Power grid topology identification method and system, terminal and storage medium

Abstract

The invention provides a power grid topology identification method and system, a terminal and a storage medium, and the method comprises the steps: obtaining the real-time data of a power grid, the real-time data of the power grid comprises the real-time operation data and the real-time topology data of the power grid, and the real-time operation data comprises the real-time operation information of each node and the line fault information; the real-time topological data comprises a connection relationship of each node; the power grid real-time data is input to a preset topology recognition model, so that the topology recognition model carries out topology recognition on the power grid according to the power grid real-time data, a topology recognition result is output, and the topology recognition result comprises a topology structure diagram, node state information, fault positioning information and future topology state information. And the accuracy and comprehensiveness of power grid topology identification are effectively improved.

Description

Technical Field

[0001] The present application relates to the technical field of power grid topology identification, and in particular to a power grid topology identification method, system, terminal and storage medium. Background Art

[0002] With the development of smart grids, the complexity and dynamic characteristics of power grids are increasing, and the demand for real-time identification and monitoring of power grid topology is becoming more urgent. Graph convolutional networks (GCNs) have gradually become a research hotspot for power grid topology identification due to their superiority in processing graph structured data, and can effectively capture the relationships and characteristics between power grid nodes.

[0003] Existing technical solutions mainly include topology recognition based on traditional machine learning methods, such as support vector machines (SVM) and decision trees; these methods often rely on manual feature extraction, and tend to ignore the correlation between nodes when dealing with complex power grid structures, resulting in reduced recognition accuracy. However, existing technologies have limited recognition capabilities under dynamic changes and abnormal conditions, making it difficult to adapt to the real-time monitoring needs of power grids. Summary of the invention

[0004] In response to the above technical problems, the present application provides a power grid topology identification method, system, terminal and storage medium, which effectively improve the accuracy and comprehensiveness of power grid topology identification.

[0005] In a first aspect, an embodiment of the present application provides a method for identifying a power grid topology, including:

[0006] Acquire real-time data of the power grid, wherein the real-time data of the power grid includes real-time operation data and real-time topology data of the power grid, wherein the real-time operation data includes real-time operation information of each node and line fault information, and the real-time topology data includes connection relationships of each node;

[0007] Inputting the real-time data of the power grid into a preset topology recognition model, so that the topology recognition model performs topology recognition on the power grid according to the real-time data of the power grid, and outputs a topology recognition result, wherein the topology recognition result includes a topology structure diagram, node status information, fault location information, and future topology status information;

[0008] Among them, the topology recognition model is obtained by combining a graph convolutional network model, a time series model and an abnormal state recognition model, and the graph convolutional network model and the time series model are obtained by training based on historical power grid data.

[0009] The embodiment of the present application provides a power grid topology identification method, which realizes the identification of the current topological structure of the power grid and the prediction of the future topological state of the power grid through a preset topological identification model. In the embodiment of the present application, since the graph convolutional network model represents a neural network structure specially designed to process graphic data, it can capture the complex relationship between nodes and the structural characteristics of the graph by performing convolution operations on the graph, and is therefore particularly suitable for processing topological data. At the same time, considering that the topological state of the power grid has characteristics such as dynamic changes over time and accompanied by abnormal states, the embodiment of the present application further combines the time series model and the abnormal state identification model to construct the topological identification model, so that the topological identification model has both the time series feature identification and abnormal state identification capabilities, which effectively improves the accuracy and comprehensiveness of the power grid topology identification.

[0010] Furthermore, the obtaining of real-time power grid data includes:

[0011] Acquiring the real-time operation data through a plurality of preset sensors;

[0012] The real-time topological data is obtained through a preset geographic information system.

[0013] In a possible implementation, the topology identification model performs topology identification on the power grid according to the real-time data of the power grid and outputs a topology identification result, including:

[0014] Inputting the real-time topology data in the real-time data of the power grid into the time series model, so that the time series model performs real-time inference according to the real-time topology data, and then predicts the future state of the power grid, and generates the current time series characteristics of the power grid and the future topology state information;

[0015] Inputting the real-time operation data in the real-time data of the power grid into the abnormal state recognition model, so that the abnormal state recognition model analyzes the sudden abnormality of the fault or overload in the real-time operation data, identifies and locates the problem area, and then generates the fault location information;

[0016] Performing format conversion on the real-time data of the power grid to obtain a corresponding graph data set, wherein the graph data set includes an adjacency matrix and a feature matrix, wherein the adjacency matrix is ​​used to store the connection relationship between each node in the power grid, and the feature matrix is ​​used to store the feature of each node;

[0017] Performing embedding processing on the graph data set, encoding the features of each node in the graph data set, and generating a corresponding embedding vector;

[0018] Inputting the embedding vector and the time series features into the graph convolutional network model, so that the graph convolutional network model extracts and aggregates the embedding vector and the time series features through a multi-layer convolution processing architecture and an adaptive feature weight optimization mechanism, thereby generating the topological structure diagram and the node status information;

[0019] The topology identification result is outputted by combining the topology structure diagram, node status information, fault location information and future topology status information.

[0020] The embodiment of the present application provides a method for topological identification of a power grid using a topological identification model. First, the future state of the power grid is inferred in real time based on the real-time topological data through a time series model, and the corresponding time series features and future topological state information are generated. Then, the real-time operation data of the power grid is analyzed using an abnormal state identification model to generate fault location information. Finally, the real-time data of the power grid is converted into an embedded vector, and the embedded vector and time series features are extracted and aggregated using the multi-layer convolution processing and adaptive feature weight optimization mechanism of the graph convolutional network model to generate a topological structure diagram and node state information. By combining time series features and real-time power grid data, the embodiment of the present application effectively integrates various data sources, can more accurately identify the topological structure of the power grid, provide a reliable basis for power grid operation management, reduce potential risks caused by topological identification errors, and improve the accuracy and comprehensiveness of power grid topological identification.

[0021] In a possible implementation, the step of obtaining the graph convolutional network model through training based on historical power grid data includes:

[0022] Acquiring historical data of the power grid, wherein the historical data of the power grid includes historical operation data and historical topology data of the power grid;

[0023] Based on the multi-layer graph convolutional network architecture and adaptive feature weight optimization mechanism, the initial graph convolutional network model is constructed;

[0024] The initial graph convolutional network model is trained according to the historical data of the power grid by using a supervised learning method to obtain the graph convolutional network model.

[0025] An embodiment of the present application provides a method for training a graph convolutional network model. First, an initial graph convolutional network model is constructed through a multi-layer graph convolutional network architecture and an adaptive feature weight optimization mechanism. The multi-layer graph convolutional network (GCN) architecture can process topological information of different levels of the power grid layer by layer. Each layer of GCN processes the topological data of the corresponding level and transmits information between layers. The bottom layer captures the connection relationship of local nodes, and the high layer analyzes the global power grid connectivity and energy flow. The adaptive feature weight optimization mechanism can calculate the importance score of each feature and dynamically adjust the weight of the corresponding feature, thereby improving the accuracy of power grid topology recognition.

[0026] Furthermore, the supervised learning method is used to train the initial graph convolutional network model according to the historical data of the power grid to obtain the graph convolutional network model, including:

[0027] Labeling the historical data of the power grid to obtain a number of corresponding labeled training data;

[0028] Input each of the training data into the initial graph convolutional network model so that the initial graph convolutional network model outputs corresponding prediction results, and calculates the difference between each of the prediction results and the corresponding labels according to a preset cross entropy loss function and a mean square error loss function, and then updates the model parameters according to the difference between each of the prediction results and the corresponding labels to obtain a first graph convolutional network model;

[0029] If the performance of the first graph convolutional network model meets the preset conditions, the first graph convolutional network model is used as the graph convolutional network model; otherwise, the first graph convolutional network model continues to be trained until the trained model meets the preset conditions.

[0030] In an embodiment of the present application, the initial graph convolutional network model is trained by a supervised learning method to obtain the graph convolutional network model. Specifically, a method of iteratively updating model parameters is adopted in the training process. In each iteration, the model is made to predict the training data, and then the difference between each prediction result and the corresponding label is calculated according to the cross entropy loss function and the mean square error loss function, and then the model parameters are updated according to the difference between each prediction result and the corresponding label, until the performance of the model meets the preset conditions, and the graph convolutional network model is obtained. The iterative training method proposed in the embodiment of the present application can make full use of the training data and avoid data waste, so that the final graph convolutional network model can accurately reflect the characteristics of the power grid, thereby improving the accuracy and efficiency of the entire power grid topology identification process.

[0031] In a possible implementation, the step of training the time series model according to historical power grid data includes:

[0032] Acquiring historical topology data from the power grid historical data;

[0033] Construct an initial time series model based on long short-term memory networks and gated recurrent units;

[0034] Constructing a time series input data set according to the state information of each node and edge in the historical topological data;

[0035] The time series input data set is input into the initial time series model, so that the initial time series model captures the dependency of the power grid topology changing with time according to the time series input data set, thereby obtaining the time series model.

[0036] The embodiment of the present application provides a training method for a time series model. First, a long short-term memory network and a gated recursive unit are selected to construct an initial time series model. This is because the long short-term memory network and the gated recursive unit are better at processing time series type data, and can effectively solve the problem of gradient disappearance or gradient explosion during the training process, and can learn long-term dependencies to improve the efficiency of model training. During the training process, the model identifies the dynamic evolution trend of the power grid structure by learning patterns in historical data. The fully trained model can perform real-time inference on the newly input power grid data, thereby predicting the future state of the power grid topology and improving the comprehensiveness of power grid topology recognition.

[0037] Furthermore, when the real-time data of the power grid is missing, incomplete or noisy, a preset noise reduction algorithm and uncertainty processing technology are used to complete the real-time data of the power grid.

[0038] In a second aspect, accordingly, an embodiment of the present application provides a power grid topology identification system, including an acquisition module and an identification module;

[0039] The acquisition module is used to acquire real-time data of the power grid, wherein the real-time data of the power grid includes real-time operation data and real-time topology data of the power grid, wherein the real-time operation data includes real-time operation information of each node and line fault information, and the real-time topology data includes connection relationships of each node;

[0040] The recognition module is used to input the real-time data of the power grid into a preset topology recognition model, so that the topology recognition model performs topology recognition on the power grid according to the real-time data of the power grid, and outputs a topology recognition result, wherein the topology recognition result includes a topology structure diagram, node status information, fault location information, and future topology status information;

[0041] Among them, the topology recognition model is obtained by combining a graph convolutional network model, a time series model and an abnormal state recognition model, and the graph convolutional network model and the time series model are obtained by training based on historical power grid data.

[0042] In a third aspect, an embodiment of the present application provides a terminal, comprising 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, it implements any one of the power grid topology identification methods described in the embodiments of the present application.

[0043] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium includes 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 any one of the power grid topology identification methods described in the embodiments of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 A flowchart of a power grid topology identification method provided in an embodiment of the present application.

[0045] Figure 2 A schematic diagram of a process of performing topology identification using a topology identification model in a power grid topology identification method provided in an embodiment of the present application.

[0046] Figure 3 A schematic diagram of the structure of a graph convolutional topology identification system provided in an embodiment of the present application.

[0047] Figure 4 A schematic diagram of the structure of a power grid topology identification system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0048] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0049] It should be noted that the step numbers in the text are only for the convenience of explanation of the specific embodiments and do not serve to limit the order in which the steps are executed. In the description of this application, 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. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features.

[0050] Embodiment 1:

[0051] like Figure 1 As shown, embodiment 1 provides a power grid topology identification method, including step S1 and step S2:

[0052] Step S1, acquiring real-time data of the power grid, wherein the real-time data of the power grid includes real-time operation data and real-time topology data of the power grid, wherein the real-time operation data includes real-time operation information of each node and line fault information, and the real-time topology data includes connection relationships of each node;

[0053] Step S2: inputting the real-time data of the power grid into a preset topology recognition model, so that the topology recognition model performs topology recognition on the power grid according to the real-time data of the power grid, and outputs a topology recognition result, wherein the topology recognition result includes a topology structure diagram, node status information, fault location information, and future topology status information;

[0054] Among them, the topology recognition model is obtained by combining a graph convolutional network model, a time series model and an abnormal state recognition model, and the graph convolutional network model and the time series model are obtained by training based on historical power grid data.

[0055] In a preferred embodiment, the topological structure diagram of the power grid is used to display an overall diagram of the power grid nodes and connection relationships; the node status information is used to reflect the real-time operating status of each node, including parameters such as voltage and current; and the fault location information is used to identify the node or line location where a fault occurs in the power grid.

[0056] The embodiment of the present application provides a power grid topology identification method, which realizes the identification of the current topological structure of the power grid and the prediction of the future topological state of the power grid through a preset topological identification model. In the embodiment of the present application, since the graph convolutional network model represents a neural network structure specially designed to process graphic data, it can capture the complex relationship between nodes and the structural characteristics of the graph by performing convolution operations on the graph, and is therefore particularly suitable for processing topological data. At the same time, considering that the topological state of the power grid has characteristics such as dynamic changes over time and accompanied by abnormal states, the embodiment of the present application further combines the time series model and the abnormal state identification model to construct the topological identification model, so that the topological identification model has both the time series feature identification and abnormal state identification capabilities, which effectively improves the accuracy and comprehensiveness of the power grid topology identification.

[0057] Furthermore, in step S1, obtaining real-time data of the power grid includes:

[0058] Acquiring the real-time operation data through a plurality of preset sensors;

[0059] The real-time topological data is obtained through a preset geographic information system.

[0060] In a possible implementation, in step S2, the topology recognition model performs topology recognition on the power grid according to the real-time data of the power grid, and outputs a topology recognition result, such as Figure 2 As shown, steps S201-S206 are included:

[0061] Step S201: input the real-time topology data in the real-time data of the power grid into the time series model, so that the time series model performs real-time inference according to the real-time topology data, and then predicts the future state of the power grid, and generates the current time series characteristics of the power grid and the future topology state information;

[0062] Step S202: inputting the real-time operation data in the real-time data of the power grid into the abnormal state recognition model, so that the abnormal state recognition model analyzes the sudden abnormality of the fault or overload in the real-time operation data, identifies and locates the problem area, and then generates the fault location information;

[0063] Step S203: convert the format of the real-time data of the power grid to obtain a corresponding graph data set, wherein the graph data set includes an adjacency matrix and a feature matrix, wherein the adjacency matrix is ​​used to store the connection relationship between each node in the power grid, and the feature matrix is ​​used to store the feature of each node;

[0064] Step S204: embedding the graph data set, encoding the features of each node in the graph data set, and generating a corresponding embedding vector;

[0065] Step S205: input the embedding vector and the time series features into the graph convolutional network model, so that the graph convolutional network model extracts and aggregates the embedding vector and the time series features through a multi-layer convolution processing architecture and an adaptive feature weight optimization mechanism, thereby generating the topological structure diagram and the node status information;

[0066] Step S206: output the topology recognition result by combining the topology structure diagram, node status information, fault location information and future topology status information.

[0067] The embodiment of the present application provides a method for topological identification of a power grid using a topological identification model. First, the future state of the power grid is inferred in real time based on the real-time topological data through a time series model, and the corresponding time series features and future topological state information are generated. Then, the real-time operation data of the power grid is analyzed using an abnormal state identification model to generate fault location information. Finally, the real-time data of the power grid is converted into an embedded vector, and the embedded vector and time series features are extracted and aggregated using the multi-layer convolution processing and adaptive feature weight optimization mechanism of the graph convolutional network model to generate a topological structure diagram and node state information. By combining time series features and real-time power grid data, the embodiment of the present application effectively integrates various data sources, can more accurately identify the topological structure of the power grid, provide a reliable basis for power grid operation management, reduce potential risks caused by topological identification errors, and improve the accuracy and comprehensiveness of power grid topological identification.

[0068] In a preferred embodiment, in step S203, the nodes and edges of the power grid are first determined to construct the adjacency matrix and feature matrix of the power grid according to the real-time data of the power grid. The nodes represent the various devices or sites in the power grid, and the edges represent the connection relationship between the nodes. The adjacency matrix is ​​a square matrix, in which each element indicates whether there is a direct connection between two nodes, usually represented by 0 and 1. The feature matrix is ​​used to store the characteristics of each node, such as load, power generation capacity, voltage and other information, each row corresponds to a node, and each column corresponds to a feature. By combining these two matrices, the topological structure of the power grid and its related features can be effectively represented, providing basic data for subsequent graph convolution network analysis. Furthermore, while acquiring data, the real-time state of the power grid is periodically monitored, the current state is analyzed in real time, and the node and edge features are adjusted according to the monitoring results to achieve adaptive adjustment.

[0069] In a preferred embodiment, in step S204, the embedding vector of the node is specifically expressed as a vector, which encodes the features of each node through a graph convolutional network, compressing high-dimensional node features into low-dimensional vector representations, which not only retain the node's own information, but also contain the structural information and feature relationships of other nodes connected to it. The final embedding form is usually a dense vector with a lower dimension, each vector representing the position of the node in the graph network and its feature correlation, thereby facilitating subsequent model training and inference.

[0070] In a preferred embodiment, in step S205, the time series features output by the time series model are passed as input to the graph convolutional network model, combined with the embedding vector, and the recognition process of the power grid topology structure is dynamically adjusted to generate the topology structure diagram and the node status information.

[0071] The topology recognition model provided in the embodiment of the present application integrates a time series model, an abnormal state recognition model and a graph convolutional network model, and can support real-time monitoring and management, and output the current state information of the power grid, including the recognition results of the topology structure diagram, node state information and fault location information, so as to help operation and maintenance personnel quickly understand the current state of the power grid and take corresponding measures; at the same time, the topology recognition model can also perform real-time inference on the input power grid data, so as to predict the future state of the power grid topology, including potential topology changes, node state trends and possible fault locations, and output these prediction results together with the current state information.

[0072] In a preferred embodiment, the topology recognition model can be further combined with spatiotemporal feature modeling to perform convolution processing on the spatiotemporal data of the power grid to capture the characteristics of the power grid changing with time and space; the specific operation is to combine the time series data with the spatial data, and extract the spatiotemporal features through convolution operations, so as to more accurately identify dynamic changes.

[0073] In a possible implementation, the step of obtaining the graph convolutional network model through training based on historical power grid data includes:

[0074] Acquiring historical data of the power grid, wherein the historical data of the power grid includes historical operation data and historical topology data of the power grid;

[0075] Based on the multi-layer graph convolutional network architecture and adaptive feature weight optimization mechanism, the initial graph convolutional network model is constructed;

[0076] The initial graph convolutional network model is trained according to the historical data of the power grid by using a supervised learning method to obtain the graph convolutional network model.

[0077] An embodiment of the present application provides a method for training a graph convolutional network model. First, an initial graph convolutional network model is constructed through a multi-layer graph convolutional network architecture and an adaptive feature weight optimization mechanism. The multi-layer graph convolutional network (GCN) architecture can process topological information of different levels of the power grid layer by layer. Each layer of GCN processes the topological data of the corresponding level and transmits information between layers. The bottom layer captures the connection relationship of local nodes, and the high layer analyzes the global power grid connectivity and energy flow. The adaptive feature weight optimization mechanism can calculate the importance score of each feature and dynamically adjust the weight of the corresponding feature, thereby improving the accuracy of power grid topology recognition.

[0078] In a preferred embodiment, further, the supervised learning method is used to train the initial graph convolutional network model according to the historical data of the power grid to obtain the graph convolutional network model, including:

[0079] Labeling the historical data of the power grid to obtain corresponding labeled training data;

[0080] Input each of the training data into the initial graph convolutional network model so that the initial graph convolutional network model outputs corresponding prediction results, and calculates the difference between each of the prediction results and the corresponding labels according to a preset cross entropy loss function and a mean square error loss function, and then updates the model parameters according to the difference between each of the prediction results and the corresponding labels to obtain a first graph convolutional network model;

[0081] If the performance of the first graph convolutional network model meets the preset conditions, the first graph convolutional network model is used as the graph convolutional network model; otherwise, the first graph convolutional network model continues to be trained until the trained model meets the preset conditions.

[0082] In an embodiment of the present application, the initial graph convolutional network model is trained by a supervised learning method to obtain the graph convolutional network model. Specifically, a method of iteratively updating model parameters is adopted in the training process. In each iteration, the model is made to predict the training data, and then the difference between each prediction result and the corresponding label is calculated according to the cross entropy loss function and the mean square error loss function, and then the model parameters are updated according to the difference between each prediction result and the corresponding label, until the performance of the model meets the preset conditions, and the graph convolutional network model is obtained. The iterative training method proposed in the embodiment of the present application can make full use of the training data and avoid data waste, so that the final graph convolutional network model can accurately reflect the characteristics of the power grid, thereby improving the accuracy and efficiency of the entire power grid topology identification process.

[0083] In a possible implementation, the step of training the time series model according to historical power grid data includes:

[0084] Acquiring historical topology data from the power grid historical data;

[0085] Construct an initial time series model based on long short-term memory networks and gated recurrent units;

[0086] Constructing a time series input data set according to the state information of each node and edge in the historical topological data;

[0087] The time series input data set is input into the initial time series model, so that the initial time series model captures the dependency of the power grid topology changing with time according to the time series input data set, thereby obtaining the time series model.

[0088] The embodiment of the present application provides a training method for a time series model. First, a long short-term memory network and a gated recursive unit are selected to construct an initial time series model. This is because the long short-term memory network and the gated recursive unit are better at processing time series type data, and can effectively solve the problem of gradient disappearance or gradient explosion during the training process, and can learn long-term dependencies to improve the efficiency of model training. During the training process, the model identifies the dynamic evolution trend of the power grid structure by learning patterns in historical data. The fully trained model can perform real-time inference on the newly input power grid data, thereby predicting the future state of the power grid topology and improving the comprehensiveness of power grid topology recognition.

[0089] Furthermore, when the real-time data of the power grid is missing, incomplete or noisy, a preset noise reduction algorithm and uncertainty processing technology are used to complete the real-time data of the power grid.

[0090] In summary, the embodiments of the present application have the following beneficial effects:

[0091] 1. The embodiment of the present application calculates the feature importance score and adjusts the weight through an adaptive feature weight optimization mechanism; at the same time, the multi-level topology structure processes information at different levels, performs all-round analysis from local to global, effectively integrates various data sources, and can more accurately identify the power grid topology structure, providing a reliable basis for power grid operation management, and reducing potential risks caused by topology identification errors.

[0092] 2. The embodiments of the present application combine spatiotemporal feature modeling, time series modeling, and abnormal state recognition capabilities; they can accurately capture the power grid status as it changes dynamically over time and space, and use relevant technologies to effectively handle abnormal situations such as data missing and noise, maintain efficient topology recognition, and ensure the effectiveness and stability of real-time monitoring and management of the power grid.

[0093] 3. The embodiment of the present application acquires data through multiple channels to construct an adjacency matrix and a feature matrix, which comprehensively covers the power grid related information; the modules collaborate to make full use of these data in different processing stages, such as accurately reflecting the power grid characteristics when training the model, avoiding data waste, and thus improving the accuracy and efficiency of the entire power grid topology identification process.

[0094] Embodiment 2:

[0095] like Figure 3 As shown, Example 2 provides a graph convolution topology identification system, which integrates the power grid topology identification method as described in Example 1, including a data collection module, a construction and training module, and a topology identification module. By combining a graph convolutional network, an adaptive feature weight optimization mechanism, multi-level topology structure processing, and time series modeling, it can accurately identify and predict the topological structure of the power grid and its dynamic evolution.

[0096] The graph convolution topology identification system described in Example 2 acquires data through a data collection module; the module deploys a large number of sensors, and combines historical operation data and geographic information systems (GIS) to comprehensively collect node and edge features in the power grid; nodes usually represent power stations, substations and load centers in the power grid, while edges represent the connection relationship between these nodes; in order to characterize the topological structure of the power grid, the data collection module constructs two core data structures: an adjacency matrix and a feature matrix; the adjacency matrix is ​​a square matrix used to describe the connection relationship between power grid nodes, where 1 indicates that there is a connection between nodes and 0 indicates that there is no connection; the feature matrix is ​​used to store the feature information of each node, such as voltage, current, power, etc.; at the same time, the data collection module cleans and standardizes the collected data, removes outliers and noise, and ensures the accuracy and consistency of the data.

[0097] After data collection is completed, the data is input into the construction and training module. The core of this module is a multi-level graph convolutional network architecture. First, the adjacency matrix and feature matrix are input into the graph convolutional network for feature extraction. The graph convolutional network consists of multiple convolutional layers. Each layer captures the local information of nodes and edges through convolution operations, and transmits global information layer by layer. The activation function is introduced into the convolutional layer to enhance the nonlinear expression ability of the model. During the convolution process, the adaptive feature weight optimization mechanism dynamically adjusts the feature weight of each node and edge to improve the model's ability to recognize the topological structure of the power grid. The working principle of this optimization mechanism is to calculate the importance score of each feature through the attention mechanism, and dynamically adjust the weight according to these scores, so that under different power grid conditions, the model can flexibly adjust its focus.

[0098] Then it enters the training phase of the supervised learning model. The construction and training module uses the cross entropy loss function or the mean square error loss function to train the model. By comparing the difference between the predicted output and the actual topology label, the model parameters are continuously adjusted. The training process is an iterative process. The model learns more grid structure features in each iteration and updates the parameters in real time until the model error converges to an acceptable range. Through this training, it is possible to learn how to infer the topological structure of the grid from the input data and identify its dynamic evolution.

[0099] After the training of the graph convolutional topology recognition system is completed, the model is integrated into the topology recognition module. This module combines time series modeling and abnormal state recognition capabilities to monitor and analyze the topological structure of the power grid in real time. It can not only output the current topological structure of the power grid, but also capture the dynamic changes of the power grid over time. By analyzing the historical topological data, it can predict the future state of the power grid. Especially in the event of emergencies or anomalies, the module can quickly identify the problem and output the topological structure diagram, node status information and fault location information; it can help operation and maintenance personnel quickly understand the current state of the power grid and take appropriate measures to prevent potential power grid failures.

[0100] During the recognition process of the topology recognition module, it is necessary to obtain a variety of power grid data in real time, including operating parameters of node voltage, current, power and load information, edge connection status, load conditions and topology change data, and also collect environmental data such as temperature, weather impact and line fault information. These data are updated in real time through the sensor network and monitoring system and transmitted to the recognition module.

[0101] During the recognition process of the topology recognition module, the power grid data obtained from the data collection module will be sent as input to the trained graph convolutional network model, and then the format conversion will be performed first to convert the real-time data of the power grid into a graph data structure, including an adjacency matrix and a feature matrix, and update the structure of the graph according to the changes in nodes and edges; the model embeds these input data and converts the state features of nodes and edges into representation vectors in a high-dimensional space; then, the input graph data is feature extracted and aggregated through multi-layer convolution processing and an adaptive feature weight optimization mechanism, and the recognition results, including the current topology structure and node status, are output.

[0102] When the graph convolutional network model is combined with time series modeling and abnormal state recognition capabilities, the LSTM and GRU time series models are first used to model historical data to capture the time dependency of the power grid operation status. The output time series features are then passed into the graph convolutional network as input and combined with real-time power grid data to dynamically adjust the recognition process of the power grid topology structure. The abnormal state recognition part analyzes sudden anomalies such as faults or overloads in the input data, identifies and locates the problem area, and generates a comprehensive recognition result including a topology diagram, node status information, and fault location information.

[0103] Embodiment three:

[0104] The method used in the third embodiment is similar to that used in the second embodiment, except that the graph convolution topology identification system in the third embodiment is particularly suitable for processing multi-level features of regional power grids, urban power grids and local distribution networks.

[0105] It is mainly reflected in the fact that the embodiment of the present application adopts a multi-level graph convolutional network model to process the multi-level features of regional power grids, urban power grids and local distribution networks; in specific operations, by designing independent graph convolutional network layers for each level, the power grid information of different levels is processed layer by layer: for the global power grid layer, the model focuses on capturing the connectivity of power grids in a large range; for the urban power grid layer, the model pays more attention to the connection relationship of medium-sized nodes and the steady-state of local networks; and in the local distribution network layer, the model refines the detailed connection and load characteristics between local nodes; and the model dynamically adjusts the feature weights of nodes and edges in each level through an adaptive feature weight optimization mechanism to adapt to the power grid needs at different levels, thereby improving the recognition accuracy of multi-level features.

[0106] The graph convolution topology identification system provided in the embodiment of the present application uses sensors deployed at different levels through data collection modules to comprehensively monitor the power grid and collect topological data of regional power grids, urban power grids and local distribution networks; the data collection module adopts a periodic monitoring method to ensure that the power grid status at each level can be obtained in real time; by combining sensor data, historical operation records and geographic information system (GIS) data, the adjacency matrix and feature matrix of the power grid can be effectively constructed; similar to the first implementation example, the adjacency matrix is ​​used to describe the connection relationship between nodes in the power grid, and the feature matrix is ​​used to store the voltage, power and load attributes of each node; during the data collection process, data cleaning and standardization will be automatically performed to ensure that the power grid data at each level is of good quality.

[0107] Then the graph convolutional topology recognition system inputs the collected data into the construction and training module, in which a multi-level GCN model is used to process the topological information of the power grid; first, the adjacency matrix and feature matrix are input into the underlying GCN network, which is responsible for capturing the node connection relationship of the local distribution network. As the convolutional layer is passed layer by layer, it will gradually process higher-level power grid topology information, such as the connectivity of urban power grids and regional power grids; the characteristic of the layer GCN model is that it can simultaneously capture the fine-grained connection relationship between local nodes and the global characteristics of the power grid as a whole; each layer of the GCN network focuses on the power grid data at a specific level, and integrates global and local information through inter-layer information transmission; in this way, it can maintain efficient recognition capabilities when processing complex multi-level topological structures.

[0108] The adaptive feature weight optimization mechanism in this module further improves the recognition accuracy of the model; by calculating the feature importance scores of nodes and edges at each level, the weights of the corresponding features can be adjusted according to the scores, so as to flexibly adjust the focus of the model at different power grid levels; in this way, when dealing with regional power grids, more attention can be paid to the global connectivity of the power grid, and when dealing with local distribution networks, more attention can be paid to the specific connection relationship between nodes.

[0109] The supervised learning method is used in the training process, and the model is trained through the labeled power grid topology data. The model adjusts the model parameters through the cross entropy loss or mean square error loss function to gradually improve the model's prediction ability. After each iteration, the model will update the parameters according to the difference between the predicted results and the actual results. Finally, the model can accurately identify the topological structure of multi-level power grids.

[0110] After training, the topology recognition module is responsible for real-time recognition and management of the topological structure of the power grid. This module combines time series modeling and can capture the dynamic changes of the power grid by analyzing the topological data of the power grid that changes over time. Through the modeling of historical data, the future evolution of the power grid topology can be predicted. In addition, the module also has the ability to recognize abnormal states and can respond quickly when sudden failures occur in the power grid, outputting detailed topological structure diagrams, node status information and fault location information. The information provided can help power grid operation and maintenance personnel to promptly discover problems, perform power grid maintenance and management, and ensure the safe operation of the power grid.

[0111] Embodiment 4:

[0112] like Figure 4 As shown, the fourth embodiment provides a power grid topology identification system, including an acquisition module 10 and an identification module 20;

[0113] The acquisition module 10 is used to acquire real-time data of the power grid, wherein the real-time data of the power grid includes real-time operation data and real-time topology data of the power grid, wherein the real-time operation data includes real-time operation information of each node and line fault information, and the real-time topology data includes connection relationships of each node;

[0114] The identification module 20 is used to input the real-time data of the power grid into a preset topology identification model, so that the topology identification model performs topology identification on the power grid according to the real-time data of the power grid, and outputs a topology identification result, wherein the topology identification result includes a topology structure diagram, node status information, fault location information, and future topology status information;

[0115] Among them, the topology recognition model is obtained by combining a graph convolutional network model, a time series model and an abnormal state recognition model, and the graph convolutional network model and the time series model are obtained by training based on historical power grid data.

[0116] Furthermore, the acquisition module 10 acquires real-time data of the power grid, including:

[0117] Acquiring the real-time operation data through a plurality of preset sensors;

[0118] The real-time topological data is obtained through a preset geographic information system.

[0119] In a possible implementation, the identification module 20 performs topology identification on the power grid according to the real-time data of the power grid and outputs a topology identification result, including:

[0120] Inputting the real-time topology data in the real-time data of the power grid into the time series model, so that the time series model performs real-time inference according to the real-time topology data, and then predicts the future state of the power grid, and generates the current time series characteristics of the power grid and the future topology state information;

[0121] Inputting the real-time operation data in the real-time data of the power grid into the abnormal state recognition model, so that the abnormal state recognition model analyzes the sudden abnormality of the fault or overload in the real-time operation data, identifies and locates the problem area, and then generates the fault location information;

[0122] Performing format conversion on the real-time data of the power grid to obtain a corresponding graph data set, wherein the graph data set includes an adjacency matrix and a feature matrix, wherein the adjacency matrix is ​​used to store the connection relationship between each node in the power grid, and the feature matrix is ​​used to store the feature of each node;

[0123] Performing embedding processing on the graph data set, encoding the features of each node in the graph data set, and generating a corresponding embedding vector;

[0124] Inputting the embedding vector and the time series features into the graph convolutional network model, so that the graph convolutional network model extracts and aggregates the embedding vector and the time series features through a multi-layer convolution processing architecture and an adaptive feature weight optimization mechanism, thereby generating the topological structure diagram and the node status information;

[0125] The topology identification result is outputted by combining the topology structure diagram, node status information, fault location information and future topology status information.

[0126] In a possible implementation, the step of obtaining the graph convolutional network model through training based on historical power grid data includes:

[0127] Acquiring historical data of the power grid, wherein the historical data of the power grid includes historical operation data and historical topology data of the power grid;

[0128] Based on the multi-layer graph convolutional network architecture and adaptive feature weight optimization mechanism, the initial graph convolutional network model is constructed;

[0129] The initial graph convolutional network model is trained according to the historical data of the power grid by using a supervised learning method to obtain the graph convolutional network model.

[0130] Furthermore, the supervised learning method is used to train the initial graph convolutional network model according to the historical data of the power grid to obtain the graph convolutional network model, including:

[0131] Labeling the historical data of the power grid to obtain corresponding labeled training data;

[0132] Input each of the training data into the initial graph convolutional network model so that the initial graph convolutional network model outputs corresponding prediction results, and calculates the difference between each of the prediction results and the corresponding labels according to a preset cross entropy loss function and a mean square error loss function, and then updates the model parameters according to the difference between each of the prediction results and the corresponding labels to obtain a first graph convolutional network model;

[0133] If the performance of the first graph convolutional network model meets the preset conditions, the first graph convolutional network model is used as the graph convolutional network model; otherwise, the first graph convolutional network model continues to be trained until the trained model meets the preset conditions.

[0134] In a possible implementation, the step of training the time series model according to historical power grid data includes:

[0135] Acquiring historical topology data from the power grid historical data;

[0136] Construct an initial time series model based on long short-term memory networks and gated recurrent units;

[0137] Constructing a time series input data set according to the state information of each node and edge in the historical topological data;

[0138] The time series input data set is input into the initial time series model, so that the initial time series model captures the dependency of the power grid topology changing with time according to the time series input data set, thereby obtaining the time series model.

[0139] In one possible implementation, the power grid topology identification system also includes a data completion module, which is used to complete the real-time power grid data using a preset noise reduction algorithm and uncertainty processing technology when the real-time power grid data is missing, incomplete or noisy.

[0140] The embodiment of the present application provides a power grid topology identification system, which can identify the current topological structure of the power grid and predict the future topological state of the power grid through a preset topological identification model. In the embodiment of the present application, since the graph convolutional network model represents a neural network structure specially designed to process graphic data, it can capture the complex relationship between nodes and the structural characteristics of the graph by performing convolution operations on the graph, so it is particularly suitable for processing topological data. At the same time, considering that the topological state of the power grid has characteristics such as dynamic changes over time and accompanied by abnormal states, the embodiment of the present application further combines the time series model and the abnormal state identification model to construct the topological identification model, so that the topological identification model has both time series feature identification and abnormal state identification capabilities, which effectively improves the accuracy and comprehensiveness of power grid topology identification.

[0141] The more detailed working principle and step flow of this embodiment can refer to, but are not limited to, the relevant records of Embodiment 1.

[0142] Embodiment five:

[0143] Embodiment 5 provides a terminal, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor implements the power grid topology identification method described in the present application when executing the computer program.

[0144] The processor may be a central processing unit (CPU), or 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. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the terminal, and various interfaces and lines are used to connect various parts of the entire terminal.

[0145] The memory can be used to store the computer program, and the processor realizes various functions of the terminal by running or executing the computer program stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (SmartMedia Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0146] Embodiment six:

[0147] Embodiment 6 provides a computer-readable storage medium, wherein the computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the power grid topology identification method described in the present application.

[0148] Wherein, if the module integrated in the power grid topology identification method 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 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 may 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), electrical carrier signal, telecommunication signal and software distribution medium, etc.

[0149] The specific embodiments described above further describe the purpose, technical solutions and beneficial effects of the present application in detail. It should be understood that the above description is only a specific embodiment of the present application and is not intended to limit the scope of protection of the present application. It is particularly pointed out that for those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of protection of the present application.

Claims

1. A method for identifying power grid topology, characterized in that: include: Acquire real-time data of the power grid, wherein the real-time data of the power grid includes real-time operation data and real-time topology data of the power grid, wherein the real-time operation data includes real-time operation information of each node and line fault information, and the real-time topology data includes connection relationships of each node; Inputting the real-time data of the power grid into a preset topology recognition model, so that the topology recognition model performs topology recognition on the power grid according to the real-time data of the power grid, and outputs a topology recognition result, wherein the topology recognition result includes a topology structure diagram, node status information, fault location information, and future topology status information; Among them, the topology recognition model is obtained by combining a graph convolutional network model, a time series model and an abnormal state recognition model, and the graph convolutional network model and the time series model are obtained by training based on historical power grid data.

2. A method for identifying power grid topology according to claim 1, characterized in that: The obtaining of real-time power grid data includes: Acquiring the real-time operation data through a plurality of preset sensors; The real-time topological data is obtained through a preset geographic information system.

3. A method for identifying power grid topology according to claim 1, characterized in that: The topology recognition model performs topology recognition on the power grid according to the real-time data of the power grid and outputs a topology recognition result, including: Inputting the real-time topology data in the real-time data of the power grid into the time series model, so that the time series model performs real-time inference according to the real-time topology data, and then predicts the future state of the power grid, and generates the current time series characteristics of the power grid and the future topology state information; Inputting the real-time operation data in the real-time data of the power grid into the abnormal state recognition model, so that the abnormal state recognition model analyzes the sudden abnormality of the fault or overload in the real-time operation data, identifies and locates the problem area, and then generates the fault location information; Performing format conversion on the real-time data of the power grid to obtain a corresponding graph data set, wherein the graph data set includes an adjacency matrix and a feature matrix, wherein the adjacency matrix is ​​used to store the connection relationship between each node in the power grid, and the feature matrix is ​​used to store the feature of each node; Performing embedding processing on the graph data set, encoding the features of each node in the graph data set, and generating a corresponding embedding vector; Inputting the embedding vector and the time series features into the graph convolutional network model, so that the graph convolutional network model extracts and aggregates the embedding vector and the time series features through a multi-layer convolution processing architecture and an adaptive feature weight optimization mechanism, thereby generating the topological structure diagram and the node status information; The topology identification result is outputted by combining the topology structure diagram, node status information, fault location information and future topology status information.

4. A method for identifying power grid topology according to claim 1, characterized in that: The step of obtaining the graph convolutional network model through training based on historical power grid data includes: Acquiring historical data of the power grid, wherein the historical data of the power grid includes historical operation data and historical topology data of the power grid; Based on the multi-layer graph convolutional network architecture and adaptive feature weight optimization mechanism, the initial graph convolutional network model is constructed; The initial graph convolutional network model is trained according to the historical data of the power grid by using a supervised learning method to obtain the graph convolutional network model.

5. A method for identifying power grid topology according to claim 4, characterized in that: The supervised learning method is used to train the initial graph convolutional network model according to the historical data of the power grid to obtain the graph convolutional network model, including: Labeling the historical data of the power grid to obtain a number of corresponding labeled training data; Input each of the training data into the initial graph convolutional network model so that the initial graph convolutional network model outputs corresponding prediction results, and calculates the difference between each of the prediction results and the corresponding labels according to a preset cross entropy loss function and a mean square error loss function, and then updates the model parameters according to the difference between each of the prediction results and the corresponding labels to obtain a first graph convolutional network model; If the performance of the first graph convolutional network model meets the preset conditions, the first graph convolutional network model is used as the graph convolutional network model; otherwise, the first graph convolutional network model continues to be trained until the trained model meets the preset conditions.

6. A method for identifying power grid topology according to claim 1, characterized in that: The step of obtaining the time series model by training according to historical power grid data includes: Acquiring historical topology data from the power grid historical data; Construct an initial time series model based on long short-term memory networks and gated recurrent units; Constructing a time series input data set according to the state information of each node and edge in the historical topological data; The time series input data set is input into the initial time series model, so that the initial time series model captures the dependency of the power grid topology changing with time according to the time series input data set, thereby obtaining the time series model.

7. A method for identifying power grid topology according to claim 1, characterized in that: When the real-time data of the power grid is missing, incomplete or noisy, a preset noise reduction algorithm and uncertainty processing technology are used to complete the real-time data of the power grid.

8. A power grid topology identification system, characterized in that: It includes an acquisition module and an identification module; The acquisition module is used to acquire real-time data of the power grid, wherein the real-time data of the power grid includes real-time operation data and real-time topology data of the power grid, wherein the real-time operation data includes real-time operation information of each node and line fault information, and the real-time topology data includes connection relationships of each node; The recognition module is used to input the real-time data of the power grid into a preset topology recognition model, so that the topology recognition model performs topology recognition on the power grid according to the real-time data of the power grid, and outputs a topology recognition result, wherein the topology recognition result includes a topology structure diagram, node status information, fault location information, and future topology status information; Among them, the topology recognition model is obtained by combining a graph convolutional network model, a time series model and an abnormal state recognition model, and the graph convolutional network model and the time series model are obtained by training based on historical power grid data.

9. A terminal, characterized in that: The method comprises 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, a power grid topology identification method as claimed in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute a power grid topology identification method according to any one of claims 1 to 7.

Images