substation area.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, an embodiment of the present invention provides a method for perceiving distribution network data in a substation area based on a spatiotemporal graph neural network, comprising:

[0008] According to the distribution network of the substation, an undirected graph is constructed to describe the topological relationship between the electrical branches and electrical nodes, an adjacency matrix is ​​formed, the adjacency matrix is ​​normalized, and a Laplace matrix is ​​constructed;

[0009] Constructing a sequential graph neural network model based on the Laplace matrix;

[0010] Through the data acquisition system, the input feature data and state estimation data of different nodes in the distribution network of the substation area are recorded, labeled training data is constructed, and stored in the historical database for training the time series graph neural network model;

[0011] Based on the historical database, train the time series graph neural network model;

[0012] Use the trained time-series graph neural network model to estimate the state of the distribution network in the substation area;

[0013] The grid operation status is monitored in real time through the cloud server, and the data is collected through the integrated terminal. The monitored time series data is stored in the historical database. When the collected new data reaches the set threshold, the time series graph neural network model is supplemented with training.

[0014] As a preferred solution for the data perception method of the distribution network in the substation area based on the spatiotemporal graph neural network, the following is proposed:

[0015] The forming of the adjacency matrix comprises:

[0016] Let the adjacency matrix be A. For each element A in A ij , if nodes i and j are adjacent, A ij =1, otherwise A ij =0;

[0017] The space-time graph is represented as Gt=(V, Et, At), where V is the node set, Et and At represent the edge set and adjacency matrix at time t respectively.

[0018] As a preferred solution for the data perception method of the distribution network in the substation area based on the spatiotemporal graph neural network, the following is proposed:

[0019] The normalizing of the adjacency matrix comprises:

[0020] By multiplying both sides of the adjacency matrix by the square root of the node degree and then taking the inverse, for each node i and j, the normalized matrix is The elements in are represented as:

[0021]

[0022] Among them, deg(v i ), deg(v j ) are the degrees of nodes i and j respectively, that is, the values ​​of the degree matrix at nodes i and j.

[0023] As a preferred solution for the data perception method of the distribution network in the substation area based on the spatiotemporal graph neural network, the following is proposed:

[0024] The constructing of the Laplace matrix comprises:

[0025] For a graph G = (V, E), the Laplacian matrix is ​​defined as L = DA, where L is the Laplacian matrix, D = diag (d) is the degree matrix of the vertices, The elements on the diagonal are the degrees of the vertices.

[0026] As a preferred solution for the data perception method of the distribution network in the substation area based on the spatiotemporal graph neural network, the following is proposed:

[0027] The time series graph neural network model includes 1 input layer, 1 graph convolution layer, 1 pooling layer and 1 output layer; the characteristics of each node in the input layer include input characteristics and state estimation data: the node input characteristics corresponding to the power grid node include active injection power and reactive injection power; based on the input characteristics of each node, a feature input matrix is ​​formed, and the active injection power and reactive injection power of each node are spliced ​​to form a feature input matrix; the state estimation data includes voltage amplitude and voltage phase angle.

[0028] As a preferred solution for the data perception method of the distribution network in the substation area based on the spatiotemporal graph neural network, the following is proposed:

[0029] The training time series graph neural network model includes:

[0030] The cross entropy loss function is used as the loss function. The formula of the cross entropy loss function is:

[0031]

[0032] Among them, y i is the actual label value, is the label value predicted by the model, and N is the number of samples;

[0033] The parameters for training include node features and edge weights.

[0034] As a preferred solution for the data perception method of the distribution network in the substation area based on the spatiotemporal graph neural network, the following is proposed:

[0035] The state estimation of the distribution network in the substation area using the trained time sequence graph neural network model includes:

[0036] Based on the trained timing graph neural network model, the active injection power and reactive injection power of each node in the power grid are input; the voltage amplitude and voltage phase angle of the candidate nodes in the power grid are output.

[0037] In a second aspect, an embodiment of the present invention provides a substation distribution network data perception system based on a spatiotemporal graph neural network, including:

[0038] The matrix construction module is used to construct an undirected graph based on the distribution network of the substation to describe the topological relationship between the electrical branches and electrical nodes, form an adjacency matrix, normalize the adjacency matrix, and construct a Laplace matrix;

[0039] A model building module, used for building a time series graph neural network model based on the Laplace matrix;

[0040] The acquisition module is used to record the input feature data and state estimation data of different nodes in the distribution network of the substation through the data acquisition system, construct labeled training data, and store it in the historical database for training the time series graph neural network model;

[0041] The training module is used to train the time series graph neural network model based on the historical database;

[0042] The state estimation module is used to estimate the state of the distribution network in the substation using the trained time series graph neural network model;

[0043] The supplementary training module is used to monitor the operation status of the power grid in real time through the cloud server, collect data through the integrated terminal, and store the monitored time series data in the historical database. When the collected new data reaches the set threshold, the time series graph neural network model is supplemented with training.

[0044] In a third aspect, an embodiment of the present invention provides a computing device, including:

[0045] Memory and processor;

[0046] The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the one or more programs are executed by the one or more processors, the one or more processors implement the substation distribution network data perception method based on spatiotemporal graph neural network as described in any embodiment of the present invention.

[0047] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the method for perceiving substation distribution network data based on a spatiotemporal graph neural network.

[0048] Beneficial effects of the present invention: The present invention adopts a spatiotemporal graph neural network model, which combines the advantages of a graph neural network and a sequence model, the former being used to process dependencies between time series, and the latter being used to learn time patterns. Benefited from the joint modeling of spatial and temporal patterns. The distribution network data of the substation area are spatiotemporal, and they are not only related to location, but also change over time. The spatiotemporal graph neural network used in the present invention takes into account the topological structure of the power grid, the association between the nodes of the power grid, and the mutual influence between the nodes during the construction and training process, and constructs this information in a graph convolutional neural network, so as to more accurately calculate the perception data of each node. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0050] Figure 1 It is an overall flow chart of the method for perceiving distribution network data in a substation based on a spatiotemporal graph neural network according to the first embodiment of the present invention;

[0051] Figure 2 It is a schematic diagram of the spatiotemporal graph neural network model of the substation distribution network data perception method based on the spatiotemporal graph neural network described in the first embodiment of the present invention. DETAILED DESCRIPTION

[0052] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0053] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0054] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0055] Example 1

[0056] Reference Figure 1 , which is the first embodiment of the present invention, and provides a method for perceiving distribution network data in a substation area based on a spatiotemporal graph neural network, comprising:

[0057] S1: Construct an undirected graph based on the distribution network of the substation to describe the topological relationship between the electrical branches and electrical nodes, form an adjacency matrix, normalize the adjacency matrix, and construct a Laplace matrix;

[0058] In the embodiment of the present application, let the adjacency matrix be A, for each element A in A ij , if nodes i and j are adjacent, A ij =1, otherwise A ij =0;

[0059] The spatiotemporal graph is represented as Gt = (V, Et, At), where V is the node set, Et and At represent the edge set and adjacency matrix at time t respectively;

[0060] Normalize the adjacency matrix A: multiply both sides of the adjacency matrix by the square root of the node degree and then take the inverse. For each node i and j, the normalized matrix is The elements in are represented as:

[0061]

[0062] Among them, deg(v i ), deg(v j ) are the degrees of nodes i and j respectively, that is, the values ​​of the degree matrix at nodes i and j.

[0063] Construct the Laplacian matrix: For a graph G = (V, E), the Laplacian matrix is ​​defined as L = DA, where L is the Laplacian matrix, D = diag (d) is the degree matrix of the vertex, The elements on the diagonal are the degrees of each vertex in turn;

[0064] S2: constructing a time series graph neural network model based on the Laplace matrix;

[0065] In an embodiment of the present application, a timing graph neural network model includes 1 input layer, 1 graph convolution layer, 1 pooling layer and 1 output layer; the characteristics of each node in the input layer include input characteristics and state estimation data: the node input characteristics corresponding to the power grid node include active injection power and reactive injection power; a feature input matrix is ​​formed based on the input characteristics of each node, and the active injection power and reactive injection power of each node are spliced ​​to form a feature input matrix; the state estimation data includes voltage amplitude and voltage phase angle.

[0066] Specifically, the graph convolution layer contains multiple diffusion convolution layers, and the diffusion convolution layers are connected by activation functions. After repeated experiments, the activation function Leaky ReLU has the best effect. Leaky ReLU gives all negative values ​​a non-zero slope. The mathematical expression is as follows:

[0067]

[0068] According to the results of multiple experiments, the best training effect is achieved when the parameter α is between 0.08 and 0.36.

[0069] The pooling layer is used to improve the robustness of the extracted features;

[0070] The output layer includes the state estimation output data of the candidate nodes in the distribution network of the substation, which is used for distribution network data perception.

[0071] S3: Through the data acquisition system, the input feature data and state estimation data of different nodes in the distribution network of the substation area are recorded, labeled training data is constructed, and stored in the historical database for training the time series graph neural network model;

[0072] It should be noted that by systematically collecting and storing the input features and state estimation data of each node in the substation distribution network, high-quality, labeled training data can be provided for the timing graph neural network model, thereby improving the accuracy and generalization ability of the model, and further enhancing the perception and analysis of distribution network data.

[0073] S4: Based on the historical database, train the time series graph neural network model;

[0074] In the embodiment of the present application, a cross entropy loss function is used as a loss function, and the formula of the cross entropy loss function is:

[0075]

[0076] Among them, y i is the actual label value, is the label value predicted by the model, and N is the number of samples;

[0077] It should be noted that the cross entropy loss function measures the performance of the model by calculating the difference between the actual value and the predicted value. When the predicted value is closer to the actual value, the value of the cross entropy loss function is smaller.

[0078] The parameters for training include node features and edge weights.

[0079] It should also be noted that node features and edge weights are two important parameters, which represent the attributes of each node in the graph and the connection relationship between nodes respectively. Node features can be multidimensional, containing various attribute information of nodes, while edge weights describe the strength of interaction between nodes. These parameters need to be optimized and adjusted during the training process through optimization algorithms such as gradient descent and back propagation algorithms to minimize prediction errors and improve the prediction accuracy of the model.

[0080] S5: Use the trained time series graph neural network model to estimate the state of the distribution network in the substation area;

[0081] In an embodiment of the present application, based on a trained timing graph neural network model, the active injection power and reactive injection power of each node in the power grid are input; and the voltage amplitude and voltage phase angle of the candidate nodes in the power grid are output.

[0082] S6: Monitor the grid operation status in real time through the cloud server, collect data through the integrated terminal, and store the monitored time series data in the historical database. When the collected new data reaches the set threshold, the time series graph neural network model is supplemented with training;

[0083] Specifically, in order to improve the accuracy of the model and adapt to new changes in grid parameters, the time series graph neural network can be retrained using the data stored in the historical database, or the model can be updated using the incremental training method. Depending on the actual situation, 10 days, 1 month, 3 months, etc. can be selected.

[0084] In another possible implementation, a substation distribution network data perception system is provided, which is composed of a cloud server, an integrated terminal and related sensors.

[0085] The cloud server includes cloud computing services, historical database services, real-time data cache services, message response services, etc. It is responsible for training the time-series graph neural network, data perception and calculation of the distribution network in the substation area, collecting real-time data, fault alarms, and four-remote data information uploaded by each integrated terminal, and relaying the communication between each integrated terminal;

[0086] The integrated terminal includes an intelligent micro-fusion terminal unit, an edge computing unit, a communication unit, and a fault recording unit, etc., which is responsible for the collection and communication services of data from metering devices, photovoltaics, wind turbines and other new energy equipment and other controllable equipment and sensors in the power grid.

[0087] The above is a schematic scheme of the method for perceiving data of a distribution network in a substation based on a spatiotemporal graph neural network in this embodiment. It should be noted that the technical scheme of the system for perceiving data of a distribution network in a substation based on a spatiotemporal graph neural network and the technical scheme of the method for perceiving data of a distribution network in a substation based on a spatiotemporal graph neural network belong to the same concept. For the details not described in detail in the technical scheme of the system for perceiving data of a distribution network in a substation based on a spatiotemporal graph neural network in this embodiment, please refer to the description of the technical scheme of the method for perceiving data of a distribution network in a substation based on a spatiotemporal graph neural network.

[0088] The substation distribution network data perception system based on the spatiotemporal graph neural network in this embodiment includes:

[0089] The matrix construction module is used to construct an undirected graph based on the distribution network of the substation to describe the topological relationship between the electrical branches and electrical nodes, form an adjacency matrix, normalize the adjacency matrix, and construct a Laplace matrix;

[0090] A model building module, used for building a time series graph neural network model based on the Laplace matrix;

[0091] The acquisition module is used to record the input feature data and state estimation data of different nodes in the distribution network of the substation through the data acquisition system, construct labeled training data, and store it in the historical database for training the time series graph neural network model;

[0092] The training module is used to train the time series graph neural network model based on the historical database;

[0093] The state estimation module is used to estimate the state of the distribution network in the substation using the trained time series graph neural network model;

[0094] The supplementary training module is used to monitor the operation status of the power grid in real time through the cloud server, collect data through the integrated terminal, and store the monitored time series data in the historical database. When the collected new data reaches the set threshold, the time series graph neural network model is supplemented with training.

[0095] This embodiment further provides a computing device, which is applicable to the case of a method for perceiving distribution network data in a substation area based on a spatiotemporal graph neural network, including:

[0096] Memory and processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the substation distribution network data perception method based on spatiotemporal graph neural network as proposed in the above embodiment.

[0097] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for perceiving substation distribution network data based on a spatiotemporal graph neural network as proposed in the above embodiment is implemented.

[0098] The storage medium proposed in this embodiment and the substation distribution network data perception method based on spatiotemporal graph neural network proposed in the above embodiment belong to the same inventive concept. The technical details not described in detail in this embodiment can be referred to the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0099] Example 2

[0100] The following is an embodiment of the present invention, which provides a method for perceiving distribution network data in a substation based on a spatiotemporal graph neural network. In order to verify the beneficial effects of the present invention, a scientific demonstration is carried out through simulation experiments.

[0101] 1. Graph construction and data preparation:

[0102] According to the electrical network topology of a certain distribution network, an undirected graph is constructed to describe the topological relationship between electrical branches and nodes.

[0103] Suppose the adjacency matrix is ​​a 5x5 matrix, which represents the connection between the five nodes in the graph.

[0104] Normalize the adjacency matrix. Assume the degree of the nodes is [2,3,2,1,2][2,3,2,1,2], then calculate the normalized matrix and construct the Laplacian matrix.

[0105] 2. Construction of time series graph neural network model:

[0106] Build a temporal graph neural network model consisting of 1 input layer, 1 graph convolution layer, 1 pooling layer, and 1 output layer.

[0107] The input layer includes the active injected power, reactive injected power, voltage amplitude and voltage phase angle of the node. Assume that the input feature matrix size is 5x4.

[0108] The graph convolution layer contains two diffusion convolution layers, each of which is followed by a Leaky ReLU activation function. The parameter α of the Leaky ReLU activation function is set to 0.1.

[0109] The pooling layer is used to enhance the robustness of the features. The output layer is used to estimate the status of candidate nodes in the distribution network, such as voltage amplitude and voltage phase angle.

[0110] 3. Data collection and training:

[0111] Use the data acquisition system to record the input feature data and state estimation data of different nodes. Construct labeled training data and store it in the historical database.

[0112] The collected data include node characteristics and status data, and 1 month of data records are stored in the historical database.

[0113] The time series graph neural network model is trained based on the historical database. The cross entropy loss function is used as the loss function, and the training data contains 10,000 samples. During the training process, the node features and edge weights are adjusted to minimize the prediction error.

[0114] 4. State estimation and model updating:

[0115] Use the trained model to estimate the state of the distribution network in the substation area. Input the active and reactive injected power of the node, and output the predicted voltage amplitude and voltage phase angle.

[0116] The cloud server monitors the power grid status in real time, and uses the integrated terminal to collect new time series data. When the new data reaches the set threshold for 1 month, the time series graph neural network model is supplemented or incrementally trained to improve the model accuracy.

[0117] Supplementary training:

[0118] Every month, the model is incrementally trained or retrained based on one month’s newly collected data. The updated model can better adapt to changes and new characteristics of the power grid.

[0119] Through simulation verification, this method can effectively improve the data perception accuracy and real-time performance of the substation distribution network, accurately predict voltage status and power demand changes, and ensure the stability and safety of power grid operation.

[0120] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for perceiving distribution network data in a substation area based on a spatiotemporal graph neural network, characterized in that: include: According to the distribution network of the substation, an undirected graph is constructed to describe the topological relationship between the electrical branches and electrical nodes, an adjacency matrix is ​​formed, the adjacency matrix is ​​normalized, and a Laplace matrix is ​​constructed; Constructing a sequential graph neural network model based on the Laplace matrix; Through the data acquisition system, the input feature data and state estimation data of different nodes in the distribution network of the substation area are recorded, labeled training data is constructed, and stored in the historical database for training the time series graph neural network model; Based on the historical database, train the time series graph neural network model; Use the trained time-series graph neural network model to estimate the state of the distribution network in the substation area; The grid operation status is monitored in real time through the cloud server, and the data is collected through the integrated terminal. The monitored time series data is stored in the historical database. When the collected new data reaches the set threshold, the time series graph neural network model is supplemented with training.

2. The method for perceiving distribution network data based on spatiotemporal graph neural network according to claim 1, characterized in that: The forming of the adjacency matrix comprises: Let the adjacency matrix be A. For each element A in A ij , if nodes i and j are adjacent, A ij =1, otherwise A ij =0; The space-time graph is represented as Gt=(V,Et,At), where V is the node set, Et and At represent the edge set and adjacency matrix at time t respectively.

3. The method for perceiving distribution network data based on spatiotemporal graph neural network according to claim 2, characterized in that: The normalizing of the adjacency matrix comprises: By multiplying both sides of the adjacency matrix by the square root of the node degree and then taking the inverse, for each node i and j, the normalized matrix is The elements in are represented as: Among them, deg(v i ),deg(v j ) are the degrees of nodes i and j respectively, that is, the values ​​of the degree matrix at nodes i and j.

4. The method for perceiving distribution network data based on spatiotemporal graph neural network according to claim 3, characterized in that: The constructing of the Laplace matrix comprises: For a graph G = (V, E), the Laplacian matrix is ​​defined as L = DA, where L is the Laplacian matrix and D = diag (d) is the degree matrix of the vertices. The elements on the diagonal are the degrees of the vertices.

5. The method for perceiving distribution network data based on spatiotemporal graph neural network according to claim 4, characterized in that: The time series graph neural network model includes 1 input layer, 1 graph convolution layer, 1 pooling layer and 1 output layer; the characteristics of each node in the input layer include input characteristics and state estimation data: the node input characteristics corresponding to the power grid node include active injection power and reactive injection power; based on the input characteristics of each node, a feature input matrix is ​​formed, and the active injection power and reactive injection power of each node are spliced ​​to form a feature input matrix; the state estimation data includes voltage amplitude and voltage phase angle.

6. The method for perceiving distribution network data based on spatiotemporal graph neural network according to claim 5, characterized in that: The training time series graph neural network model includes: The cross entropy loss function is used as the loss function. The formula of the cross entropy loss function is: Among them, y i is the actual label value, is the label value predicted by the model, and N is the number of samples; The parameters for training include node features and edge weights.

7. The method for perceiving distribution network data based on spatiotemporal graph neural network according to claim 6, characterized in that: The state estimation of the distribution network in the substation area using the trained time sequence graph neural network model includes: Based on the trained timing graph neural network model, the active injection power and reactive injection power of each node in the power grid are input; the voltage amplitude and voltage phase angle of the candidate nodes in the power grid are output.

8. A system using the method for sensing data of a distribution network in a substation area based on a spatiotemporal graph neural network as claimed in any one of claims 1 to 7, characterized in that: include: The matrix construction module is used to construct an undirected graph based on the distribution network of the substation to describe the topological relationship between the electrical branches and electrical nodes, form an adjacency matrix, normalize the adjacency matrix, and construct a Laplace matrix; A model building module, used for building a time series graph neural network model based on the Laplace matrix; The acquisition module is used to record the input feature data and state estimation data of different nodes in the distribution network of the substation through the data acquisition system, construct labeled training data, and store it in the historical database for training the time series graph neural network model; The training module is used to train the time series graph neural network model based on the historical database; The state estimation module is used to estimate the state of the distribution network in the substation using the trained time series graph neural network model; The supplementary training module is used to monitor the operation status of the power grid in real time through the cloud server, collect data through the integrated terminal, and store the monitored time series data in the historical database. When the collected new data reaches the set threshold, the time series graph neural network model is supplemented with training.

9. A computing device comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the substation distribution network data perception method based on spatiotemporal graph neural network as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the method for perceiving substation distribution network data based on a spatiotemporal graph neural network as described in any one of claims 1 to 7.

Images