Correction method and device of power grid topological structure, terminal equipment and storage medium

Abstract

The invention discloses a power grid topological structure correction method and device, terminal equipment and a storage medium, and the method comprises the steps: inputting a remote signaling adjacency matrix used for representing the connection state between nodes in a current power grid topological structure, and a remote measurement matrix used for representing the voltage value of each node into a preset neural network model; according to the method and the system, the neural network model is based on the extracted node features, then the nodes with the wrong connection state in the remote signaling adjacency matrix can be modified according to the similarity comparison result of the node features between every two adjacent nodes, the recognition and correction process is automatic, manual intervention is not needed, and the recognition efficiency is improved. According to the method, the error correction efficiency and accuracy are improved, and the connection state of each node in the remote signaling data can be accurately detected and corrected in real time based on the modified remote signaling adjacent matrix, so that the topology analysis of the power grid is more accurate, and the stable operation of the power grid is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of topology correction, and in particular to a method, device, terminal equipment and storage medium for correcting a power grid topology structure. Background Art

[0002] An accurate topology can ensure that the power grid can operate stably under various working conditions and avoid power failures and blackouts caused by topology errors. In smart grids, power dispatching needs to make decisions based on the real-time status of the power grid. If the topology of the power grid is wrong, the analysis will be inaccurate, which will lead to deviations in dispatching decisions and may trigger a chain reaction, further affecting the safety and stability of the power grid.

[0003] In the actual operation process, due to data collection errors, equipment failures and other reasons, telesignaling data may be erroneous, resulting in inaccurate topology analysis results. Therefore, it is necessary to timely analyze and correct the errors in telesignaling data to provide an accurate basis for subsequent power grid management and decision-making. However, when processing telesignaling data and telemetry data of power grids, the existing technology often manually judges the accuracy and completeness of the data. When errors or abnormal conditions are found, professionals manually correct them. However, this method is not only time-consuming and labor-intensive, but also prone to errors, so it is impossible to detect and correct the connection status of each node in the telesignaling data in real time and accurately, which reduces the accuracy of power grid topology analysis and cannot guarantee the stable operation of the power grid. Summary of the invention

[0004] The embodiment of the present invention provides a method, an apparatus, a terminal device and a storage medium for correcting the topology structure of a power grid. By automatically identifying and correcting nodes with incorrect connection states in a telesignaling adjacency matrix, the connection states of each node in the telesignaling data can be detected and corrected in real time and accurately, so that the topological analysis of the power grid will be more accurate. This can effectively solve the problem in the prior art that the accuracy and completeness of the data are often judged manually, and the correction is performed manually, resulting in the inability to detect and correct the telesignaling data in real time and accurately, resulting in low accuracy of the power grid topology analysis.

[0005] An embodiment of the present invention provides a method for correcting a power grid topology structure, comprising:

[0006] Obtaining a telesignaling adjacency matrix and a telemetering matrix; wherein the telesignaling adjacency matrix is ​​used to characterize the connection status between nodes in the current power grid topology; and the telemetering matrix is ​​used to characterize the voltage value of each node;

[0007] The telesignaling adjacency matrix and the telemetering matrix are input into a preset neural network model, so that the neural network model extracts the node features corresponding to each node according to the connection status of the nodes in the telesignaling adjacency matrix and the voltage values ​​of the nodes in the telemetering matrix, and according to the node feature similarity comparison results between the nodes, the incorrect connection status in the telesignaling adjacency matrix is ​​modified, and then the corrected telesignaling adjacency matrix is ​​generated;

[0008] Correcting the current power grid topology according to the modified remote signal adjacency matrix;

[0009] The training process of the preset neural network model includes:

[0010] The telesignaling adjacency matrix sample data and the corresponding telemetering matrix sample data are used as corresponding training samples, and the actual corrected telesignaling adjacency matrix of each training sample is set;

[0011] Taking each training sample and the actual corrected telesignal adjacency matrix of each training sample as input and the corrected telesignal adjacency matrix predicted by each training sample as output, the neural network model to be trained is iteratively trained until the neural network model converges to generate a preset neural network model.

[0012] Preferably, the neural network model extracts node features corresponding to each node according to the connection status of the nodes in the telesignaling adjacency matrix and the voltage values ​​of the nodes in the telemetry matrix, including:

[0013] The neural network model converts the remote signaling adjacency matrix and the remote sensing matrix into a graph structure;

[0014] Based on the connection status and voltage value of each node in the graph structure, the node feature of each node is generated; wherein the node feature is used to characterize the node's own attribute information and the interaction information between nodes.

[0015] Preferably, the node features include: node pair features, node degree features and node status features;

[0016] The node features of each node used to characterize the node's own attribute information and the interaction information between nodes are generated based on the connection status and voltage value of each node in the graph structure, including:

[0017] According to the connection status between each node in the graph structure, a node pair feature is generated in the graph structure for characterizing the strength of the relationship between two connected nodes;

[0018] The connection status of each node in the graph structure and the voltage value of each node are aggregated to generate node degree features and node status features; wherein the node degree feature is used to characterize the quantity characteristics of a node connected to other nodes; the node status feature is used to characterize the degree of voltage difference between a node and other adjacent nodes.

[0019] Preferably, the step of modifying the incorrect connection state in the remote signaling adjacency matrix according to the node feature similarity comparison result between the nodes, and then generating the modified remote signaling adjacency matrix, comprises:

[0020] For each node, generating an embedding vector representation corresponding to the node according to an attention weight between the node and a target node used to represent a connection with the node, and a node feature of the node;

[0021] The embedding vector representation of the node and the embedding vector representation corresponding to the target node are respectively compared for similarity, and then according to each similarity comparison result, the connection state between the node and the target node in the remote signal adjacency matrix is ​​modified, and then a modified remote signal adjacency matrix is ​​generated.

[0022] Preferably, generating an embedding vector representation corresponding to the node according to an attention weight between the node and a target node used to characterize a connection with the node, and a node feature of the node, comprises:

[0023] The node connected to the node is taken as the target node, and the attention weights corresponding to the node and each target node are calculated through the attention mechanism;

[0024] Each attention weight, as well as the node pair feature, node degree feature and node state feature of the node are aggregated to generate an embedding vector representation corresponding to the node.

[0025] Preferably, the connection status includes: connection exists or connection does not exist;

[0026] The embedding vector representation of the node and the embedding vector representation corresponding to the target node are respectively compared for similarity, and then according to each similarity comparison result, the connection state between the node and the target node in the remote signaling adjacency matrix is ​​modified, including:

[0027] For each target node, determining a cosine similarity value between an embedding vector representation of the target node and an embedding vector representation of the node;

[0028] Comparing the cosine similarity value with a preset similarity threshold;

[0029] When it is determined that the cosine similarity value is less than a preset similarity threshold, if the connection state between the target node and the node is that there is a connection, the connection state between the target node and the node in the remote signaling adjacency matrix is ​​modified to be that there is no connection; if the connection state between the target node and the node is that there is no connection, the connection state between the target node and the node in the remote signaling adjacency matrix is ​​kept unchanged;

[0030] When it is determined that the cosine similarity value is greater than a preset similarity threshold, if the connection state between the target node and the node is that there is no connection, the connection state between the target node and the node in the remote signal adjacency matrix is ​​modified to be connected; if the connection state between the target node and the node is that there is a connection, the connection state between the target node and the node in the remote signal adjacency matrix is ​​kept unchanged.

[0031] Preferably, during each iterative training, a training sample is input into the neural network model, so that the neural network model generates a predicted corrected telesignaling adjacency matrix corresponding to the training sample according to the node features in the training sample;

[0032] The predicted corrected telesignaling adjacency matrix is ​​compared with the actual corrected telesignaling adjacency matrix corresponding to the training samples, and the network parameters of the neural network model are adjusted according to the comparison results.

[0033] Based on the above method embodiments, the present invention provides corresponding device embodiments.

[0034] An embodiment of the present invention provides a device for correcting a power grid topology structure, comprising: a matrix acquisition module, a matrix correction module, and a topology structure correction module;

[0035] The matrix acquisition module is used to acquire a telesignaling adjacency matrix and a telemetering matrix; wherein the telesignaling adjacency matrix is ​​used to characterize the connection status between nodes in the current power grid topology; and the telemetering matrix is ​​used to characterize the voltage value of each node;

[0036] The matrix correction module is used to input the telesignaling adjacency matrix and the telemetry matrix into a preset neural network model, so that the neural network model extracts the node features corresponding to each node according to the connection status of the nodes in the telesignaling adjacency matrix and the voltage values ​​of the nodes in the telemetry matrix, and according to the node feature similarity comparison results between the nodes, the incorrect connection status in the telesignaling adjacency matrix is ​​modified, and then the corrected telesignaling adjacency matrix is ​​generated;

[0037] The topology correction module is used to correct the current power grid topology according to the repaired remote signal adjacency matrix;

[0038] The training process of the preset neural network model includes:

[0039] The training process of the preset neural network model includes:

[0040] The telesignaling adjacency matrix sample data and the corresponding telemetering matrix sample data are used as corresponding training samples, and the actual corrected telesignaling adjacency matrix of each training sample is set;

[0041] Taking each training sample and the actual corrected telesignal adjacency matrix of each training sample as input and the corrected telesignal adjacency matrix predicted by each training sample as output, the neural network model to be trained is iteratively trained until the neural network model converges to generate a preset neural network model.

[0042] Based on the above method embodiments, the present invention provides corresponding terminal device embodiments.

[0043] Another embodiment of the present invention provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and when the processor executes the computer program, a method for correcting a power grid topology structure described in the above-mentioned embodiment of the invention is implemented.

[0044] Based on the above method embodiments, the present invention provides corresponding storage medium item embodiments.

[0045] Another embodiment of the present invention provides a 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 a method for correcting a power grid topology structure as described in the above-mentioned embodiment of the invention.

[0046] The following beneficial effects are achieved by implementing the present invention:

[0047] The embodiment of the present invention provides a method, device, terminal device and storage medium for correcting the topological structure of a power grid. The present invention can input the telesignaling adjacency matrix and the telemetering matrix into a preset neural network model, so that the neural network model can modify the nodes with wrong connection status in the telesignaling adjacency matrix based on the extracted node features according to the node feature similarity comparison results between two adjacent nodes, that is, by comparing the similarity results, it can be identified which nodes' connection status may be wrong, so that the nodes with wrong connection status in the telesignaling adjacency matrix can be modified and corrected according to the results of the similarity comparison, and the identification and correction process of the present invention is automatic, without manual intervention, and the error correction efficiency and accuracy are improved. Moreover, based on the modified adjacency correction matrix, the connection status of each node in the telesignaling data can be detected and corrected in real time and accurately, so that the topological analysis of the power grid will be more accurate, and the stable operation of the power grid is guaranteed. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a flowchart of a method for correcting a power grid topology structure provided by an embodiment of the present invention.

[0049] Figure 2 It is a structural schematic diagram of a power grid topology correction device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0050] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] like Figure 1 FIG. 1 is a flow chart of a method for correcting a power grid topology structure provided by an embodiment of the present invention. The method for correcting a power grid topology structure includes:

[0052] Step S1: obtaining a telesignaling adjacency matrix and a telemetering matrix; wherein the telesignaling adjacency matrix is ​​used to characterize the connection status between nodes in the current power grid topology; and the telemetering matrix is ​​used to characterize the voltage value of each node;

[0053] Step S2: inputting the telesignaling adjacency matrix and the telemetering matrix into a preset neural network model, so that the neural network model extracts the node features corresponding to each node according to the connection status of the nodes in the telesignaling adjacency matrix and the voltage values ​​of the nodes in the telemetering matrix, and according to the node feature similarity comparison results between the nodes, modifying the wrong connection status in the telesignaling adjacency matrix, and then generating a corrected telesignaling adjacency matrix;

[0054] The training process of the preset neural network model includes:

[0055] The telesignaling adjacency matrix sample data and the corresponding telemetering matrix sample data are used as corresponding training samples, and the actual corrected telesignaling adjacency matrix of each training sample is set;

[0056] Taking each training sample and the actual corrected telesignal adjacency matrix of each training sample as input, taking the corrected telesignal adjacency matrix predicted by each training sample as output, iteratively training the neural network model to be trained until the neural network model converges, and generating a preset neural network model;

[0057] Step S3: Correct the current power grid topology according to the modified remote signaling adjacency matrix.

[0058] For step S1, in a preferred embodiment, the deep neural network model of the present invention can output a corrected adjacency matrix based on the input of the telesignaling adjacency matrix and the telemetry matrix.

[0059] Indicatively, the dimension of the input telesignaling adjacency matrix is ​​N×N, and the dimension of the telemetry matrix is ​​N×M, where N is the number of nodes in the power grid and M is the number of telemetry indicators. The dimension of the output corrected adjacency matrix is ​​also N×N.

[0060] Assuming that a small power system has 5 nodes, the corresponding telesignaling adjacency matrix and the corrected adjacency matrix of the power grid topology have dimensions of 5 × 5. The telemetry matrix contains 1 telemetry indicator for each node, with a dimension of 5 × 1. Therefore, through the training of the neural network model, the neural network model can learn the mapping relationship between the telesignaling adjacency matrix and the telemetry matrix, so that the errors in the adjacency matrix can be accurately calculated and corrected in the future.

[0061] For step S2, in a preferred embodiment, the training process of the preset neural network model includes:

[0062] First, each telesignaling adjacency matrix sample data and the telemetering matrix sample data corresponding to each telesignaling adjacency matrix sample data are used as corresponding training samples, and each training sample corresponds to an actual corrected telesignaling adjacency matrix. It can be understood that when the telesignaling adjacency matrix in the training sample has no error, the corresponding actual corrected telesignaling adjacency matrix is ​​the telesignaling adjacency matrix itself; when the telesignaling adjacency matrix in the training sample has an error, the corresponding actual corrected telesignaling adjacency matrix is ​​the modified telesignaling adjacency matrix corresponding to the telesignaling adjacency matrix.

[0063] Then, several training samples and the actual corrected telesignaling adjacency matrix corresponding to each training sample are used as input, and the predicted corrected telesignaling adjacency matrix of each training sample is used as output, and the neural network model to be trained is iteratively trained until the neural network model converges and generates a preset neural network model. Specifically, the telesignaling data and telemetering data of the power grid are collected, and features are extracted from the telesignaling data and telemetering data, including the opening and closing status of the line and the voltage value of the node, etc., to generate telesignaling adjacency matrix samples and telemetering matrix samples, so as to mark the known correct or incorrect connection status in the telesignaling adjacency matrix samples, and the marked actual corrected telesignaling adjacency matrix is ​​used as the input of the model, and the telesignaling adjacency matrix samples and telemetering matrix samples are used as the input of the model, so that when the predicted corrected telesignaling adjacency matrix of each training sample is used as the output, iterative training of the neural network model can be realized.

[0064] In a preferred embodiment, during each iterative training, a training sample is input into the neural network model so that the neural network model generates a predicted corrected telesignaling adjacency matrix corresponding to the training sample according to the node features in the training sample;

[0065] The predicted corrected telesignaling adjacency matrix is ​​compared with the actual corrected telesignaling adjacency matrix corresponding to the training samples, and the network parameters of the neural network model are adjusted according to the comparison results.

[0066] Indicatively, the hyperparameters of the neural network model (such as learning rate, batch size, number of iterations, etc.) can be adjusted through methods such as cross-validation, and during and after the training process, the performance of the neural network model can be evaluated using a test set, including indicators such as accuracy, recall, and F1 score.

[0067] Through the above steps, a neural network model can be trained to accurately correct the incorrect connection status in the power grid topology and applied to the actual power grid topology correction task.

[0068] In a preferred embodiment, the neural network model is composed of a graph neural network, a convolutional neural network, and a recurrent neural network. Specifically, a graph neural network (GNN) can be used to process graph structure data, and a convolutional neural network (CNN) and a recurrent neural network (RNN) can be combined to extract spatial and temporal features, thereby extracting node features corresponding to each node, and based on the calculated node feature similarity, the comparison between each node is achieved, so that the incorrect connection state in the remote signal adjacency matrix can be modified, and the corrected matrix can be output.

[0069] Furthermore, based on the trained neural network model, the node features of each node can be extracted from the input telesignaling adjacency matrix and telemetry matrix. The specific extraction process is as follows:

[0070] The neural network model converts the remote signaling adjacency matrix and the remote sensing matrix into a graph structure;

[0071] Based on the connection status and voltage value of each node in the graph structure, the node feature of each node is generated; wherein the node feature is used to characterize the node's own attribute information and the interaction information between nodes.

[0072] Specifically, the node features include: node pair features, node degree features, and node state features; when the graph structure is identified and convoluted, the corresponding node features can be obtained according to the connection status between the nodes in the graph structure and the voltage value of each node, and then:

[0073] According to the connection status between each node in the graph structure, a node pair feature is generated in the graph structure to characterize the strength of the relationship between two connected nodes; schematically, the node pair feature can directly reflect the actual connection status and electrical interaction between nodes in the power grid, so that the incorrect connection status can be more accurately identified based on the feature in the future.

[0074] The connection status of each node in the graph structure and the voltage value of each node are aggregated to generate node degree features and node status features; it can be understood that the node degree feature is used to characterize the quantity characteristics of a node connected to other nodes; the node status feature is used to characterize the degree of voltage difference between a node and other adjacent nodes.

[0075] Schematically, for each node in the graph structure, the number of directly connected neighboring nodes is counted, which is the degree of the node and is used as the node degree feature. The node degree feature is a basic and important attribute in the graph structure. By calculating the number of connections, it can reflect the importance and influence of the node in the network.

[0076] In the graph structure, for each node, the average, maximum, minimum and other statistics of the voltage difference between it and all adjacent nodes are calculated to obtain the node state characteristics. The node state characteristics can reflect the electrical state and operating conditions of the node in the power grid. By considering the degree of voltage difference, the node state characteristics can provide more detailed electrical information, helping the neural network model to more accurately understand and correct errors in the power grid topology.

[0077] Therefore, by extracting node pair features, node degree features, and node status features, rich node feature information can be provided for the neural network model to achieve accurate correction and optimization of the power grid topology.

[0078] Furthermore, after the above-mentioned node feature extraction is completed, the neural network model of the present invention can also obtain the embedded vector representation of each node based on each node feature, thereby realizing similarity comparison and matrix modification.

[0079] Specifically, the method of modifying the incorrect connection state in the remote signaling adjacency matrix according to the node feature similarity comparison result between the nodes, and then generating the modified remote signaling adjacency matrix, specifically includes:

[0080] For each node, generating an embedding vector representation corresponding to the node according to an attention weight between the node and a target node used to represent a connection with the node, and a node feature of the node;

[0081] Specifically, the node connected to the node is first used as the target node, so as to calculate the attention weights corresponding to the node and each target node respectively through the attention mechanism;

[0082] After aggregating each attention weight, as well as the node pair features, node degree features, and node state features of the node, an embedding vector representation corresponding to the node can be generated.

[0083] It can be understood that by generating embedded vector representations, high-dimensional features (such as node pair features, node degree features, and node state features) can be mapped into low-dimensional space while retaining key information and relationships. This not only reduces the computational complexity, but also makes the comparison between features more efficient and intuitive.

[0084] Moreover, the embedded vector can capture the complex relationships between nodes. After the attention weights are calculated by the attention mechanism, the present invention can further emphasize the neighboring nodes that have a greater impact on specific nodes, thereby better reflecting these relationships in the embedded vector and achieving accurate representation of the vector.

[0085] In the actual operation of the power grid, the state and connection relationship of the nodes may change. By generating an embedded vector representation, the characteristics of the nodes can be easily updated and the embedded vector can be recalculated, thereby realizing dynamic correction of the power grid topology.

[0086] In a preferred embodiment, the node features in the graph can be weighted and aggregated, that is, by introducing the attention mechanism, the weights between different nodes can be dynamically adjusted, so as to more accurately capture important neighborhood information. In the previous step, the model has extracted the node features, so the introduction of the attention mechanism and further feature aggregation operations can more accurately capture the dependencies between nodes, thereby generating an embedded vector. The embedded vector representation h of each node can be generated according to the following formula i :

[0087] h i =σ(∑ j∈N(i) α ij W*h j );

[0088] Among them, h i represents the embedding vector representation of the i-th node, α ij is the attention weight α between node i and target node j ij , N is the number of target nodes connected to the node, σ is the activation function, and W is the weight matrix.

[0089] α ij The weight reflects the importance of target node j to node i. A higher attention weight α ijThis indicates that target node j is more important to node i and may need to be connected, so the similarity between the two will be higher.

[0090] Furthermore, the embedding vector representation of the node and the embedding vector representation corresponding to the target node can be compared for similarity, and then according to each similarity comparison result, the connection state between the node and the target node in the remote signal adjacency matrix is ​​modified, and then a modified adjacency correction matrix is ​​generated.

[0091] Specifically, the connection status between nodes includes: whether a connection exists or not;

[0092] Then for each target node, firstly calculate and determine the cosine similarity value between the embedding vector representation of the target node and the embedding vector representation of the node;

[0093] Next, the cosine similarity value is compared with a preset similarity threshold;

[0094] When it is determined that the cosine similarity value is less than a preset similarity threshold, if the connection state between the target node and the node is that there is a connection, the connection state between the target node and the node in the remote signaling adjacency matrix is ​​modified to be that there is no connection; if the connection state between the target node and the node is that there is no connection, the connection state between the target node and the node in the remote signaling adjacency matrix is ​​kept unchanged;

[0095] When it is determined that the cosine similarity value is greater than a preset similarity threshold, if the connection state between the target node and the node is that there is no connection, the connection state between the target node and the node in the remote signal adjacency matrix is ​​modified to be connected; if the connection state between the target node and the node is that there is a connection, the connection state between the target node and the node in the remote signal adjacency matrix is ​​kept unchanged.

[0096] Schematically, by calculating the cosine similarity between the embedding vectors of the nodes, the similarity between the nodes can be quantified, thereby taking into account the comprehensive influence between the features and being able to effectively represent the relationship between the nodes in a low-dimensional space.

[0097] When the similarity is lower than the preset threshold, it means that the similarity between the two nodes is not high, indicating that there is no connection between the two nodes. If the connection status between the target node and the node is connected, the corresponding connection status needs to be modified at this time. The connection status between the target node and the node in the remote signal adjacency matrix can be modified to non-connected.

[0098] When the similarity is higher than the preset threshold, it means that the similarity between the two nodes is very high, indicating that there should be a connection relationship between the two. If the connection status between the target node and the node is that there is no connection, then the corresponding connection status needs to be modified. In this case, the connection status between the target node and the node in the remote signal adjacency matrix can be modified to exist.

[0099] In a preferred embodiment, it is assumed that there is an adjacency matrix A of a power grid, and the matrix represents the connection relationship between nodes in the power grid (0 represents no connection, and 1 represents connection).

[0100] The embedded features of nodes 1 and 2 are h1 and h2 respectively. If the similarity of the embedded feature vectors h1 and h2 of two nodes is very high, it can be inferred that the two nodes are highly related in function or should be connected to each other, but nodes 1 and 2 are not directly connected in the adjacency matrix (i.e., A 12 =0), so the adjacency matrix is ​​updated to A, so that A 12 =1.

[0101] On the contrary, if the similarity of the embedded feature vectors h1 and h2 of two nodes is very low, it can be inferred that the two nodes are highly unrelated in function, but nodes 1 and 2 have a direct connection in the adjacency matrix (i.e., A 12 =1), then update the adjacency matrix to A, so that A 12 =0.

[0102] Thus, through the above-mentioned method, the similarity comparison is closely combined with the modification of the adjacency matrix, so that the finally generated adjacency matrix is ​​more consistent with the actual network structure and node status.

[0103] Therefore, the neural network model of the present invention processes the data of the power grid topology structure by combining the graph neural network (GNN), the convolutional neural network (CNN) and the recurrent neural network (RNN), that is, converts and identifies the telesignaling adjacency matrix and the telemetry matrix to extract and compare the node features, thereby correcting the erroneous connection status.

[0104] The present invention can accurately identify the possible erroneous connection states in the telesignaling adjacency matrix by calculating the cosine similarity between node embedding vectors and combining the preset similarity threshold. Moreover, the whole process is automated and does not require human intervention, that is, the node features can be automatically learned through the neural network model, the embedding vectors can be generated, and the similarity comparison and connection state modification can be performed, thereby improving the efficiency and accuracy of the power grid topology correction and reducing the possibility of human error.

[0105] For step S3, in a preferred embodiment, the present invention can compare the original adjacency matrix and the corrected adjacency matrix to find out the differences between the two, such as new connections that need to be added and erroneous connections that need to be deleted, so as to actually modify the current power grid topology based on the differences between the two.

[0106] If the connection status between two nodes in the corrected matrix is ​​1 (indicating that there is a connection) but is 0 (indicating that there is no connection) in the original adjacency matrix, a new connection is added between the two nodes.

[0107] If the connection status between two nodes in the correction matrix is ​​0 (indicating that there is no connection) but is 1 (indicating that there is a connection) in the original adjacency matrix, the incorrect connection between the two nodes is deleted.

[0108] The corrected grid topology is then updated to the grid topology database for subsequent grid management and analysis to ensure that the database is updated in a timely and accurate manner so that subsequent applications can work based on the latest grid topology.

[0109] Therefore, through the above steps, the current power grid topology can be effectively corrected according to the adjacency correction matrix, thereby improving the reliability and operation efficiency of the power grid.

[0110] Therefore, the present invention can ensure the reliability and stability of the power grid by timely correcting errors in the topology structure, reduce the risk of failures and power outages caused by topology errors, and is of great significance for ensuring the safe and stable operation of the power system.

[0111] like Figure 2 As shown, based on the above-mentioned various embodiments of the method for correcting the topology of the power grid, the present invention provides corresponding device embodiments;

[0112] An embodiment of the present invention provides a device for correcting a power grid topology structure, comprising: a matrix acquisition module, a matrix correction module, and a topology structure correction module;

[0113] The matrix acquisition module is used to acquire a telesignaling adjacency matrix and a telemetering matrix; wherein the telesignaling adjacency matrix is ​​used to characterize the connection status between nodes in the current power grid topology; and the telemetering matrix is ​​used to characterize the voltage value of each node;

[0114] The matrix correction module is used to input the telesignaling adjacency matrix and the telemetry matrix into a preset neural network model, so that the neural network model extracts the node features corresponding to each node according to the connection status of the nodes in the telesignaling adjacency matrix and the voltage values ​​of the nodes in the telemetry matrix, and according to the node feature similarity comparison results between the nodes, the incorrect connection status in the telesignaling adjacency matrix is ​​modified, and then the corrected telesignaling adjacency matrix is ​​generated;

[0115] The topology correction module is used to correct the current power grid topology according to the repaired remote signal adjacency matrix;

[0116] The training process of the preset neural network model includes:

[0117] The training process of the preset neural network model includes:

[0118] The telesignaling adjacency matrix sample data and the corresponding telemetering matrix sample data are used as corresponding training samples, and the actual corrected telesignaling adjacency matrix of each training sample is set;

[0119] Taking each training sample and the actual corrected telesignal adjacency matrix of each training sample as input and the corrected telesignal adjacency matrix predicted by each training sample as output, the neural network model to be trained is iteratively trained until the neural network model converges to generate a preset neural network model.

[0120] It should be noted that the device embodiments described above are merely schematic, wherein the modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical modules, and may be located in one place or distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. Those of ordinary skill in the art can understand and implement it without paying creative labor.

[0121] Those skilled in the art can clearly understand that, for the sake of convenience and brevity, the specific working process of the device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0122] Based on the above-mentioned embodiments of the method for correcting various power grid topologies, the present invention provides corresponding embodiments of terminal equipment items.

[0123] An embodiment of the present invention provides a terminal device, including 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 a method for correcting a power grid topology structure as described in any method embodiment of the present invention.

[0124] The terminal device may be a computing terminal device such as a desktop computer, a notebook, a palm computer, a cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0125] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the terminal device, and uses various interfaces and lines to connect various parts of the entire terminal device.

[0126] The memory can be used to store the computer program, and the processor realizes various functions of the terminal device 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, at least one application required for a function, etc.; the data storage area can store data created according to the use of the mobile phone, 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 (Smart Media 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 device.

[0127] Based on the above-mentioned various embodiments of the method for correcting the topology of the power grid, the present invention provides corresponding storage medium embodiments.

[0128] An embodiment of the present invention provides a storage medium, which 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 a method for correcting a power grid topology structure as described in any method embodiment of the present invention.

[0129] The storage medium is a computer-readable storage medium, and the computer program is stored in the computer-readable storage medium. When the computer program is executed by the processor, the steps of each method embodiment described above can be implemented. Among them, 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, USB flash drive, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0130] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for correcting a power grid topology structure, characterized in that: include: Obtaining a telesignaling adjacency matrix and a telemetering matrix; wherein the telesignaling adjacency matrix is ​​used to characterize the connection status between nodes in the current power grid topology; and the telemetering matrix is ​​used to characterize the voltage value of each node; The telesignaling adjacency matrix and the telemetering matrix are input into a preset neural network model, so that the neural network model extracts the node features corresponding to each node according to the connection status of the nodes in the telesignaling adjacency matrix and the voltage values ​​of the nodes in the telemetering matrix, and according to the node feature similarity comparison results between the nodes, the incorrect connection status in the telesignaling adjacency matrix is ​​modified, and then the corrected telesignaling adjacency matrix is ​​generated; Correcting the current power grid topology according to the modified remote signal adjacency matrix; The training process of the preset neural network model includes: The telesignaling adjacency matrix sample data and the corresponding telemetering matrix sample data are used as corresponding training samples, and the actual corrected telesignaling adjacency matrix of each training sample is set; Taking each training sample and the actual corrected telesignal adjacency matrix of each training sample as input and the corrected telesignal adjacency matrix predicted by each training sample as output, the neural network model to be trained is iteratively trained until the neural network model converges to generate a preset neural network model.

2. A method for correcting a power grid topology structure according to claim 1, characterized in that: The neural network model extracts node features corresponding to each node according to the connection status of the nodes in the telesignaling adjacency matrix and the voltage values ​​of the nodes in the telemetry matrix, including: The neural network model converts the remote signaling adjacency matrix and the remote sensing matrix into a graph structure; Based on the connection status and voltage value of each node in the graph structure, the node feature of each node is generated; wherein the node feature is used to characterize the node's own attribute information and the interaction information between nodes.

3. A method for correcting a power grid topology structure as claimed in claim 2, characterized in that: The node features include: node pair features, node degree features, and node status features; The node features of each node used to characterize the node's own attribute information and the interaction information between nodes are generated based on the connection status and voltage value of each node in the graph structure, including: According to the connection status between each node in the graph structure, a node pair feature is generated in the graph structure for characterizing the strength of the relationship between two connected nodes; The connection status of each node in the graph structure and the voltage value of each node are aggregated to generate node degree features and node status features; wherein the node degree feature is used to characterize the quantity characteristics of a node connected to other nodes; the node status feature is used to characterize the degree of voltage difference between a node and other adjacent nodes.

4. A method for correcting a power grid topology structure as claimed in claim 3, characterized in that: The method of modifying the incorrect connection state in the remote signaling adjacency matrix according to the node feature similarity comparison result between the nodes, and then generating the modified remote signaling adjacency matrix, includes: For each node, generating an embedding vector representation corresponding to the node according to an attention weight between the node and a target node used to represent a connection with the node, and a node feature of the node; The embedding vector representation of the node and the embedding vector representation corresponding to the target node are respectively compared for similarity, and then according to each similarity comparison result, the connection state between the node and the target node in the remote signal adjacency matrix is ​​modified, and then a modified remote signal adjacency matrix is ​​generated.

5. A method for correcting a power grid topology structure as claimed in claim 4, characterized in that: The step of generating an embedded vector representation corresponding to the node according to an attention weight between the node and a target node used to represent a connection with the node, and a node feature of the node, comprises: The node connected to the node is taken as the target node, and the attention weights corresponding to the node and each target node are calculated through the attention mechanism; Each attention weight, as well as the node pair feature, node degree feature and node state feature of the node are aggregated to generate an embedding vector representation corresponding to the node.

6. A method for correcting a power grid topology structure according to claim 5, characterized in that: The connection status includes: whether a connection exists or not; The embedding vector representation of the node and the embedding vector representation corresponding to the target node are respectively compared for similarity, and then according to each similarity comparison result, the connection state between the node and the target node in the remote signaling adjacency matrix is ​​modified, including: For each target node, determining a cosine similarity value between an embedding vector representation of the target node and an embedding vector representation of the node; Comparing the cosine similarity value with a preset similarity threshold; When it is determined that the cosine similarity value is less than a preset similarity threshold, if the connection state between the target node and the node is that there is a connection, the connection state between the target node and the node in the remote signaling adjacency matrix is ​​modified to be that there is no connection; if the connection state between the target node and the node is that there is no connection, the connection state between the target node and the node in the remote signaling adjacency matrix is ​​kept unchanged; When it is determined that the cosine similarity value is greater than a preset similarity threshold, if the connection state between the target node and the node is that there is no connection, the connection state between the target node and the node in the remote signal adjacency matrix is ​​modified to be connected; if the connection state between the target node and the node is that there is a connection, the connection state between the target node and the node in the remote signal adjacency matrix is ​​kept unchanged.

7. A method for correcting a power grid topology structure according to claim 1, characterized in that: In each iterative training, a training sample is input into the neural network model, so that the neural network model generates a predicted and corrected telesignaling adjacency matrix corresponding to the training sample according to the node features in the training sample; The predicted corrected telesignaling adjacency matrix is ​​compared with the actual corrected telesignaling adjacency matrix corresponding to the training samples, and the network parameters of the neural network model are adjusted according to the comparison results.

8. A device for correcting a power grid topology, characterized in that: include: Matrix acquisition module, matrix correction module and topology correction module; The matrix acquisition module is used to acquire a telesignaling adjacency matrix and a telemetering matrix; wherein the telesignaling adjacency matrix is ​​used to characterize the connection status between nodes in the current power grid topology; and the telemetering matrix is ​​used to characterize the voltage value of each node; The matrix correction module is used to input the telesignaling adjacency matrix and the telemetry matrix into a preset neural network model, so that the neural network model extracts the node features corresponding to each node according to the connection status of the nodes in the telesignaling adjacency matrix and the voltage values ​​of the nodes in the telemetry matrix, and according to the node feature similarity comparison results between the nodes, the incorrect connection status in the telesignaling adjacency matrix is ​​modified, and then the corrected telesignaling adjacency matrix is ​​generated; The topology correction module is used to correct the current power grid topology according to the repaired remote signal adjacency matrix; The training process of the preset neural network model includes: The training process of the preset neural network model includes: The telesignaling adjacency matrix sample data and the corresponding telemetering matrix sample data are used as corresponding training samples, and the actual corrected telesignaling adjacency matrix of each training sample is set; Taking each training sample and the actual corrected telesignal adjacency matrix of each training sample as input and the corrected telesignal adjacency matrix predicted by each training sample as output, the neural network model to be trained is iteratively trained until the neural network model converges to generate a preset neural network model.

9. A terminal device, characterized in that: The invention 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 method for correcting a power grid topology structure as claimed in any one of claims 1 to 7 is implemented.

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

Images