Distributed photovoltaic cluster dynamic division method and device

Abstract

The invention provides a distributed photovoltaic cluster dynamic division method and device, and belongs to the field of distributed photovoltaic regulation, and the method comprises the steps: constructing a first division similarity matrix based on a comprehensive index of electrical characteristics, and constructing a Jaccard similarity matrix based on the first division similarity matrix; performing first-layer division on the distributed photovoltaic nodes based on the Jaccard similarity matrix to obtain at least one cluster; each cluster comprises two distributed photovoltaic nodes, one node is the node with the maximum Jaccard similarity of the other node, and the number of common neighbor nodes owned by the two nodes is greater than or equal to 2; constructing a cluster network taking a cluster as a network node, and obtaining a second division similarity matrix; and performing distributed cluster division on the plurality of distributed photovoltaic nodes by using a Louvain algorithm in combination with the improved modularity gain of the reactive balance degree to obtain a division result of the plurality of distributed photovoltaic nodes.

Description

Technical Field

[0001] The present invention relates to the field of distributed photovoltaic control, and in particular to a method and device for dynamically dividing a distributed photovoltaic cluster. Background Art

[0002] Distributed photovoltaic power generation is connected to the distribution network using cluster technology for dispatching and management. The distribution network is divided into sub-networks for dispatching and control, and real-time monitoring and management of the operating status of the sub-networks is achieved. This can improve management efficiency, reduce operation and maintenance costs and human resource requirements, and make the overall operation of the system more stable and efficient.

[0003] If the division of distributed photovoltaic clusters is unreasonable, the divided clusters will be unfavorable for coordinated scheduling and will also reduce the stability of power generation in the power grid. Summary of the invention

[0004] The present invention provides a distributed photovoltaic cluster dynamic division method and device, which are used to improve the stability of power generation in a power grid.

[0005] The present invention provides a distributed photovoltaic cluster dynamic division method, comprising:

[0006] Based on the comprehensive index of electrical characteristics, constructing a first partition similarity matrix of multiple distributed photovoltaic nodes, the comprehensive index of electrical characteristics including at least one of the following: an electrical distance matrix based on reactive sensitivity, active regulation capacity, reactive regulation capacity, load rate and net load;

[0007] Based on the first partition similarity matrix, obtaining the adjacency relationship of the distributed photovoltaic nodes;

[0008] Based on the adjacency relationship of the distributed photovoltaic nodes, constructing a Jaccard similarity matrix of the distributed photovoltaic nodes;

[0009] Based on the Jaccard similarity matrix, the distributed photovoltaic nodes are divided into a first layer to obtain at least one cluster; wherein different clusters include different distributed photovoltaic nodes, each cluster includes two distributed photovoltaic nodes, one of the two distributed photovoltaic nodes is a distributed photovoltaic node with the largest Jaccard similarity to the other distributed photovoltaic node, and the number of common neighbor nodes owned by the two distributed photovoltaic nodes is greater than or equal to 2;

[0010] constructing a cluster network with the at least one cluster as at least one network node, and obtaining a second partition similarity matrix of the at least one network node;

[0011] Based on the second partition similarity matrix and in combination with the improved modularity gain of reactive power balance, the Louvain algorithm is used to perform distributed cluster partitioning on the multiple distributed photovoltaic nodes to obtain a partitioning result of the multiple distributed photovoltaic nodes.

[0012] According to a distributed photovoltaic cluster dynamic partitioning method provided by the present invention, the partitioning similarity matrix of distributed photovoltaic nodes is constructed based on the comprehensive index of electrical characteristics, including:

[0013] Normalizing the active regulation capacity, reactive regulation capacity, load rate and net load of each photovoltaic node to obtain a normalized regulation capacity index;

[0014] Based on the normalized index of the regulation capacity, the electrical distance matrix based on reactive sensitivity is fused to construct the partition similarity matrix;

[0015] Among them, the electrical distance matrix based on reactive sensitivity is expressed as: d ij represents the electrical distance between node i and node j, and the element S in the reactive sensitivity matrix S ij The element S in the reactive sensitivity matrix S represents the change in the voltage of node i corresponding to the unit reactive power injected into node j. jj It represents the change value of node j voltage corresponding to the unit reactive power injected into node j. The relevant expression of reactive sensitivity matrix S is ΔV=SΔQ, where ΔV represents the change amount of voltage amplitude and ΔQ represents the change amount of reactive power.

[0016] Active power regulation capacity P of PV node i i Expressed as: P i ∈[0,P max ], P max Indicates the active output of the photovoltaic power generation system under the control of the maximum power point tracking system;

[0017] Reactive power regulation capacity Q of PV node i j It is expressed as:

[0018]

[0019] S max is the grid-connected capacity of the photovoltaic inverter corresponding to the photovoltaic cluster, P max represents the active output of the photovoltaic power generation system where the photovoltaic cluster is located under the control of the maximum power point tracking system, λ max Indicates the upper limit of the power factor, λ min Indicates the lower limit of the power factor.

[0020] According to a distributed photovoltaic cluster dynamic division method provided by the present invention, the active regulation capacity, reactive regulation capacity, load rate and net load of each photovoltaic node are normalized to obtain a normalized regulation capacity index, including:

[0021] By using the first formula, the active regulation capacity, reactive regulation capacity, load rate and net load of each photovoltaic node are normalized to obtain a normalized index of the regulation capacity of each photovoltaic node;

[0022] The first formula is expressed as:

[0023]

[0024] Among them, x i,m ′ represents the normalized index of the mth regulation capacity of PV node i, x i,m represents the mth index of photovoltaic node i, which includes active regulation capacity, reactive regulation capacity, load rate and net load, x 1 / 2,m represents the median value of the mth indicator, x 1 / 4,m Represents the one-quarter digit value of the m-th indicator.

[0025] According to a distributed photovoltaic cluster dynamic partitioning method provided by the present invention, based on the normalized index of the regulation capacity, the electrical distance matrix based on reactive sensitivity is fused to construct the partition similarity matrix, including:

[0026] Based on the regulation capacity normalization index, the electrical distance matrix based on reactive sensitivity is fused through a second formula to obtain the partition similarity matrix;

[0027] The second formula is expressed as:

[0028]

[0029] Among them, s ij Represents the element in the i-th row and j-th column of the similarity matrix, α m represents the weight of the mth indicator, x i,m ′ represents the normalized index of the mth regulation capacity of PV node i, x j,m ′ represents the normalized index of the mth regulation capacity of PV node j, d ij represents the electrical distance between node i and node j based on reactive sensitivity, d max represents the maximum value in the electrical distance matrix based on reactive sensitivity, λ 1 With λ 2 Represent the weights of comprehensive index and electrical distance respectively.

[0030] According to a distributed photovoltaic cluster dynamic partitioning method provided by the present invention, the Jaccard similarity matrix of the distributed photovoltaic nodes is constructed based on the adjacency relationship of the distributed photovoltaic nodes, including:

[0031] Based on the adjacency relationship of the distributed photovoltaic nodes, a Jaccard similarity matrix of the distributed photovoltaic nodes is constructed by a third formula;

[0032] The third formula is expressed as:

[0033]

[0034] Among them, Sim jaccard，ij Represents the element in the i-th row and j-th column of the Jaccard similarity matrix, Sim jaccard，ij represents the Jaccard similarity between photovoltaic node i and photovoltaic node j, Y i represents the neighbor node set of distributed photovoltaic node i, Y j represents the neighbor node set of distributed photovoltaic node j, Ψ(Y i , Y j ) means when Y i With Y j If the intersection is not empty, the modulus of the intersection is taken; otherwise, the value is 1.

[0035] According to a distributed photovoltaic cluster dynamic partitioning method provided by the present invention, the step of constructing a cluster network with the at least one cluster as at least one network node and obtaining a second partition similarity matrix of the at least one network node includes:

[0036] Using the at least one cluster as at least one network node, constructing the cluster network;

[0037] Using a zero matrix as an initial similarity matrix of the cluster network;

[0038] Based on the sum of the pairwise similarities of the nodes in the at least one cluster, updating the diagonal elements of the initial similarity matrix;

[0039] Based on the fourth formula, calculating the pairwise partition similarity of at least one cluster, updating the initial similarity matrix, and obtaining a second partition similarity matrix;

[0040] The fourth formula is expressed as:

[0041]

[0042] in, represents the partition similarity between cluster t and cluster c in at least one cluster. Distributed photovoltaic node i and distributed photovoltaic node j are nodes in cluster t and cluster c respectively. Si,j Represents the partition similarity between distributed photovoltaic nodes i and j.

[0043] According to a distributed photovoltaic cluster dynamic partitioning method provided by the present invention, based on the second partition similarity matrix and in combination with the improved modularity gain of reactive power balance, the Louvain algorithm is used to perform distributed cluster partitioning on the multiple distributed photovoltaic nodes to obtain the partitioning results of the multiple distributed photovoltaic nodes, including:

[0044] Perform the first process at least once until the division results of the multiple distributed photovoltaic nodes obtained in the two most recent first processes remain unchanged, and use the division result of the multiple distributed photovoltaic nodes obtained in the most recent first process as the division result of the multiple distributed photovoltaic nodes;

[0045] Wherein, the first process once includes:

[0046] Traversing multiple distributed photovoltaic nodes until the second process is performed once for each distributed photovoltaic node;

[0047] The second process includes:

[0048] For photovoltaic node i, calculating the improved modularity gain of the reactive power balance degree of each photovoltaic node adjacent to the photovoltaic node i;

[0049] Select the photovoltaic node j corresponding to the maximum value of the improved modularity gain of the reactive power balance degree of the photovoltaic node i to each photovoltaic node adjacent to the photovoltaic node i, and add the photovoltaic node i to the cluster where the photovoltaic node j is located;

[0050] Among them, the improved modularity gain of reactive power balance is expressed as: ΔM new =β 1 ΔM+β 2 Δω; ΔM represents the improved modularity gain of the mobile node, Δω represents the reactive power balance gain of the mobile node, β 1 With β 2 Represent the weights of improved modularity gain and reactive power balance gain respectively.

[0051] The present invention also provides a distributed photovoltaic cluster dynamic division device, comprising:

[0052] A first construction module is used to construct a first partition similarity matrix of multiple distributed photovoltaic nodes based on a comprehensive index of electrical characteristics, wherein the comprehensive index of electrical characteristics includes at least one of the following: an electrical distance matrix based on reactive sensitivity, active regulation capacity, reactive regulation capacity, load rate, and net load;

[0053] An adjacency relationship acquisition module, configured to obtain the adjacency relationship of the distributed photovoltaic nodes based on the first partition similarity matrix;

[0054] A second construction module is used to construct a Jaccard similarity matrix of the distributed photovoltaic nodes based on the adjacency relationship of the distributed photovoltaic nodes;

[0055] A first partitioning module is used to perform a first-layer partitioning of the distributed photovoltaic nodes based on the Jaccard similarity matrix to obtain at least one cluster; wherein different clusters include different distributed photovoltaic nodes, each cluster includes two distributed photovoltaic nodes, one of the two distributed photovoltaic nodes is a distributed photovoltaic node with the largest Jaccard similarity to the other distributed photovoltaic node, and the number of common neighbor nodes owned by the two distributed photovoltaic nodes is greater than or equal to 2;

[0056] A third construction module is used to construct a cluster network with the at least one cluster as at least one network node, and obtain a second partition similarity matrix of the at least one network node;

[0057] The second partitioning module is used to perform distributed cluster partitioning on the multiple distributed photovoltaic nodes based on the second partitioning similarity matrix and the improved modularity gain of the reactive balance degree using the Louvain algorithm to obtain the partitioning results of the multiple distributed photovoltaic nodes.

[0058] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of any of the above-mentioned methods for dynamically dividing distributed photovoltaic clusters are implemented.

[0059] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of any of the above-mentioned methods for dynamically dividing a distributed photovoltaic cluster are implemented.

[0060] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the steps of any of the above-mentioned methods for dynamically dividing a distributed photovoltaic cluster are implemented.

[0061] The distributed photovoltaic cluster dynamic partitioning method and device provided by the present invention perform hierarchical partitioning by using a comprehensive index system based on electrical characteristics and constructing two similarity matrices based on position characteristics respectively; the timeliness of the Louvain algorithm is improved by hierarchical improvement of the Louvain algorithm, wherein the first layer of partitioning is based on position characteristics, and at the same time, it is considered that the partitioning only allows two nodes in a cluster and each node is only partitioned once, so that the timeliness is improved while also ensuring the effectiveness of the partitioning, and reducing the influence of the first layer of partitioning results on subsequent partitioning; the second layer of partitioning is based on the improved Louvain algorithm partitioning of electrical characteristics, and by combining the improved modularity gain of the reactive balance and introducing the reactive balance for optimization, the partitioning effect is improved, the effectiveness and rationality of the dynamic partitioning of the distributed photovoltaic cluster is ensured, and the timeliness of the dynamic partitioning algorithm of the distributed photovoltaic cluster is also improved, and the stability of the power generation work of the power grid is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0063] Figure 1 It is a schematic diagram of the process of the distributed photovoltaic cluster dynamic partitioning method provided by the present invention;

[0064] Figure 2 It is a schematic diagram of the first layer division provided by the present invention;

[0065] Figure 3 It is a flow chart of the cluster network construction method provided by the present invention;

[0066] Figure 4 A flowchart of using the Louvain algorithm for a new network in combination with improved modularity gain of reactive balance in the present invention;

[0067] Figure 5 A schematic diagram of the network structure of the data set provided by the present invention;

[0068] Figure 6 A schematic diagram of the Louvain algorithm provided by the present invention for dividing the network structure;

[0069] Figure 7 A schematic diagram of the improved hierarchical Louvain algorithm for dividing the network structure provided by the present invention;

[0070] Figure 8It is a structural schematic diagram of a distributed photovoltaic cluster dynamic partitioning device provided by the present invention;

[0071] Fig. 9 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0072] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are 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.

[0073] The distributed photovoltaic cluster dynamic partitioning method and device of the present invention are described below with reference to the accompanying drawings.

[0074] Figure 1 is a flow chart of the distributed photovoltaic cluster dynamic partitioning method provided by the present invention, such as Figure 1 As shown, including:

[0075] Step 100, constructing a first partition similarity matrix of multiple distributed photovoltaic nodes based on comprehensive indicators of electrical characteristics, wherein the comprehensive indicators of electrical characteristics include at least one of the following: an electrical distance matrix based on reactive sensitivity, active regulation capacity, reactive regulation capacity, load rate, and net load;

[0076] Specifically, the comprehensive index system based on electrical characteristics of the present invention mainly consists of five parts, namely, an electrical distance matrix based on reactive sensitivity, active regulation capacity, reactive regulation capacity, load rate and net load.

[0077] Specifically, the electrical distance is used to characterize the degree of electrical coupling between two distributed PV nodes;

[0078] Specifically, the existence of reactive power regulation capacity can adjust the reactive output of the generator to maintain the voltage stability of the power system; the existence of active power regulation capacity can adjust the active power output of the generator to maintain the power balance and frequency stability of the power system; the regulation capacity can adjust the output power of the generator according to the demand of load changes to meet the load demand of the power system. When a system failure occurs, the regulation capacity can provide additional power support to maintain the stable operation of the power system; the load rate can help ensure that each distributed photovoltaic node in the cluster is allocated with similar load demands, which helps to ensure the balance, reliability and sustainability of load distribution between distributed photovoltaic nodes in the distributed photovoltaic cluster. The present invention can maximize the efficiency and stability of the power system through reasonable load balancing.

[0079] Specifically, net load reflects the actual load demand of the power system and is crucial to maintaining the stability of the system. Similar net load means that the power demand patterns of the two nodes are similar. In distributed photovoltaic clusters, it can help achieve load balancing, that is, it is easier to match photovoltaic power generation with load demand. It helps to reduce excess or insufficient power supply and improve the stability of the power system.

[0080] Specifically, the electrical distance matrix based on reactive sensitivity mainly focuses on the impact of reactive power on the stability of the power system. Photovoltaic power generation systems mainly generate active power, while reactive power is relatively small, especially under standard conditions. Therefore, the electrical distance matrix of reactive sensitivity can better reflect the reactive voltage stability problem of the photovoltaic cluster. The electrical distance matrix of reactive sensitivity can be used to evaluate the reactive power interaction between the photovoltaic cluster and other components in the power system (such as reactive compensation equipment, transformers, etc.) to ensure voltage stability. This is very important for voltage control and reactive regulation in the cluster to avoid voltage sag or voltage peak problems. Although active power also affects the stability of the power system to a certain extent, the management of reactive power is more urgent in photovoltaic clusters. The electrical distance matrix of reactive sensitivity provides information about the flow of reactive power, which helps to accurately evaluate reactive voltage stability and solve related problems. The first partition similarity matrix constructed by combining the above five indicators well reflects the electrical characteristics between distributed photovoltaic nodes and can improve the rationality of the partition results.

[0081] Step 110, obtaining the adjacency relationship of the distributed photovoltaic nodes based on the first partition similarity matrix;

[0082] Specifically, after obtaining the first partition similarity matrix constructed by integrating the above five indicators, the adjacency relationship of distributed photovoltaic nodes can be further determined;

[0083] It should be noted that any method for obtaining the adjacency relationship of the distributed photovoltaic nodes based on the first partition similarity matrix is ​​applicable to the present invention and is not limited here.

[0084] Step 120, constructing a Jaccard similarity matrix of the distributed photovoltaic nodes based on the adjacency relationship of the distributed photovoltaic nodes;

[0085] Specifically, in order to divide the distributed photovoltaic nodes, the Jaccard similarity matrix of the distributed photovoltaic nodes can be constructed based on the adjacency relationship of the distributed photovoltaic nodes;

[0086] It should be noted that any method of constructing a Jaccard similarity matrix of distributed photovoltaic nodes is applicable to the present invention as long as the adjacency relationship based on the distributed photovoltaic nodes can be realized, and is not limited here.

[0087] Step 130, based on the Jaccard similarity matrix, the distributed photovoltaic nodes are divided into a first layer to obtain at least one cluster; wherein different clusters include different distributed photovoltaic nodes, each cluster includes two distributed photovoltaic nodes, one of the two distributed photovoltaic nodes is a distributed photovoltaic node with the largest Jaccard similarity to the other distributed photovoltaic node, and the number of common neighbor nodes owned by the two distributed photovoltaic nodes is greater than or equal to 2;

[0088] Figure 2 is a schematic diagram of the first layer division provided by the present invention, such as Figure 2 As shown, the first-level division of distributed photovoltaic nodes of a distributed photovoltaic cluster using the distributed photovoltaic node Jaccard similarity matrix is ​​as follows:

[0089] (1) Each individual distributed photovoltaic node is a cluster. All distributed photovoltaic nodes are traversed. When traversing to distributed photovoltaic node i, if the number of nodes in the cluster where node i is located is not 1, skip traversing the next node, otherwise go to step 2);

[0090] (2) Find the distributed photovoltaic node j with the greatest Jaccard similarity to the distributed photovoltaic node i from the Jaccard similarity matrix. If the number of common neighbor nodes between the node j with the greatest Jaccard similarity and node i is greater than or equal to 2, then add node i to the cluster of the node j with the greatest Jaccard similarity, otherwise skip traversing the next node;

[0091] (3) After the traversal is completed, the first layer of division is completed, and the first layer of distributed photovoltaic cluster division results are saved.

[0092] In the present invention, the first-level partitioning only allows two nodes in a cluster and the two nodes have more than two common neighbor nodes, and each node is partitioned only once. In order to prevent excessive partitioning based on geographical characteristics, which will seriously affect the effectiveness of subsequent partitioning, this partitioning repeatedly utilizes geographical characteristics while greatly reducing the impact on subsequent partitioning, ensuring the effectiveness of the partitioning and effectively improving the timeliness of subsequent partitioning.

[0093] Step 140, constructing a cluster network with the at least one cluster as at least one network node, and obtaining a second partition similarity matrix of the at least one network node;

[0094] Specifically, after the distributed photovoltaic nodes of the distributed photovoltaic cluster are divided into the first layer and at least one cluster is obtained, the at least one cluster can be divided with one cluster as one network node.

[0095] Step 150: Based on the second partition similarity matrix and in combination with the improved modularity gain of reactive power balance, the Louvain algorithm is used to perform distributed cluster partitioning on the plurality of distributed photovoltaic nodes to obtain a partitioning result of the plurality of distributed photovoltaic nodes.

[0096] Optionally, in the present invention, a distributed photovoltaic cluster dynamic partitioning algorithm based on an improved modularity gain hierarchical Louvain algorithm comprises the following steps:

[0097] (1) Constructing the partition similarity matrix of distributed photovoltaic nodes by combining the comprehensive index system based on electrical characteristics;

[0098] (2) Construct the Jaccard similarity matrix of distributed photovoltaic nodes based on the adjacency relationship, i.e., location characteristics of distributed photovoltaic nodes;

[0099] (3) Use the Jaccard similarity matrix of distributed photovoltaic nodes to perform the first-level division of distributed photovoltaic clusters;

[0100] (4) Reconstructing the distributed photovoltaic node network based on the first-layer division results, that is, constructing a cluster network and obtaining the second division similarity matrix. Each node in the cluster network is a cluster after the first-layer division;

[0101] (5) Based on the second partition similarity matrix, the Louvain algorithm is used to divide the cluster network into distributed photovoltaic clusters using the improved modularity gain combined with reactive balance.

[0102] The commonly used distributed photovoltaic cluster dynamic partitioning algorithms in related technologies mainly include clustering algorithms, intelligent optimization algorithms and complex community discovery algorithms; among clustering algorithms, the K-means clustering algorithm is the most common, but the K-means clustering needs to determine the number of clusters and the center point, which reduces the effect and timeliness of the partitioning; the intelligent optimization algorithm constructs a comprehensive evaluation index system and constructs a corresponding objective function to optimize it to obtain the final cluster partitioning result. Compared with the K-means clustering, there is no need to determine the number of clusters and the center point, but the comprehensive evaluation index system is difficult to construct and the optimization problem is prone to fall into local optimality; the complex community discovery algorithm constructs a similarity matrix based on the electrical index system, and then divides the community discovery algorithm based on the optimization modularity on the basis of the similarity matrix. There is no need to consider the construction of the objective function or the determination of the number of clusters. The Louvain algorithm is one of the most efficient complex community discovery algorithms. The present invention is also designed based on the application idea of ​​the complex community discovery algorithm in the distribution of photovoltaic clusters, and is theoretically feasible.

[0103] For the division of large-scale distributed photovoltaic clusters, the Louvain algorithm is not very effective in division timeliness. The present invention improves the Louvain algorithm in layers to improve the timeliness of the Louvain algorithm. The first-layer division is based on location characteristics, and at the same time, it is considered that only two nodes in a cluster are allowed and each node is only divided once. This division not only improves the timeliness but also ensures the effectiveness of the division, reducing the impact of the first-layer division results on subsequent divisions.

[0104] The second level of division is based on the improved Louvain algorithm division of electrical characteristics. By combining the improved modularity gain of reactive balance, the traditional modularity gain does not calculate the modularity gain brought by the node removal when the node moves, which will lead to poor subsequent division effect. The present invention improves this and introduces reactive balance for optimization. Reactive balance is an indicator to measure the balance degree of reactive power in the power system. It indicates whether the distribution of reactive power in the power system is balanced, so that the division effect is improved.

[0105] The present invention uses a comprehensive index system based on electrical characteristics and constructs two similarity matrices based on position characteristics for hierarchical division. The present invention can also ensure the effectiveness and rationality of the dynamic division of distributed photovoltaic clusters and improve the timeliness of the dynamic division algorithm of distributed photovoltaic clusters.

[0106] The purpose of the present invention is to achieve reasonable division of distributed photovoltaic clusters and improve the efficiency of cluster control scheduling; at the same time, the timeliness of the distributed photovoltaic cluster dynamic division algorithm is increased, and the effectiveness of short-time time scale control scheduling is improved. The present invention provides a distributed photovoltaic cluster dynamic division algorithm, which designs a division similarity matrix between distributed photovoltaic nodes based on a comprehensive index system of electrical characteristics, and a Jaccard similarity matrix based on position characteristics, and combines the electrical characteristics of distributed photovoltaics with the position characteristics for cluster division; at the same time, the algorithm improves the Louvain algorithm in layers, and finally realizes the dynamic division of distributed photovoltaic clusters by optimizing the modularity and reactive balance of the cluster.

[0107] The distributed photovoltaic cluster dynamic partitioning method provided by the present invention performs hierarchical partitioning by using a comprehensive indicator system based on electrical characteristics and constructing two similarity matrices based on position characteristics respectively; the timeliness of the Louvain algorithm is improved by hierarchical improvement of the Louvain algorithm, wherein the first-layer partitioning is based on the position characteristics, and at the same time, it is considered that the partitioning only allows two nodes in a cluster and each node is only partitioned once, so that the timeliness is improved while also ensuring the effectiveness of the partitioning, and reducing the influence of the first-layer partitioning result on the subsequent partitioning; the second-layer partitioning is based on the improved Louvain algorithm partitioning of electrical characteristics, and by combining the improved modularity gain of the reactive balance and introducing the reactive balance for optimization, the partitioning effect is improved, the effectiveness and rationality of the dynamic partitioning of the distributed photovoltaic cluster is ensured, and at the same time, the timeliness of the dynamic partitioning algorithm of the distributed photovoltaic cluster is improved, and the stability of the power generation work of the power grid is improved.

[0108] In some optional embodiments, the construction of a partition similarity matrix of distributed photovoltaic nodes based on the comprehensive index of electrical characteristics includes:

[0109] Normalizing the active regulation capacity, reactive regulation capacity, load rate and net load of each photovoltaic node to obtain a normalized regulation capacity index;

[0110] Based on the normalized index of the regulation capacity, the electrical distance matrix based on reactive sensitivity is fused to construct the partition similarity matrix;

[0111] Among them, the electrical distance matrix based on reactive sensitivity is expressed as: d ij represents the electrical distance between node i and node j, and the element S in the reactive sensitivity matrix S ij The element S in the reactive sensitivity matrix S represents the change in the voltage of node i corresponding to the unit reactive power injected into node j. jjIt represents the change value of node j voltage corresponding to the unit reactive power injected into node j. The relevant expression of reactive sensitivity matrix S is ΔV=SΔQ, where ΔV represents the change amount of voltage amplitude and ΔQ represents the change amount of reactive power.

[0112] Active power regulation capacity P of PV node i i Expressed as: P i ∈[0,P max ], P max Indicates the active output of the photovoltaic power generation system under the control of the maximum power point tracking system;

[0113] Reactive power regulation capacity Q of PV node i j It is expressed as:

[0114]

[0115] S max is the grid-connected capacity of the photovoltaic inverter corresponding to the photovoltaic cluster, P max represents the active output of the photovoltaic power generation system where the photovoltaic cluster is located under the control of the maximum power point tracking system, λ max Indicates the upper limit of the power factor, λ min Indicates the lower limit of the power factor.

[0116] Specifically, the steps for constructing the partition similarity matrix of distributed photovoltaic nodes based on the comprehensive indicator system of electrical characteristics are as follows:

[0117] 1) The comprehensive indicators based on electrical characteristics mainly include five parts, namely, the electrical distance matrix based on reactive sensitivity, active regulation capacity, reactive regulation capacity, load rate and net load.

[0118] 2) Calculate the reactive sensitivity matrix using the following formula:

[0119] ΔV=SΔQ (1)

[0120] Among them, ΔV represents the change in voltage amplitude; ΔQ represents the change in reactive power; S is the sensitivity matrix, and S in the matrix ij The change in voltage of node i corresponding to the unit reactive power injected into node j is expressed as follows:

[0121]

[0122] Among them, d ij Represents the electrical distance between node i and node j. The electrical distance is used to characterize the degree of electrical coupling between two nodes. The closer the distance, the more similar the electrical characteristics of the two nodes are, and the farther the distance, the greater the difference in the electrical characteristics of the two nodes.

[0123] 3) Specifically, the existence of active power regulation capacity can adjust the active power output of the generator to maintain the power balance and frequency stability of the power system. The active power regulation capacity is calculated by the following formula:

[0124] P i ∈[0,P max ](3)

[0125] Among them, P i ∈[0,P max ] represents the active regulation capacity of PV node i, P max Indicates the active output of the photovoltaic power generation system under the control of the maximum power point tracking system;

[0126] 4) The existence of reactive power regulation capacity can adjust the reactive power output of the generator to maintain the power balance and frequency stability of the power system. The reactive power regulation capacity is calculated by the following formula:

[0127]

[0128] Among them, Q j represents the reactive power regulation capacity of PV node j, S max is the grid-connected capacity of the photovoltaic inverter, P max represents the active output of the photovoltaic power generation system under the control of the maximum power point tracking system, λ max and λ min Respectively represent the upper and lower limits of the power factor.

[0129] The regulation capacity can adjust the output power of the generator according to the needs of load changes to meet the load requirements of the power system. When a system failure occurs, the regulation capacity can provide additional power support to maintain the stable operation of the power system.

[0130] 5) The photovoltaic node load rate reflects the power generation efficiency of the photovoltaic node within a certain time range and is used to describe the degree to which the photovoltaic system meets the node load. The formula for calculating the load rate is as follows:

[0131] L r,i =L d,i / L f,i (5)

[0132] Among them, L r,i represents the load rate of PV node i, L d,i represents the actual power generation of photovoltaic node i, L f,i Represents the maximum power generation of PV node i.

[0133] 6) The net load of the photovoltaic node takes into account the contribution of photovoltaic power generation, which involves the power balance of the photovoltaic power generation system and its impact on the overall power system. The formula for calculating the net load is as follows:

[0134] H i =H o,i -H l,i (6)

[0135] Among them, H i represents the net load of PV node i, H o,i represents the photovoltaic output of photovoltaic node i, H l,i Represents the load of PV node i.

[0136] In some optional embodiments, the active regulation capacity, reactive regulation capacity, load rate and net load of each photovoltaic node are normalized to obtain a normalized regulation capacity index, including:

[0137] By using the first formula, the active regulation capacity, reactive regulation capacity, load rate and net load of each photovoltaic node are normalized to obtain a normalized index of the regulation capacity of each photovoltaic node;

[0138] The first formula is expressed as:

[0139]

[0140] Among them, x i,m ′ represents the normalized index of the mth regulation capacity of PV node i, x i,m represents the mth index of photovoltaic node i, which includes active regulation capacity, reactive regulation capacity, load rate and net load, x 1 / 2,m represents the median value of the mth indicator, x 1 / 4,m Represents the one-quarter digit value of the m-th indicator.

[0141] Specifically, when constructing the partition similarity matrix of distributed photovoltaic nodes, the active regulation capacity, reactive regulation capacity, load rate and net load can be normalized first:

[0142]

[0143] Among them, x i,m ′ represents the normalized mth index of PV node i, x i,m represents the mth index of photovoltaic node i, x 1 / 2,m represents the median value of the mth indicator, x 1 / 4,m Represents the one-quarter digit value of the m-th index.

[0144] In some optional embodiments, based on the normalized index of the regulation capacity, the electrical distance matrix based on reactive sensitivity is fused to construct the partition similarity matrix, including:

[0145] Based on the regulation capacity normalization index, the electrical distance matrix based on reactive sensitivity is fused through a second formula to obtain the partition similarity matrix;

[0146] The second formula is expressed as:

[0147]

[0148] Among them, x i,m represents the mth indicator of photovoltaic node i, which may be active regulation capacity, reactive regulation capacity, load rate and net load, s ij Represents the element in the i-th row and j-th column of the similarity matrix, α m represents the weight of the mth indicator, x i,m ′ represents the normalized index of the mth regulation capacity of PV node i, x j,m ′ represents the normalized index of the mth regulation capacity of PV node j, d ij represents the electrical distance between node i and node j based on reactive sensitivity, d max represents the maximum value in the electrical distance matrix based on reactive sensitivity, λ 1 With λ 2 Represent the weights of comprehensive index and electrical distance respectively.

[0149] Specifically, when constructing the partition similarity matrix of distributed photovoltaic nodes, after normalizing the active regulation capacity, reactive regulation capacity, load rate and net load, the electrical distance matrix based on reactive sensitivity can be fused to construct the partition similarity matrix. Specifically, when the electrical distance based on reactive sensitivity is fused to construct the partition similarity matrix, the fusion method is as follows:

[0150]

[0151] Among them, s ij Represents the element in the i-th row and j-th column of the similarity matrix, α m represents the weight of the mth indicator, x i,m ′ represents the normalized mth index of PV node i, x j,m ′ represents the normalized mth index of PV node j, d ij represents the electrical distance between node i and node j based on reactive sensitivity, d max represents the maximum value in the electrical distance matrix based on reactive sensitivity, λ 1 With λ 2Represent the weights of comprehensive index and electrical distance respectively.

[0152] When calculating the division similarity above, the normalized active regulation capacity and reactive regulation capacity are vector dot product to reflect the similarity of the active regulation capacity and reactive regulation capacity indicators of the two nodes corresponding to the i-th row and j-th column element of the similarity matrix; since the larger the electrical distance, the smaller the similarity between the two nodes, the electrical distance is negative; the range of similarity is generally between 0 and 1, so the calculated similarity indicators are all greater than their maximum values.

[0153] In some optional embodiments, the Jaccard similarity matrix of the distributed photovoltaic nodes is constructed based on the adjacency relationship of the distributed photovoltaic nodes, including:

[0154] Based on the adjacency relationship of the distributed photovoltaic nodes, a Jaccard similarity matrix of the distributed photovoltaic nodes is constructed by a third formula;

[0155] The third formula is expressed as:

[0156]

[0157] Among them, Sim jaccard，ij Represents the element in the i-th row and j-th column of the Jaccard similarity matrix, Sim jaccard，ij represents the Jaccard similarity between photovoltaic node i and photovoltaic node j, Y i represents the neighbor node set of distributed photovoltaic node i, Y j represents the neighbor node set of distributed photovoltaic node j, Ψ(Y i , Y j ) means when Y i With Y j If the intersection is not empty, the modulus of the intersection is taken; otherwise, the value is 1.

[0158] Specifically, the Jaccard similarity matrix of distributed photovoltaic nodes is constructed by combining the adjacency relationship of distributed photovoltaic nodes, that is, the location characteristics. The calculation method is as follows:

[0159]

[0160] Among them, Sim jaccard，ij represents the Jaccard similarity matrix of distributed photovoltaic nodes, the Jaccard similarity between node i and node j, Y i represents the neighbor node set of distributed photovoltaic node i, Y j represents the neighbor node set of distributed photovoltaic node j, Ψ(Y i , Y j ) means when Y i With Yj If the intersection is not empty, take the modulus of the intersection; otherwise, take 1.

[0161] Photovoltaic nodes with similar geographical locations have similar electrical characteristics. The present invention utilizes this to perform a rough division and then a detailed division to improve the efficiency and timeliness of the division.

[0162] In some optional embodiments, the step of constructing a cluster network with the at least one cluster as at least one network node and obtaining a second partition similarity matrix of the at least one network node includes:

[0163] Using the at least one cluster as at least one network node, constructing the cluster network;

[0164] Using a zero matrix as an initial similarity matrix of the cluster network;

[0165] Based on the sum of the pairwise similarities of the nodes in the at least one cluster, updating the diagonal elements of the initial similarity matrix;

[0166] Based on the fourth formula, calculating the pairwise partition similarity of at least one cluster, updating the initial similarity matrix, and obtaining a second partition similarity matrix;

[0167] The fourth formula is expressed as:

[0168]

[0169] in, represents the partition similarity between cluster t and cluster c in at least one cluster. Distributed photovoltaic node i and distributed photovoltaic node j are nodes in cluster t and cluster c respectively. S i,j Represents the partition similarity between distributed photovoltaic nodes i and j.

[0170] Figure 3 It is a flow chart of the cluster network construction method provided by the present invention, such as Figure 3 As shown in the figure, the specific steps for reconstructing the distributed photovoltaic node network based on the first-layer division results are as follows:

[0171] 1) Obtain all distributed photovoltaic clusters obtained after the first layer of division, build a new network (i.e., cluster network) and the partition similarity matrix of the network, i.e., the second partition similarity matrix. Each node of the cluster network is represented as a cluster of the first layer of division results (i.e., the divided cluster i is abstracted as a new node j). The second partition similarity matrix is ​​initialized to a zero matrix.

[0172] 2) The sum of the similarities between any two nodes in the cluster is assigned to the diagonal of the new partition similarity matrix; that is, for each cluster i (node ​​j in the cluster network) after partition, the sum of the similarities between any two nodes in cluster i is taken as the self-loop weight edge of the new node j.

[0173] 3) Traverse each cluster and calculate the similarity between the cluster and other clusters according to the following formula:

[0174]

[0175] in, represents the new partition similarity between cluster t and cluster c. Distributed photovoltaic nodes i and j are nodes in cluster t and cluster c respectively. S i,j Represents the partition similarity between distributed photovoltaic nodes i and j.

[0176] 4) The partition similarity between each cluster is corresponded to the partition similarity between each node of the new network (i.e., cluster network).

[0177] In some optional embodiments, the method of performing distributed cluster division on the plurality of distributed photovoltaic nodes based on the second partition similarity matrix and combining the improved modularity gain of reactive power balance using the Louvain algorithm to obtain the division results of the plurality of distributed photovoltaic nodes includes:

[0178] Perform the first process at least once until the division results of the multiple distributed photovoltaic nodes obtained in the two most recent first processes remain unchanged, and use the division result of the multiple distributed photovoltaic nodes obtained in the most recent first process as the division result of the multiple distributed photovoltaic nodes;

[0179] Wherein, the first process once includes:

[0180] Traversing multiple distributed photovoltaic nodes until the second process is performed once for each distributed photovoltaic node;

[0181] The second process includes:

[0182] For photovoltaic node i, calculating the improved modularity gain of the reactive power balance degree of each photovoltaic node adjacent to the photovoltaic node i;

[0183] Select the photovoltaic node j corresponding to the maximum value of the improved modularity gain of the reactive power balance degree of the photovoltaic node i to each photovoltaic node adjacent to the photovoltaic node i, and add the photovoltaic node i to the cluster where the photovoltaic node j is located;

[0184] Among them, the improved modularity gain of reactive power balance is expressed as: ΔM new =β1 ΔM+β 2 Δω; ΔM represents the improved modularity gain of the mobile node, Δω represents the reactive power balance gain of the mobile node, β 1 With β 2 Represent the weights of improved modularity gain and reactive power balance gain respectively.

[0185] Specifically, the improved modularity gain combined with reactive balance is composed of reactive balance gain and improved modularity gain; wherein, the calculation formula of improved modularity gain is as follows:

[0186]

[0187] Where ΔM represents the improved modularity gain, cluster t and cluster c are the cluster where the i distributed photovoltaic node originally belongs and the cluster to be moved to, respectively, s t represents the sum of the partition similarities of all distributed photovoltaic nodes in cluster t, s i represents the sum of the partition similarities between node i and other distributed photovoltaic nodes, s i，t represents the sum of the partition similarities from node i to the distributed photovoltaic nodes in cluster t;

[0188] The reactive balance calculation is shown as follows:

[0189]

[0190]

[0191] Among them, ω sup,i represents the maximum reactive power output of cluster i, ω need,i represents the reactive power demand value of cluster i, ω i represents the reactive balance of cluster i, c represents the total number of clusters, and ω represents the reactive balance of the overall PV node network.

[0192] The improved modularity gain calculation formula combined with reactive balance is as follows:

[0193] ΔM new =β 1 ΔM+β 2 Δω (14)

[0194] Among them, ΔM new represents the improved modularity gain combined with reactive power balance, Δω represents the reactive power balance gain of the mobile node, β 1 With β 2 Respectively represent the weights of improved modularity gain and reactive power balance gain;

[0195] The traditional modularity gain of the Louvain algorithm does not calculate the modularity gain brought by the node removal when the node moves, which will lead to poor subsequent partitioning effect and instability. The improved modularity gain increases the stability and efficiency of subsequent partitioning.

[0196] Reactive power balance plays an important role and significance in evaluating system stability, predicting potential problems, optimizing reactive power distribution, guiding power system operation strategies and energy efficiency evaluation. Adding reactive power balance to the optimization index for optimization will help improve the reliability, stability and operation quality of the power system.

[0197] Figure 4 The flowchart of the present invention is to use the Louvain algorithm for the new network in combination with the improved modularity gain of reactive balance; specifically, a new distributed photovoltaic node network is calculated, and the new network is divided into distributed photovoltaic clusters using the Louvain algorithm based on the partition similarity matrix using the improved modularity gain combined with reactive balance. The specific steps are as follows:

[0198] 1) Traverse and select a distributed photovoltaic node i to calculate the improved modularity gain combined with reactive power balance for all neighboring distributed photovoltaic nodes of the node;

[0199] 2) For node i, select the node j with the largest improved modularity gain combined with reactive balance, add node i to the distributed photovoltaic cluster where node j is located, and complete the second process of node i. Repeat step 1) until all distributed photovoltaic nodes are selected once, and complete the first process once;

[0200] 3) If all distributed photovoltaic nodes are traversed once and no node movement occurs, the cluster partitioning algorithm ends, otherwise the distributed photovoltaic node network is reconstructed and the process goes to step 1);

[0201] 4) The cluster division based on the Louvain algorithm with improved modularity gain is completed, and the final distributed photovoltaic cluster division result is obtained.

[0202] In one embodiment, Figure 5 This is a schematic diagram of the network structure of the data set provided by the present invention. In order to verify the effectiveness and timeliness of the improved hierarchical Louvain algorithm and simplify the experimental steps, the present invention is based on Matlab Figure 5 The adjacency matrix distribution of the network structure shown tests the Louvain algorithm and the improved hierarchical Louvain algorithm, where the weights of the edges are all set to 1, and the program outputs the running time, the list of partitioned clusters, and the modularity.

[0203] The Louvain algorithm and the improved hierarchical Louvain algorithm were tested 10 times, and the cluster modularity and running time of each test result were counted. The test results are shown in Table 1:

[0204] Table 1 Test results

[0205] Partitioning Algorithm Average modularity Average running time Louvain algorithm 0.40206 0.92651 Improved hierarchical Louvain algorithm 0.40549 0.45216

[0206] After analyzing the results of 10 test partitions, the improved hierarchical Louvain algorithm increased the average modularity by 0.243%, making the clusters more compact after partitioning, and reducing the average running time by nearly half, proving that the hierarchical improved Louvain algorithm ensures the efficiency of cluster partitioning while greatly increasing the timeliness. The number of network nodes used in this test case is 34, and the improvement in timeliness will be more significant for larger networks.

[0207] Figure 6 A schematic diagram of the Louvain algorithm provided by the present invention for dividing the network structure; Figure 7 The schematic diagram of the improved hierarchical Louvain algorithm provided by the present invention for dividing the network structure is shown in FIG. The optimal division results of the Louvain algorithm and the improved hierarchical Louvain algorithm for 10 test division results are analyzed. Figure 6 , Figure 7 As shown in Table 2:

[0208] Table 2 Experimental partition results

[0209]

[0210] The number of clusters divided by the two algorithms is 4 clusters. The difference in the division results lies in node 10. Node 10 of the Louvain algorithm is in cluster 1, and the improved hierarchical Louvain algorithm is in cluster 4. Due to the movement of node 10, the improved hierarchical Louvain algorithm improves the modularity by 0.091% compared with the Louvain algorithm, making the cluster structure more coupled internally, and reducing the running time by more than half. Compared with the average modularity, the improvement of the optimal result in modularity is reduced, and the running time is increased, which proves that the division results of the improved hierarchical Louvain algorithm are stable.

[0211] Therefore, the algorithm of the present invention greatly reduces the running time of the algorithm for dynamic partitioning while ensuring that the cluster structure of the cluster partitioning result has strong internal coupling and weak external coupling, and improves the potential for short-time scale control and scheduling of distributed photovoltaic clusters. Short-time scale control scheduling is achieved, and real-time and fast distributed photovoltaic cluster partitioning is achieved, and control and scheduling efficiency is improved. While reducing the running time of the distributed photovoltaic cluster partitioning algorithm, the internal strong coupling and external weak coupling of the distributed photovoltaic cluster are improved, and the working effect of cluster technology is further improved.

[0212] The distributed photovoltaic cluster dynamic division device provided by the present invention is described below. The distributed photovoltaic cluster dynamic division device described below and the distributed photovoltaic cluster dynamic division method described above can be referred to each other.

[0213] Figure 8 is a schematic diagram of the structure of the distributed photovoltaic cluster dynamic partitioning device provided by the present invention, such as Figure 8 As shown, the device 800 includes:

[0214] A first construction module 810 is used to construct a first partition similarity matrix of multiple distributed photovoltaic nodes based on a comprehensive index of electrical characteristics, wherein the comprehensive index of electrical characteristics includes at least one of the following: an electrical distance matrix based on reactive sensitivity, active regulation capacity, reactive regulation capacity, load rate, and net load;

[0215] An adjacency relationship acquisition module 820, configured to obtain the adjacency relationship of the distributed photovoltaic nodes based on the first partition similarity matrix;

[0216] A second construction module 830 is used to construct a Jaccard similarity matrix of the distributed photovoltaic nodes based on the adjacency relationship of the distributed photovoltaic nodes;

[0217] A first partitioning module 840 is configured to perform a first-layer partitioning of the distributed photovoltaic nodes based on the Jaccard similarity matrix to obtain at least one cluster; wherein different clusters include different distributed photovoltaic nodes, each cluster includes two distributed photovoltaic nodes, one of the two distributed photovoltaic nodes is a distributed photovoltaic node with the largest Jaccard similarity to the other distributed photovoltaic node, and the number of common neighbor nodes owned by the two distributed photovoltaic nodes is greater than or equal to 2;

[0218] A third construction module 850 is used to construct a cluster network with the at least one cluster as at least one network node, and obtain a second partition similarity matrix of the at least one network node;

[0219] The second partitioning module 860 is used to perform distributed cluster partitioning on the multiple distributed photovoltaic nodes using the Louvain algorithm based on the second partitioning similarity matrix and the improved modularity gain of the reactive balance degree to obtain the partitioning results of the multiple distributed photovoltaic nodes.

[0220] It should be noted that the distributed photovoltaic cluster dynamic division device of the present invention can implement the various embodiments of the distributed photovoltaic cluster dynamic division method described above and achieve the same technical effects, which will not be described in detail herein.

[0221] The distributed photovoltaic cluster dynamic partitioning device provided by the present invention performs hierarchical partitioning by using a comprehensive indicator system based on electrical characteristics and constructing two similarity matrices based on position characteristics respectively; the timeliness of the Louvain algorithm is improved by hierarchical improvement of the Louvain algorithm, wherein the first layer of partitioning is based on position characteristics, and at the same time, it is considered that the partitioning only allows two nodes in a cluster and each node is only partitioned once, so that the timeliness is improved while also ensuring the effectiveness of the partitioning, and reducing the influence of the first layer of partitioning results on subsequent partitioning; the second layer of partitioning is based on the improved Louvain algorithm partitioning of electrical characteristics, and by combining the improved modularity gain of the reactive balance and introducing the reactive balance for optimization, the partitioning effect is improved, the effectiveness and rationality of the dynamic partitioning of the distributed photovoltaic cluster is ensured, and the timeliness of the dynamic partitioning algorithm of the distributed photovoltaic cluster is also improved.

[0222] Fig. 9 An example of a physical structure diagram of an electronic device is shown in FIG. Fig. 9 As shown, the electronic device may include: a processor 910, a communication interface 920, a memory 930 and a communication bus 940, wherein the processor 910, the communication interface 920 and the memory 930 communicate with each other through the communication bus 940. The processor 910 may call the logic instructions in the memory 930 to execute the distributed photovoltaic cluster dynamic division method, including:

[0223] Based on the comprehensive index of electrical characteristics, constructing a first partition similarity matrix of multiple distributed photovoltaic nodes, the comprehensive index of electrical characteristics including at least one of the following: an electrical distance matrix based on reactive sensitivity, active regulation capacity, reactive regulation capacity, load rate and net load;

[0224] Based on the first partition similarity matrix, obtaining the adjacency relationship of the distributed photovoltaic nodes;

[0225] Based on the adjacency relationship of the distributed photovoltaic nodes, constructing a Jaccard similarity matrix of the distributed photovoltaic nodes;

[0226] Based on the Jaccard similarity matrix, the distributed photovoltaic nodes are divided into a first layer to obtain at least one cluster; wherein different clusters include different distributed photovoltaic nodes, each cluster includes two distributed photovoltaic nodes, one of the two distributed photovoltaic nodes is a distributed photovoltaic node with the largest Jaccard similarity to the other distributed photovoltaic node, and the number of common neighbor nodes owned by the two distributed photovoltaic nodes is greater than or equal to 2;

[0227] constructing a cluster network with the at least one cluster as at least one network node, and obtaining a second partition similarity matrix of the at least one network node;

[0228] Based on the second partition similarity matrix and in combination with the improved modularity gain of reactive power balance, the Louvain algorithm is used to perform distributed cluster partitioning on the multiple distributed photovoltaic nodes to obtain a partitioning result of the multiple distributed photovoltaic nodes.

[0229] In addition, the logic instructions in the above-mentioned memory 930 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0230] On the other hand, the present invention further provides a computer program product, the computer program product includes a computer program, the computer program can be stored in a non-transitory computer-readable storage medium, and when the computer program is executed by a processor, the computer can execute the distributed photovoltaic cluster dynamic division method provided by the above methods, including:

[0231] Based on the comprehensive index of electrical characteristics, constructing a first partition similarity matrix of multiple distributed photovoltaic nodes, the comprehensive index of electrical characteristics including at least one of the following: an electrical distance matrix based on reactive sensitivity, active regulation capacity, reactive regulation capacity, load rate and net load;

[0232] Based on the first partition similarity matrix, obtaining the adjacency relationship of the distributed photovoltaic nodes;

[0233] Based on the adjacency relationship of the distributed photovoltaic nodes, constructing a Jaccard similarity matrix of the distributed photovoltaic nodes;

[0234] Based on the Jaccard similarity matrix, the distributed photovoltaic nodes are divided into a first layer to obtain at least one cluster; wherein different clusters include different distributed photovoltaic nodes, each cluster includes two distributed photovoltaic nodes, one of the two distributed photovoltaic nodes is a distributed photovoltaic node with the largest Jaccard similarity to the other distributed photovoltaic node, and the number of common neighbor nodes owned by the two distributed photovoltaic nodes is greater than or equal to 2;

[0235] constructing a cluster network with the at least one cluster as at least one network node, and obtaining a second partition similarity matrix of the at least one network node;

[0236] Based on the second partition similarity matrix and in combination with the improved modularity gain of reactive power balance, the Louvain algorithm is used to perform distributed cluster partitioning on the multiple distributed photovoltaic nodes to obtain a partitioning result of the multiple distributed photovoltaic nodes.

[0237] In another aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which is implemented when the computer program is executed by a processor to execute the distributed photovoltaic cluster dynamic division method provided by the above methods, including:

[0238] Based on the comprehensive index of electrical characteristics, constructing a first partition similarity matrix of multiple distributed photovoltaic nodes, the comprehensive index of electrical characteristics including at least one of the following: an electrical distance matrix based on reactive sensitivity, active regulation capacity, reactive regulation capacity, load rate and net load;

[0239] Based on the first partition similarity matrix, obtaining the adjacency relationship of the distributed photovoltaic nodes;

[0240] Based on the adjacency relationship of the distributed photovoltaic nodes, constructing a Jaccard similarity matrix of the distributed photovoltaic nodes;

[0241] Based on the Jaccard similarity matrix, the distributed photovoltaic nodes are divided into a first layer to obtain at least one cluster; wherein different clusters include different distributed photovoltaic nodes, each cluster includes two distributed photovoltaic nodes, one of the two distributed photovoltaic nodes is a distributed photovoltaic node with the largest Jaccard similarity to the other distributed photovoltaic node, and the number of common neighbor nodes owned by the two distributed photovoltaic nodes is greater than or equal to 2;

[0242] constructing a cluster network with the at least one cluster as at least one network node, and obtaining a second partition similarity matrix of the at least one network node;

[0243] Based on the second partition similarity matrix and in combination with the improved modularity gain of reactive power balance, the Louvain algorithm is used to perform distributed cluster partitioning on the multiple distributed photovoltaic nodes to obtain a partitioning result of the multiple distributed photovoltaic nodes.

[0244] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0245] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0246] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A distributed photovoltaic cluster dynamic division method, characterized in that: include: Based on the comprehensive index of electrical characteristics, constructing a first partition similarity matrix of multiple distributed photovoltaic nodes, wherein the comprehensive index of electrical characteristics includes at least one of the following: an electrical distance matrix based on reactive sensitivity, active regulation capacity, reactive regulation capacity, load rate and net load; Based on the first partition similarity matrix, obtaining the adjacency relationship of the distributed photovoltaic nodes; Based on the adjacency relationship of the distributed photovoltaic nodes, constructing a Jaccard similarity matrix of the distributed photovoltaic nodes; Based on the Jaccard similarity matrix, the distributed photovoltaic nodes are divided into a first layer to obtain at least one cluster; wherein different clusters include different distributed photovoltaic nodes, each cluster includes two distributed photovoltaic nodes, one of the two distributed photovoltaic nodes is a distributed photovoltaic node with the largest Jaccard similarity to the other distributed photovoltaic node, and the number of common neighbor nodes owned by the two distributed photovoltaic nodes is greater than or equal to 2; constructing a cluster network with the at least one cluster as at least one network node, and obtaining a second partition similarity matrix of the at least one network node; Based on the second partition similarity matrix and in combination with the improved modularity gain of reactive power balance, the Louvain algorithm is used to perform distributed cluster partitioning on the multiple distributed photovoltaic nodes to obtain a partitioning result of the multiple distributed photovoltaic nodes.

2. The distributed photovoltaic cluster dynamic division method according to claim 1, characterized in that: The comprehensive index based on electrical characteristics is used to construct a partition similarity matrix of distributed photovoltaic nodes, including: Normalizing the active regulation capacity, reactive regulation capacity, load rate and net load of each photovoltaic node to obtain a normalized regulation capacity index; Based on the normalized index of the regulation capacity, the electrical distance matrix based on reactive sensitivity is fused to construct the partition similarity matrix; Among them, the electrical distance matrix based on reactive sensitivity is expressed as: d ij represents the electrical distance between node i and node j, and the element S in the reactive sensitivity matrix S ij The element S in the reactive sensitivity matrix S represents the change in the voltage of node i corresponding to the unit reactive power injected into node j. jj It represents the change value of node j voltage corresponding to the unit reactive power injected into node j. The relevant expression of reactive sensitivity matrix S is ΔV=SΔQ, where ΔV represents the change amount of voltage amplitude and ΔQ represents the change amount of reactive power. Active power regulation capacity P of PV node i i Expressed as: P i ∈[0,P max ], P max Indicates the active output of the photovoltaic power generation system under the control of the maximum power point tracking system; Reactive power regulation capacity Q of PV node i j It is expressed as: S max is the grid-connected capacity of the photovoltaic inverter corresponding to the photovoltaic cluster, P max represents the active output of the photovoltaic power generation system where the photovoltaic cluster is located under the control of the maximum power point tracking system, λ max Indicates the upper limit of the power factor, λ min Indicates the lower limit of the power factor.

3. The distributed photovoltaic cluster dynamic division method according to claim 2 is characterized in that: The active regulation capacity, reactive regulation capacity, load rate and net load of each photovoltaic node are normalized to obtain a normalized regulation capacity index, including: By using the first formula, the active regulation capacity, reactive regulation capacity, load rate and net load of each photovoltaic node are normalized to obtain a normalized index of the regulation capacity of each photovoltaic node; The first formula is expressed as: Among them, x i,m ′ represents the normalized index of the mth regulation capacity of PV node i, x i,m represents the mth index of photovoltaic node i, which includes active regulation capacity, reactive regulation capacity, load rate and net load, x 1 / 2,m represents the median value of the mth indicator, x 1 / 4,m Represents the one-quarter digit value of the m-th indicator.

4. The distributed photovoltaic cluster dynamic division method according to claim 3 is characterized in that: Based on the normalized index of the regulation capacity, the electrical distance matrix based on reactive sensitivity is fused to construct the partition similarity matrix, including: Based on the regulation capacity normalization index, the electrical distance matrix based on reactive sensitivity is fused through a second formula to obtain the partition similarity matrix; The second formula is expressed as: Among them, s ij Represents the element in the i-th row and j-th column of the similarity matrix, α m represents the weight of the mth indicator, x i,m ′ represents the normalized index of the mth regulation capacity of PV node i, x j,m ′ represents the normalized index of the mth regulation capacity of PV node j, d ij represents the electrical distance between node i and node j based on reactive sensitivity, d max It represents the maximum value in the electrical distance matrix based on reactive sensitivity, and λ1 and λ2 represent the weights of the comprehensive index and electrical distance, respectively.

5. The distributed photovoltaic cluster dynamic division method according to claim 1, characterized in that: The Jaccard similarity matrix of the distributed photovoltaic nodes is constructed based on the adjacency relationship of the distributed photovoltaic nodes, including: Based on the adjacency relationship of the distributed photovoltaic nodes, a Jaccard similarity matrix of the distributed photovoltaic nodes is constructed by a third formula; The third formula is expressed as: Among them, Sim jaccard，ij Represents the element in the i-th row and j-th column of the Jaccard similarity matrix, Sim jaccard，ij represents the Jaccard similarity between photovoltaic node i and photovoltaic node j, Y i represents the neighbor node set of distributed photovoltaic node i, Y j represents the neighbor node set of distributed photovoltaic node j, Ψ(Y i , Y j ) means when Y i With Y j If the intersection is not empty, the modulus of the intersection is taken; otherwise, the value is 1.

6. The distributed photovoltaic cluster dynamic division method according to claim 1, characterized in that: The constructing a cluster network using the at least one cluster as at least one network node and obtaining a second partition similarity matrix of the at least one network node includes: Using the at least one cluster as at least one network node, constructing the cluster network; Using a zero matrix as an initial similarity matrix of the cluster network; Based on the sum of the pairwise similarities of the nodes in the at least one cluster, updating the diagonal elements of the initial similarity matrix; Based on the fourth formula, calculating the pairwise partition similarity of at least one cluster, updating the initial similarity matrix, and obtaining a second partition similarity matrix; The fourth formula is expressed as: in, represents the partition similarity between cluster t and cluster c in at least one cluster. Distributed photovoltaic node i and distributed photovoltaic node j are nodes in cluster t and cluster c respectively. S i,j Represents the partition similarity between distributed photovoltaic nodes i and j.

7. The distributed photovoltaic cluster dynamic division method according to claim 1, characterized in that: The method of performing distributed cluster division on the plurality of distributed photovoltaic nodes based on the second partition similarity matrix and combining the improved modularity gain of reactive power balance using the Louvain algorithm to obtain the division results of the plurality of distributed photovoltaic nodes includes: Perform the first process at least once until the division results of the multiple distributed photovoltaic nodes obtained in the two most recent first processes remain unchanged, and use the division result of the multiple distributed photovoltaic nodes obtained in the most recent first process as the division result of the multiple distributed photovoltaic nodes; Wherein, the first process once includes: Traversing multiple distributed photovoltaic nodes until the second process is performed once for each distributed photovoltaic node; The second process includes: For photovoltaic node i, calculating the improved modularity gain of the reactive power balance degree of each photovoltaic node adjacent to the photovoltaic node i; Select the photovoltaic node j corresponding to the maximum value of the improved modularity gain of the reactive power balance degree of the photovoltaic node i to each photovoltaic node adjacent to the photovoltaic node i, and add the photovoltaic node i to the cluster where the photovoltaic node j is located; Among them, the improved modularity gain of reactive power balance is expressed as: ΔM new =β1ΔM+β2Δω; ΔM represents the improved modularity gain of the mobile node, Δω represents the reactive power balance gain of the mobile node, and β1 and β2 represent the weights of the improved modularity gain and the reactive power balance gain, respectively.

8. A distributed photovoltaic cluster dynamic division device, characterized in that: include: A first construction module is used to construct a first partition similarity matrix of multiple distributed photovoltaic nodes based on a comprehensive index of electrical characteristics, wherein the comprehensive index of electrical characteristics includes at least one of the following: an electrical distance matrix based on reactive sensitivity, active regulation capacity, reactive regulation capacity, load rate, and net load; An adjacency relationship acquisition module, configured to obtain the adjacency relationship of the distributed photovoltaic nodes based on the first partition similarity matrix; A second construction module is used to construct a Jaccard similarity matrix of the distributed photovoltaic nodes based on the adjacency relationship of the distributed photovoltaic nodes; A first partitioning module is used to perform a first-layer partitioning of the distributed photovoltaic nodes based on the Jaccard similarity matrix to obtain at least one cluster; wherein different clusters include different distributed photovoltaic nodes, each cluster includes two distributed photovoltaic nodes, one of the two distributed photovoltaic nodes is a distributed photovoltaic node with the largest Jaccard similarity to the other distributed photovoltaic node, and the number of common neighbor nodes owned by the two distributed photovoltaic nodes is greater than or equal to 2; A third construction module is used to construct a cluster network with the at least one cluster as at least one network node, and obtain a second partition similarity matrix of the at least one network node; The second partitioning module is used to perform distributed cluster partitioning on the multiple distributed photovoltaic nodes based on the second partitioning similarity matrix and the improved modularity gain of the reactive balance degree using the Louvain algorithm to obtain the partitioning results of the multiple distributed photovoltaic nodes.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the steps of the method for dynamically dividing a distributed photovoltaic cluster as claimed in any one of claims 1 to 7 are implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for dynamically dividing a distributed photovoltaic cluster as claimed in any one of claims 1 to 7 are implemented.

Images