A power distribution network partitioning and verification method, system, device and medium

Abstract

The application provides a power distribution network partitioning and verification method, system, device and medium, comprising: constructing an embedding matrix through feature decomposition and feature extraction based on historical voltage time series of each node of a power distribution network topology; obtaining an initial partitioning scheme by clustering the embedding matrix using a nearest neighbor propagation algorithm; performing feasibility partitioning verification on the initial partitioning scheme based on connectivity constraints and source-load balance constraints, adjusting the partitioning that does not pass the verification until the verification is satisfied, and outputting the partitioning that passes the verification as a final partitioning result. The application describes the dynamic correlation of distributed power supply access to the power distribution network by introducing historical voltage time series of nodes and the feasibility verification of connectivity and source-load balance constraints, improves the adaptability of the partitioning result to changes in the operating state, and provides reliable technical support for the partitioning operation control, optimal scheduling and intelligent auxiliary decision-making of the power distribution network.

Description

Technical Field

[0001] This invention relates to the field of power distribution network zoning, and specifically to a power distribution network zoning and verification method, system, equipment and medium. Background Technology

[0002] Against the backdrop of "dual-carbon" goals and the construction of new power systems, distributed resources, represented by photovoltaics, energy storage, and electric vehicle charging loads, are widely integrated into medium- and low-voltage distribution networks, resulting in significant randomness and uncertainty in the operation of these networks. The correlation of voltage fluctuations between different nodes has significantly increased and exhibited marked differences. The electrical coupling relationships of the distribution system are no longer determined solely by the physical topology but are simultaneously influenced by operating conditions and load behavior. In this context, how to rationally partition the distribution network while satisfying system connectivity and operational constraints, ensuring similar node operating characteristics within the same partition and decoupling between partitions, becomes a key issue supporting hierarchical control and optimized scheduling.

[0003] Existing research primarily relies on network topology or electrical distance metrics to statically partition distribution networks using information such as branch impedance, electrical distance, or power flow sensitivity. Some studies combine graph theory methods with spectral clustering techniques, constructing graph models and performing feature decomposition to achieve structured partitioning of distribution networks. These methods have improved the rationality of the partitioning results to some extent, providing technical support for the partitioned operation of distribution networks. However, some existing partitioning methods based on static topology or single electrical metrics struggle to reflect the dynamic characteristics of the distribution network's operating state after the integration of distributed generation, resulting in insufficient adaptability of the partitioning results to changes in operating conditions. Summary of the Invention

[0004] To address the shortcomings of existing distribution network zoning technologies in reflecting the dynamic characteristics of distribution network operation after the integration of distributed generation and in adapting to changes in operating conditions, this invention proposes a distribution network zoning and verification method, including: Based on the historical voltage time series of each node in the distribution network topology, an embedding matrix is ​​constructed through feature decomposition and feature extraction. The embedding matrix is ​​clustered using the nearest neighbor propagation algorithm to obtain an initial partitioning scheme; The initial partitioning scheme is validated for feasibility based on connectivity constraints and source-load balance constraints. Partitions that fail the validation are adjusted until the validation is satisfied. All partitions that pass the validation are output as the final partitioning result.

[0005] Preferably, the historical voltage time series of each node based on the distribution network topology is used to construct an embedding matrix through feature decomposition and feature extraction, including: The historical voltage time series of each node in the distribution network topology are used to form an observation matrix. The dynamic similarity between any two non-equilibrium nodes in the observation matrix is ​​calculated, and a symmetric similarity matrix is ​​constructed. The symmetric similarity matrix is ​​used as a weighted adjacency matrix, and combined with the degree matrix to construct a Laplace matrix; Dimensionality reduction decomposition is performed based on the Laplacian matrix to obtain the eigenvalues ​​and the corresponding eigenvectors. The cumulative eigenvalue ratio method is used to calculate the proportion of non-zero eigenvalues, and the number of non-zero eigenvalues ​​that meet the preset threshold proportion is counted as the embedding dimension. Extract the feature vectors corresponding to the feature values ​​of the embedding dimension, and form an embedding matrix; The historical voltage time series of each node includes historical data and real-time data.

[0006] Preferably, the step of clustering the embedding matrix using the nearest neighbor propagation algorithm to obtain the initial partitioning scheme includes: The mean of the squared negative Euclidean distance between any two different nodes in the embedding matrix is ​​calculated as the global preference value. Based on the global preference value and the initialized responsibility and availability, the responsibility and availability are iteratively updated alternately until the change in responsibility and availability is less than the preset convergence value or the maximum number of iterations is reached. Cluster centers are determined based on maximizing responsibility and availability. Nodes with the same cluster center are grouped into the same cluster to obtain the initial partitioning scheme.

[0007] Preferably, the feasibility partitioning verification of the initial partitioning scheme based on connectivity constraints and source-load balance constraints includes: The undirected graph formed by the node set and edge set in the distribution network topology is used as a connectivity constraint. All nodes in the current partition, the edges of all nodes in the current partition in the undirected graph, and the power nodes are used to form a subgraph. Starting from the power nodes, the connectivity of all nodes in the subgraph is traversed using breadth-first search. If all nodes in the subgraph are connected, the current partition passes the connectivity constraint check; otherwise, the current partition fails the connectivity constraint check and is considered a disconnected partition. The average load of all nodes in the current partition is combined with the imbalance threshold as the source-load balance constraint. The imbalance of the current partition is calculated based on the historical average output and average load of the distributed power sources of all nodes in the current partition. Compare the imbalance of the current partition with the source load balance constraint. If the imbalance of the current partition is greater than the source load balance constraint, the current partition fails the source load balance constraint check and is a source load imbalance partition. Otherwise, the current partition passes the source load balance constraint check. The current partition that passes both connectivity and source-load balance constraints is considered a feasible partition.

[0008] Preferably, adjusting the partitions that failed the verification includes: For disconnected partitions, a graph search algorithm is used to retrieve all connected component sets, and all connected components in the connected component sets except for the largest connected component are merged into the target connected partition adjacent to the connected component. The boundary nodes in the source-load imbalance partition are migrated sequentially to the target source-load balance partition adjacent to the node.

[0009] Preferably, the determination of the target connected component includes: All feasible partitions adjacent to the connected component to be merged are considered as connected candidate partitions. Based on the weights of the edges between nodes in the connected component to be merged and nodes in each candidate connected partition, the average connection strength between the connected component to be merged and each candidate connected partition is calculated, and the candidate connected partition with the largest average connection strength is determined as the target connected partition.

[0010] Preferably, the determination of the target source-load balancing partition includes: All feasible partitions adjacent to the boundary node to be migrated are selected as source payload candidate partitions; Based on the imbalance amount of the boundary node to be migrated out by the source load imbalance partition and the imbalance amount of the boundary node to be migrated into by the source load candidate partition, the migration index is calculated. The candidate source-load partition that satisfies the source-load balance constraint and has the smallest migration index is determined as the target source-load balance partition.

[0011] Based on the same inventive concept, the present invention also provides a power distribution network zoning and verification system, comprising: The matrix construction module is used to construct an embedding matrix based on the historical voltage time series of each node in the distribution network topology through feature decomposition and feature extraction. The initial partitioning module is used to cluster the embedding matrix using the nearest neighbor propagation algorithm to obtain an initial partitioning scheme. The final partitioning module is used to perform feasibility partitioning verification on the initial partitioning scheme based on connectivity constraints and source-load balance constraints. Partitions that fail the verification are adjusted until the verification is satisfied, and all partitions that pass the verification are output as the final partitioning result.

[0012] Preferably, the matrix construction module includes: The similarity matrix construction unit is used to form an observation matrix from the historical voltage time series of each node in the distribution network topology, calculate the dynamic similarity between any two non-equilibrium nodes in the observation matrix, and construct a symmetric similarity matrix. The Laplacian matrix construction unit is used to construct a Laplacian matrix by combining the symmetric similarity matrix as a weighted adjacency matrix with the degree matrix. An embedding matrix construction unit is used to perform dimensionality reduction decomposition based on the Laplacian matrix to obtain eigenvalues ​​and corresponding eigenvectors; calculate the numerical proportion of non-zero eigenvalues ​​using the cumulative eigenvalue ratio method; count the number of non-zero eigenvalues ​​whose numerical proportions meet a preset threshold as the embedding dimension; and extract the eigenvectors corresponding to the eigenvalues ​​of the embedding dimension to form an embedding matrix. The historical voltage time series of each node includes historical data and real-time data.

[0013] Preferably, the initial partitioning module is specifically used for: The mean of the squared negative Euclidean distance between any two different nodes in the embedding matrix is ​​calculated as the global preference value. Based on the global preference value and the initialized responsibility and availability, the responsibility and availability are iteratively updated alternately until the change in responsibility and availability is less than the preset convergence value or the maximum number of iterations is reached. Cluster centers are determined based on maximizing responsibility and availability. Nodes with the same cluster center are grouped into the same cluster to obtain the initial partitioning scheme.

[0014] Preferably, the final partitioning module includes: A verification unit is used to perform a feasibility partitioning verification on the initial partitioning scheme based on connectivity constraints and source-load balance constraints. The adjustment unit is used to adjust partitions that fail the verification until the verification is satisfied; The output unit is used to output all partitions that pass the verification as the final partition result.

[0015] Preferably, the verification unit includes: The connectivity constraint verification subunit is used to take the undirected graph composed of the node set and edge set in the distribution network topology as the connectivity constraint. It constructs a subgraph by taking all nodes in the current partition, the edges of all nodes in the current partition in the undirected graph, and the power supply nodes. Starting from the power supply nodes, it uses breadth-first search to traverse the connectivity of all nodes in the subgraph. If all nodes in the subgraph are connected, the current partition passes the connectivity constraint verification; otherwise, the current partition fails the connectivity constraint verification and is considered a disconnected partition. The source-load balance constraint verification subunit is used to combine the average load of all nodes in the current partition with the imbalance threshold as the source-load balance constraint. It calculates the imbalance of the current partition based on the historical average output and average load of the distributed power sources of all nodes in the current partition. It compares the imbalance of the current partition with the source-load balance constraint. If the imbalance of the current partition is greater than the source-load balance constraint, the current partition fails the source-load balance constraint verification and belongs to the source-load imbalance partition. Otherwise, the current partition passes the source-load balance constraint verification. The current partition that passes both connectivity and source-load balance constraints is considered a feasible partition.

[0016] Preferably, the adjustment unit includes: The disconnected partition adjustment subunit is used to retrieve all connected component sets for a disconnected partition using a graph search algorithm, and merge the connected components in the connected component set except for the largest connected component into the target connected partition adjacent to the connected component. The source-load imbalance partition adjustment subunit is used to sequentially migrate the boundary nodes in the source-load imbalance partition to the target source-load balance partition adjacent to the node.

[0017] Preferably, the determination of the target connected component includes: All feasible partitions adjacent to the connected component to be merged are considered as connected candidate partitions. Based on the weights of the edges between nodes in the connected component to be merged and nodes in each candidate connected partition, the average connection strength between the connected component to be merged and each candidate connected partition is calculated, and the candidate connected partition with the largest average connection strength is determined as the target connected partition.

[0018] Preferably, the determination of the target source-load balancing partition includes: All feasible partitions adjacent to the boundary node to be migrated are selected as source payload candidate partitions; Based on the imbalance amount of the boundary node to be migrated out by the source load imbalance partition and the imbalance amount of the boundary node to be migrated into by the source load candidate partition, the migration index is calculated. The candidate source-load partition that satisfies the source-load balance constraint and has the smallest migration index is determined as the target source-load balance partition.

[0019] In another aspect, the present invention also provides an electronic device, comprising: at least one processor and a memory; the memory and the processor are connected via a bus; The memory is used to store one or more programs; When the one or more programs are executed by the at least one processor, a power distribution network partitioning and verification method as described above is implemented.

[0020] In another aspect, the present invention also provides a readable storage medium having an executable program stored thereon, which, when executed, implements a power distribution network partitioning and verification method as described above.

[0021] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention provides a distribution network partitioning and verification method, comprising: constructing an embedding matrix based on the historical voltage time series of each node in the distribution network topology through feature decomposition and feature extraction; clustering the embedding matrix using the nearest neighbor propagation algorithm to obtain an initial partitioning scheme; performing feasibility partitioning verification on the initial partitioning scheme based on connectivity constraints and source-load balance constraints; adjusting partitions that fail the verification until the verification is satisfied; and outputting all partitions that pass the verification as the final partitioning result. This invention, by introducing the historical voltage time series of nodes and the feasibility verification of connectivity and source-load balance constraints, characterizes the dynamic correlation of distributed generation access to the distribution network, improves the adaptability of the partitioning result to changes in operating status, and provides reliable technical support for the partitioned operation control, optimized scheduling, and intelligent auxiliary decision-making of the distribution network. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of a power distribution network zoning and verification method according to the present invention; Figure 2 This is a flowchart of a power distribution network zoning and verification method according to the present invention; Figure 3 This is a schematic diagram of an electronic device for a power distribution network zoning and verification method according to the present invention. Detailed Implementation

[0023] This invention proposes a distribution network zoning and verification method, system, equipment, and medium. The specific embodiments of this invention will be further described in detail below with reference to the accompanying drawings.

[0024] Example 1: A method for zoning and verifying a power distribution network, such as Figure 1 As shown, it includes: Step S1: Based on the historical voltage time series of each node in the distribution network topology, construct an embedding matrix through feature decomposition and feature extraction; Step S2: Cluster the embedding matrix using the nearest neighbor propagation algorithm to obtain an initial partitioning scheme; Step S3: Perform a feasibility partitioning verification on the initial partitioning scheme based on connectivity constraints and source-load balance constraints. Adjust the partitions that fail the verification until the verification is satisfied. Output all partitions that pass the verification as the final partitioning result.

[0025] Specifically, step S1: Based on the historical voltage time series of each node in the distribution network topology, an embedding matrix is ​​constructed through feature decomposition and feature extraction, including: The historical voltage time series of each node in the distribution network topology are used to form an observation matrix. The dynamic similarity between any two non-equilibrium nodes in the observation matrix is ​​calculated, and a symmetric similarity matrix is ​​constructed. The symmetric similarity matrix is ​​used as a weighted adjacency matrix, and combined with the degree matrix to construct a Laplace matrix; Dimensionality reduction decomposition is performed based on the Laplacian matrix to obtain the eigenvalues ​​and the corresponding eigenvectors. The cumulative eigenvalue ratio method is used to calculate the proportion of non-zero eigenvalues, and the number of non-zero eigenvalues ​​that meet the preset threshold proportion is counted as the embedding dimension. Extract the feature vectors corresponding to the feature values ​​of the embedding dimension, and form an embedding matrix; The historical voltage time series of each node includes historical data and real-time data.

[0026] In this embodiment, the standard IEEE 33-node example is used as an example, such as Figure 2 As shown, the static topology information of the power distribution network topology, i.e., the IEEE 33-node system, is read.

[0027] Define a set of nodes for:

[0028] edge set for:

[0029] For example , And so on, resulting in a total of 32 edges. Thus, an undirected graph is constructed based on the set of nodes and the set of edges. This serves as the physical constraint basis for subsequent partitioning.

[0030] S101: Data Acquisition and Observation Matrix Construction. Collect historical voltage time series data for each node, including both historical and real-time data. For example, the SCADA system collects voltage amplitude data for each observable node over 7 days, with a sampling interval of 15 minutes; therefore, the total number of time segments is... T for T =672 groups. Record nodes. i, j At any moment t ( t =1,2,3,... T The voltage amplitudes are respectively , The voltage data from all nodes form the observation matrix. :

[0031] S102: To characterize the similarity between nodes, a node voltage fluctuation similarity matrix is ​​constructed. For any two non-equilibrium nodes... Calculate the Pearson correlation coefficient of its voltage time series. As a dynamic similarity:

[0032] in, Represents a node i average voltage, ; Represents a node j The average voltage.

[0033] The closer the value is to 1, the more synchronized the voltage fluctuations of the two nodes. Let the diagonal elements... This ultimately forms a 32×32 symmetric similarity matrix. .

[0034] S103: Construct the graph Laplacian matrix to further describe the relationships between all nodes. Treat the 32 unbalanced nodes as vertices of the graph, and... As edge weights, a weighted adjacency matrix is ​​defined. :

[0035]

[0036] Calculate the degree matrix , is a diagonal matrix, and its diagonal elements are: , i =2,3,...,33.

[0037] This leads to the construction of the unnormalized Graph Laplace matrix. : .

[0038] S104: Perform eigenvalue decomposition on the graph Laplacian matrix and calculate the embedding dimension. This is a real symmetric positive semi-definite matrix, i.e., a graph Laplacian matrix. Perform eigenvalue decomposition: , k =1,2,...,32; in, Let eigenvalues ​​be eigenvalues, and let eigenvalues ​​satisfy eigenvalues ​​? ; This represents the eigenvector.

[0039] Embedding dimension is determined using the cumulative eigenvalue ratio method. Before calculation The percentage of non-zero eigenvalues :

[0040] choose The smallest , making ,in Indicates the preset threshold. For example, taking values This ensures that the selected dimension retains at least 85% of the effective structural information in the graph Laplace.

[0041] S105: Based on Embedding Dimension Extract the corresponding feature values The eigenvectors form the embedding matrix. :

[0042] Specifically, step S2: using the nearest neighbor propagation algorithm to cluster the embedding matrix to obtain an initial partitioning scheme includes: The mean of the squared negative Euclidean distance between any two different nodes in the embedding matrix is ​​calculated as the global preference value. Based on the global preference value and the initialized responsibility and availability, the responsibility and availability are iteratively updated alternately until the change in responsibility and availability is less than the preset convergence value or the maximum number of iterations is reached. Cluster centers are determined based on maximizing responsibility and availability. Nodes with the same cluster center are grouped into the same cluster to obtain the initial partitioning scheme.

[0043] In this embodiment, the Affinity Propagation (AP) algorithm is used for clustering. The embedding matrix is... As input, the AP clustering algorithm is executed.

[0044] S201: Construct the input preference matrix. Calculate the embedding matrix. Any two nodes The squared negative Euclidean distance in spectral space is used as the preference degree. enter:

[0045] in, Represents the embedding matrix of the first... i The feature vector of each node; Represents the embedding matrix of the first... j The feature vectors of each node. Take the embedding matrix. All The mean of the diagonal elements. The global preference value.

[0046] S202: Iterative Message Passing. The AP algorithm updates responsibility alternately. and availability ,in Represents a node i The node is considered h The degree to which it is suitable as its cluster center. Represents a node h Willing to accept nodes i The degree to which one considers oneself the center of a cluster. Level of responsibility in the initial stage. and availability All values ​​are initialized to zero, and the algorithm enters an iterative update phase. In each iteration, the algorithm first checks the current availability. Calculate the new level of responsibility Its expression is:

[0047] in, Represents a node i With nodes h The degree of preference between them; Indicates traversal of the division h All other candidate centers besides the node. This formula measures the node's performance after excluding other more attractive candidates. i For nodes h The degree of recognition as its cluster center. The new availability is calculated using the updated responsibility level, with rules divided into two cases, when… hour:

[0048] Traversal i and h All nodes except those This indicates that only the sum of pairs is accumulated. hA high degree of responsibility with a positive evaluation; Represents a node h The intention of itself as a cluster center is taken to the minimum value of 0 to prevent excessive accumulation of availability; when hour:

[0049] node h Its availability is equal to the sum of the positive responsibility of all other nodes towards it as the center, reflecting the degree of its widespread support. Subsequently, the responsibility obtained in this round of updates... and availability This information is used for the next iteration, and so on, updating alternately to gradually correct each node's judgment of the cluster center. The iteration terminates when the change in responsibility and availability is less than the preset convergence threshold or when the maximum number of iterations is reached.

[0050] S203: Determine and assign cluster centers. After the above iterations converge or terminate, for each node, determine its corresponding cluster center. By maximizing responsibility and availability The sum is used to determine this, and the specific formula is as follows:

[0051] in, An index representing the candidate cluster centers; This indicates that the nodes were measured. i With candidate centers q The overall matching strength between them; Indicates all possible candidate centers q Choose the one that maximizes the matching strength. q This is the final cluster center of the node. All nodes with the same cluster center... Nodes with values ​​are grouped into the same cluster, thus partitioning the data. The output is a partition label vector. Clustering labels , This indicates the number of partitions automatically determined by the AP algorithm.

[0052] This invention introduces historical voltage time series of nodes, constructs a node voltage fluctuation similarity matrix to characterize the dynamic correlation between nodes, and combines graph Laplacian matrix and spectral embedding techniques to uniformly map the physical structure information and operating status information of the distribution network into a low-dimensional feature space. Based on this, an affinity propagation clustering algorithm is used to automatically identify partition centers, avoiding the dependence of traditional clustering methods on the number of clusters and initial conditions, thus obtaining more objective and stable initial partitioning results.

[0053] Specifically, in step S3: a feasibility partitioning verification is performed on the initial partitioning scheme based on connectivity constraints and source-load balance constraints, including: The undirected graph formed by the node set and edge set in the distribution network topology is used as a connectivity constraint. All nodes in the current partition, the edges of all nodes in the current partition in the undirected graph, and the power nodes are used to form a subgraph. Starting from the power nodes, the connectivity of all nodes in the subgraph is traversed using breadth-first search. If all nodes in the subgraph are connected, the current partition passes the connectivity constraint check; otherwise, the current partition fails the connectivity constraint check and is considered a disconnected partition. The average load of all nodes in the current partition is combined with the imbalance threshold as the source-load balance constraint. The imbalance of the current partition is calculated based on the historical average output and average load of the distributed power sources of all nodes in the current partition. Compare the imbalance of the current partition with the source load balance constraint. If the imbalance of the current partition is greater than the source load balance constraint, the current partition fails the source load balance constraint check and is a source load imbalance partition. Otherwise, the current partition passes the source load balance constraint check. The current partition that passes both connectivity and source-load balance constraints is considered a feasible partition.

[0054] In this embodiment, the feasibility of the partitioning results is verified. If all partitions... n ( n = If the connectivity and source-load balance constraints are not met, a lightweight partitioning adjustment strategy is executed; if they are met, the final partitioning result is output.

[0055] S301: After completing the clustering and partitioning, the feasibility of the partitioning results must be verified to ensure that each partition meets the actual operational constraints at the physical or system level.

[0056] Define each partition n Node set for: ,in, Indicates the first i The cluster label of the node belongs to the first node. n One partition; Indicates the first n The set of partition nodes contains all nodes that have been partitioned to the nth partition. n The node index of each partition.

[0057] The following two constraints will be checked in sequence: (1) Connectivity constraints: With power nodes and their properties in an undirected graph The edges in the graph form a subgraph Starting from the power node, a breadth-first search is used to traverse the network. The connectivity of all nodes in the array, if If not connected, mark the partition. n This is not feasible.

[0058] (2) Source load balance constraint: Assume nodes i The historical average output of distributed power sources is The average load is The imbalance of the partition for:

[0059] The summation in the formula covers all nodes within partition n. This reflects the net power difference between the power source and the load within the partition. If the following conditions are met:

[0060] in, This represents the imbalance threshold. If so, the source load of that partition is determined to be unbalanced.

[0061] Specifically, in step S3: adjusting the partitions that failed the verification includes: For disconnected partitions, a graph search algorithm is used to retrieve all connected component sets, and all connected components in the connected component sets except for the largest connected component are merged into the target connected partition adjacent to the connected component. The boundary nodes in the source-load imbalance partition are migrated sequentially to the target source-load balance partition adjacent to the node.

[0062] The determination of the target connected partition includes: All feasible partitions adjacent to the connected component to be merged are considered as connected candidate partitions. Based on the weights of the edges between nodes in the connected component to be merged and nodes in each candidate connected partition, the average connection strength between the connected component to be merged and each candidate connected partition is calculated, and the candidate connected partition with the largest average connection strength is determined as the target connected partition.

[0063] The determination of the target source load balancing partition includes: All feasible partitions adjacent to the boundary node to be migrated are selected as source payload candidate partitions; Based on the imbalance amount of the boundary node to be migrated out by the source load imbalance partition and the imbalance amount of the boundary node to be migrated into by the source load candidate partition, the migration index is calculated. The candidate source-load partition that satisfies the source-load balance constraint and has the smallest migration index is determined as the target source-load balance partition.

[0064] In this embodiment, partitions that do not meet the constraints are corrected for both the "connectivity constraint" and the "source-load balance constraint". The partition set must be updated after each adjustment. And perform a re-verification: S302: Handling disconnected partitions: For partitions that do not satisfy connectivity constraints In an undirected graph Extract its subgraph And obtain its set of connected components using a graph search algorithm (such as breadth-first search): ,in The number of connected components.

[0065] Define the connected component with the most nodes as the principal component. : ,in k From 1 to Q The index to be traversed. Except for the principal component. Other connected components Let these be the connected components to be merged: For each Perform the following merge operation: S302-1: Candidate Target Partition Set: Definition and A set of adjacent feasible partitions with physical connections ;

[0066] S302-2: Candidate Partition Selection Criteria: For each candidate partition... Calculate the average connection strength ;

[0067] in, Indicates the connected components to be merged nodes i Adjacent Feasibility Zones nodes j The weight of the edges between them; S302-3: Select target partition: Select the partition with the highest average connection strength. As the target partition;

[0068] S302-4: Partition merging, performing partition node set update: ,gather Update to itself and Union, set Update to subtract from itself The remaining part.

[0069] S303: Source-load imbalance zone handling: For partitions that satisfy connectivity but not source-load balance constraints Define the set of boundary nodes for the source load imbalance partition. for: ,in, Includes all at least with Nodes connected to other nodes within the partition (i.e., nodes located inside the partition but potentially interacting with the outside) are only allowed to interact with other nodes from outside the partition. Select a node to migrate.

[0070] S303-1: Construction of the candidate migration set. For each boundary node... Define its set of adjacent feasible partitions for migration. For the candidate migration set:

[0071] S303-2: Node migration effect evaluation. For arbitrary boundary node migration operations... Calculate the partition imbalance after the boundary node migration: Imbalance quantity of boundary nodes migrating out of source load imbalance partitions for:

[0072] in, Represents a node The historical average output of distributed power sources; Represents a node The average load; Imbalance in the migration of boundary nodes into candidate target partitions for:

[0073] in, Represents a node j The historical average output of distributed power sources; Represents a node j Average load; Define the optimal migration optimization index for:

[0074] in, This represents the weighting coefficient.

[0075] S303-3: Optimal migration decision. The optimal migration operation is:

[0076] The migration is performed only if the following connectivity constraints are met: After the source load imbalance partition migrates out of the boundary node In undirected graphs Maintain connectivity; After the candidate target partition is migrated into the boundary node In undirected graphs Maintain connectivity; If the conditions are not met, the candidate operation is eliminated, and the suboptimal solution is selected.

[0077] S303-4: Node migration, performing partition node set update: ,gather Update to the boundary node between itself and the migrated node. Union, set Update to itself minus the migrated boundary nodes. The remaining part.

[0078] The adjustment will stop when any of the following conditions are met: All partitions satisfy connectivity constraints and source-load imbalance constraints; The absence of Feasible migration operations for descent.

[0079] Otherwise, continue with connectivity constraint and source-load imbalance constraint verification.

[0080] The final output includes all feasible partition results, including the partition set and the set of interconnecting branches (interconnecting switches).

[0081] This invention proposes a distribution network partitioning and verification method. On one hand, by introducing historical voltage time series of nodes and constructing a voltage fluctuation similarity matrix, it effectively characterizes the dynamic correlation between distribution network nodes under distributed generation access conditions, improving the adaptability of the partitioning results to changes in operating status. It does not require precise static data of the distribution system; it can partition the distribution system under complex operating environments based on existing system measurement data. On the other hand, by combining topological constraints, spectral embedding, and nearest neighbor propagation clustering, and through connectivity and source-load balance verification and adjustment mechanisms, it avoids the problems of disconnected partitions or operational infeasibility in traditional clustering methods. This invention has low parameter dependence and can provide reliable technical support for hierarchical partitioning operation control, optimized scheduling, and intelligent auxiliary decision-making in distribution networks.

[0082] Example 2: Based on the same inventive concept, the present invention also provides a power distribution network zoning and verification system, including: The matrix construction module is used to construct an embedding matrix based on the historical voltage time series of each node in the distribution network topology through feature decomposition and feature extraction. The initial partitioning module is used to cluster the embedding matrix using the nearest neighbor propagation algorithm to obtain an initial partitioning scheme. The final partitioning module is used to perform feasibility partitioning verification on the initial partitioning scheme based on connectivity constraints and source-load balance constraints. Partitions that fail the verification are adjusted until the verification is satisfied, and all partitions that pass the verification are output as the final partitioning result.

[0083] The matrix construction module includes: The similarity matrix construction unit is used to form an observation matrix from the historical voltage time series of each node in the distribution network topology, calculate the dynamic similarity between any two non-equilibrium nodes in the observation matrix, and construct a symmetric similarity matrix. The Laplacian matrix construction unit is used to construct a Laplacian matrix by combining the symmetric similarity matrix as a weighted adjacency matrix with the degree matrix. An embedding matrix construction unit is used to perform dimensionality reduction decomposition based on the Laplacian matrix to obtain eigenvalues ​​and corresponding eigenvectors; calculate the numerical proportion of non-zero eigenvalues ​​using the cumulative eigenvalue ratio method; count the number of non-zero eigenvalues ​​whose numerical proportions meet a preset threshold as the embedding dimension; and extract the eigenvectors corresponding to the eigenvalues ​​of the embedding dimension to form an embedding matrix. The historical voltage time series of each node includes historical data and real-time data.

[0084] The initial partitioning module is specifically used for: The mean of the squared negative Euclidean distance between any two different nodes in the embedding matrix is ​​calculated as the global preference value. Based on the global preference value and the initialized responsibility and availability, the responsibility and availability are iteratively updated alternately until the change in responsibility and availability is less than the preset convergence value or the maximum number of iterations is reached. Cluster centers are determined based on maximizing responsibility and availability. Nodes with the same cluster center are grouped into the same cluster to obtain the initial partitioning scheme.

[0085] The final partitioning module includes: A verification unit is used to perform a feasibility partitioning verification on the initial partitioning scheme based on connectivity constraints and source-load balance constraints. The adjustment unit is used to adjust partitions that fail the verification until the verification is satisfied; The output unit is used to output all partitions that pass the verification as the final partition result.

[0086] The verification unit includes: The connectivity constraint verification subunit is used to take the undirected graph composed of the node set and edge set in the distribution network topology as the connectivity constraint. It constructs a subgraph by taking all nodes in the current partition, the edges of all nodes in the current partition in the undirected graph, and the power supply nodes. Starting from the power supply nodes, it uses breadth-first search to traverse the connectivity of all nodes in the subgraph. If all nodes in the subgraph are connected, the current partition passes the connectivity constraint verification; otherwise, the current partition fails the connectivity constraint verification and is considered a disconnected partition. The source-load balance constraint verification subunit is used to combine the average load of all nodes in the current partition with the imbalance threshold as the source-load balance constraint. It calculates the imbalance of the current partition based on the historical average output and average load of the distributed power sources of all nodes in the current partition. It compares the imbalance of the current partition with the source-load balance constraint. If the imbalance of the current partition is greater than the source-load balance constraint, the current partition fails the source-load balance constraint verification and belongs to the source-load imbalance partition. Otherwise, the current partition passes the source-load balance constraint verification. The current partition that passes both connectivity and source-load balance constraints is considered a feasible partition.

[0087] The adjustment unit includes: The disconnected partition adjustment subunit is used to retrieve all connected component sets for a disconnected partition using a graph search algorithm, and merge the connected components in the connected component set except for the largest connected component into the target connected partition adjacent to the connected component. The source-load imbalance partition adjustment subunit is used to sequentially migrate the boundary nodes in the source-load imbalance partition to the target source-load balance partition adjacent to the node.

[0088] The determination of the target connected partition includes: All feasible partitions adjacent to the connected component to be merged are considered as connected candidate partitions. Based on the weights of the edges between nodes in the connected component to be merged and nodes in each candidate connected partition, the average connection strength between the connected component to be merged and each candidate connected partition is calculated, and the candidate connected partition with the largest average connection strength is determined as the target connected partition.

[0089] The determination of the target source load balancing partition includes: All feasible partitions adjacent to the boundary node to be migrated are selected as source payload candidate partitions; Based on the imbalance amount of the boundary node to be migrated out by the source load imbalance partition and the imbalance amount of the boundary node to be migrated into by the source load candidate partition, the migration index is calculated. The candidate source-load partition that satisfies the source-load balance constraint and has the smallest migration index is determined as the target source-load balance partition.

[0090] This invention proposes a distribution network zoning and verification system. On one hand, by introducing historical voltage time series of nodes and constructing a voltage fluctuation similarity matrix, it effectively characterizes the dynamic correlation between distribution network nodes under distributed generation access conditions, improving the adaptability of the zoning results to changes in operating status. It does not require precise static data of the distribution system; it can zonify the distribution system under complex operating environments based on existing system measurement data. On the other hand, by combining topological constraints, spectral embedding, and nearest neighbor propagation clustering, and through connectivity and source-load balance verification and adjustment mechanisms, it avoids the problems of disconnected zoning or operational infeasibility in traditional clustering methods. This invention has low parameter dependence and can provide reliable technical support for hierarchical zoning operation control, optimized scheduling, and intelligent auxiliary decision-making in distribution networks.

[0091] Example 3: like Figure 3 As shown, the present invention also provides an electronic device, which may be a computer device, a microcontroller device, a smart mobile device, etc. The electronic device in this embodiment may include a processor, a memory, a transceiver component, etc. The memory, processor, and transceiver component are connected via a bus; the memory can be used to store executable programs, and an exemplary executable program may include instructions; the processor is used to execute the instructions stored in the memory. The memory can also be used to store data, which can be accessed and / or modified when instructions are executed.

[0092] The processor may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, and it is suitable for implementing one or more instructions. Specifically, it is suitable for loading and executing one or more instructions in the storage medium to implement the corresponding method flow or corresponding function, so as to realize the steps of the power distribution network partitioning and verification method in the above embodiments.

[0093] Example 4: Based on the same inventive concept, this invention also provides a readable storage medium, specifically an electronic device readable storage medium (Memory). This readable storage medium is a memory device within an electronic device used to store programs and data. It is understood that the storage medium here can include both built-in storage media within the electronic device and extended storage media supported by the electronic device. The storage medium provides storage space, which stores the terminal's operating system. Furthermore, this storage space also stores one or more instructions suitable for loading and execution by a processor. These instructions can be one or more executable programs (including program code). It should be noted that the storage medium here can be high-speed RAM or non-volatile memory, such as at least one disk storage device. Loading and executing one or more instructions stored in the storage medium by the processor can implement the steps of a power distribution network partitioning and verification method in the above embodiments.

[0094] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0095] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0096] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0097] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0098] The above are merely embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention are included within the scope of the claims of the present invention pending approval.

Claims

1. A method for dividing and verifying a power distribution network, characterized in that, include: Based on the historical voltage time series of each node in the distribution network topology, an embedding matrix is ​​constructed through feature decomposition and feature extraction. The embedding matrix is ​​clustered using the nearest neighbor propagation algorithm to obtain an initial partitioning scheme; The initial partitioning scheme is validated for feasibility based on connectivity constraints and source-load balance constraints. Partitions that fail the validation are adjusted until the validation is satisfied. All partitions that pass the validation are output as the final partitioning result.

2. The method according to claim 1, characterized in that, The historical voltage time series of each node based on the distribution network topology is used to construct an embedding matrix through feature decomposition and feature extraction, including: The historical voltage time series of each node in the distribution network topology are used to form an observation matrix. The dynamic similarity between any two non-equilibrium nodes in the observation matrix is ​​calculated, and a symmetric similarity matrix is ​​constructed. The symmetric similarity matrix is ​​used as a weighted adjacency matrix, and combined with the degree matrix to construct a Laplace matrix; Dimensionality reduction decomposition is performed based on the Laplacian matrix to obtain the eigenvalues ​​and the corresponding eigenvectors. The cumulative eigenvalue ratio method is used to calculate the proportion of non-zero eigenvalues, and the number of non-zero eigenvalues ​​that meet the preset threshold proportion is counted as the embedding dimension. Extract the feature vectors corresponding to the feature values ​​of the embedding dimension, and form an embedding matrix; The historical voltage time series of each node includes historical data and real-time data.

3. The method according to claim 2, characterized in that, The step of clustering the embedding matrix using the nearest neighbor propagation algorithm to obtain the initial partitioning scheme includes: The mean of the squared negative Euclidean distance between any two different nodes in the embedding matrix is ​​calculated as the global preference value. Based on the global preference value and the initialized responsibility and availability, the responsibility and availability are iteratively updated alternately until the change in responsibility and availability is less than the preset convergence value or the maximum number of iterations is reached. Cluster centers are determined based on maximizing responsibility and availability. Nodes with the same cluster center are grouped into the same cluster to obtain the initial partitioning scheme.

4. The method according to claim 1, characterized in that, The feasibility partitioning verification of the initial partitioning scheme based on connectivity constraints and source-load balance constraints includes: The undirected graph formed by the node set and edge set in the distribution network topology is used as a connectivity constraint. All nodes in the current partition, the edges of all nodes in the current partition in the undirected graph, and the power nodes are used to form a subgraph. Starting from the power nodes, the connectivity of all nodes in the subgraph is traversed using breadth-first search. If all nodes in the subgraph are connected, the current partition passes the connectivity constraint check; otherwise, the current partition fails the connectivity constraint check and is considered a disconnected partition. The average load of all nodes in the current partition is combined with the imbalance threshold as the source-load balance constraint. The imbalance of the current partition is calculated based on the historical average output and average load of the distributed power sources of all nodes in the current partition. Compare the imbalance of the current partition with the source load balance constraint. If the imbalance of the current partition is greater than the source load balance constraint, the current partition fails the source load balance constraint check and is a source load imbalance partition. Otherwise, the current partition passes the source load balance constraint check. The current partition that passes both connectivity and source-load balance constraints is considered a feasible partition.

5. The method according to claim 4, characterized in that, The adjustment of partitions that failed the verification includes: For disconnected partitions, a graph search algorithm is used to retrieve all connected component sets, and all connected components in the connected component sets except for the largest connected component are merged into the target connected partition adjacent to the connected component. The boundary nodes in the source-load imbalance partition are migrated sequentially to the target source-load balance partition adjacent to the node.

6. The method according to claim 5, characterized in that, The determination of the target connected partition includes: All feasible partitions adjacent to the connected component to be merged are considered as connected candidate partitions. Based on the weights of the edges between nodes in the connected component to be merged and nodes in each candidate connected partition, the average connection strength between the connected component to be merged and each candidate connected partition is calculated, and the candidate connected partition with the largest average connection strength is determined as the target connected partition.

7. The method according to claim 5, characterized in that, The determination of the target source load balancing partition includes: All feasible partitions adjacent to the boundary node to be migrated are selected as source payload candidate partitions; Based on the imbalance amount of the boundary node to be migrated out by the source load imbalance partition and the imbalance amount of the boundary node to be migrated into by the source load candidate partition, the migration index is calculated. The candidate source-load partition that satisfies the source-load balance constraint and has the smallest migration index is determined as the target source-load balance partition.

8. A power distribution network zoning and verification system, characterized in that, include: The matrix construction module is used to construct an embedding matrix based on the historical voltage time series of each node in the distribution network topology through feature decomposition and feature extraction. The initial partitioning module is used to cluster the embedding matrix using the nearest neighbor propagation algorithm to obtain an initial partitioning scheme. The final partitioning module is used to perform feasibility partitioning verification on the initial partitioning scheme based on connectivity constraints and source-load balance constraints. Partitions that fail the verification are adjusted until the verification is satisfied, and all partitions that pass the verification are output as the final partitioning result.

9. An electronic device, characterized in that, include: At least one processor and memory; The memory and processor are connected via a bus; The memory is used to store one or more programs; When the one or more programs are executed by the at least one processor, a power distribution network partitioning and verification method as described in any one of claims 1 to 8 is implemented.

10. A readable storage medium, characterized in that, It contains an execution program, which, when executed, implements a power distribution network partitioning and verification method as described in any one of claims 1 to 8.

Images