Method and apparatus for voltage partitioning, non-volatile storage medium

Abstract

This application discloses a method and apparatus for voltage partitioning, as well as a non-volatile storage medium. The method includes: acquiring real-time operating data of each node in a power grid under test, where each node represents an electrical connection point in the power grid; for each node, constructing a first state vector corresponding to the node based on the real-time operating data; determining a second state vector corresponding to each node based on the topology information of the power grid under test and multiple first state vectors, wherein the dimension of the second state vector is smaller than the dimension of the first state vector; determining the geodesic distance between every two nodes based on the multiple second state vectors; and clustering all nodes in the power grid under test based on the multiple geodesic distances to obtain multiple voltage partitions of the power grid under test. This application solves the technical problems of partitioning rigidity and control mismatch caused by relying on static parameters for point-based voltage partitioning in related technologies.

Description

Technical Field

[0001] This application relates to the field of power system automation technology, and more specifically, to a method and apparatus for voltage partitioning and a non-volatile storage medium. Background Technology

[0002] In related technologies, the operating state of distribution networks exhibits strong nonlinearity and time-varying characteristics. Voltage control typically adopts a static zoning strategy, relying on static parameters to set fixed zoning for the power grid. However, once the power grid operating mode changes (such as load fluctuations or changes in renewable energy output), the fixed zoning boundaries are no longer applicable. Therefore, there are problems such as rigid zoning, delayed response, and inability to adapt to the dynamic changes of the power grid.

[0003] There is currently no effective solution to the above problems. Summary of the Invention

[0004] This application provides a method and apparatus for voltage partitioning, as well as a non-volatile storage medium, to at least solve the technical problems of partitioning rigidity and control mismatch caused by relying on static parameters for point-to-point voltage partitioning in related technologies.

[0005] According to one aspect of the embodiments of this application, a method for voltage partitioning is provided, comprising: acquiring real-time operating data of each node in a power grid under test, wherein the node represents an electrical connection point in the power grid under test; for each node, constructing a first state vector corresponding to the node based on the real-time operating data, wherein the first state vector is used to reflect the operating state of the node; determining a second state vector corresponding to each node based on the topology information of the power grid under test and multiple first state vectors, wherein the dimension of the second state vector is smaller than the dimension of the first state vector; determining the geodesic distance between every two nodes based on the multiple second state vectors, wherein the geodesic distance is used to quantify the strength of electrical coupling between the two nodes corresponding to the geodesic distance; and performing clustering processing on all nodes included in the power grid under test based on the multiple geodesic distances to obtain multiple voltage partitions of the power grid under test.

[0006] Optionally, determining the second state vector corresponding to each node based on the topology information of the power grid under test and multiple first state vectors includes: generating a target matrix based on the topology information, wherein the target matrix is ​​used to record the electrical connection relationship between each node; encoding the target matrix and the first state matrix to obtain a second state matrix, wherein the first state matrix is ​​constructed based on multiple first state vectors, and each element in the second state matrix is ​​a second state vector.

[0007] Optionally, generating a target matrix based on topology information includes: determining the node identification information as the row and column identifiers of the target matrix; for each element in the target matrix, determining the target row identifier and target column identifier associated with the element; determining the electrical connection relationship in the power grid under test between the first node indicated by the target row identifier and the second node indicated by the target column identifier based on the topology information, wherein the electrical connection relationship includes: direct connection and indirect connection, where a direct connection indicates that there are no other devices on the connection line between the two nodes, and an indirect connection indicates that there are other devices on the connection line between the two nodes; when the electrical connection relationship is a direct connection, assigning the element a first value, and when the electrical connection relationship is an indirect connection, assigning the element a second value.

[0008] Optionally, the target matrix and the first state matrix are encoded to obtain a second state matrix, including: for each node, determining the directly connected adjacent nodes based on the target matrix, and determining the state vector of the adjacent nodes based on the first state matrix; fusing the first state vector and the state vector of the adjacent nodes in the graph neural network layer of the encoder to obtain a fused vector; processing the fused vector and the first state vector in the fully connected neural network layer of the encoder to obtain Gaussian distribution parameters output by the fully connected neural network layer, wherein the Gaussian distribution parameters are used to quantize the distribution characteristics of the first state vector in the low-dimensional space; resampling the Gaussian distribution parameters to obtain the second state vector corresponding to the node, wherein the dimension of the second state vector is the same as the dimension of the low-dimensional space; and generating a second state matrix based on multiple second state vectors.

[0009] Optionally, determining the geodesic distance between every two nodes based on multiple second state vectors includes: determining the Euclidean distance between every two nodes based on the second state vector corresponding to each node; for each node, determining a preset number of target Euclidean distances in ascending order from the multiple Euclidean distances corresponding to the node, and determining the set of nodes corresponding to the preset number of target Euclidean distances as the set of adjacent nodes corresponding to the node; for the first node and the second node, if the first node's set of first adjacent nodes contains the second node, and the second node's set of second adjacent nodes contains the first node, establishing an edge between the first node and the second node, where the first node and the second node are any two different nodes in the power grid under test; generating a weighted undirected graph based on multiple nodes and multiple edges, where the weight of an edge is the Euclidean distance between the two nodes associated with the edge; determining the shortest distance from the first node to the second node based on the weighted undirected graph, where the shortest distance is the geodesic distance between the first node and the second node.

[0010] Optionally, clustering is performed on all nodes in the power grid under test based on multiple geodesic distances, including: determining the local curvature of each node, where the local curvature is used to quantify the electrical behavior coupling strength of the node in its region; determining the cluster radius of the node based on the local curvature, determining the cluster range of the node based on the cluster radius, and determining the number of actual nodes contained within the cluster range; determining the target number of actual nodes greater than the minimum number of nodes, and determining the nodes corresponding to the target number of actual nodes as core nodes, where the minimum number of nodes is the lower limit of the number of nodes contained within the cluster range; and clustering is performed based on the core nodes, the cluster radius, and the target topology graph, where the target topology graph reflects the electrical structure of the power grid under test, the vertices of the target topology graph are nodes, and the edges in the target topology graph are generated based on the geodesic distances.

[0011] Optionally, clustering is performed based on core nodes, cluster radius, and target topology, including: in each round of clustering, determining the cluster starting point, identifying clusterable nodes among the nodes within the cluster range of the cluster starting point in the target topology, and adding the clusterable nodes to the cluster set until all clusterable nodes are added to the cluster set. Here, clusterable nodes are nodes not included in the cluster set. In the first round of clustering, the cluster starting point is any core node. In subsequent rounds of clustering, the cluster starting point is a newly added node in the cluster set. The area indicated by each cluster set in the target topology is a voltage partition.

[0012] Optionally, after obtaining multiple voltage zones of the power grid under test, the method further includes: determining a center state vector based on the second state vector of each node contained in the voltage zone, wherein the center state vector is used to represent the geometric center of the voltage zone; for each node contained in the voltage zone, determining the distance value between the node and the center state vector based on the second state vector corresponding to the node; and determining the node corresponding to the minimum distance value as the control node of the voltage zone, wherein the control node is the voltage control point of the voltage zone.

[0013] According to another aspect of the embodiments of this application, a voltage partitioning apparatus is also provided, comprising: an acquisition module, configured to acquire real-time operating data of each node in a power grid under test, wherein the node represents an electrical connection point in the power grid under test; a vector construction module, configured to construct a first state vector corresponding to each node based on the real-time operating data, wherein the first state vector is used to reflect the operating state of the node; a first determination module, configured to determine a second state vector corresponding to each node based on the topology information of the power grid under test and multiple first state vectors, wherein the dimension of the second state vector is smaller than the dimension of the first state vector; a second determination module, configured to determine the geodesic distance between every two nodes based on multiple second state vectors, wherein the geodesic distance is used to quantify the strength of electrical coupling between the two nodes corresponding to the geodesic distance; and a clustering module, configured to perform clustering processing on all nodes included in the power grid under test based on multiple geodesic distances to obtain multiple voltage partitions of the power grid under test.

[0014] According to another aspect of the embodiments of this application, a non-volatile storage medium is also provided, in which a computer program is stored, wherein the above-described voltage partitioning method is executed by running the computer program in the device where the non-volatile storage medium is located.

[0015] According to another aspect of the embodiments of this application, an electronic device is also provided, including a memory and a processor, wherein the memory stores a computer program and the processor is configured to execute the above-described voltage partitioning method through the computer program.

[0016] According to another aspect of the embodiments of this application, a computer program product is also provided, including computer instructions that, when executed by a processor, implement the steps of the above-described voltage partitioning method.

[0017] In this embodiment, real-time operational data of each node in the power grid under test is acquired, where each node represents an electrical connection point in the power grid under test. For each node, a first state vector is constructed based on the real-time operational data, reflecting the node's operational state. A second state vector is determined for each node based on the topology information of the power grid under test and multiple first state vectors, wherein the dimension of the second state vector is smaller than the dimension of the first state vector. The geodesic distance between any two nodes is determined based on multiple second state vectors, where the geodesic distance is used to quantify the strength of the electrical coupling between the two nodes corresponding to the geodesic distance. The geodetic distance method clusters all nodes in the power grid under test to obtain multiple voltage partitions. By jointly modeling the power grid's operating state and topology in a low-dimensional space and adaptively identifying areas of strong and weak electrical coupling based on geometric distance, it achieves the goal of automatically generating voltage control partitions that match the current operating conditions when partitioning the power grid, taking into account the dynamic changes of the power grid. This realizes the technical effects of automating voltage partitioning and improving the matching between voltage partitions and power grid operating conditions, thereby solving the technical problems of partitioning rigidity and control mismatch caused by relying on static parameters for point-based voltage partitioning in related technologies. Attached Figure Description

[0018] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0019] Figure 1 This is a hardware structure block diagram of a computer terminal for implementing a voltage partitioning method according to an embodiment of this application;

[0020] Figure 2 This is a flowchart illustrating the steps of a voltage partitioning method according to an embodiment of this application;

[0021] Figure 3 This is a schematic diagram illustrating the determination of geodetic distance according to an embodiment of this application;

[0022] Figure 4 This is a structural diagram of a voltage partitioning device according to an embodiment of this application;

[0023] Figure 5 This is a flowchart of a voltage partitioning apparatus performing a voltage partitioning method according to an embodiment of this application. Detailed Implementation

[0024] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0025] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0026] To better understand the embodiments of this application, the technical terms involved in the embodiments of this application are explained below:

[0027] Electrical connection point: In the embodiments of this application, the electrical connection point refers to a key access point or bus node that has a clear electrical connection relationship and can measure state quantities such as voltage, current, and power, such as load access point, distributed photovoltaic grid connection point, energy storage system connection point, reactive power compensation device installation point, or substation feeder bus, etc.

[0028] Three-phase voltage: The instantaneous voltage value of each phase of a three-phase alternating current in a power system relative to the neutral point or reference point. It can be used to characterize the voltage amplitude and phase balance state of a node.

[0029] Frequency: refers to the rate of periodic change of alternating current, measured in Hertz (Hz), and reflects the overall active power balance of the power grid.

[0030] Power angle: refers to the phase difference between the voltage phasor of a generator or grid node and the reference axis (such as the bus voltage of an infinite system), used to characterize the electromagnetic power transmission characteristics of the node.

[0031] Active power: refers to the power that actually does work in a circuit and is converted into other forms of energy (such as mechanical energy and thermal energy). The unit is kilowatt (kW), which can reflect the power consumption or generation load level of a node.

[0032] Reactive power: The power that is exchanged between the power source and the load without being consumed.

[0033] Manifold: In the latent representation space (i.e., low-dimensional space), the node states naturally form a "surface" or "structure" as the power grid operation mode changes; on this manifold, nodes that are close to each other have similar electrical behaviors and are tightly coupled, while nodes that are far apart have weak electrical connections.

[0034] In related technologies, voltage zoning employs a static method based on fixed grid parameters (such as impedance or voltage-reactive power sensitivity). This method relies on an unchanging topology and manually set clustering thresholds, considering only the linear relationship between voltage and reactive power, and neglecting changes in grid operating conditions and the coupling characteristics of multiple physical quantities such as frequency and power angle. Therefore, the zoning is rigid and unable to adapt to dynamic changes in the grid, leading to control mismatch problems when the grid dynamically changes. To address this issue, this application provides relevant solutions in its embodiments, which are detailed below.

[0035] According to an embodiment of this application, a method embodiment for voltage partitioning is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0036] The methods and embodiments provided in this application can be executed on mobile terminals, computer terminals, or similar computing devices. Figure 1 A hardware block diagram of a computer terminal for implementing a voltage partitioning method is shown. Figure 1 As shown, the computer terminal 10 may include one or more processors 102 (shown as 102a, 102b, ..., 102n in the figure) 102 (processor 102 may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.), a memory 104 for storing data, and a transmission device 106 for communication functions. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of a BUS bus), a network interface, a power supply, and / or a camera. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the aforementioned electronic device. For example, computer terminal 10 may also include... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0037] It should be noted that the aforementioned one or more processors 102 and / or other data processing circuits are generally referred to herein as "data processing circuits". These data processing circuits may be embodied, in whole or in part, in software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuits may be a single, independent processing module, or may be integrated, in whole or in part, into any other element within the computer terminal 10. As involved in the embodiments of this application, the data processing circuits serve as processor control (e.g., selection of a variable resistor termination path connected to an interface).

[0038] The memory 104 can be used to store software programs and modules for application software, such as the program instructions / data storage device corresponding to the voltage partitioning method in this embodiment. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, thereby implementing the voltage partitioning method described above. The memory 104 may include high-speed random access memory and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to the computer terminal 10 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0039] The transmission device 106 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the communication provider of the computer terminal 10. In one example, the transmission device 106 includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device 106 may be a Radio Frequency (RF) module, used for wireless communication with the Internet.

[0040] The display can be, for example, a touchscreen liquid crystal display (LCD) that allows the user to interact with the user interface of the computer terminal 10.

[0041] This application provides a voltage partitioning method that can operate under the above-described operating environment. Figure 2 This is a flowchart of the voltage partitioning method provided in the embodiments of this application, as follows: Figure 2 As shown, the method includes the following steps:

[0042] Step S202: Obtain real-time operating data of each node in the power grid under test, where a node represents an electrical connection point in the power grid under test.

[0043] This application provides a method for voltage partitioning based on the multi-dimensional operating status and topology of the power grid. By taking into account the dynamic operation of the power grid, the voltage partitioning is updated in a timely manner, thereby improving the adaptability of the voltage partitioning to the power grid status.

[0044] In step S202, real-time operating data of each electrical connection point is collected in real time through distributed terminals or Supervisory Control and Data Acquisition (SCADA) systems deployed at each node of the distribution network under test. The electrical connection points (i.e., nodes) in the network under test are key access points or busbar nodes with electrical connections that allow for the measurement of voltage, current, etc., such as load access points and distributed photovoltaic grid connection points. The real-time operating data of each electrical connection point describes its instantaneous electrical state. For example, the real-time operating data of each electrical connection point includes: three-phase voltage (… ),frequency( ), work angle ( ), active power ( ) and reactive power ( In this embodiment, real-time operating data of each node in the power grid under test is collected according to a preset sampling frequency of not less than 1Hz, so as to comprehensively reflect the electrical status of the node under the current operating conditions.

[0045] Step S204: For each node, construct a first state vector corresponding to the node based on real-time running data, wherein the first state vector is used to reflect the running state of the node.

[0046] In step S204, for each electrical node (i.e., node) in the power grid under test, a high-dimensional state vector (i.e., the first state vector) of the node is constructed based on the real-time operating data of the node collected in step S202; for example, for electrical node i, the following real-time operating data of node i were collected in step S202: three-phase voltage amplitude ( ), system frequency ( ), work angle ( ), active power ( ) and reactive power ( As can be seen from the above, the high-dimensional state vector (i.e., the first state vector) of each node is directly composed of the multi-dimensional real-time running data of that node. That is, the high-dimensional state vector of node i... Therefore, the high-dimensional state vector (i.e. the first state vector) of each node can quantitatively characterize the comprehensive electrical state of the node at the current operating moment.

[0047] Step S206: Determine the second state vector corresponding to each node based on the topology information of the power grid under test and multiple first state vectors, wherein the dimension of the second state vector is smaller than the dimension of the first state vector.

[0048] In step S206, based on the topology information of the power grid under test, the high-dimensional vector (i.e., the first state vector) constructed for each electrical node (i.e., the node) using the method provided in step S204 is subjected to dimensionality reduction processing to obtain a low-dimensional state vector (i.e., the second state vector) for each node. The topology information includes the connection relationship between each node and other nodes in the power grid under test. Therefore, the low-dimensional vector (i.e., the second state vector) obtained by processing the high-dimensional vector (i.e., the first state vector) with the topology information contains implicit features. These implicit features are used to reflect the real electrical connection relationship between nodes in the power grid under test.

[0049] According to some optional embodiments of this application, determining the second state vector corresponding to each node based on the topology information of the power grid under test and multiple first state vectors includes: generating a target matrix based on the topology information, wherein the target matrix is ​​used to record the electrical connection relationship between each node; encoding the target matrix and the first state matrix to obtain a second state matrix, wherein the first state matrix is ​​constructed based on multiple first state vectors, and each element in the second state matrix is ​​a second state vector.

[0050] In this embodiment, when performing dimensionality reduction processing on the high-dimensional state vector (i.e., the first state vector) constructed using the method provided in step S204 based on the topology information of the power grid under test, the high-dimensional state vector (i.e., the first state vector) of the node and the power grid topology graph A are used as the joint input of a graph variational autoencoder (Graph VAE). The pre-trained graph variational autoencoder (Graph VAE) is used to jointly encode the high-dimensional state vector (i.e., the first state vector) of the node and the power grid topology graph A into a low-dimensional latent space z, thereby obtaining the latent coordinates of each node. (i.e., the second state vector). The power grid topology diagram A records the topology information of the power grid under test, such as the electrical connections between nodes in the power grid under test. In this embodiment, when the power grid topology diagram A is used as input to the graph variational autoencoder, it is converted into a matrix (i.e., the target matrix) used to record the electrical connections between nodes.

[0051] Similarly, when using the high-dimensional state vectors of nodes (i.e., the first state vectors) as input to a Graph Variational Autoencoder (GraphVAE), it is also necessary to construct the node state matrix (i.e., the first state matrix) using the high-dimensional state vectors of all electrical connection points in the power grid under test as matrix elements. Therefore, the Graph VAE can jointly encode the matrix containing the power grid topology information (i.e., the target matrix) and the matrix containing the high-dimensional state vectors of the nodes (i.e., the first state matrix) in the graph convolutional layer. The Graph VAE jointly encodes the input information into a low-dimensional latent space z to obtain the latent coordinates of each node (i.e., the second state vector). Output the latent coordinates of the nodes. The low-dimensional state matrix (i.e., the second state matrix) serves as the elements.

[0052] The method provided in this embodiment achieves semantic fusion of topology and operating state by jointly encoding the multidimensional state vectors of nodes in the power grid and the power grid topology information. This enables the latent representation (i.e., the second state vector) of each node obtained by dimensionality reduction to not only reflect its own electrical state, but also its structural position in the power grid topology, thereby improving the ability of low-dimensional representation to model the overall coupling characteristics of the power grid.

[0053] Optionally, generating a target matrix based on topology information includes: determining the node identification information as the row and column identifiers of the target matrix; for each element in the target matrix, determining the target row identifier and target column identifier associated with the element; determining the electrical connection relationship in the power grid under test between the first node indicated by the target row identifier and the second node indicated by the target column identifier based on the topology information, wherein the electrical connection relationship includes: direct connection and indirect connection, where a direct connection indicates that there are no other devices on the connection line between the two nodes, and an indirect connection indicates that there are other devices on the connection line between the two nodes; when the electrical connection relationship is a direct connection, assigning the element a first value, and when the electrical connection relationship is an indirect connection, assigning the element a second value.

[0054] As mentioned in the above embodiments, when the power grid topology diagram A is used as input to the graph variational autoencoder (Graph VAE), it is converted into a matrix (i.e., a target matrix) that records the electrical connections between nodes. In this embodiment, the value of each element in the target matrix indicates the electrical connection relationship between the two nodes indicated by the (target) row identifier and (target) column identifier in the power grid topology diagram. For example, each element in the matrix (i.e., the target matrix) obtained by converting the power grid topology diagram A is represented as... ,if, =1, indicating that the electrical connection between node i and node j is a direct electrical connection; if =0 indicates that there is no direct connection between node i and node j. Therefore, converting the power grid topology information into a matrix (i.e., the target matrix) can be achieved through two steps: identification and assignment. Specifically, first, the topology of the power grid under test is identified, and the node numbers are used as the row and column identifiers of the target matrix. Elements representing node pairs with direct electrical connections (i.e., directly connected through lines or switches) are assigned one type of value (i.e., the first value, such as 1), and elements representing node pairs without direct electrical connections (not connected, or indirectly connected through intermediate equipment) are assigned another type of value (i.e., the second value, such as 0). This yields a binary adjacency matrix (i.e., the target matrix). The target matrix accurately reflects the physical topology connections of the power grid under test. The dimension of the target matrix is ​​determined by the number of electrical connection points (i.e., nodes) in the power grid under test. For example, if the power grid under test contains N electrical connection points (such as distribution transformer nodes, distributed power source access points, load nodes, etc.), then the dimension of the target matrix is... .

[0055] In the above scheme, a direct electrical connection refers to a connection between two nodes without any intermediate equipment; for example, in the power grid topology diagram A, if there are no other nodes on the connection line between node 1 (i=1) and node 2 (j=2), then... =1; however, the absence of a direct electrical connection between two nodes does not mean that there is absolutely no connection between them. Two nodes without a direct electrical connection may have an indirect electrical connection (i.e., an indirect connection), for example, =0, node 3 (i=3) and node 4 (j=4) may be indirectly connected. In topology graph A, node 3 and node 4 may be indirectly connected through intermediate devices such as transformers and capacitors.

[0056] The method provided in this embodiment assigns valid values ​​only to directly electrically connected node pairs, which avoids interference from indirect coupling on joint coding. This is beneficial to improving the latent space representation (i.e., the second state vector)'s ability to restore the actual electrical coupling relationship, thereby improving the accuracy of voltage partitioning.

[0057] According to some optional embodiments of this application, encoding the target matrix and the first state matrix to obtain a second state matrix includes: for each node, determining adjacent nodes with directly connected electrical connections based on the target matrix, and determining the state vector of the adjacent nodes based on the first state matrix; fusing the first state vector and the state vector of the adjacent nodes in the graph neural network layer of the encoder to obtain a fused vector; processing the fused vector and the first state vector in the fully connected neural network layer of the encoder to obtain Gaussian distribution parameters output by the fully connected neural network layer, wherein the Gaussian distribution parameters are used to quantize the distribution characteristics of the first state vector in the low-dimensional space; resampling the Gaussian distribution parameters to obtain the second state vector corresponding to the node, wherein the dimension of the second state vector is the same as the dimension of the low-dimensional space; and generating a second state matrix based on multiple second state vectors.

[0058] The graph variational autoencoder (GraphVAE) used in this embodiment is a conditional graph variational autoencoder (C-GraphVAE). This encoder consists of a graph neural network layer and a fully connected layer. During the joint encoding process, the graph neural network layer aggregates the target matrix layer by layer. Specifically, for each node, the high-dimensional state vector of the node (i.e., the first state vector) and the high-dimensional state vectors of its neighboring nodes are fused to obtain a fused feature (i.e., a fused vector) that records the local structure perception. The aforementioned neighboring nodes are determined based on the information recorded in the target matrix. When the target matrix records that the electrical connection between two nodes is a direct connection, these two nodes are determined to be each other's neighboring nodes. For example, if node 1 and node 2 in the previous embodiment have a direct connection relationship, then node 1 is the neighboring node of node 2, and node 2 is the neighboring node of node 1.

[0059] The graph neural network layer feeds the fused features (i.e., the fused vector) into the fully connected layer (i.e., the fully connected neural network layer), obtaining the Gaussian distribution parameters output by the fully connected layer. These Gaussian distribution parameters are used to quantify the probability distribution characteristics (i.e., distribution features) of each node in the low-dimensional latent space, thereby achieving compression of the high-dimensional state vector. For example, the Gaussian distribution parameters include: the mean value following a Gaussian distribution (…). ) and variance ( ), where the above mean and variance It is obtained by performing a two-layer fully connected transformation (including a linear transformation and a non-linear transformation based on a non-linear activation function (such as ReLU)) on the fused vector output by the fully connected layer of the graph neural network layer. The fully connected layer performs linear and non-linear transformations on the fused vector in sequence, mapping the fused vector into two independent output vectors, one of which is used as the mean. One as variance .

[0060] Furthermore, the Gaussian distribution parameters (mean) of the fully connected layer output are... and variance Reparameterized sampling is performed to obtain the low-dimensional latent vector (i.e., the second state vector) of each node. The process of reparameterized sampling can be described by the formula: ,in, It is a standard normal noise that follows a standard normal distribution. The value of is determined during the training process of the graph variational autoencoder. Ultimately, the low-dimensional state vector of the node (i.e., the second state vector) can be obtained. The matrix formed by the two states (i.e., the second state matrix) has the same number of high-dimensional state vectors (i.e., the first state vector) and low-dimensional state vectors (i.e., the second state vector) as the number of nodes remains unchanged during dimensionality reduction. Therefore, the first and second state matrices have the same dimensions. The dimension of the dimensionality-reduced state vectors (i.e., the second state vector) is the same as the dimension of the latent space. The dimension of the latent space (i.e., the low-dimensional space) can be preset or determined during the training of the graph variational autoencoder. In this embodiment, the dimension (d) of the low-dimensional space is 3. This ensures that the graph variational autoencoder can reduce the dimension of the vectors while fully preserving the electrical states and topological coupling information of the nodes, avoiding the feature loss problem caused by dimensionality reduction.

[0061] When training the graph variational autoencoder used in this application embodiment, historical operating data of the power grid under test (including different load levels and new energy output scenarios) is collected as training data. The training objective is to reconstruct node features and the adjacency matrix (i.e., the second state matrix). The loss function used in the training process includes reconstruction loss and information divergence (KL divergence). The trained graph variational autoencoder can map the node state at any time to the latent space (i.e., the low-dimensional space) to obtain the node manifold distribution under the current operating mode. This online mapping is achieved through one forward propagation, with a computational complexity of ( )for ,in, For the number of nodes, For example, the latent space dimension (e.g., 3). The hidden layer width is the number of neurons per layer, so even when dealing with a power grid with hundreds of nodes, dimensionality reduction can be completed in milliseconds.

[0062] In this embodiment, the graph variational autoencoder (GUE) can be loaded into memory. For example, the raw data of the GUE can be loaded from non-volatile memory into volatile memory so that the processor can run the GUE. The raw data of the GUE refers to unprocessed data, which typically includes the parameters and structural data of the GUE. The structural data can be the computational relationships based on the parameters, such as the forward propagation computational relationships between intermediate layers and between neurons. Specifically, the structural data can include structure-related code of the GUE, such as code used to perform related computations between intermediate layers and between neurons.

[0063] In one implementation, a region in memory can be partitioned for loading the graph variational autoencoder, which may include a structure data storage region and a parameter storage region. The structure data storage region stores structure-related code, and the parameters referenced by it can be pointed to by pointers to the addresses of specific parameters in the parameter storage region. During the training process of the graph variational autoencoder, it may be necessary to frequently update the parameters; in this case, updating the parameter values ​​in the parameter storage region is sufficient.

[0064] The method provided in this embodiment maps high-dimensional, high-noise original electrical quantities to a low-dimensional continuous manifold space while preserving the nonlinear relationship between topological constraints and operating states. It introduces dynamic change information, reduces the complexity of data processing, overcomes the limitations of static voltage partitioning in related technologies, and improves the speed of voltage partitioning.

[0065] Step S208: Determine the geodesic distance between every two nodes based on multiple second state vectors, wherein the geodesic distance is used to quantify the strength of the electrical coupling between the two nodes corresponding to the geodesic distance.

[0066] After mapping the high-dimensional vector to the low-dimensional space using the method provided in step S206, in step S208, a weighted undirected graph is constructed based on the local similarity of the three-dimensional latent state vectors (i.e., the second state vectors) of all nodes in the low-dimensional latent space. The shortest path algorithm is then used on the constructed weighted undirected graph to determine the shortest path length between every two nodes in the power grid under test. In this embodiment, the shortest path length between two nodes determined by the shortest path algorithm is the geodesic distance between the two nodes. Since the state vector (i.e., the second state vector) representing the node records various electrical information when determining the geodesic distance between two nodes, the geodesic distance between two nodes can comprehensively reflect the degree of electrical coupling between the nodes on the nonlinear manifold composed of voltage, frequency, power angle, active power, and reactive power. The smaller the geodesic distance between two nodes, the stronger the electrical correlation between the two nodes.

[0067] Optionally, determining the geodesic distance between every two nodes based on multiple second state vectors includes: determining the Euclidean distance between every two nodes based on the second state vector corresponding to each node; for each node, determining a preset number of target Euclidean distances in ascending order from the multiple Euclidean distances corresponding to the node, and determining the set of nodes corresponding to the preset number of target Euclidean distances as the set of adjacent nodes corresponding to the node; for the first node and the second node, if the first node's set of first adjacent nodes contains the second node, and the second node's set of second adjacent nodes contains the first node, establishing an edge between the first node and the second node, where the first node and the second node are any two different nodes in the power grid under test; generating a weighted undirected graph based on multiple nodes and multiple edges, where the weight of an edge is the Euclidean distance between the two nodes associated with the edge; determining the shortest distance from the first node to the second node based on the weighted undirected graph, where the shortest distance is the geodesic distance between the first node and the second node.

[0068] As mentioned in the above embodiments, the method provided in this application combines an undirected weighted graph and a shortest path algorithm to determine the geodesic distance between two nodes. The undirected weighted graph is constructed based on the Euclidean distance between nodes. The specific method is as follows: In the latent space, the low-dimensional feature vector (i.e., the second state vector) of each node i is used. For each node, the k-nearest neighbor (kNN) algorithm is used to construct an adjacency graph (i.e., an undirected weighted graph). Specifically, the kNN algorithm determines the edges connecting every two nodes. For each node, the Euclidean distance between it and all other nodes is determined, and then the k nodes with the smallest Euclidean distance are selected as its k-nearest neighbors. After determining the k-nearest neighbors of each node using the above method, further judgments are made based on these neighbors: an edge is established between two nodes only if they are each other's k-nearest neighbors, ultimately resulting in a symmetric weighted undirected graph. In the undirected weighted graph constructed using this scheme, the weight of each edge is the Euclidean distance between the two nodes connected by that edge in the latent space. In the kNN algorithm, k represents a preset number. Node i is selected as a k-nearest neighbor based on the Euclidean distance between node i and all other nodes. For example, if the total number of nodes is N and k is 10, the Euclidean distance between node i and the remaining (N-1) nodes j is determined. Among the (N-1) Euclidean distances, select the 10 smaller (target) Euclidean distances in ascending order of their values. The other nodes j corresponding to these 10 target Euclidean distances are the k nearest neighbors (or the set of adjacent nodes) of node i.

[0069] After constructing the undirected weighted graph, the shortest path algorithm is used to determine the shortest path length between any two nodes i and j in the undirected weighted graph. (i.e., the shortest distance). The shortest distance between two nodes determined by the shortest path algorithm described above is the geodesic distance between these two nodes. In this embodiment, the shortest path algorithm can specifically be Dijkstra's algorithm. When using Dijkstra's algorithm to determine the geodesic distance, for every two nodes, taking one node as the starting point and the other node as the ending point, the shortest path length from the starting point to the ending point is searched in the undirected weighted graph. The shortest distance is used as the geodesic distance between the two nodes. Figure 3 This is a schematic diagram for determining geodetic distance, such as... Figure 3 As shown, in an undirected weighted graph, there are two paths from node m to node n. When determining the geodesic distance between nodes m and n, node m is used as the starting point and node n as the ending point. The path with the shortest distance between nodes m and n is searched. In this embodiment, the weight of each edge is used as the length of that edge. Since the weight of each edge is the Euclidean distance between the two nodes connected by that edge, the path with the shortest distance is the path with the smallest sum of weights. Specifically... Figure 3 In the scenario shown, since there are two paths from node m to node n, the shortest path is determined by comparing the weighted sum of path 1 (node ​​m → node p → node n) and path 2 (node ​​m → node Q → node n). If the length of path 1 (the Euclidean distance between node m and node p plus the Euclidean distance between node p and node n) is less than the length of path 2 (the Euclidean distance between node m and node Q plus the Euclidean distance between node Q and node n), then the geodesic distance between node m and node n is the length of path 1 (the Euclidean distance between node m and node p plus the Euclidean distance between node p and node n).

[0070] The geodetic distance determined by the method provided in this embodiment can reflect the interrelationship of the inherent electrical behavior patterns of electrical nodes under actual operating conditions; the stronger the electrical coupling between nodes, the smaller the geodetic distance, which improves the matching degree between voltage zoning and the actual operating conditions of the power grid, and improves the accuracy of voltage zoning.

[0071] Step S210: Cluster all nodes in the power grid under test according to multiple geodetic distances to obtain multiple voltage partitions of the power grid under test.

[0072] In step S210, a clustering algorithm is used in the low-dimensional latent space, with the geodesic distance between nodes as the similarity measure, to cluster the nodes in the power grid to be tested, and the nodes with close geodesic distance and high density are grouped into the same cluster; among them, the nodes that are grouped into a cluster are nodes with tight electrical coupling and similar operating states, and each cluster corresponds to a voltage control area, while isolated nodes that are not grouped into any cluster are regarded as areas that require special attention.

[0073] Optionally, clustering is performed on all nodes in the power grid under test based on multiple geodesic distances, including: determining the local curvature of each node, where the local curvature is used to quantify the electrical behavior coupling strength of the node in its region; determining the cluster radius of the node based on the local curvature, determining the cluster range of the node based on the cluster radius, and determining the number of actual nodes contained within the cluster range; determining the target number of actual nodes greater than the minimum number of nodes, and determining the nodes corresponding to the target number of actual nodes as core nodes, where the minimum number of nodes is the lower limit of the number of nodes contained within the cluster range; and clustering is performed based on the core nodes, the cluster radius, and the target topology graph, where the target topology graph reflects the electrical structure of the power grid under test, the vertices of the target topology graph are nodes, and the edges in the target topology graph are generated based on the geodesic distances.

[0074] In this embodiment, when performing clustering based on geodesic distance, nodes with close geodesic distances and high density are clustered into the same voltage control region. The clustering radius used in the clustering process is determined based on the local curvature of the nodes. In this embodiment, the local curvature of each node i is... The local curvature of node i can be determined based on its K-nearest neighbor topology in the undirected weighted graph described above. For example, the variance of the k Euclidean distances between node i and its k-nearest neighbors can be used to determine the local curvature of node i. Alternatively, one can first determine the angles between the vectors of node i and its K nearest neighbors, and then use the standard deviation or entropy of the k angles as the local curvature of node i. Determine the local curvature based on the above. The method can be used to determine the local curvature of node i. It can reflect the degree of drastic change in the operating state of the electrical region where node i is located. For example, when determining the local curvature based on the variance, the larger the variance, the more uneven the neighborhood structure and the more curved the local manifold.

[0075] In this embodiment, the DBSCAN algorithm can be used for clustering. DBSCAN does not require pre-specifying the number of clusters; only the neighborhood radius needs to be set. And the minimum number of points minPts (i.e., the minimum number of nodes). In this embodiment, the neighborhood radius... Based on local curvature Determine the neighborhood radius. The method can be described as a formula = ,in, As the reference radius, This is the sensitivity coefficient. The value can be determined based on the historical operating data of the power grid under test; for example, it can be set to 0.2. The value of is preset, for example, it can be set to 1.5; using the above method, the radius will be appropriately enlarged for regions with large local curvature (manifold bending), which can avoid over-segmentation. The minimum number of points minPts = 2 × d, where d is the dimension of the low-dimensional state vector (i.e., the second state vector).

[0076] Since the method provided in this application embodiment clusters nodes with short geodetic distances and high density into the same voltage control region, the minimum number of nodes minPts is used as the threshold during clustering, and the adaptive neighborhood radius is used. Each node's neighborhood is defined. When the actual number of nodes contained in a node's neighborhood (i.e., the actual number of nodes) is greater than minPts, this node is determined as a core node. Further, starting from the core nodes, clusters of core nodes are formed by expanding along the connected graph (i.e., the target topology graph) defined by the geodesic distance between the core nodes. The connected graph described above uses each node as a vertex, and edges are generated based on the geodesic distance between nodes. In the target topology graph, two nodes with geodesic distances are connected by an edge, and the length of each edge is the geodesic distance between the two nodes connected by that edge.

[0077] The method provided in this embodiment adaptively adjusts the clustering radius based on the local curvature perception of the nodes, performs clustering based on the clustering radius, and completes the subsequent voltage partitioning of the power grid under test based on the clustering results. This method is different from the static voltage partitioning based on the preset number of partitions in related technologies, which is based on the operating status of the power grid. This improves the matching between the voltage partitioning and the operating scenario of the power grid under test.

[0078] According to some optional embodiments of this application, clustering is performed based on core nodes, clustering radius, and target topology graph, including: in each round of clustering, determining the clustering starting point, identifying clusterable nodes among the nodes located within the clustering range of the clustering starting point in the target topology graph, and adding the clusterable nodes to the cluster set until all clusterable nodes are added to the cluster set. Here, clusterable nodes are nodes not included in the cluster set. In the first round of clustering, the clustering starting point is any core node. In subsequent rounds of clustering, the clustering starting point is a newly added node in the cluster set. The area indicated by each cluster set in the target topology graph is a voltage partition.

[0079] In this embodiment, multiple rounds of clustering are required to expand the connected graph (i.e., the target topology graph) along the geodesic distance defined by the core point to form clusters. In the first round of clustering, a core node is selected as the starting point for clustering, and its adjacent nodes in the target topology graph are traversed. Clusters whose geodesic distance from the core node in the target topology graph is less than or equal to the cluster radius are selected. Nodes not yet assigned to any partition are added to the cluster set of the starting point. Subsequently, each newly added node (i.e., the new node) is used as a new starting point for subsequent rounds of clustering (i.e., non-first round clustering). During subsequent clustering, all unvisited adjacent nodes of this new starting point are recursively searched. If any of these unvisited adjacent nodes has a geodesic distance from the new starting point that is less than the cluster radius, then the cluster is considered a new starting point. If a node is found, it is added to the cluster set of the starting point of the first round of clustering, until all density-reachable nodes in the connected region of the starting point of the first round of clustering have been added to its cluster set, and no new nodes can be added.

[0080] After clustering a core node using the above method, if there are multiple remaining nodes that are not added to the cluster set, and these remaining nodes include a core node, then the above method is repeated, starting with one of the core nodes among the remaining nodes. This process continues until, after a certain clustering iteration, no more core nodes remain among the remaining nodes. These remaining nodes are then marked as noise and can be considered as regions requiring special attention. After clustering is complete, a closed voltage control region (i.e., voltage partition) can be determined in the target topology map based on each generated cluster set.

[0081] The method provided in this embodiment achieves accurate identification of irregularly shaped partitions through traversal recursive expansion. It clusters based on geodesic distance rather than according to a preset number of partitions, which improves the matching between voltage partitions and power grid operating status and improves the accuracy of partitioning results.

[0082] According to some optional embodiments of this application, after obtaining multiple voltage zones of the power grid under test, the method further includes: determining a center state vector based on a second state vector of each node contained in the voltage zone, wherein the center state vector is used to represent the geometric center of the voltage zone; for each node contained in the voltage zone, determining a distance value between the node and the center state vector based on the second state vector corresponding to the node; and determining the node corresponding to the minimum distance value as the control node of the voltage zone, wherein the control node is the voltage control point of the voltage zone.

[0083] After clustering is completed using the method provided in the above embodiments, based on The defined rules determine the geometric center in each cluster set (i.e., voltage partition). (i.e., the central state vector), where cluster represents the set of all nodes contained in the cluster set of node i (i.e., a voltage partition). For each voltage partition, a distance is selected. The nearest node with controllable resources (such as reactive power compensation devices, photovoltaic inverters, and energy storage devices) is designated as the key control point (i.e., the control node, or pilot bus) for this zone. This control node has characteristics that represent the overall electrical behavior of the zone, and since the zone is determined by the current operating state of the power grid, the control point remains stable over time. Voltage control of the control node determined using the above method can effectively regulate the voltage of the entire voltage zone where the control node is located.

[0084] The method provided in this embodiment is based on locating voltage control nodes at the manifold geometric center. It completes voltage partitioning by taking into account the actual electrical conditions of the power grid, which can improve the response efficiency and coordination accuracy of partition control and avoid the problem of partition control lag caused by manually specifying control points in related technologies.

[0085] Through the above steps, dynamic voltage partitioning of the power grid can be achieved based on the real-time multidimensional operating status and topology of the power grid, overcoming the limitations of related technologies that rely on static parameters and manually set thresholds for voltage partitioning. By compressing high-dimensional electrical quantities into low-dimensional manifold representations, determining geodesic distances based on low-dimensional manifold representations to quantify the nonlinear electrical coupling strength between nodes, and automatically performing voltage partitioning based on electrical coupling strength clustering, the response speed to dynamic operating conditions such as energy fluctuations and load changes in the power grid is improved, the accuracy of voltage partitioning is enhanced, the synchronization between voltage control strategies and real-time operating status is ensured, and the voltage stability of the distribution network is strengthened.

[0086] Figure 4 This is a structural diagram of a voltage partitioning device provided according to an embodiment of this application, such as... Figure 4 As shown, the voltage partitioning device includes: an acquisition module 40, used to acquire real-time operating data of each node in the power grid under test, wherein the node represents an electrical connection point in the power grid under test; a vector construction module 42, used to construct a first state vector corresponding to each node based on the real-time operating data, wherein the first state vector is used to reflect the operating state of the node; a first determination module 44, used to determine a second state vector corresponding to each node based on the topology information of the power grid under test and multiple first state vectors, wherein the dimension of the second state vector is smaller than the dimension of the first state vector; a second determination module 46, used to determine the geodesic distance between every two nodes based on multiple second state vectors, wherein the geodesic distance is used to quantify the strength of electrical coupling between the two nodes corresponding to the geodesic distance; and a clustering module 48, used to perform clustering processing on all nodes included in the power grid under test based on multiple geodesic distances to obtain multiple voltage partitions of the power grid under test.

[0087] Figure 5 This is a flowchart of the voltage partitioning device performing the voltage partitioning method, such as... Figure 5As shown, when the voltage partitioning device executes the voltage partitioning method provided in this application embodiment, it receives real-time operating data of each node in the power grid under test through the acquisition module 40. For example, the power grid under test may be a 10 kV distribution network containing 60 nodes and with a large number of distributed photovoltaic connections. The acquisition module 40 collects real-time operating data such as voltage, frequency, power angle, active power, and reactive power of each node through distributed terminals deployed at key nodes of the power grid. The vector construction module 42 constructs a high-dimensional vector for each node based on the data output by the acquisition module 40. (i.e., the first state vector). The first determination module 44 uses an offline-trained graph variational autoencoder (Graph VAE) to jointly map the high-dimensional vectors of the nodes (i.e., the first state vectors) output by the vector construction module 42 with the topological information of the power grid under test to a low-dimensional latent space, and outputs the low-dimensional state vectors of the nodes. (i.e., the second state vector), such as Figure 5 As shown, the Graph Variational Autoencoder (Graph VAE) achieves processing of high-dimensional vectors through forward propagation. Dimensionality reduction processing; in this embodiment, the graph variational autoencoder (Graph VAE) is trained using historical data of the power grid under test. Further, the second determining module 46 determines the geodesic distance between any two nodes based on the low-dimensional state vector of the nodes (i.e., the second state vector). For example, the geodesic distance between any two nodes i and j can be determined using the k-nearest neighbor graph construction and Dijkstra's algorithm described in the above embodiment. In this embodiment, geodesic distance can quantify the electrical coupling strength between nodes, reflecting the true electrical correlation strength of nodes on a nonlinear manifold. The clustering module 48 clusters nodes based on geodesic distance, for example, using an adaptive density clustering (DBSCAN) method to cluster nodes into multiple voltage zones. During the clustering process, the clustering radius is automatically adjusted based on the local curvature of the nodes. As before... Figure 5 As shown, the method provided in this application embodiment can determine the geometric center within each partition after clustering, and select the node closest to the geometric center and possessing controllable resources as the key control point (Pilot Bus) of that partition. In this embodiment, if the step of identifying the key control point (Pilot Bus) is performed, the partitioning result and the key control point of each partition are output; if the step of identifying the key control point (Pilot Bus) is not performed, only the partitioning result is output.

[0088] It should be noted that, Figure 4 Preferred embodiments of the shown examples can be found in [reference needed]. Figure 2 The relevant descriptions of the embodiments shown will not be repeated here.

[0089] For example, when implementing the voltage zoning method provided in this application for a 10kV distribution network containing 60 nodes, real-time operating data such as node voltage, frequency, power angle, and power are first collected through distributed terminals deployed at key nodes of the power grid. On a certain morning, when the power grid experiences a voltage spike, the current cross-sectional data is collected to construct a high-dimensional state vector (i.e., the first state vector) for the 60 nodes. The latent coordinates are then obtained through forward propagation using an encoder. (i.e., the second state vector), this process takes approximately 2 milliseconds (ms). The latent coordinates output above... The dimension is determined during encoder training. For example, if the encoder (GraphVAE) is trained using historical data from the previous year's power grid (including photovoltaic fluctuations and load changes), after training, the encoder maps the input information to a 3-dimensional latent space. Therefore, when using the encoder in an application, the encoder output... The dimension is 3D. Further, the geodesic distance between every two nodes in the 60 nodes is determined, and DBSCAN clustering is performed on these 60 nodes based on the geodesic distance, taking approximately 8ms. The clustering results show that the nodes are divided into 4 clusters: Region 1 (near photovoltaic nodes), Region 2 (industrial load area), Region 3 (residential area), and Region 4 (weakly connected end area). Region 1 has high node density, large curvature, adaptively enlarged cluster radius, and clear boundaries with neighboring regions. Next, the geometric center of Region 1 can be determined, and the node closest to the geometric center and directly connected or equipped with a photovoltaic inverter is selected as the key control point. For Region 2, the substation bus closest to the geometric center and directly connected or connected according to the Static Var Generator (SVG) is selected as the key control point; since four zones are clustered, a total of 4 key control points are identified. Finally, the zoning results and key control points are sent to the Automatic Voltage Control (AVC) system. In Zone 1, the primary focus is on regulating the photovoltaic inverters, while in Zone 2, the primary focus is on regulating the SVG (Static Var Generator). After coordinated control, the voltage is restored to a safe range. If the grid experiences an increase in load and a decrease in photovoltaic output in the afternoon, resulting in a change in system operating status, the system will automatically re-execute the zoning (time <15ms) to obtain new zoning results. This method ensures that the control strategy always matches the current operating conditions, improving the accuracy of the zoning.

[0090] This application also provides a non-volatile storage medium storing a computer program, wherein the device containing the non-volatile storage medium executes the above-mentioned voltage partitioning method by running the computer program.

[0091] The aforementioned non-volatile storage medium is used to store a program that performs the following functions: acquiring real-time operating data of each node in the power grid under test, wherein each node represents an electrical connection point in the power grid under test; for each node, constructing a first state vector corresponding to the node based on the real-time operating data, wherein the first state vector is used to reflect the operating state of the node; determining a second state vector corresponding to each node based on the topology information of the power grid under test and multiple first state vectors, wherein the dimension of the second state vector is smaller than the dimension of the first state vector; determining the geodesic distance between every two nodes based on multiple second state vectors, wherein the geodesic distance is used to quantify the strength of electrical coupling between the two nodes corresponding to the geodesic distance; and clustering all nodes in the power grid under test based on multiple geodesic distances to obtain multiple voltage partitions of the power grid under test.

[0092] This application also provides an electronic device, including a memory and a processor. The memory stores a computer program, and the processor is configured to execute the above-mentioned voltage partitioning method through the computer program.

[0093] The processor in the aforementioned electronic device is used to run a program that performs the following functions: acquiring real-time operating data of each node in the power grid under test, wherein each node represents an electrical connection point in the power grid under test; for each node, constructing a first state vector corresponding to the node based on the real-time operating data, wherein the first state vector is used to reflect the operating state of the node; determining a second state vector corresponding to each node based on the topology information of the power grid under test and multiple first state vectors, wherein the dimension of the second state vector is smaller than the dimension of the first state vector; determining the geodesic distance between every two nodes based on multiple second state vectors, wherein the geodesic distance is used to quantify the strength of electrical coupling between the two nodes corresponding to the geodesic distance; and clustering all nodes in the power grid under test based on multiple geodesic distances to obtain multiple voltage partitions of the power grid under test.

[0094] According to an embodiment of this application, a computer program product is also provided, including computer instructions, which, when executed by a processor, implement the steps of the above-described voltage partitioning method.

[0095] It should be noted that each module in the voltage partitioning device can be a program module (e.g., a set of program instructions to implement a specific function) or a hardware module. For the latter, it can be manifested in the following forms, but is not limited to them: each of the above modules is manifested as a processor, or the functions of each of the above modules are implemented by a processor.

[0096] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0097] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0098] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0099] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0100] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0101] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to related technologies, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0102] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method of voltage partitioning, the method comprising: include: Acquire real-time operating data of each node in the power grid under test, wherein the node represents an electrical connection point in the power grid under test; For each node, a first state vector is constructed based on the real-time operating data, wherein the first state vector is used to reflect the operating state of the node; A second state vector corresponding to each node is determined based on the topology information of the power grid under test and multiple first state vectors, wherein the dimension of the second state vector is smaller than the dimension of the first state vector. The geodesic distance between each pair of nodes is determined based on a plurality of second state vectors, wherein the geodesic distance is used to quantify the strength of the electrical coupling between the two nodes corresponding to the geodesic distance; Clustering is performed on all nodes in the power grid under test based on multiple geodetic distances to obtain multiple voltage zones of the power grid under test.

2. The method of claim 1, wherein, Determining a second state vector corresponding to each node based on the topology information of the power grid under test and multiple first state vectors includes: A target matrix is ​​generated based on the topology information, wherein the target matrix is ​​used to record the electrical connection relationships between each of the nodes; The target matrix and the first state matrix are encoded to obtain a second state matrix, wherein the first state matrix is ​​constructed based on multiple first state vectors, and each element in the second state matrix is ​​a second state vector.

3. The method of claim 2, wherein, Generating a target matrix based on the topology information includes: The identification information of the node is determined as the row identifier and column identifier of the target matrix; For each element in the target matrix, determine the target row identifier and target column identifier associated with that element; The electrical connection relationship between the first node indicated by the target row identifier and the second node indicated by the target column identifier in the power grid under test is determined based on the topology information. The electrical connection relationship includes direct connection and indirect connection. The direct connection indicates that there are no other devices on the connection line between the two nodes, and the indirect connection indicates that there are other devices on the connection line between the two nodes. When the electrical connection is a direct connection, the element is assigned a first value; when the electrical connection is an indirect connection, the element is assigned a second value.

4. The method of claim 2, wherein, The target matrix and the first state matrix are encoded to obtain the second state matrix, including: For each node, the adjacent nodes with which the electrical connection with the node is directly connected are determined according to the target matrix, and the state vector of the adjacent node is determined according to the first state matrix; The first state vector and the state vectors of the neighboring nodes are fused in the graph neural network layer of the encoder to obtain a fused vector; The fused vector and the first state vector are processed in the fully connected neural network layer of the encoder to obtain the Gaussian distribution parameters output by the fully connected neural network layer, wherein the Gaussian distribution parameters are used to quantify the distribution characteristics of the first state vector in the low-dimensional space. The Gaussian distribution parameters are resampled to obtain the second state vector corresponding to the node, wherein the dimension of the second state vector is the same as the dimension of the low-dimensional space; The second state matrix is ​​generated based on multiple second state vectors.

5. The method of claim 1, wherein, Determining the geodesic distance between every two nodes based on multiple second state vectors includes: The Euclidean distance between any two nodes is determined based on the second state vector corresponding to each node. For each node, a preset number of target Euclidean distances are determined in ascending order from the multiple Euclidean distances corresponding to the node, and the set consisting of a preset number of nodes corresponding to the preset number of target Euclidean distances is determined as the set of adjacent nodes corresponding to the node. For the first node and the second node, if the first node's first adjacent node set contains the second node, and the second node's second adjacent node set contains the first node, an edge is established between the first node and the second node, wherein the first node and the second node are any two different nodes in the power grid under test. A weighted undirected graph is generated based on the plurality of nodes and the plurality of edges, wherein the weight of the edge is the Euclidean distance between the two nodes associated with the edge; The shortest distance from the first node to the second node is determined based on the weighted undirected graph, wherein the shortest distance is the geodesic distance between the first node and the second node.

6. The method of claim 1, wherein, Clustering is performed on all nodes in the power grid under test based on multiple geodetic distances, including: Determine the local curvature of each node, wherein the local curvature is used to quantify the electrical behavior coupling strength of the node in its region; The clustering radius of the node is determined based on the local curvature, the clustering range of the node is determined based on the clustering radius, and the actual number of nodes contained within the clustering range is determined. Determine the target actual number of nodes that is greater than the minimum number of nodes, and identify the nodes corresponding to the target actual number of nodes as core nodes, wherein the minimum number of nodes is the lower limit of the number of nodes included in the clustering range; Clustering is performed based on the core node, the cluster radius, and the target topology graph, wherein the target topology graph is used to reflect the electrical structure of the power grid under test, the vertices of the target topology graph are the nodes, and the edges of the target topology graph are generated based on the geodesic distance.

7. The method of claim 6, wherein, Clustering is performed based on the core nodes, the cluster radius, and the target topology graph, including: In each round of clustering, a clustering starting point is determined. Among the nodes located within the clustering range of the clustering starting point in the target topology graph, clusterable nodes are identified and added to the cluster set. This process continues until all clusterable nodes are added to the cluster set. Clusterable nodes are nodes not included in the cluster set. In the first round of clustering, the clustering starting point is any of the core nodes. In subsequent rounds of clustering, the clustering starting point is a newly added node in the cluster set. Each cluster set indicates a voltage partition in the target topology graph.

8. The method of claim 1, wherein, After obtaining multiple voltage zones of the power grid under test, the method further includes: A center state vector is determined based on the second state vector of each node contained in the voltage partition, wherein the center state vector is used to represent the geometric center of the voltage partition; For each node contained in the voltage partition, the distance value between the node and the center state vector is determined according to the second state vector corresponding to the node; The node corresponding to the minimum distance value is determined as the control node of the voltage partition, wherein the control node is the voltage control point of the voltage partition.

9. An apparatus for voltage partitioning, the apparatus comprising: include: The acquisition module is used to acquire real-time operating data of each node in the power grid under test, wherein the node represents the electrical connection point in the power grid under test; A vector construction module is used to construct a first state vector corresponding to each node based on the real-time running data, wherein the first state vector is used to reflect the running state of the node; The first determining module is used to determine a second state vector corresponding to each node based on the topology information of the power grid under test and multiple first state vectors, wherein the dimension of the second state vector is smaller than the dimension of the first state vector. The second determining module is used to determine the geodesic distance between every two nodes based on a plurality of second state vectors, wherein the geodesic distance is used to quantify the strength of the electrical coupling between the two nodes corresponding to the geodesic distance; The clustering module is used to cluster all nodes in the power grid under test according to multiple geodesic distances to obtain multiple voltage partitions of the power grid under test.

10. A non-volatile storage medium, comprising: The non-volatile storage medium stores a computer program, wherein the device containing the non-volatile storage medium executes the voltage partitioning method according to any one of claims 1 to 8 by running the computer program.

11. An electronic device comprising a memory and a processor, characterized in that The memory stores a computer program, and the processor is configured to execute the voltage partitioning method according to any one of claims 1 to 8 through the computer program.

12. A computer program product comprising computer instructions, characterized in that, When the computer instructions are executed by the processor, they implement the steps of the voltage partitioning method according to any one of claims 1 to 8.

Images