Power grid real-time topology identification method and system, medium and terminal
By using a virtual complete graph and a physical information fusion graph neural network model, the problem of poor robustness in power grid topology identification is solved, achieving high-precision and high-efficiency power grid topology identification and meeting the real-time requirements of large-scale power grids.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- WILLFAR INFORMATION TECH CO LTD
- Filing Date
- 2025-12-31
- Publication Date
- 2026-05-19
AI Technical Summary
Existing power grid topology identification methods have poor robustness, low efficiency and accuracy, and are difficult to meet the real-time requirements of large-scale power grids.
A virtual complete graph construction and a physical information fusion graph neural network model are adopted. Edge feature vectors are constructed by using the spatiotemporal correlation features and instantaneous difference features of measurement data. Topological reasoning is performed using a pre-trained physical information fusion graph neural network model, and the physical constraint loss function is combined to ensure the physical rationality of the results.
It achieves high-precision and high-efficiency power grid topology identification, and can maintain identification capability even when measurement data is incomplete or noisy, significantly improving the reliability and security of the results, and realizing real-time topology identification at the second or even millisecond level.
Smart Images
Figure CN122065002A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of power Internet of Things technology, and particularly relates to a real-time power grid topology identification method and system, medium and terminal. Background Technology
[0002] Power grid topology identification is a fundamental step in power system operation and dispatch, and its accuracy directly affects the effectiveness and security of advanced applications such as state estimation, power flow calculation, and fault location. In traditional power grid dispatch systems, topology information typically relies on manual configuration or monitoring of remote signaling signals from switches. However, these signals may become erroneous during transmission due to communication delays, signal loss, or false alarms, thus threatening the safe operation of the power grid. Currently, mainstream power grid topology identification methods are mainly divided into rule-based reasoning methods and residual analysis methods based on state estimation. Rule-based reasoning methods heavily rely on the completeness (measurements are required for each node and branch) and accuracy of measurements, are highly sensitive to measurement noise and missing data, have poor robustness, and cannot provide probabilistic results. Residual analysis methods based on state estimation involve enormous computational costs, especially when there are many candidate topologies, making it difficult to meet the real-time requirements of large-scale power grids.
[0003] Patent application CN118200153A discloses a method for identifying large-scale network router-level topology based on knowledge reasoning, comprising: S1, obtaining structural information of the interface-level network topology through the active probing tool Traceroute, and performing preliminary IP interface alias resolution and preliminary anonymous router identification to obtain high-confidence results, while collecting multimodal data including the subordinate relationship between IP interfaces and autonomous systems (AS), and the geographical location information corresponding to IP interfaces; S2, based on step S1, establishing a comprehensive network topology knowledge graph and extracting high-dimensional features of entities and relationships; S3, based on step S2, establishing an evaluation system for IP alias relationship triples and an evaluation system for triples belonging to the same anonymous router relationship, and identifying and completing the topology through heuristic algorithms; S4, based on step S3, supplementing and improving the comprehensive network topology knowledge graph, and extracting high-dimensional features of entities and relationships; S5, based on step S4, establishing an evaluation system for router connection relationship triples, and identifying and completing the topology through heuristic algorithms, thereby identifying a complete and accurate router-level network topology. The topology identification method in this patent application is also based on rule-based reasoning, which has poor robustness and suffers from the same drawbacks as existing technologies.
[0004] Therefore, how to overcome the shortcomings of existing technologies and achieve highly robust, efficient, and accurate power grid topology identification is a problem that urgently needs to be solved by researchers in this field. Summary of the Invention
[0005] To address the shortcomings of existing technologies, the present invention aims to provide a real-time power grid topology identification method to solve the problems of poor robustness, low efficiency, and low accuracy in existing power grid topology identification methods. In addition, the present invention also provides a real-time power grid topology identification system, medium, and terminal.
[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:
[0007] In a first aspect, the present invention provides a method for real-time topology identification of power grids, comprising the following steps:
[0008] S10. Treat all measurement points in the power grid as nodes, and construct virtual edges between any two nodes to obtain a virtual complete graph;
[0009] S20. Based on the electrical measurement data of each measurement point, construct the node feature vector. Based on the spatiotemporal correlation characteristics and instantaneous difference / ratio characteristics between the measurement data of paired measurement points, construct the edge feature vector of the corresponding virtual edge.
[0010] S30. The virtual complete graph is inferred by a pre-trained physical information fusion graph neural network model, and the probability that each virtual edge is a real physical connection is output.
[0011] S40. Determine the true topology of the power grid based on the probability.
[0012] Furthermore, in S20, the spatiotemporal correlation characteristics include the Pearson correlation coefficient of the voltage sequence, the average value of the voltage difference, and the standard deviation of the voltage difference; the instantaneous difference / ratio includes the absolute difference of the voltage amplitude at the current moment and the absolute difference of the voltage phase angle at the current moment.
[0013] Furthermore, in S30, the physical information fusion graph neural network model includes an encoder, a multi-layer message passing network, and a decoder. The encoder is used to map the node feature vector and the edge feature vector to a high-dimensional embedding space to obtain the initial node embedding and the initial edge embedding, respectively. The multi-layer message passing network is used to iteratively update the node embedding and the edge embedding. The decoder is used to predict the topological state of each virtual edge based on the final edge embedding.
[0014] Furthermore, in the multi-layer message passing network:
[0015] Node information aggregation: For each node, aggregate all information from its neighboring nodes;
[0016] Node state update: Each node updates its own state by combining its own state with the aggregated messages;
[0017] Edge state update: Each edge updates its own state by combining the updated states of the two nodes it connects to.
[0018] Furthermore, the training process of the physical information fusion graph neural network model includes calculating a total loss function, which includes a supervisory loss and a physical constraint loss. The supervisory loss is a cross-entropy loss, used to measure the difference between the predicted probability and the true label. The physical constraint loss is used to penalize predictions that violate Kirchhoff's current law.
[0019] Secondly, the present invention also provides a real-time power grid topology identification system, comprising:
[0020] The virtual graph construction module is used to construct power grid measurement points as nodes and build virtual edges between any two nodes to obtain a virtual complete graph;
[0021] The feature extraction module is used to generate node feature vectors and edge feature vectors based on measurement data;
[0022] The physical information fusion graph neural network module is used to receive the virtual complete graph and its features, and output the probability that each virtual edge is a real physical connection;
[0023] The output module is used to output the actual topology of the power grid based on the probability.
[0024] Furthermore, the physical information fusion graph neural network module includes an encoder, a multi-layer message passing network, and a decoder. The encoder is used to map the node feature vectors and edge feature vectors to a high-dimensional embedding space to obtain initial node embeddings and initial edge embeddings. The multi-layer message passing network is used to iteratively update the node embeddings and edge embeddings. The decoder is used to predict the topological state of each virtual edge based on the final edge embedding.
[0025] Thirdly, the present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method described above.
[0026] Fourthly, the present invention also provides an electronic terminal, comprising: a processor and a memory; the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory to cause the terminal to perform the method described above.
[0027] Compared with existing technologies, the real-time power grid topology identification method, system, medium, and terminal provided by this invention have at least the following advantages:
[0028] Existing mainstream power grid topology identification methods are highly sensitive to measurement noise and missing data, exhibit poor robustness, and cannot provide probabilistic results. They also involve enormous computational demands, especially when the number of candidate topologies is large, making it difficult to meet the real-time requirements of large-scale power grids. This invention offers a simple and convenient process. By constructing a virtual complete graph and utilizing a Geometric Node (GNN) for learning, it successfully transforms the topology identification problem into a graph structure learning problem, thus avoiding reliance on prior physical connections. Since GNNs are highly robust to irregular data and noise, this invention maintains high accuracy and efficiency even with incomplete or noisy measurement data. Furthermore, by incorporating physical constraints, the model's predictions are not only data-driven but also conform to the physical laws of power grid operation, significantly improving the reliability and security of the results. Ultimately, this method can achieve real-time topology identification at the second or even millisecond level and automatically associate physical locations, providing a solid technical foundation for intelligent operation and maintenance and real-time control of power grids. Attached Figure Description
[0029] To more clearly illustrate the solution of the present invention, a brief introduction will be given to the drawings used in the description of the embodiments below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.
[0030] Figure 1 A flowchart of a real-time power grid topology identification method provided in an embodiment of the present invention;
[0031] Figure 2 This is an architecture diagram of a real-time power grid topology identification system provided in an embodiment of the present invention. Detailed Implementation
[0032] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Preferred embodiments of the invention are shown in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a thorough and complete understanding of the disclosure of the invention.
[0033] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.
[0034] This invention provides a real-time power grid topology identification method, applied to the identification of topology structures in power systems. The real-time power grid topology identification method includes the following steps:
[0035] S10. Treat all measurement points in the power grid as nodes and construct virtual edges between any two nodes to obtain a virtual complete graph; S20. Based on the electrical measurement data of each measurement point, construct node feature vectors, and based on the spatiotemporal correlation features and instantaneous difference / ratio features between the measurement data of paired measurement points, construct edge feature vectors for the corresponding virtual edges; S30. Use a pre-trained physical information fusion graph neural network model to infer the virtual complete graph and output the probability that each virtual edge is a real physical connection; S40. Determine the real topology of the power grid based on the probability.
[0036] The present invention has a simple process and is easy to operate, achieving high-precision identification of power grid topology and ensuring the reliability and security of the identification results.
[0037] To enable those skilled in the art to better understand the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.
[0038] This invention provides a real-time power grid topology identification method, applied to the identification of topology structures in power systems. By modeling an unknown power grid topology as a "virtual complete graph," it utilizes a graph neural network to automatically learn and infer the true physical connectivity from the time-series correlation of measurement data. Furthermore, a physical constraint loss function ensures the physical plausibility of the results. Figure 1 As shown, in this embodiment, the real-time power grid topology identification method includes the following steps:
[0039] S10. Treat all measurement points in the power grid as nodes, and construct virtual edges between any two nodes to obtain a virtual complete graph.
[0040] S20. Based on the electrical measurement data of each measurement point, construct the node feature vector. Based on the spatiotemporal correlation characteristics and instantaneous difference / ratio characteristics between the measurement data of paired measurement points, construct the edge feature vector of the corresponding virtual edge.
[0041] Specifically, in this embodiment, the node Construction: Basic electrical characteristics (such as voltage, current, and power) can be used as the constituent elements of the node feature vector.
[0042] Specifically, in this embodiment, virtual edges Construction: Assuming that there may be an edge between any two measurement points, for a system with N measurement points, construct a complete graph containing N(N−1) / 2 edges.
[0043] Furthermore, in this embodiment, the virtual edge has the following characteristics: due to the physical current Since the problem is unknown, it is necessary to construct features that can indirectly reflect the electrical relationship between the two points. These features can come from:
[0044] Time series correlation: Calculate the measurement values (such as voltage amplitude) at two nodes. and Statistical characteristics within a certain time window. Specifically, these include:
[0045] Pearson correlation coefficient of voltage series; connected node voltages typically have high correlation.
[0046] The average value of the voltage difference.
[0047] Standard deviation of voltage difference.
[0048] Instantaneous difference / ratio of physical quantities:
[0049] : The absolute difference in voltage amplitude at the current moment.
[0050] : The absolute difference in voltage phase angle at the current moment (if available).
[0051] Example of virtual edge feature vector:
[0052] .
[0053] S30. Reason about the virtual complete graph using a pre-trained physical information fusion graph neural network model, and output the probability that each virtual edge is a real physical connection.
[0054] Specifically, in this embodiment, the physical information fusion graph neural network model includes an encoder (feature embedding), a multi-layer message passing network, and a decoder. The encoder is used to map the node feature vectors and edge feature vectors to a high-dimensional embedding space, respectively, to enhance the model's expressive power and obtain the initial node embedding and the initial edge embedding. The multi-layer message passing network is used to iteratively update the node embedding and the edge embedding. The decoder is used to predict the topological state of each virtual edge based on the final edge embedding.
[0055] In this embodiment, the initial node embedding expression in the encoder is as follows:
[0056]
[0057] The initial edge embedding expression is as follows:
[0058]
[0059] Message passing along the edge: For each edge Its eigenvectors can be represented by the following formula:
[0060]
[0061] MLP is a multilayer perceptron.
[0062] Furthermore, in this embodiment, in a multi-layer messaging network:
[0063] Node information aggregation: For each node j, aggregate all information from neighboring nodes. Information;
[0064]
[0065] Node state update: Each node updates its own state by combining its own state with the aggregated messages;
[0066]
[0067] Edge state update: Each edge updates its own state by combining the updated states of the two nodes it connects to.
[0068] .
[0069] Furthermore, in this embodiment, in the decoder (topology state classification), after L-layer message passing, the final edge embedding vector is used to predict the topology state.
[0070]
[0071] .
[0072] Furthermore, in this embodiment, the loss function is designed as follows:
[0073] Monitoring losses Cross-entropy loss measures the difference between the predicted probability and the true label.
[0074]
[0075] Physical constraint loss: penalizes predictions that violate Kirchhoff's current law. For each node j:
[0076]
[0077] in, It is the model's predicted value (0 or 1) for the edge state, which can be used during training. The gradient is calculated using a smooth approximation (such as Straight-Through Gumbel-Softmax). This loss term drives the model to predict topological topologies that make the sum of node currents close to zero.
[0078] Total loss: , where λ is a hyperparameter used to balance the weights of the two losses.
[0079] S40. Determine the true topology of the power grid based on probability.
[0080] This invention also provides a real-time power grid topology identification system, deployed in a power grid dispatch center, for receiving measurement data from field equipment and outputting the identification results to upper-level power grid application software, such as... Figure 2 As shown, in this embodiment, the real-time power grid topology identification system includes:
[0081] The virtual graph construction module is used to construct power grid measurement points as nodes and build virtual edges between any two nodes to obtain a virtual complete graph;
[0082] The feature extraction module is used to generate node feature vectors and edge feature vectors based on measurement data;
[0083] The physical information fusion graph neural network module is used to receive the virtual complete graph and its features, and output the probability that each virtual edge is a real physical connection;
[0084] The output module is used to output the actual topology of the power grid based on probability.
[0085] Furthermore, in this embodiment, the physical information fusion graph neural network module is a pre-trained model, including an encoder, a multi-layer message passing network, and a decoder. The encoder is used to map the node feature vector and the edge feature vector to a high-dimensional embedding space to obtain the initial node embedding and the initial edge embedding, respectively. The multi-layer message passing network is used to iteratively update the node embedding and the edge embedding, and the decoder is used to predict the topological state of each virtual edge based on the final edge embedding.
[0086] This invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the methods in this embodiment.
[0087] This invention also provides an electronic terminal, including: a processor and a memory; the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory, so that the terminal performs any of the methods in this embodiment.
[0088] As will be understood by those skilled in the art, the computer-readable storage medium described in this embodiment allows for the implementation of all or part of the steps in the above method embodiments by computer program-related hardware. The aforementioned computer program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.
[0089] The electronic terminal provided in this embodiment includes a processor, a memory, a transceiver, and a communication interface. The memory and the communication interface are connected to the processor and the transceiver and complete communication between them. The memory is used to store computer programs, the communication interface is used to perform communication, and the processor and the transceiver are used to run the computer programs, so that the electronic terminal performs the steps of the above method.
[0090] Compared with existing technologies, the real-time power grid topology identification method, system, medium, and terminal described in the above embodiments are significantly improved. Existing mainstream power grid topology identification methods are highly sensitive to measurement noise and missing data, exhibit poor robustness, cannot provide probabilistic results, and involve enormous computational loads. This is especially problematic when the number of candidate topologies is large, making it difficult to meet the real-time requirements of large-scale power grids. This invention, however, is simple in process and easy to operate. By constructing a virtual complete graph and utilizing a graph neural network (GNN) for learning, it successfully transforms the topology identification problem into a graph structure learning problem, thus avoiding dependence on prior physical connections. Since GNNs are highly robust to irregular data and noise, this invention maintains high accuracy and efficiency even with incomplete or noisy measurement data. Furthermore, by incorporating physical constraints, the model's prediction results are not only data-driven but also conform to the physical laws of power grid operation, significantly improving the reliability and security of the results. Ultimately, this method can achieve real-time topology identification at the second or even millisecond level and automatically associate physical locations, providing a solid technical foundation for intelligent operation and maintenance and real-time control of the power grid.
[0091] Obviously, the embodiments described above are merely preferred embodiments of the present invention, and not all embodiments. The accompanying drawings illustrate preferred embodiments of the present invention, but do not limit the scope of the patent. The present invention can be implemented in many different forms; rather, these embodiments are provided to provide a more thorough and complete understanding of the disclosure of the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing specific embodiments, or make equivalent substitutions for some of the technical features. Any equivalent structures made using the content of this specification and drawings, directly or indirectly applied to other related technical fields, are similarly within the scope of patent protection of this invention.
Claims
1. A method for real-time topology identification of a power grid, characterized in that, Includes the following steps: S10. Treat all measurement points in the power grid as nodes, and construct virtual edges between any two nodes to obtain a virtual complete graph; S20. Based on the electrical measurement data of each measurement point, construct the node feature vector. Based on the spatiotemporal correlation characteristics and instantaneous difference / ratio characteristics between the measurement data of paired measurement points, construct the edge feature vector of the corresponding virtual edge. S30. The virtual complete graph is inferred by a pre-trained physical information fusion graph neural network model, and the probability that each virtual edge is a real physical connection is output. S40. Determine the true topology of the power grid based on the probability.
2. The real-time power grid topology identification method according to claim 1, characterized in that, In S20, the spatiotemporal correlation characteristics include the Pearson correlation coefficient of the voltage sequence, the average value of the voltage difference, and the standard deviation of the voltage difference; the instantaneous difference / ratio includes the absolute difference of the voltage amplitude at the current moment and the absolute difference of the voltage phase angle at the current moment.
3. The real-time power grid topology identification method according to claim 1, characterized in that, In S30, the physical information fusion graph neural network model includes an encoder, a multi-layer message passing network, and a decoder. The encoder is used to map the node feature vector and the edge feature vector to a high-dimensional embedding space to obtain the initial node embedding and the initial edge embedding. The multi-layer message passing network is used to iteratively update the node embedding and the edge embedding. The decoder is used to predict the topological state of each virtual edge based on the final edge embedding.
4. The real-time power grid topology identification method according to claim 3, characterized in that, In the multi-layer message passing network: Node information aggregation: For each node, aggregate all information from its neighboring nodes; Node state update: Each node updates its own state by combining its own state with the aggregated messages; Edge state update: Each edge updates its own state by combining the updated states of the two nodes it connects to.
5. The real-time power grid topology identification method according to claim 3, characterized in that, The training process of the physical information fusion graph neural network model includes calculating a total loss function, which includes a supervisory loss and a physical constraint loss. The supervisory loss is a cross-entropy loss, used to measure the difference between the predicted probability and the true label. The physical constraint loss is used to penalize predictions that violate Kirchhoff's current law.
6. A system employing the method as described in any one of claims 1 to 5, characterized in that, include: The virtual graph construction module is used to construct power grid measurement points as nodes and build virtual edges between any two nodes to obtain a virtual complete graph; The feature extraction module is used to generate node feature vectors and edge feature vectors based on measurement data; The physical information fusion graph neural network module is used to receive the virtual complete graph and its features, and output the probability that each virtual edge is a real physical connection; The output module is used to output the actual topology of the power grid based on the probability.
7. The system according to claim 6, characterized in that, The physical information fusion graph neural network module includes an encoder, a multi-layer message passing network, and a decoder. The encoder is used to map the node feature vectors and edge feature vectors to a high-dimensional embedding space to obtain initial node embeddings and initial edge embeddings. The multi-layer message passing network is used to iteratively update the node embeddings and edge embeddings. The decoder is used to predict the topological state of each virtual edge based on the final edge embedding.
8. The system according to claim 6 or 7, characterized in that, The system is deployed in the power grid dispatch center to receive measurement data from the field equipment layer and output the identification results to the upper-level application software of the power grid.
9. A computer-readable storage medium, characterized in that, The storage medium stores a computer program that, when executed by a processor, implements the method as described in any one of claims 1 to 5.
10. An electronic terminal, characterized in that, include: Processor and memory; The memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory to cause the terminal to perform the method as described in any one of claims 1 to 5.