Layering, clustering and partitioning method and system for reactive voltage control

Through the hierarchical clustering partitioning method for reactive voltage control, the agglomeration hierarchical clustering algorithm and the Calinski-Harabasz index are used to solve the problem of insufficient scientificity and objectivity of the traditional power system partitioning method, and efficient voltage partitioning and system are achieved Control effects.

CN118970971APending Publication Date: 2024-11-15NANJING UNIV OF SCI & TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410900533.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-05
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

The traditional power system partitioning method is affected by administrative division, lacks scientificity and objectivity, and cannot fully consider the electrical characteristics of the system and the needs of layered control, resulting in defects in system analysis and operation control.

Method used

Using a hierarchical clustering partitioning method for reactive voltage control, by obtaining grid node parameters, calculating the admission matrix and Jacobian matrix, generating control space, using the condensed hierarchical clustering algorithm and Ward distance for clustering, combined with Calinski-Harabasz The index determines the optimal number of partitions to achieve high-efficiency voltage partitioning.

Benefits of technology

It realizes efficient voltage partitioning, avoids the difficulty of manually specifying the number of partitions in traditional methods, improves the scientificity and objectivity of partitions, and enhances the system's analysis and operation control effect.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118970971A_ABST
    Figure CN118970971A_ABST
Patent Text Reader

Abstract

The invention discloses a hierarchical clustering zoning method and system oriented to reactive voltage control, and relates to the technical field of power distribution network control application, and the method comprises the steps: obtaining power grid node parameters, measuring and calculating an admittance matrix of a whole network through the power grid node parameters, and calculating the admittance matrix of the whole network through a preset node power equation; carrying out load flow calculation by utilizing the admittance matrix of the whole network to obtain a Jacobian matrix; according to the sensitivity of considering the quasi-steady-state physical response of the PV node in the Jacobian matrix, generating a control space of each reactive power source node as a coordinate axis, and obtaining an electrical distance according to the control space and the control capability of the reactive power source to the load node; the method comprises the following steps: selecting a Ward distance according to an electrical distance, carrying out inter-class distance calculation by using an agglomerated hierarchical clustering algorithm according to the electrical distance to obtain a clustering result, inputting the clustering result into a pre-established partition model, and outputting to obtain a partition result, the partition model being constructed based on a Clinski-Harabasz index.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of distribution network control applications, and in particular to a hierarchical clustering partitioning method and system for reactive power voltage control. Background Art

[0002] As the scale of the power grid continues to expand, how to accurately partition the power system has become a key challenge for the reactive power / voltage optimization control of the power system. In traditional power system dispatching and control, the power system dispatching organization is usually set up according to the division of regions and administrative agencies, resulting in reactive power and voltage zoning control often being affected by administrative divisions. This approach has certain limitations, cannot fully consider the electrical characteristics of the system and the needs of hierarchical regulation, and lacks objectivity and scientificity. Compared with the ideal zoning scheme, the traditional method has a certain gap, which may lead to defects in system analysis and operation control, and needs further optimization.

[0003] In order to solve the problem of power system partitioning, a clustering-based reactive voltage control partitioning method has been proposed, in which the clustering method includes hierarchical clustering and non-hierarchical clustering. Considering that the actual goal of power grid partitioning is to ensure strong coupling between nodes within sub-regions, and low coupling between nodes in different sub-regions, and an appropriate number of sub-regions is required. Due to the relatively small number of nodes in the power grid, the K-means method is greatly affected by subjective factors and is more suitable for big data processing. Summary of the invention

[0004] In order to solve the deficiencies mentioned in the above background technology, the purpose of the present invention is to provide a hierarchical clustering partitioning method and system for reactive power voltage control, which can realize efficient voltage partitioning.

[0005] In a first aspect, the purpose of the present invention can be achieved by the following technical solution: a hierarchical clustering partitioning method for reactive power voltage control, the method comprising the following steps:

[0006] Obtain grid node parameters, use the grid node parameters to calculate the admittance matrix of the entire grid, and use the preset node power equation to calculate the power flow using the admittance matrix of the entire grid to obtain the Jacobian matrix;

[0007] According to the sensitivity of the quasi-steady-state physical response of the PV node in the Jacobian matrix, a control space with each reactive source node as a coordinate axis is generated, and the electrical distance is obtained according to the control space and the control ability of the reactive source on the load node;

[0008] An agglomerative hierarchical clustering algorithm is used to select Ward distance for inter-class distance calculation according to electrical distance to obtain clustering results, which are input into a pre-established partition model to output a partition result, wherein the partition model is constructed based on the Calinski-Harabasz index.

[0009] In combination with the first aspect, in some implementations of the first aspect, the method further includes: a calculation process of measuring the admittance matrix of the entire network using the grid node parameters:

[0010]

[0011] Where, self-admittance Y ii Indicates that a unit voltage U is added to node i i When all other nodes are grounded, the current injected into the network through point i is numerically equal to the sum of the admittances of all branches connected to the node. The mutual admittance Y ij It means that when a unit voltage is injected into node i and other nodes are grounded, the current injected into the network through node j is equal to the negative value of the admittance of the branch element connected between nodes i and j. When there is no branch directly connected between nodes i and j, Y ij =Y ji .

[0012] In combination with the first aspect, in some implementations of the first aspect, the method further includes: a process of generating a control space with each reactive source node as a coordinate axis according to the sensitivity of the quasi-steady-state physical response of the PV node considered in the Jacobian matrix:

[0013] Set the node type, set the j-th generator node in set G to a PQ node, and for other generator nodes in set G, if they participate in PVC, they are set to PV nodes, otherwise they are set to PQ nodes;

[0014] When calculating the sensitivity, the diagonal elements corresponding to the PV nodes in the full-dimensional B” matrix containing the PV nodes are increased to ensure that the node voltage remains unchanged, thereby retaining the quasi-steady-state control response characteristics of the PVC device;

[0015] Based on the constructed B”, the control sensitivity of the injected reactive power of node j to the voltage of each node in the set L is solved. For load node i, where i∈L, the sensitivity is the quasi-steady-state sensitivity, expressed as S ij ;

[0016]

[0017] Where S ij It is expressed as the voltage control capability of the jth generator node’s reactive injection on the load node i, for each generator node belonging to the set G;

[0018] The control capability of each generator node is regarded as one dimension in the coordinate space. Thus, a g-dimensional coordinate space is constructed from the set G as the reactive source control space, using a g-dimensional vector (x i1 ,x i2 ,…,x ig ) is used as the coordinate to describe the load node i, where the jth coordinate component is defined as:

[0019] x ij = -lg|S ij ∣

[0020] Among them, x ij It uses coordinates to describe the degree of coupling between reactive power and voltage.

[0021] In combination with the first aspect, in some implementations of the first aspect, the method further includes: a process of obtaining the electrical distance according to the control space and the control capability of the reactive source over the load node:

[0022] In the g-dimensional space, the coordinate vector (x i1 ,x i2 ,…,x ig ) is the control sensitivity vector of all generator nodes to it. The jth component of the coordinate reflects the sensitivity of the reactive power injection of the jth generator node to control the voltage of the node. The two nodes to be partitioned m and n correspond to the coordinate vector (x m1 ,x m2 ,…,x mg ) and (x n1 ,x n2 ,…,x ng ), the distance between two points is directly defined using the Euclidean distance:

[0023]

[0024] In combination with the first aspect, in some implementations of the first aspect, the method further includes: using the agglomerative hierarchical clustering algorithm to select the Ward distance to calculate the inter-class distance according to the electrical distance to obtain the clustering result:

[0025] The nodes with large coupling degree are divided into one category, and the nodes with small coupling degree are separated. Cluster analysis is performed to obtain the cluster analysis diagram. Each node is first regarded as a cluster. The average distance method is used to calculate the inter-cluster distance. The inter-cluster distance selects the Ward distance for inter-class distance calculation. The Ward distance is the difference between the sum of squares of the deviations between two clusters Δ(C i ,C j ), the calculation formula is as follows:

[0026] Sum of squared deviations:

[0027] Δ(C i ,C j )=ESS(C i ∪C j )-ESS(C i )-ESS(C j )

[0028] After calculating the Ward distance between each cluster, merge the two clusters with the closest distance; repeat the above two steps until all clusters are merged to obtain the clustering result.

[0029] In combination with the first aspect, in some implementations of the first aspect, the method further includes: a process of establishing the pre-established partition model:

[0030] Cluster the data set X with a capacity of N into K classes, measure the closeness within the class by calculating the sum of the squares of the distances between each point in the class and the class center, as the intra-class distance, and measure the separation of the data set by the sum of the squares of the distances between each class center point and the center point of the data set, as the inter-class distance;

[0031] The calculation formula of CH index is:

[0032]

[0033] Among them B k is the covariance matrix between classes, W k The covariance matrix of the intra-class data;

[0034] B k The calculation formula is as follows:

[0035]

[0036] where c q represents the center point of class q, c e Represents the center point of the data set, n q represents the number of data in class q, C q Representing a data set of class q, the larger the score of the Calinski-Harabasz index, the better it is, among which the smaller the covariance within the category, the better, and the larger the covariance between categories, the better.

[0037] In combination with the first aspect, in some implementations of the first aspect, the method further includes: the partition effect of the pre-established partition model is calculated by using the silhouette coefficient s(x i ) for evaluation:

[0038] The silhouette coefficient of the data point x is defined as:

[0039]

[0040] a(x i ), b(x i ) is calculated as follows: a(x i ) is the data point x i The average distance to other points in the same cluster, b(x i ) is the data point x i Average distance to the nearest cluster:

[0041]

[0042] Among them, the higher the silhouette coefficient value, the better the point matches its own cluster and the worse it matches the neighboring clusters, which means the better the cluster partitioning effect.

[0043] In a second aspect, in order to achieve the above-mentioned object, the present invention discloses a hierarchical clustering partitioning system for reactive power and voltage control, comprising:

[0044] The matrix calculation module is used to obtain the grid node parameters, use the grid node parameters to calculate the admittance matrix of the whole grid, and use the preset node power equation to calculate the power flow using the admittance matrix of the whole grid to obtain the Jacobian matrix;

[0045] The distance calculation module is used to generate a control space with each reactive source node as a coordinate axis according to the sensitivity of the quasi-steady-state physical response of the PV node in the Jacobian matrix, and obtain the electrical distance according to the control space and the control ability of the reactive source on the load node;

[0046] The clustering partitioning module is used to use the agglomerative hierarchical clustering algorithm to select the Ward distance to calculate the inter-class distance according to the electrical distance to obtain the clustering result, input the clustering result into a pre-established partitioning model, and output the partitioning result, wherein the partitioning model is constructed based on the Calinski-Harabasz index.

[0047] In conjunction with the second aspect, in some implementations of the second aspect, the system further includes: a calculation process of calculating the admittance matrix of the entire network using the grid node parameters in the matrix calculation module:

[0048]

[0049] Where, self-admittance Y ii Indicates that a unit voltage U is added to node i i When all other nodes are grounded, the current injected into the network through point i is numerically equal to the sum of the admittances of all branches connected to the node. The mutual admittance Y ijIt means that when a unit voltage is injected into node i and other nodes are grounded, the current injected into the network through node j is equal to the negative value of the admittance of the branch element connected between nodes i and j. When there is no branch directly connected between nodes i and j, Y ij =Y ji ;

[0050] The distance calculation module generates the control space process with each reactive source node as the coordinate axis according to the sensitivity of the quasi-steady-state physical response of the PV node in the Jacobian matrix:

[0051] Set the node type, set the j-th generator node in set G to a PQ node, and for other generator nodes in set G, if they participate in PVC, they are set to PV nodes, otherwise they are set to PQ nodes;

[0052] When calculating the sensitivity, the diagonal elements corresponding to the PV nodes in the full-dimensional B” matrix containing the PV nodes are increased to ensure that the node voltage remains unchanged, thereby retaining the quasi-steady-state control response characteristics of the PVC device;

[0053] Based on the constructed B”, the control sensitivity of the injected reactive power of node j to the voltage of each node in the set L is solved. For load node i, where i∈L, the sensitivity is the quasi-steady-state sensitivity, expressed as S ij ;

[0054]

[0055] Where S ij It is expressed as the voltage control capability of the jth generator node’s reactive injection on the load node i, for each generator node belonging to the set G;

[0056] The control capability of each generator node is regarded as one dimension in the coordinate space. Thus, a g-dimensional coordinate space is constructed from the set G as the reactive source control space, using a g-dimensional vector (x i1 ,x i2 ,…,x ig ) is used as the coordinate to describe the load node i, where the jth coordinate component is defined as:

[0057] x ij = -lg|S ij ∣

[0058] Among them, x ij It uses coordinates to describe the tightness of coupling between reactive power and voltage;

[0059] The process of obtaining the electrical distance in the distance calculation module based on the control space and the control capability of the reactive source on the load node:

[0060] In the g-dimensional space, the coordinate vector (x i1 ,x i2 ,…,x ig ) is the control sensitivity vector of all generator nodes to it. The jth component of the coordinate reflects the sensitivity of the reactive power injection of the jth generator node to control the voltage of the node. The two nodes to be partitioned m and n correspond to the coordinate vector (x m1 ,x m2 ,…,x mg ) and (x n1 ,x n2 ,…,x ng ), the distance between two points is directly defined using the Euclidean distance:

[0061]

[0062] The clustering partitioning module uses the agglomerative hierarchical clustering algorithm to select the Ward distance to calculate the distance between classes based on the electrical distance to obtain the clustering result:

[0063] The nodes with large coupling degree are divided into one category, and the nodes with small coupling degree are separated. Cluster analysis is performed to obtain the cluster analysis diagram. Each node is first regarded as a cluster. The average distance method is used to calculate the inter-cluster distance. The inter-cluster distance selects the Ward distance for inter-class distance calculation. The Ward distance is the difference between the sum of squares of the deviations between two clusters Δ(C i ,C j ), the calculation formula is as follows:

[0064] Sum of squared deviations:

[0065] Δ(C i ,C j )=ESS(C i ∪C j )-ESS(C i )-ESS(C j )

[0066] After calculating the Ward distance between each cluster, merge the two clusters with the closest distance; repeat the above two steps until all clusters are merged to obtain the clustering result;

[0067] The process of establishing the pre-established partition model in the cluster partition module:

[0068] Cluster the data set X with a capacity of N into K classes, measure the closeness within the class by calculating the sum of the squares of the distances between each point in the class and the class center, as the intra-class distance, and measure the separation of the data set by the sum of the squares of the distances between each class center point and the center point of the data set, as the inter-class distance;

[0069] The calculation formula of CH index is:

[0070]

[0071] Among them B k is the covariance matrix between classes, W k The covariance matrix of the intra-class data;

[0072] B k The calculation formula is as follows:

[0073]

[0074] where c q represents the center point of class q, c e Represents the center point of the data set, n q represents the number of data in class q, C q Represents a data set of class q. The larger the score of the Calinski-Harabasz index, the better it is. The smaller the covariance within the class, the better, and the larger the covariance between classes, the better.

[0075] The partitioning effect of the pre-established partitioning model in the cluster partitioning module is measured by the silhouette coefficient s(x i ) for evaluation:

[0076] The silhouette coefficient of the data point x is defined as:

[0077]

[0078] a(x i ), b(x i ) is calculated as follows: a(x i ) is the data point x i The average distance to other points in the same cluster, b(x i ) is the data point x i Average distance to the nearest cluster:

[0079]

[0080] Among them, the higher the silhouette coefficient value, the better the point matches its own cluster and the worse it matches the neighboring clusters, which means the better the cluster partitioning effect.

[0081] In another aspect of the present invention, in order to achieve the above-mentioned purpose, a terminal device is disclosed, including a memory, a processor, and a computer program stored in the memory and capable of running on the processor, wherein the memory stores a computer program capable of running on the processor, and when the processor loads and executes the computer program, a hierarchical clustering partitioning method for reactive voltage control as described above is adopted.

[0082] Beneficial effects of the present invention:

[0083] The invention constructs reactive source control space by analyzing the power flow characteristics of distribution network, defines new electrical distance based on sensitivity information, adopts agglomerative hierarchical clustering algorithm and Ward distance for clustering, uses Calinski-Harabasz index to determine the optimal number of partitions, and uses silhouette coefficient to evaluate the partition effect, thereby realizing efficient voltage partitioning. BRIEF DESCRIPTION OF THE DRAWINGS

[0084] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0085] Figure 1 It is a schematic flow chart of the method of the present invention;

[0086] Figure 2 It is a schematic diagram of the clustering process of the IEEE39 node system of the present invention;

[0087] Figure 3 It is a schematic diagram of the process flow of the agglomerative hierarchical clustering algorithm of the present invention;

[0088] Figure 4 It is a schematic diagram of the system structure of the present invention;

[0089] Figure 5 This is a diagram of online zoning calculation results of a provincial power grid according to the present invention. DETAILED DESCRIPTION

[0090] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0091] Embodiment 1:

[0092] The following is an introduction to the relevant terms involved in the embodiments of the present application:

[0093] Distribution network: Distribution network refers to the power grid that receives electric energy from the transmission network or regional power plants and distributes it locally or step by step according to voltage to various users through distribution facilities. It is composed of overhead lines, cables, towers, distribution transformers, disconnectors, reactive power compensators and some ancillary facilities, and plays an important role in distributing electric energy in the power grid.

[0094] Clustering: The process of dividing a collection of physical or abstract objects into multiple classes consisting of similar objects is called clustering. The clusters generated by clustering are a set of data objects that are similar to the objects in the same cluster and different from the objects in other clusters.

[0095] like Figure 1 As shown, a hierarchical clustering partitioning method for reactive power and voltage control comprises the following steps:

[0096] S101: Obtain grid node parameters, use the grid node parameters to calculate the admittance matrix of the entire grid, and use the preset node power equation and the admittance matrix of the entire grid to perform power flow calculation to obtain a Jacobian matrix;

[0097] The calculation process of measuring the admittance matrix of the entire network using the grid node parameters is as follows:

[0098] The node admittance matrix is ​​constructed as shown in (1) to (4), where formula (3) is the self-admittance matrix, which is the diagonal element of the node admittance matrix. The self-admittance Yii means that when the unit voltage Ui is added to the node i and all other nodes are grounded, the current injected into the network through point i is numerically equal to the sum of the admittances of all branches connected to the node. Formula (4) is the mutual admittance matrix, which refers to the other non-diagonal elements Yij (i≠j) in the node admittance matrix. The logistic meaning of mutual admittance Yij is the current injected into the network through node j when the unit voltage is injected into the node i and other nodes are grounded. In terms of numerical value, the mutual admittance is equal to the negative value of the admittance of the branch elements connected between nodes i and j. It is easy to know that when there is no branch directly connected between nodes i and j, Yij=Yji.

[0099]

[0100]

[0101] The Newton-Raphson method is used to calculate the power flow of the power grid and obtain the voltage and power information of each node in the system. Formula (5) is the power flow correction equation of the Newton-Raphson method power system expressed in polar coordinate form, Formula (6) is the expression of the active and reactive unbalanced quantity in Formula (5), and Formulas (7) to (10) are the elements of the Jacobian matrix.

[0102]

[0103]

[0104] S102: generating a control space with each reactive source node as a coordinate axis according to the sensitivity of the quasi-steady-state physical response of the PV node in the Jacobian matrix, and obtaining an electrical distance according to the control space and the control capability of the reactive source on the load node;

[0105] The process of generating the control space with each reactive source node as the coordinate axis according to the sensitivity of the quasi-steady-state physical response of the PV node in the Jacobian matrix:

[0106] Set the node type, set the j-th generator node in set G to a PQ node, and for other generator nodes in set G, if they participate in PVC, they are set to PV nodes, otherwise they are set to PQ nodes;

[0107] When calculating the sensitivity, the diagonal elements corresponding to the PV nodes in the full-dimensional B” matrix containing the PV nodes are increased to ensure that the node voltage remains unchanged, thereby retaining the quasi-steady-state control response characteristics of the PVC device;

[0108] Based on the constructed B”, the control sensitivity of the injected reactive power of node j to the voltage of each node in the set L is solved. For load node i, where i∈L, the sensitivity is the quasi-steady-state sensitivity, expressed as S ij ;

[0109]

[0110] Where S ij It is expressed as the voltage control capability of the jth generator node’s reactive injection on the load node i, for each generator node belonging to the set G;

[0111] The control capability of each generator node is regarded as one dimension in the coordinate space. Thus, a g-dimensional coordinate space is constructed from the set G as the reactive source control space, using a g-dimensional vector (x i1 ,x i2 ,…,x ig ) is used as the coordinate to describe the load node i, where the jth coordinate component is defined as:

[0112] x ij = -lg|S ij ∣

[0113] Among them, x ij It uses coordinates to describe the degree of coupling between reactive power and voltage.

[0114] The process of obtaining the electrical distance according to the control space and the control capability of the reactive source on the load node is as follows:

[0115] In the g-dimensional space, the coordinate vector (x i1 ,x i2 ,…,x ig) is the control sensitivity vector of all generator nodes to it. The jth component of the coordinate reflects the sensitivity of the reactive power injection of the jth generator node to control the voltage of the node. The two nodes to be partitioned m and n correspond to the coordinate vector (x m1 ,x m2 ,…,x mg ) and (x n1 ,x n2 ,…,x ng ), the distance between two points is directly defined using the Euclidean distance:

[0116]

[0117] S103: using an agglomerative hierarchical clustering algorithm, selecting Ward distance to calculate the distance between classes according to the electrical distance, obtaining a clustering result, inputting the clustering result into a pre-established partition model, and outputting a partition result, wherein the partition model is constructed based on the Calinski-Harabasz index.

[0118] Using the agglomerative hierarchical clustering algorithm, according to the electrical distance, Ward distance is selected to calculate the distance between classes and obtain the clustering result:

[0119] The nodes with large coupling degree are divided into one category, and the nodes with small coupling degree are separated. Cluster analysis is performed to obtain the cluster analysis diagram. Each node is first regarded as a cluster. The average distance method is used to calculate the inter-cluster distance. The inter-cluster distance selects the Ward distance for inter-class distance calculation. The Ward distance is the difference between the sum of squares of the deviations between two clusters Δ(C i ,C j ), the calculation formula is as follows:

[0120] Sum of squared deviations:

[0121] Δ(C i ,C j )=ESS(C i ∪C j )-ESS(C i )-ESS(C j )

[0122] After calculating the Ward distance between each cluster, merge the two clusters with the closest distance; repeat the above two steps until all clusters are merged to obtain the clustering result.

[0123] The process of building a pre-built partition model:

[0124] Cluster the data set X with a capacity of N into K classes, measure the closeness within the class by calculating the sum of the squares of the distances between each point in the class and the class center, as the intra-class distance, and measure the separation of the data set by the sum of the squares of the distances between each class center point and the center point of the data set, as the inter-class distance;

[0125] The calculation formula of CH index is:

[0126]

[0127] Among them B k is the covariance matrix between classes, W k The covariance matrix of the intra-class data;

[0128] B k The calculation formula is as follows:

[0129]

[0130] where c q represents the center point of class q, c e Represents the center point of the data set, n q represents the number of data in class q, C q Representing a data set of class q, the larger the score of the Calinski-Harabasz index, the better it is, among which the smaller the covariance within the category, the better, and the larger the covariance between categories, the better.

[0131] The partitioning effect of the pre-established partitioning model is expressed by the silhouette coefficient s(x i ) for evaluation:

[0132] The silhouette coefficient of the data point x is defined as:

[0133]

[0134] a(x i ), b(x i ) is calculated as follows: a(x i ) is the data point x i The average distance to other points in the same cluster, b(x i ) is the data point x i Average distance to the nearest cluster:

[0135]

[0136] Among them, the higher the silhouette coefficient value, the better the point matches its own cluster and the worse it matches the neighboring clusters, which means the better the cluster partitioning effect.

[0137] Here, firstly, the online partition calculation results of a provincial power grid are given for the power grid data of a certain province. Figure 5The clustering process curve when using Ward inter-class distance is shown. From the curve, it can be found that a long plateau period occurs in the process of changing from 4 partitions to 3 partitions. Therefore, the best decoupling can be achieved when the power grid is decomposed into 4 zones, thus avoiding the difficulty of manually specifying the number of partitions in the traditional partitioning algorithm and making online adaptive partitioning possible.

[0138] Embodiment 2: In a second aspect, in order to achieve the above-mentioned purpose, the present invention discloses a hierarchical clustering partitioning system for reactive power and voltage control, comprising:

[0139] The matrix calculation module 11 is used to obtain the grid node parameters, calculate the admittance matrix of the whole grid using the grid node parameters, and calculate the Jacobian matrix using the admittance matrix of the whole grid through the preset node power equation;

[0140] The distance calculation module 12 is used to generate a control space with each reactive source node as a coordinate axis according to the sensitivity of the quasi-steady-state physical response of the PV node in the Jacobian matrix, and obtain the electrical distance according to the control space and the control ability of the reactive source on the load node;

[0141] The clustering partitioning module 13 is used to use the agglomerative hierarchical clustering algorithm to select the Ward distance to calculate the inter-class distance according to the electrical distance to obtain the clustering result, input the clustering result into a pre-established partitioning model, and output the partitioning result, wherein the partitioning model is constructed based on the Calinski-Harabasz index.

[0142] Combined with the second aspect, such as Figure 4 As shown, in some implementations of the second aspect, the system further includes: a calculation process of calculating the admittance matrix of the entire network using the grid node parameters in the matrix calculation module 11:

[0143]

[0144] Where, self-admittance Y ii Indicates that a unit voltage U is added to node i i When all other nodes are grounded, the current injected into the network through point i is numerically equal to the sum of the admittances of all branches connected to the node. The mutual admittance Y ij It means that when a unit voltage is injected into node i and other nodes are grounded, the current injected into the network through node j is equal to the negative value of the admittance of the branch element connected between nodes i and j. When there is no branch directly connected between nodes i and j, Y ij =Y ji ;

[0145] The distance calculation module 12 generates a control space process with each reactive source node as a coordinate axis according to the sensitivity of the quasi-steady-state physical response of the PV node in the Jacobian matrix:

[0146] Set the node type, set the j-th generator node in set G to a PQ node, and for other generator nodes in set G, if they participate in PVC, they are set to PV nodes, otherwise they are set to PQ nodes;

[0147] When calculating the sensitivity, the diagonal elements corresponding to the PV nodes in the full-dimensional B” matrix containing the PV nodes are increased to ensure that the node voltage remains unchanged, thereby retaining the quasi-steady-state control response characteristics of the PVC device;

[0148] Based on the constructed B”, the control sensitivity of the injected reactive power of node j to the voltage of each node in the set L is solved. For load node i, where i∈L, the sensitivity is the quasi-steady-state sensitivity, expressed as S ij ;

[0149]

[0150] Where S ij It is expressed as the voltage control capability of the jth generator node’s reactive injection on the load node i, for each generator node belonging to the set G;

[0151] The control capability of each generator node is regarded as one dimension in the coordinate space. Thus, a g-dimensional coordinate space is constructed from the set G as the reactive source control space, using a g-dimensional vector (x i1 ,x i2 ,…,x ig ) is used as the coordinate to describe the load node i, where the jth coordinate component is defined as:

[0152] x ij = -lg|S ij ∣

[0153] Among them, x ij It uses coordinates to describe the tightness of coupling between reactive power and voltage;

[0154] The distance calculation module 12 obtains the process of the electrical distance according to the control space and the control capability of the reactive source on the load node:

[0155] In the g-dimensional space, the coordinate vector (x i1 ,x i2 ,…,x ig) is the control sensitivity vector of all generator nodes to it. The jth component of the coordinate reflects the sensitivity of the reactive power injection of the jth generator node to control the voltage of the node. The two nodes to be partitioned m and n correspond to the coordinate vector (x m1 ,x m2 ,…,x mg ) and (x n1 ,x n2 ,…,x ng ), the distance between two points is directly defined using the Euclidean distance:

[0156]

[0157] The clustering partitioning module 13 uses the agglomerative hierarchical clustering algorithm to select the Ward distance to calculate the distance between classes according to the electrical distance to obtain the clustering result:

[0158] The nodes with large coupling degree are divided into one category, and the nodes with small coupling degree are separated. Cluster analysis is performed to obtain the cluster analysis diagram. Each node is first regarded as a cluster. The average distance method is used to calculate the inter-cluster distance. The inter-cluster distance selects the Ward distance for inter-class distance calculation. The Ward distance is the difference between the sum of squares of the deviations between two clusters Δ(C i ,C j ), the calculation formula is as follows:

[0159] Sum of squared deviations:

[0160] Δ(C i ,C j )=ESS(C i ∪C j )-ESS(C i )-ESS(C j )

[0161] After calculating the Ward distance between each cluster, merge the two clusters with the closest distance; repeat the above two steps until all clusters are merged to obtain the clustering result;

[0162] The process of establishing the partition model pre-established in the cluster partition module 13:

[0163] Cluster the data set X with a capacity of N into K classes, measure the closeness within the class by calculating the sum of the squares of the distances between each point in the class and the class center, as the intra-class distance, and measure the separation of the data set by the sum of the squares of the distances between each class center point and the center point of the data set, as the inter-class distance;

[0164] The calculation formula of CH index is:

[0165]

[0166] Among them B k is the covariance matrix between classes, W k The covariance matrix of the intra-class data;

[0167] B k The calculation formula is as follows:

[0168]

[0169] where c q represents the center point of class q, c e Represents the center point of the data set, n q represents the number of data in class q, C q Represents a data set of class q. The larger the score of the Calinski-Harabasz index, the better it is. The smaller the covariance within the class, the better, and the larger the covariance between classes, the better.

[0170] The partitioning effect of the partitioning model pre-established in the cluster partitioning module 13 is measured by the silhouette coefficient s(x i ) for evaluation:

[0171] The silhouette coefficient of the data point x is defined as:

[0172]

[0173] a(x i ), b(x i ) is calculated as follows: a(x i ) is the data point x i The average distance to other points in the same cluster, b(x i ) is the data point x i Average distance to the nearest cluster:

[0174]

[0175] Among them, the higher the silhouette coefficient value, the better the point matches its own cluster and the worse it matches the neighboring clusters, which means the better the cluster partitioning effect.

[0176] Based on the same inventive concept, the present invention also provides a computer device, which includes: one or more processors, and a memory for storing one or more computer programs; the program includes program instructions, and the processor is used to execute the program instructions stored in the memory. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, which is used to implement one or more instructions, specifically for loading and executing one or more instructions in a computer storage medium to implement the above method.

[0177] It needs to be further explained that, based on the same inventive concept, the present invention also provides a computer storage medium, on which a computer program is stored, and the computer program is executed by a processor to execute the above method. The storage medium can adopt any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electrical, magnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples (non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium can be any tangible medium containing or storing a program, which can be used by an instruction execution system, device or device or used in combination with it.

[0178] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present disclosure. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0179] The above shows and describes the basic principles, main features and advantages of the present disclosure. Those skilled in the art should understand that the present disclosure is not limited by the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present disclosure. Without departing from the spirit and scope of the present disclosure, the present disclosure may have various changes and improvements, and these changes and improvements fall within the scope of the present disclosure to be protected.

Claims

1. A hierarchical clustering partitioning method for reactive power and voltage control, characterized in that: The method comprises the following steps: Obtain grid node parameters, use the grid node parameters to calculate the admittance matrix of the entire grid, and use the preset node power equation to calculate the power flow using the admittance matrix of the entire grid to obtain the Jacobian matrix; According to the sensitivity of the quasi-steady-state physical response of the PV node in the Jacobian matrix, a control space with each reactive source node as a coordinate axis is generated, and the electrical distance is obtained according to the control space and the control ability of the reactive source on the load node; Using an agglomerative hierarchical clustering algorithm, Ward distance is selected to calculate the inter-class distance according to the electrical distance to obtain a clustering result, and the clustering result is input into a pre-established partition model to output a partition result, wherein the partition model is constructed based on the Calinski-Harabasz index.

2. A hierarchical clustering partitioning method for reactive power and voltage control according to claim 1, characterized in that: The calculation process of measuring the admittance matrix of the whole network by using the grid node parameters is as follows: Where, self-admittance Y ii Indicates that a unit voltage U is added to node i i When all other nodes are grounded, the current injected into the network through point i is numerically equal to the sum of the admittances of all branches connected to the node. The mutual admittance Y ij It means that when a unit voltage is injected into node i and other nodes are grounded, the current injected into the network through node j is equal to the negative value of the admittance of the branch element connected between nodes i and j. When there is no branch directly connected between nodes i and j, Y ij =Y ji .

3. A hierarchical clustering partitioning method for reactive power and voltage control according to claim 1, characterized in that: The process of generating a control space with each reactive source node as a coordinate axis according to the sensitivity of the quasi-steady-state physical response of the PV node in the Jacobian matrix is ​​as follows: Set the node type, set the j-th generator node in set G as a PQ node, and for other generator nodes in set G, if they participate in PVC, they are set as PV nodes, otherwise they are set as PQ nodes; When calculating the sensitivity, the diagonal elements corresponding to the PV nodes in the full-dimensional B” matrix containing the PV nodes are increased to ensure that the node voltage remains unchanged, thereby retaining the quasi-steady-state control response characteristics of the PVC device; Based on the constructed B”, the control sensitivity of the injected reactive power of node j to the voltage of each node in the set L is solved. For load node i, where i∈L, the sensitivity is the quasi-steady-state sensitivity, expressed as S ij ; Where S ij It is expressed as the voltage control capability of the jth generator node’s reactive injection on the load node i, for each generator node belonging to the set G; The control capability of each generator node is regarded as one dimension in the coordinate space. Thus, a g-dimensional coordinate space is constructed from the set G as the reactive source control space, using a g-dimensional vector (x i1 ,x i2 ,…,x ig ) is used as the coordinate to describe the load node i, where the jth coordinate component is defined as: x ij =-lg∣S ij ∣ Among them, x ij It uses coordinates to describe the degree of coupling between reactive power and voltage.

4. A hierarchical clustering partitioning method for reactive power and voltage control according to claim 1, characterized in that: The process of obtaining the electrical distance according to the control space and the control capability of the reactive source on the load node is as follows: In the g-dimensional space, the coordinate vector (x i1 ,x i2 ,…,x ig ) is the control sensitivity vector of all generator nodes to it. The jth component of the coordinate reflects the sensitivity of the reactive power injection of the jth generator node to control the voltage of the node. The two nodes to be partitioned m and n correspond to the coordinate vector (x m1 ,x m2 ,…,x mg ) and (x n1 ,x n2 ,…,x ng ), the distance between two points is directly defined using the Euclidean distance:

5. A hierarchical clustering partitioning method for reactive power and voltage control according to claim 1, characterized in that: The process of using the agglomerative hierarchical clustering algorithm to select the Ward distance to calculate the distance between classes according to the electrical distance to obtain the clustering result is as follows: The nodes with large coupling degree are divided into one category, and the nodes with small coupling degree are separated. Cluster analysis is performed to obtain the cluster analysis diagram. Each node is first regarded as a cluster. The average distance method is used to calculate the inter-cluster distance. The inter-cluster distance selects the Ward distance for inter-class distance calculation. The Ward distance is the difference between the sum of squares of the deviations between two clusters Δ(C i ,C j ), the calculation formula is as follows: Sum of squared deviations: Δ(C i ,C j )=ESS(C i ∪C j )-ESS(C i )-ESS(C j ) After calculating the Ward distance between each cluster, merge the two clusters with the closest distance; repeat the above two steps until all clusters are merged to obtain the clustering result.

6. A hierarchical clustering partitioning method for reactive power and voltage control according to claim 1, characterized in that: The establishment process of the pre-established partition model: Cluster the data set X with a capacity of N into K classes, measure the closeness within the class by calculating the sum of the squares of the distances between each point in the class and the class center, as the intra-class distance, and measure the separation of the data set by the sum of the squares of the distances between each class center point and the center point of the data set, as the inter-class distance; The calculation formula of CH index is: Among them B k is the covariance matrix between classes, W k The covariance matrix of the intra-class data; B k The calculation formula is as follows: where c q represents the center point of class q, c e Represents the center point of the data set, n q represents the number of data in class q, C q Representing a data set of class q, the larger the score of the Calinski-Harabasz index, the better it is, among which the smaller the covariance within the category, the better, and the larger the covariance between categories, the better.

7. A hierarchical clustering partitioning method for reactive power and voltage control according to claim 6, characterized in that: The partitioning effect of the pre-established partitioning model is measured by the silhouette coefficient s(x i ) for evaluation: The silhouette coefficient of the data point x is defined as: a(x i ), b(x i ) is calculated as follows: a(x i ) is the data point x i The average distance to other points in the same cluster, b(x i ) is the data point x i Average distance to the nearest cluster: Among them, the higher the silhouette coefficient value, the better the point matches its own cluster and the worse it matches the neighboring clusters, which means the better the cluster partitioning effect.

8. A hierarchical clustering partitioning system for reactive power and voltage control, characterized in that: include: The matrix calculation module is used to obtain the grid node parameters, use the grid node parameters to calculate the admittance matrix of the whole grid, and use the preset node power equation to calculate the power flow using the admittance matrix of the whole grid to obtain the Jacobian matrix; The distance calculation module is used to generate a control space with each reactive source node as a coordinate axis according to the sensitivity of the quasi-steady-state physical response of the PV node in the Jacobian matrix, and obtain the electrical distance according to the control space and the control ability of the reactive source on the load node; The clustering partitioning module is used to use the agglomerative hierarchical clustering algorithm to select the Ward distance to calculate the inter-class distance according to the electrical distance to obtain the clustering result, input the clustering result into a pre-established partitioning model, and output the partitioning result, wherein the partitioning model is constructed based on the Calinski-Harabasz index.

9. A hierarchical clustering partitioning system for reactive power and voltage control according to claim 8, characterized in that: The calculation process of the admittance matrix of the whole network using the grid node parameters in the matrix calculation module is as follows: Where, self-admittance Y ii Indicates that a unit voltage U is added to node i i When all other nodes are grounded, the current injected into the network through point i is numerically equal to the sum of the admittances of all branches connected to the node. The mutual admittance Y ij It means that when a unit voltage is injected into node i and other nodes are grounded, the current injected into the network through node j is equal to the negative value of the admittance of the branch element connected between nodes i and j. When there is no branch directly connected between nodes i and j, Y ij =Y ji ; The distance calculation module generates the control space process with each reactive source node as the coordinate axis according to the sensitivity of the quasi-steady-state physical response of the PV node in the Jacobian matrix: Set the node type, set the j-th generator node in set G as a PQ node, and for other generator nodes in set G, if they participate in PVC, they are set as PV nodes, otherwise they are set as PQ nodes; When calculating the sensitivity, the diagonal elements corresponding to the PV nodes in the full-dimensional B” matrix containing the PV nodes are increased to ensure that the node voltage remains unchanged, thereby retaining the quasi-steady-state control response characteristics of the PVC device; Based on the constructed B”, the control sensitivity of the injected reactive power of node j to the voltage of each node in the set L is solved. For load node i, where i∈L, the sensitivity is the quasi-steady-state sensitivity, expressed as S ij ; Where S ij It is expressed as the voltage control capability of the jth generator node’s reactive injection on the load node i, for each generator node belonging to the set G; The control capability of each generator node is regarded as one dimension in the coordinate space. Thus, a g-dimensional coordinate space is constructed from the set G as the reactive source control space, using a g-dimensional vector (x i1 ,x i2 ,…,x ig ) is used as the coordinate to describe the load node i, where the jth coordinate component is defined as: x ij =-lg∣S ij ∣ Among them, x ij It uses coordinates to describe the tightness of coupling between reactive power and voltage; The process of obtaining the electrical distance in the distance calculation module based on the control space and the control capability of the reactive source on the load node: In the g-dimensional space, the coordinate vector (x i1 ,x i2 ,…,x ig ) is the control sensitivity vector of all generator nodes to it. The jth component of the coordinate reflects the sensitivity of the reactive power injection of the jth generator node to control the voltage of the node. The two nodes to be partitioned m and n correspond to the coordinate vector (x m1 ,x m2 ,…,x mg ) and (x n1 ,x n2 ,…,x ng ), the distance between two points is directly defined using the Euclidean distance: The clustering partitioning module uses the agglomerative hierarchical clustering algorithm to select the Ward distance to calculate the distance between classes based on the electrical distance to obtain the clustering result: The nodes with large coupling degree are divided into one category, and the nodes with small coupling degree are separated. Cluster analysis is performed to obtain the cluster analysis diagram. Each node is first regarded as a cluster. The average distance method is used to calculate the inter-cluster distance. The inter-cluster distance selects the Ward distance for inter-class distance calculation. The Ward distance is the difference between the sum of squares of the deviations between two clusters Δ(C i ,C j ), the calculation formula is as follows: Sum of squared deviations: Δ(C i ,C j )=ESS(C i ∪C j )-ESS(C i )-ESS(C j ) After calculating the Ward distance between each cluster, merge the two clusters with the closest distance; repeat the above two steps until all clusters are merged to obtain the clustering result; The process of establishing the pre-established partition model in the cluster partition module: Cluster the data set X with a capacity of N into K classes, measure the closeness within the class by calculating the sum of the squares of the distances between each point in the class and the class center, as the intra-class distance, and measure the separation of the data set by the sum of the squares of the distances between each class center point and the center point of the data set, as the inter-class distance; The calculation formula of CH index is: Among them B k is the covariance matrix between classes, W k The covariance matrix of the intra-class data; B k The calculation formula is as follows: where c q represents the center point of class q, c e Represents the center point of the data set, n q represents the number of data in class q, C q Represents a data set of class q. The larger the score of the Calinski-Harabasz index, the better it is. The smaller the covariance within the class, the better, and the larger the covariance between classes, the better. The partitioning effect of the pre-established partitioning model in the cluster partitioning module is measured by the silhouette coefficient s(x i ) for evaluation: The silhouette coefficient of the data point x is defined as: a(x i ), b(x i ) is calculated as follows: a(x i ) is the data point x i The average distance to other points in the same cluster, b(x i ) is the data point x i Average distance to the nearest cluster: Among them, the higher the silhouette coefficient value, the better the point matches its own cluster and the worse it matches the neighboring clusters, which means the better the cluster partitioning effect.

10. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that: The memory stores a computer program that can be run on a processor. When the processor loads and executes the computer program, a hierarchical clustering partitioning method for reactive voltage control according to any one of claims 1 to 7 is adopted.