A method and system for dynamic partitioning of active distribution network

Abstract

A method and system for dynamic partitioning of an active distribution network. When the network topology of the distribution network changes or the source and load fluctuate extremely, the method first performs preliminary partitioning on the active distribution network, determines the overlapping areas in the network, and obtains the overlapping nodes with interactive influence between the areas; then the overlapping nodes in the overlapping areas are partitioned and adjusted to obtain the area redivision result, thereby ensuring the requirement of tight coupling within the area and sparse coupling between the areas; finally, considering the balance of active and reactive power within the partition, the edge nodes in each partition are fine-tuned to obtain the final partitioning result of the active distribution network, while maintaining the stability and balance of the distribution network, ensuring that the partitioning is more accurate; and in the case of normal operation of the distribution network or small source and load fluctuation, the present invention only ensures the stability and economy of the distribution network partitioning by fine-tuning the edge nodes in each partition, thereby reducing the frequent partitioning caused by source and load fluctuation as much as possible.

Description

Technical Field

[0001] The present invention belongs to the field of distribution network partitioning, and in particular relates to a method and system for dynamic partitioning of an active distribution network. Background Art

[0002] With the rapid development of the economy, large-scale distributed renewable energy (DG), energy storage, new interactive power equipment and other load-side resources are connected to the grid, the number of nodes in the distribution network is increasing, and the topology is becoming more complex. Nowadays, traditional centralized control can hardly meet the requirements of distribution network for accuracy and speed of voltage regulation. Therefore, it is necessary to conduct zoning regulation on the distribution network, and ensure that each zone of the distribution network has sufficient power reserve and strong control capability by regulating the voltage of each independent zone, so as to achieve local balance of regional voltage fluctuations, simplify information exchange between zones, reduce communication pressure, and reduce the difficulty of voltage control in the distribution network.

[0003] In the existing technology, many clustering algorithms based on electrical distance have been proposed to solve the problem of power grid partitioning. These methods only use the sensitivity method to obtain the electrical distance, and conceptually do not consider the relationship between spatial positions, which will affect the credibility of the calculation results. In addition, distribution network partitions are mostly non-overlapping partitions, and the impact of nodes within a partition on other areas is also the focus of distribution network partition analysis. Summary of the invention

[0004] The purpose of the present invention is to provide a method and system for dynamic partitioning of an active distribution network in view of the above-mentioned problems existing in the prior art.

[0005] To achieve the above objectives, the technical solution of the present invention is as follows:

[0006] In a first aspect, the present invention provides a method for dynamic partitioning of an active power distribution network;

[0007] When the network topology changes or extreme source load fluctuations occur, the method includes:

[0008] S1. Based on the source-grid-load-storage data and network structure of the active distribution network, the active distribution network is preliminarily partitioned to determine the overlapping areas in the active distribution network;

[0009] S2, partitioning and adjusting the overlapping nodes in the overlapping area of ​​the active distribution network to obtain a regional redivision result;

[0010] S3. Based on the regional redivision results, considering the balance of active and reactive power in the partition, fine-tune the edge nodes in each partition to obtain the final partition result of the active distribution network;

[0011] When the distribution network operates normally or the source-load fluctuation is small, the method is:

[0012] Considering the balance of active and reactive power within the partition, the edge nodes in each partition are fine-tuned to obtain the final partition result of the active distribution network.

[0013] The S1 includes:

[0014] S11. The local connection similarity of nodes in the active distribution network is determined using the following formula:

[0015]

[0016] In the above formula, w ij is the local connection similarity between node i and node j, k i is the degree of node i, A ij is the weight of the edge connecting node i and node j, x is the common neighboring node of node i and node j, S ij is the set of common neighbor nodes of node i and node j, k x is the degree of node x;

[0017] S12, partitioning the distribution network nodes by fuzzy hierarchical clustering method based on the local connection similarity of the nodes in the active distribution network;

[0018] S13, use the following formula to calculate the partition similarity between each partition, if the interval similarity between two partitions is greater than the set threshold Then these two partitions are merged to complete the preliminary partitioning of the active distribution network;

[0019]

[0020] In the above formula, δ(C i , C j ) is partition C i Partition C j The similarity between i ) is partition C i The nodes included, N(C j ) is partition C j The nodes contained.

[0021] The S2 includes:

[0022] S21. Based on the electrical distance of the distribution network, the modularity of each partition is determined using the following formula:

[0023]

[0024] A ij =D′ ij ;

[0025]

[0026] In the above formula, Q v is the modularity of partition v, e vv , is the ratio of the total edge weight within partition v to the total edge weight of the entire network, a v is the ratio of the total edge weight of all nodes in partition v to the total edge weight of all nodes in the network, m is the sum of the weights of all edges in the distribution network, A ij is the edge weight between node i and node j, δ(c i , v) is partition C i and the similarity between partition v, δ(c j , v) is partition C j The similarity between the partition v and D′ ij is the electrical distance between node i and node j, k i is the sum of the weights of all edges connected to node i;

[0027] S22. The overlapping nodes in the overlapping area of ​​the distribution network determined by the preliminary partitioning are taken as independent sub-partitions, and the non-overlapping areas and the partitions without overlapping areas in a partition are taken as independent partitions. Based on the modularity index, the independent sub-partitions are merged into the adjacent independent partitions to obtain the regional redivision results, including:

[0028] S221. Perform the following operations on each independent sub-partition:

[0029] If the modularity increase after merging the independent sub-partition e into the independent partition u is greater than the modularity increase after merging e into the independent partition v, then e is merged into the independent partition u;

[0030] If the modularity reduction after merging the independent sub-partition e into the independent partition u is less than the modularity reduction after merging e into the independent partition v, then merge e into the independent partition u;

[0031] S222, calculating the reactive power reserve of each partition after the operation described in step S221, judging whether the reactive power reserve of each partition satisfies the reactive power reserve constraint, and merging the partition that does not satisfy the reactive power reserve constraint with its adjacent partition that satisfies the reactive power reserve constraint;

[0032] S223. Calculate the reactive power reserve of each partition after merging, and judge again whether the reactive power reserve of each partition meets the reactive power reserve constraint. For the partitions that still do not meet the reactive power reserve constraint, repeat step S222 until the reactive power reserve of each partition meets the reactive power reserve constraint. At this time, the regional redivision result is obtained.

[0033] In S21, the electrical distance of the distribution network is an electrical distance that characterizes the power flow relationship and spatial position relationship of the distribution network and is constructed based on the sensitivity method and the transmission impedance method. The calculation method includes:

[0034] S211. Calculate the electrical distances between the nodes of the distribution network based on the sensitivity method and the transmission impedance method respectively:

[0035] The electrical distance between the distribution network nodes based on the sensitivity method is calculated using the following formula:

[0036] D ij =1-L ij / max{L ij};

[0037]

[0038] In the above formula, D ij is the electrical distance between nodes i and j based on the sensitivity method, L ij is the influence of other nodes in the distribution network on the sensitivity of node i and node j, N is the number of nodes in the distribution network, α im is the sensitivity between node i and node m, α jm is the sensitivity between node j and node m;

[0039] The electrical distance between the nodes of the distribution network based on the transmission impedance method is calculated using the following formula:

[0040]

[0041] In the above formula, is the output power of branch ij, is the injected power of branch ij, z ij is the impedance of branch ij, representing the equivalent transmission impedance of power from node i to node j. ij | is the electrical distance between node i and node j, is the equivalent impedance of the power supply downstream of branch ij;

[0042] S212. Combining the two electrical distances, the electrical distance of the distribution network is calculated using the following formula:

[0043]

[0044] In the above formula, D′ ij is the electrical distance between node i and node j, D ij , Z ij The weight of D ij is the electrical distance between nodes i and j based on the sensitivity method, |Zij | is the electrical distance between node i and node j based on the transmission impedance method.

[0045] The balance of active power and reactive power in the partition is considered, and the edge nodes in each partition are fine-tuned to obtain the final partition result of the active distribution network, including:

[0046] S31. Based on the regional redivision results, the improved sand cat swarm optimization algorithm is used to optimize the comprehensive indicators of nodes in each partition of the distribution network, and the optimal central nodes of all partitions are output by iteratively updating particles;

[0047] The comprehensive index of the node is calculated using the following formula:

[0048] F cen =η1OBS+η2CON;

[0049] In the above formula, F cen is the comprehensive index of the node, η1 and η2 are the weights of OBS and CON respectively, OBS is the observability index of the node, and CON is the controllability index of the node;

[0050] S32, calculate the comprehensive index value of the optimal central node of each partition, and set the minimum node comprehensive index preset value F of each partition according to the comprehensive index value of the optimal central node of each partition min , keep the nodes in each partition whose comprehensive index value is greater than the preset value unchanged, take other nodes in the partition as edge nodes, and use the improved sand cat group optimization algorithm to find the optimal adjacent partition of each edge node from the outside to the inside according to the network structure of the distribution network and the connection relationship between nodes, and divide the edge node into its optimal adjacent partition;

[0051] S33, after each operation of dividing the edge nodes in the partition, the comprehensive balance index of the distribution network is calculated once, and after all the partitions of the distribution network are divided, the partition result with the largest comprehensive balance index is selected as the final partition result S of the active distribution network;

[0052] The comprehensive balance index of the distribution network is calculated using the following formula:

[0053]

[0054] In the above formula, is the total active and reactive balance index of the comprehensive distribution network, λ1 and λ2 are The weight of is the total active power balance index of the distribution network, It is the total reactive power balance index of the distribution network.

[0055] In a second aspect, the present invention provides a system for dynamic partitioning of an active power distribution network, comprising a preliminary partitioning module, a repartitioning module and a fine-tuning module;

[0056] When the network topology changes or there are extreme source load fluctuations:

[0057] The preliminary partitioning module is used to perform preliminary partitioning of the active distribution network based on the source-grid-load-storage data and network structure of the active distribution network, and determine the overlapping areas in the active distribution network;

[0058] The re-division module is used to partition and adjust the overlapping nodes in the overlapping area of ​​the active power distribution network to obtain the area re-division result;

[0059] The fine-tuning module is used to fine-tune the edge nodes in each partition based on the regional redivision result and consider the balance of active and reactive power in the partition to obtain the final partition result of the active distribution network.

[0060] When the distribution network operates normally or the source load fluctuation is small:

[0061] The fine-tuning module is used to consider the balance of active and reactive power within the partition, and to fine-tune the edge nodes within each partition to obtain the final partition result of the active distribution network.

[0062] The preliminary partitioning module includes a local similarity unit, a node partitioning unit and a partition similarity unit;

[0063] The local similarity unit is used to determine the local connection similarity of nodes in the active power distribution network using the following formula:

[0064]

[0065] In the above formula, w ij is the local connection similarity between node i and node j, k i is the degree of node i, A ij is the weight of the edge connecting node i and node j, x is the common neighboring node of node i and node j, S ij is the set of common neighbor nodes of node i and node j, k x is the degree of node x;

[0066] The node partitioning unit is used to partition the distribution network nodes by a fuzzy hierarchical clustering method based on the local connection similarity of the nodes in the active distribution network;

[0067] The partition similarity unit is used to calculate the interval similarity between partitions using the following formula. If the interval similarity between two partitions is greater than the set threshold Then these two partitions are merged to complete the preliminary partitioning of the active distribution network;

[0068]

[0069] In the above formula, δ(C i , C j ) is partition C i and partition C j The similarity between i ) is partition C i The nodes included, N(C j ) is partition C j The nodes contained.

[0070] The repartitioning module includes a modularity unit and an independent sub-partition merging unit;

[0071] The modularity unit is used to determine the modularity of each partition based on the electrical distance of the distribution network using the following formula:

[0072]

[0073] A ij =D′ ij ;

[0074]

[0075] In the above formula, Q v is the modularity of partition v, e vv is the ratio of the total edge weight within partition v to the total edge weight of the entire network, a v is the ratio of the total edge weight of all nodes in partition v to the total edge weight of all nodes in the network, m is the sum of the weights of all edges in the distribution network, A ij is the edge weight between node i and node j, δ(c i , v) is partition C i and the similarity between partition v, δ(c j , v) is partition C j The similarity between the partition v and D′ ij is the electrical distance between node i and node j, k i is the sum of the weights of all edges connected to node i;

[0076] The independent sub-partition merging unit is used to take the overlapping nodes in the distribution network overlapping area determined by the preliminary partitioning as independent sub-partitions, and the non-overlapping areas other than the overlapping areas in a partition and the partitions without overlapping areas as independent partitions, and merge the independent sub-partitions into the adjacent independent partitions based on the modularity index to obtain the regional redivision result, including:

[0077] For each independent sub-partition, perform the following operations:

[0078] If the modularity increase after merging the independent sub-partition e into the independent partition u is greater than the modularity increase after merging e into the independent partition v, then e is merged into the independent partition u;

[0079] If the modularity reduction after merging the independent sub-partition e into the independent partition u is less than the modularity reduction after merging e into the independent partition v, then merge e into the independent partition u;

[0080] Calculating the reactive power reserve of each partition after the operation is performed on each independent sub-partition, judging whether the reactive power reserve of each partition satisfies the reactive power reserve constraint, and merging the partition that does not satisfy the reactive power reserve constraint with its adjacent partition that satisfies the reactive power reserve constraint;

[0081] Calculate the reactive power reserve of each partition after merging, and judge again whether the reactive power reserve of each partition meets the reactive power reserve constraint. For the partition that still does not meet the reactive power reserve constraint, repeatedly merge the partition with its adjacent partition that meets the reactive power reserve constraint until the reactive power reserve of each partition meets the reactive power reserve constraint. At this time, the regional redivision result is obtained.

[0082] The subdivision module further comprises an electrical distance construction unit;

[0083] The electrical distance construction unit is used to construct the electrical distance between the distribution network nodes based on the sensitivity method and the transmission impedance method, and the calculation method thereof includes:

[0084] The electrical distance between the distribution network nodes based on the sensitivity method is calculated using the following formula:

[0085] D ij =1-L ij / max{L ij};

[0086]

[0087] In the above formula, D ij is the electrical distance between nodes i and j based on the sensitivity method, L ii is the influence of other nodes in the distribution network on the sensitivity of node i and node j, N is the number of nodes in the distribution network, α im is the sensitivity between node i and node m, α jm is the sensitivity between node j and node m;

[0088] The electrical distance between the nodes of the distribution network based on the transmission impedance method is calculated using the following formula:

[0089]

[0090] In the above formula, is the output power of branch ij, is the injected power of branch ij, z ij is the impedance of branch ij, representing the equivalent transmission impedance of power from node i to node j. ij | is the electrical distance between node i and node j, is the equivalent impedance of the downstream of the branch ij;

[0091] Combining the two electrical distances, the following formula is used to calculate the electrical distance of the distribution network:

[0092]

[0093] In the above formula, D′ ij is the electrical distance between node i and node j, D ij , Z ij The weight of D ij is the electrical distance between nodes i and j based on the sensitivity method, |Z ij | is the electrical distance between node i and node j based on the transmission impedance method.

[0094] The fine-tuning module includes a central node optimization unit, an edge node partitioning unit and a final partitioning unit;

[0095] The central node optimization unit is used to optimize the comprehensive indicators of nodes in each partition of the distribution network based on the regional redivision result by using the improved sand cat swarm optimization algorithm, and output the optimal central nodes of all partitions by iteratively updating particles;

[0096] The comprehensive index of the node is calculated using the following formula:

[0097] F cen =η1OBS+η2CON;

[0098] In the above formula, F cen is the comprehensive index of the node, η1 and η2 are the weights of OBS and CON respectively, OBS is the observability index of the node, and CON is the controllability index of the node;

[0099] The edge node division unit is used to calculate the comprehensive index value of the optimal central node of each partition, and set the minimum node comprehensive index preset value F of each partition according to the comprehensive index value of the optimal central node of each partition. min, keep the nodes in each partition whose comprehensive index value is greater than the preset value unchanged, take other nodes in the partition as edge nodes, and use the improved sand cat group optimization algorithm to find the optimal adjacent partition of each edge node from the outside to the inside according to the network structure of the distribution network and the connection relationship between nodes, and divide the edge node into its optimal adjacent partition;

[0100] The final partitioning unit is used to calculate the comprehensive balance index of the distribution network once after each operation of partitioning the edge nodes in the partition. After all the partitions of the distribution network are divided, the partition result with the largest comprehensive balance index is selected as the final partition result S of the active distribution network;

[0101] The comprehensive balance index of the distribution network is calculated using the following formula:

[0102]

[0103] In the above formula, is the total active and reactive balance index of the comprehensive distribution network, λ1 and λ2 are The weight of is the total active power balance index of the distribution network, It is the total reactive power balance index of the distribution network.

[0104] Compared with the prior art, the present invention has the following beneficial effects:

[0105] 1. The present invention proposes a method and system for dynamic partitioning of an active distribution network. When the network topology of the distribution network changes or the source-load fluctuation is extreme, the method first performs preliminary partitioning of the active distribution network based on the source-network-load-storage data and network structure of the active distribution network to determine the overlapping areas in the active distribution network; then the overlapping nodes in the overlapping areas of the active distribution network are partitioned and adjusted to obtain the regional redivision result; finally, based on the regional redivision result, the balance of active and reactive power within the partition is considered, and the edge nodes within each partition are fine-tuned to obtain the final partitioning result of the active distribution network. The method first obtains the overlapping nodes that interact between regions, and performs partitioning adjustment on the overlapping nodes to ensure the requirement of close coupling within the region and sparse coupling between regions, and fine-tunes the edge nodes of the partition to ensure the balance of active and reactive power within the partition. While maintaining the stability and balance of the distribution network, it ensures that the partitioning is more accurate and reduces the frequent partitioning caused by source-load fluctuations as much as possible.

[0106] 2. The present invention proposes a method and system for dynamic partitioning of an active distribution network. The method takes into account the uncertainty of the interaction between nodes in overlapping areas in the active distribution network, and constructs a gas distance that characterizes the relationship between power flow and spatial position based on the sensitivity method and the transmission impedance method, thereby ensuring the credibility of the electrical distance calculation results. The method also takes into account the voltage fluctuations caused by various disturbances in the distribution network, and constructs a constraint indicator to ensure sufficient reactive power reserves in each partition, which plays an important role in maintaining system voltage stability.

[0107] 3. The present invention proposes a method and system for dynamic partitioning of an active distribution network. Due to the randomness of photovoltaic and wind turbine output and the frequent fluctuation of load, the static power flow of a single time section cannot meet the long-term operation requirements of the active distribution network. However, in order to maintain the continuity and stability of the active distribution network scheduling, the distribution network partitioning should not change frequently. Therefore, this method adopts a dynamic partitioning method. When the distribution network operates normally or the source-load fluctuation is small, the stability and economy of the distribution network partitioning are ensured only by fine-tuning the edge nodes in each partition. BRIEF DESCRIPTION OF THE DRAWINGS

[0108] Figure 1 The figure is an overall flow chart of the method of the present invention.

[0109] Figure 2 This is a power transmission relationship diagram of a branch described in Example 1.

[0110] Figure 3 The structure diagram of the system of the present invention is shown in FIG. DETAILED DESCRIPTION

[0111] The present invention is further described in detail below in conjunction with specific implementations and drawings.

[0112] The present invention proposes a method and system for dynamic partitioning of an active distribution network. First, based on the source network load storage data and network structure of the active distribution network, the active distribution network is preliminarily partitioned to determine the overlapping areas in the active distribution network network; then the overlapping nodes in the overlapping areas of the active distribution network network are partitioned and adjusted to obtain the regional re-division result; finally, based on the regional re-division result, the balance of active power and reactive power in the partition is considered, and the edge nodes in each partition are fine-tuned to obtain the final partition result of the active distribution network. Due to the intermittent nature of the active distribution network, the power and load requirements of the distributed power sources in the distribution network are constantly changing, but the partitioning of the distribution network should not be adjusted frequently. Therefore, only when the operating state of the distribution network is adjusted to cause the network topology structure to change, or when the DG output and load demand of the distribution network fluctuate greatly with the season, the source network load storage data of the distribution network is updated, and the partitioning structure of the distribution network is readjusted to meet the power reserve demand of the distribution network, and to achieve the requirements of strong coupling within the distribution network area and weak coupling between intervals. When the distribution network is operating normally, only the edge nodes of the partition are fine-tuned.

[0113] Embodiment 1:

[0114] This embodiment takes the IEEE33 node system as the research object. When the network topology changes due to the adjustment of the operating state of the distribution network, the IEEE33 node system is dynamically partitioned. The system is connected to PV at nodes 2, 7, 11, 18, 20, 23, 26, and 32. SVC is installed at node 22 with a capacity of 200kV·A. Energy storage devices are connected at nodes 17 and 33 with capacities of 500kW·h and 400kW·h, respectively. The voltage level of the system is 12.66kV, the total active power of the load is 4174.19kW, and the total reactive power of the load is 2438kVar.

[0115] like Figure 1 As shown, a method for dynamic partitioning of an active distribution network is performed in the following steps:

[0116] 1. Construct the electrical distance that characterizes the power flow relationship and spatial position relationship of the distribution network based on the sensitivity method and the transmission impedance method;

[0117] The calculation method of the electrical distance between distribution network nodes based on the sensitivity method is as follows:

[0118] The sensitivity between nodes is calculated using the following formula:

[0119]

[0120] In the above formula, α ij is the sensitivity between node i and node j, ΔU i , ΔU j are the voltage increments of node i and node j respectively, are reactive power / voltage sensitivity, α ji is the sensitivity between node j and node i, All are reactive / voltage sensitivity;

[0121] Due to α ij ≠α ji To ensure the positivity and symmetry of the electrical distance, the following formula is used to calculate the electrical distance:

[0122] D ij =D ji =-lg(α ij α ji );

[0123] In the above formula, D ij is the electrical distance between node i and node j, D ji is the electrical distance between node j and node i;

[0124] Considering that the relationship between two nodes in the distribution network will also be affected by other nodes in the network, the following formula is used to calculate the electrical distance between distribution network nodes based on the sensitivity method:

[0125] D ij =1-L ij / max{L ij};

[0126]

[0127] In the above formula, D ij is the electrical distance between nodes i and j based on the sensitivity method, L ii is the influence of other nodes in the distribution network on the sensitivity of node i and node j, N is the number of nodes in the distribution network, α im is the sensitivity between node i and node m, α jm is the sensitivity between node j and node m;

[0128] The calculation method of the electrical distance between distribution network nodes based on the transmission impedance method is as follows:

[0129] In a single power distribution network, the power transmission relationship of a branch is as follows: Figure 2 As shown, the power relationship in the branch is calculated using the following formula:

[0130]

[0131] In the above formula, is the output power of branch ij, is the injected power of branch ij, z ijis the impedance of branch ij, which represents the power loss caused by the power transfer from node i to node j. It is called the equivalent transfer impedance. ij | as the electrical distance between node i and node j; is the equivalent impedance of the downstream of the branch ij;

[0132] Combining the two electrical distances, the following formula is used to calculate the electrical distance of the distribution network:

[0133]

[0134] In the above formula, D′ ij is the electrical distance between node i and node j in the distribution network, let D ij =A ij To represent the edge weight; D ij , Z ij The weight of Pick D ij is the electrical distance between nodes i and j based on the sensitivity method, |Z ij | is the electrical distance between node i and node j based on the transmission impedance method.

[0135] 2. Based on the source, grid, load and storage data of the active distribution network and its network structure, the active distribution network is preliminarily partitioned to determine the overlapping areas in the active distribution network;

[0136] According to the network structure of the active distribution network, the nodes are preliminarily processed: if there is a non-terminal DG node on a terminal branch of the distribution network, the terminal branch is merged with the DG node closest to it and treated as a DG node; if there is no DG node on a branch of the distribution network, all nodes of the branch are merged and treated as load nodes;

[0137] For an unweighted undirected graph G = (VW), V = {1, 2, ..., i, j, ..., n} is a node set, and the adjacency matrix is ​​represented by W = w ij (i, j = 1, 2, ..., n), the weight w between node i and node j ij Use local connection similarity to measure the importance of node i to node j:

[0138] The local connection similarity of nodes in the active distribution network is calculated using the following formula:

[0139]

[0140] In the above formula, w ij is the local connection similarity between node i and node j, k iis the degree of node i, A ij is the weight of the edge connecting node i and node j. When node i and node j are directly connected, A ij =1, when node i and node j are not connected, A ij =0, characterized by electrical distance, i.e. A ij =D′ ij ; x is the common neighboring node of node i and node j, S ij is the set of common neighbor nodes of node i and node j, k x is the degree of node x;

[0141] Based on the local connection similarity of nodes in the active distribution network, the distribution network nodes are partitioned by fuzzy hierarchical clustering method, including:

[0142] Randomly initialize the membership matrix W = w ij (i, j = 1, 2, ..., n), with 8 DGs in the distribution network as the cluster center, and the initial number of cluster partitions is k = 8;

[0143] In each iteration, first according to the current membership matrix w ij Update the cluster center, and then update the membership matrix w according to the new cluster center ij , repeat this step until convergence or the maximum number of iterations is reached;

[0144] According to the membership degree of the new matrix, the current clustering results are updated to form a node partition structure;

[0145] Based on the partition structure of the node, the partitions are merged, including:

[0146] If the interval similarity between the two partitions is greater than the set threshold When the two partitions are merged, the initial partitioning of the active distribution network is completed. It is the ratio of the number of overlapping nodes in the two partitions to the total number of nodes in the two partitions;

[0147] The following formula is used to calculate the interval similarity between partitions:

[0148]

[0149] In the above formula, δ(C i , C j ) is partition C i and partition C j The similarity between i ) is partition C i The nodes included, N(C j ) is partition C j The nodes contained.

[0150] 3. Determine the modularity function of each partition based on the electrical distance of the distribution network;

[0151] Modularity is defined as:

[0152]

[0153] In the above formula, ρ is the defined modularity, m is the sum of the weights of all edges in the distribution network, and A ij is the weight of the edge connecting node i and node j. When node i and node j are directly connected, A ij =1, when node i and node j are not connected, A ij =0, characterized by electrical distance, i.e. A ij =D ij ;k i , k j are the sum of the weights of all edges connected to node i and node j, respectively. is the expected value of the edge weight between node i and node j when the internal nodes of the network are randomly connected;

[0154] According to the high sparsity of the adjacency matrix, the following formula is used to calculate the modularity, improve the calculation efficiency of the search algorithm, and reduce storage space and calculation time:

[0155]

[0156] A ij =D′ ij ;

[0157]

[0158] In the above formula, Q v is the modularity of partition v, e vv is the ratio of the total edge weight within partition v to the total edge weight of the entire network, a v is the ratio of the total edge weight of all nodes in partition v to the total edge weight of all nodes in the network, δ(c i , v) is partition C i and the similarity between partition v, δ(c j , v) is partition C j The similarity between the partition v and D′ ij is the electrical distance between node i and node j.

[0159] 4. Perform partition adjustment on the overlapping nodes in the overlapping area of ​​the active distribution network to obtain the regional redivision result;

[0160] According to the preliminary partitioning results of the distribution network, the overlapping nodes in the overlapping areas of the distribution network determined by the preliminary partitioning are taken as independent sub-partitions, and the non-overlapping areas and the partitions without overlapping areas in a partition are taken as independent partitions. Based on the modularity index, each independent sub-partition is merged into the adjacent independent partition to obtain the regional redivision results, including:

[0161] The following operations are performed for each independent sub-area until there are no isolated overlapping nodes in the preliminary distribution network partition:

[0162] If there is an independent node e, it only increases the reactive capacity of multiple adjacent partitions, and we have:

[0163] ΔQ u，e >ΔQ v，e ;

[0164] In the above formula, ΔQ u，e , ΔQ v，e The reactive capacity increase of the node e merged in partition u and partition v is respectively, and the independent node e is merged with the partition u with a larger reactive capacity change;

[0165] If there is an independent node e, which only reduces the reactive capacity of multiple adjacent partitions, we have:

[0166] |ΔQ u，e |<|ΔQ v，e |;

[0167] In the above formula, |Δg u，e |,|ΔQ v，e | are the reactive capacity reduction of the merged node e in partitions u and v respectively, and the independent node e is merged with the partition u with a smaller reactive capacity change;

[0168] The following formula is used to calculate the reactive power reserve of each partition and its constraint conditions after each independent sub-partition performs the above operation, and to determine whether the reactive power reserve of each partition satisfies the reactive power reserve constraint:

[0169]

[0170] β k >β set ;

[0171] In the above formula, β k is the reactive power reserve of partition k, Q Gk is the dynamic reactive capacity available in zone k, Q Lk is the reactive power demand of load in partition k; β set It is the preset value of the minimum reactive power reserve margin;

[0172] For a partition that does not meet the reactive power reserve constraint, the partition is merged with its adjacent partition that meets the reactive power reserve constraint, and the merged partition satisfies:

[0173] Q u，w >Q v，w ;

[0174] In the above formula, Q u，w , Q v，w They are the reactive capacity of the zones u and v combined with the zone w. Adequate reactive reserve helps the distribution network cope with voltage fluctuations caused by various disturbances, and plays an important role in maintaining system voltage stability.

[0175] Calculate the reactive power reserve of each partition after merging, and judge whether the reactive power reserve of each partition meets the reactive power reserve constraint again. For the partition that still does not meet the reactive power reserve constraint, repeatedly merge the partition with its adjacent partition that meets the reactive power reserve constraint until the reactive power reserve of each partition meets the reactive power reserve constraint. At this time, the regional redivision result is obtained.

[0176] 5. Define the comprehensive indicators of nodes in each partition of the distribution network, as well as the comprehensive balance indicators of the distribution network;

[0177] When the load voltage in the distribution network area fluctuates, the observability of the central node in the area can be adjusted. The observability index of the distribution network node is defined by the following formula:

[0178]

[0179] In the above formula, OBS is the observability index of the node, ξ1 and ξ2 are weight coefficients, and ξ1+ξ2=1, ξ1=ξ2=0.5, V i is the voltage of node i, m is the number of load nodes, V m is the voltage of all load nodes, P m , Q m Inject active and reactive power into load nodes;

[0180] When the active and reactive injection power of the distributed generation in the distribution network area changes, the controllability of the central node in the area can be regulated to adjust the injection power to the set value. The controllability index of the distribution network node is defined by the following formula:

[0181]

[0182] In the above formula, CON is the controllability index of the node, ψ1 and ψ2 are weight coefficients, ψ1+ψ2=1, ψ1=ψ2=0.5, P DG , Q DG are respectively the active and reactive injected power of distributed generation, l DGis the number of nodes of distributed generation in the partition;

[0183] In summary, the following formula is used to calculate the comprehensive index of nodes in each partition of the distribution network:

[0184] F cen =η1OBS+η2CON;

[0185] In the above formula, F cen is the comprehensive index of the node, η1 and η2 are the weights of OBS and CON respectively, satisfying η1+η2=1, and η1=η2=0.5;

[0186] The reactive power balance index of the distribution network is defined by the following formula:

[0187]

[0188] In the above formula, is the total reactive power balance index of the distribution network, n is the number of distribution network partitions, β k is the reactive power reserve of partition k;

[0189] The following formula is used to define the active power balance index of the distribution network:

[0190]

[0191] In the above formula, is the total active power balance index of the distribution network, P k is the active reserve of partition k, T is the calculation period of the time-varying scenario, P D (t) k is the total active power of distributed power sources in partition k at time t, which is obtained from the day-ahead dispatch plan of active power sources; P L (t) is the total load power of distributed generation in partition k at time t, which is obtained from the load day-ahead forecast value;

[0192] In summary, the following formula is used to calculate the comprehensive balance index of the distribution network:

[0193]

[0194] In the above formula, is the total active and reactive balance index of the comprehensive distribution network, λ1 and λ2 are The weight satisfies λ1+λ2=1, and λ1=λ2=0.5.

[0195] 6. Based on the regional redivision results, considering the balance of active and reactive power within the partition, fine-tune the edge nodes within each partition to obtain the final partition result of the active distribution network;

[0196] The adjustment of the partition is not suitable for destroying the electrical coupling characteristics of each partition. Therefore, only fine-tuning is performed on the re-divided partition results, keeping the central node of each partition and its surrounding core nodes unchanged, and re-allocating the edge nodes of each partition according to the comprehensive balance index of observability and controllability, so that the overall comprehensive balance of the distribution network after allocation is maximized;

[0197] The improved sand cat group optimization algorithm is used to optimize the comprehensive indicators of the central nodes of each partition of the distribution network. The sand cat group algorithm has a certain degree of adaptability, allowing the sand cat to switch seamlessly between the exploration and development stages. It has strong optimization ability and fast convergence speed, reducing the frequent partitioning caused by source and load fluctuations. The obtained network structure has good convergence and partition quality.

[0198] The step of optimizing the comprehensive index of the central node of each partition of the distribution network by using the improved sand cat group optimization algorithm comprises:

[0199] Complete the initialization of the sand cat population, set the population size M = 30, the maximum number of iterations is T = 200, and calculate the position fitness value of each sand cat individual in the comprehensive index of the central node of each partition;

[0200] Based on the location fitness value of each sand cat individual, the location of the sand cat is updated according to the boot parameter R, including:

[0201] Sand cats use low-frequency noise to find prey. The sensitivity range of sand cats depends on their hearing ability in low-frequency detection and the low-frequency signals they perceive. Assuming that the hearing range of sand cats is from 2kHz to 0, their behavior is expressed by the following formula:

[0202]

[0203] In the above formula, r G For the hearing range of sand cats, the linear is changed to nonlinear. In the early iteration r G The slower change of is beneficial to the global search ability of the algorithm. G The faster change of is beneficial to the local search ability of the algorithm and the convergence of the algorithm. M To simulate the hearing of sand cats, its value is set to 2, and t is the current iteration number;

[0204] The bootstrap parameter that determines the transition from the exploration phase to the exploitation phase is R, which is calculated using the following formula:

[0205] R=2×rr G ;

[0206] r=r G ×rand(0,1);

[0207] In the above formula, R is the guiding parameter of the improved sand cat swarm optimization algorithm to control the transition between exploration and exploitation, r is the sensitivity range of each cat, and rand(0,1) is a random function between 0 and 1;

[0208] When |R|>1, i.e. the exploration phase, the cat will gradually approach its prey through the Lévy flying and wandering strategy, and will wander around the prey while approaching it. The cat will choose the best solution X according to the global optimal solution. bc (t), its current position X(t) and its sensitivity range r to update its own position;

[0209] The following formula is used to calculate the position update of the sand cat during the exploration phase:

[0210] X(t+1)=r×(X bc (t)-C×s×X(t));

[0211] In the above formula, X(t+1) is the location of the next exploration, C is a constant, let C = 0.35, s is the Lévy flight path of the sand cat, multiplied by the constant C to reduce the step length so that the sand cat can walk next to the prey as much as possible;

[0212] When |R|≤1, i.e., the development stage, the distance between the sand cat and the prey is very close, and each sand cat chooses a random angle to avoid falling into the local optimum while approaching the hunting position;

[0213] The following formula is used to calculate the position update of the sand cat during the development phase:

[0214] X(t+1)=X bc (t)-r·X rnd ·cos(θ);

[0215] X rnd =|rand(0,1)·X bc (t)-X(t)|;

[0216] In the above formula, X(t+1) is the location of the next development, X rnd is a random position vector representing the distance difference between the optimal position and the current position, and θ is a random angle;

[0217] After updating the position of the sand cat, determine whether the maximum number of iterations has been reached. If not, recalculate the position fitness value of each sand cat individual until the maximum number of iterations is reached and output the optimal central node of all partitions;

[0218] Calculate the comprehensive index value of the optimal central node of each partition, and set the minimum node comprehensive index preset value F of each partition according to the comprehensive index value of the optimal central node of each partition min, keep the nodes in each partition whose comprehensive index value is greater than the preset value unchanged. These nodes are the central nodes and surrounding core nodes of each partition. The other nodes in the partition are regarded as edge nodes. According to the network structure of the distribution network and the connection relationship between nodes, the improved sand cat group optimization algorithm is used to search for the optimal adjacent partition of each edge node from the outside to the inside, and divide the edge node into its optimal adjacent partition;

[0219] The minimum node comprehensive index preset value F of each partition min , calculated using the following formula:

[0220]

[0221] In the above formula, is the preset value of the minimum node comprehensive index of partition k, μ is the proportional coefficient, set μ = 0.8, is the comprehensive index value of the optimal central node of partition k;

[0222] After each operation of dividing the edge nodes in the partition, the comprehensive balance index of the distribution network is calculated once. After all the partitions of the distribution network are divided, the partition result with the largest comprehensive balance index is selected as the final partition result S of the active distribution network;

[0223] Embodiment 2:

[0224] This embodiment takes the IEEE 33 node system as the research object. When the source and load fluctuation of the distribution network is small, the IEEE 33 node system is dynamically partitioned, and the following steps are performed in sequence:

[0225] The fine-tuning of the partitions is not suitable for destroying the electrical coupling characteristics of each partition. Therefore, only fine-tuning is performed on the existing partition results without changing the main structure of the distribution network partitions. The edge nodes of each partition are secondary allocated according to the comprehensive balance index of observability and controllability, so that the overall comprehensive balance of the distribution network after allocation is maximized.

[0226] 1. Define the comprehensive indicators of nodes in each partition of the distribution network, as well as the comprehensive balance indicators of the distribution network;

[0227] When the load voltage in the distribution network area fluctuates, the observability of the central node in the area can be adjusted. The observability index of the distribution network node is defined by the following formula:

[0228]

[0229] In the above formula, OBS is the observability index of the node, ξ1 and ξ2 are weight coefficients, and ξ1+ξ2=1, ξ1=ξ2=0.5, V i is the voltage of node i, m is the number of load nodes, Vm is the voltage of all load nodes, P m , Q m Inject active and reactive power into load nodes;

[0230] When the active and reactive injection power of the distributed generation in the distribution network area changes, the controllability of the central node in the area can be regulated to adjust the injection power to the set value. The controllability index of the distribution network node is defined by the following formula:

[0231]

[0232] In the above formula, CON is the controllability index of the node, ψ1 and ψ2 are weight coefficients, ψ1+ψ2=1, ψ1=ψ2=0.5, P DG , Q DG are respectively the active and reactive injected power of distributed generation, l DG is the number of nodes of distributed generation in the partition;

[0233] In summary, the following formula is used to calculate the comprehensive index of nodes in each partition of the distribution network:

[0234] F cen =η1OBS+η2CON;

[0235] In the above formula, F cen is the comprehensive index of the node, η1 and η2 are the weights of OBS and CON respectively, satisfying η1+η2=1, and η1=η2=0.5;

[0236] The reactive power balance index of the distribution network is defined by the following formula:

[0237]

[0238] In the above formula, is the total reactive power balance index of the distribution network, n is the number of distribution network partitions, β k is the reactive power reserve of partition k;

[0239] The following formula is used to define the active power balance index of the distribution network:

[0240]

[0241] In the above formula, is the total active power balance index of the distribution network, P k is the active reserve of partition k, T is the calculation period of the time-varying scenario, P D (t) k is the total active power of distributed power sources in partition k at time t, which is obtained from the day-ahead dispatch plan of active power sources; P L(t) is the total load power of distributed generation in partition k at time t, which is obtained from the load day-ahead forecast value;

[0242] In summary, the following formula is used to calculate the comprehensive balance index of the distribution network:

[0243]

[0244] In the above formula, is the total active and reactive balance index of the comprehensive distribution network, λ1 and λ2 are The weight satisfies λ1+λ2=1, and λ1=λ2=0.5.

[0245] 2. Considering the balance of active and reactive power within the partition, fine-tune the edge nodes in each partition to obtain the final partition result of the active distribution network;

[0246] The improved sand cat group optimization algorithm is used to optimize the comprehensive indicators of the central nodes of each partition of the distribution network. The sand cat group algorithm has a certain degree of adaptability, allowing the sand cat to switch seamlessly between the exploration and development stages. It has strong optimization ability and fast convergence speed, reducing the frequent partitioning caused by source and load fluctuations. The obtained network structure has good convergence and partition quality.

[0247] The step of optimizing the comprehensive index of the central node of each partition of the distribution network by using the improved sand cat group optimization algorithm comprises:

[0248] Complete the initialization of the sand cat population, set the population size M = 30, the maximum number of iterations is T = 200, and calculate the position fitness value of each sand cat individual in the comprehensive index of the central node of each partition;

[0249] Based on the location fitness value of each sand cat individual, the location of the sand cat is updated according to the boot parameter R, including:

[0250] Sand cats use low-frequency noise to find prey. The sensitivity range of sand cats depends on their hearing ability in low-frequency detection and the low-frequency signals they perceive. Assuming that the hearing range of sand cats is from 2kHz to 0, their behavior is expressed by the following formula:

[0251]

[0252] In the above formula, r G For the hearing range of sand cats, the linear is changed to nonlinear. In the early iteration r G The slower change of is beneficial to the global search ability of the algorithm. G The faster change of is beneficial to the local search ability of the algorithm and the convergence of the algorithm. M To simulate the hearing of sand cats, its value is set to 2, and t is the current iteration number;

[0253] The bootstrap parameter that determines the transition from the exploration phase to the exploitation phase is R, which is calculated using the following formula:

[0254] R=2×rr G ;

[0255] r=r G ×rand(0,1);

[0256] In the above formula, R is the guiding parameter of the improved sand cat swarm optimization algorithm to control the transition between exploration and exploitation, r is the sensitivity range of each cat, and rand(0,1) is a random function between 0 and 1;

[0257] When |R|>1, i.e. the exploration phase, the cat will gradually approach its prey through the Lévy flying and wandering strategy, and will wander around the prey while approaching it. The cat will choose the best solution X according to the global optimal solution. bc (t), its current position X(t) and its sensitivity range r to update its own position;

[0258] The following formula is used to calculate the position update of the sand cat during the exploration phase:

[0259] X(t+1)=r×(X bc (t)-C×s×X(t));

[0260] In the above formula, X(t+1) is the location of the next exploration, C is a constant, let C = 0.35, s is the Lévy flight path of the sand cat, multiplied by the constant C to reduce the step length so that the sand cat can walk next to the prey as much as possible;

[0261] When |R|≤1, i.e., the development stage, the distance between the sand cat and the prey is very close, and each sand cat chooses a random angle to avoid falling into the local optimum while approaching the hunting position;

[0262] The following formula is used to calculate the position update of the sand cat during the development phase:

[0263] X(t+1)=X bc (t)-r·X rnd ·cos(θ);

[0264] X rnd =|rand(0,1)·X bc (t)-X(t)|;

[0265] In the above formula, X(t+1) is the location of the next development, X rnd is a random position vector representing the distance difference between the optimal position and the current position, and θ is a random angle;

[0266] After updating the position of the sand cat, determine whether the maximum number of iterations has been reached. If not, recalculate the position fitness value of each sand cat individual until the maximum number of iterations is reached and output the optimal central node of all partitions;

[0267] Calculate the comprehensive index value of the optimal central node of each partition, and set the minimum node comprehensive index preset value F of each partition according to the comprehensive index value of the optimal central node of each partition min , keep the nodes in each partition whose comprehensive index value is greater than the preset value unchanged, take other nodes in the partition as edge nodes, and use the improved sand cat group optimization algorithm to find the optimal adjacent partition of each edge node from the outside to the inside according to the network structure of the distribution network and the connection relationship between nodes, and divide the edge node into its optimal adjacent partition;

[0268] The minimum node comprehensive index preset value F of each partition min , calculated using the following formula:

[0269]

[0270] In the above formula, is the preset value of the minimum node comprehensive index of partition k, μ is the proportional coefficient, set μ = 0.8, is the comprehensive index value of the optimal central node of partition k;

[0271] After each operation of dividing the edge nodes in the partition, the comprehensive balance index of the distribution network is calculated once. After all the partitions of the distribution network are divided, the partition result with the largest comprehensive balance index is selected as the final partition result S of the active distribution network.

[0272] Embodiment 3:

[0273] like Figure 3 As shown, a system for dynamic partitioning of an active power distribution network includes a preliminary partitioning module, a re-partitioning module and a fine-tuning module;

[0274] When the network topology changes or the source load fluctuates extremely, the preliminary partitioning module, the repartitioning module, and the fine-tuning module all work:

[0275] The preliminary partitioning module is used to perform preliminary partitioning of the active distribution network based on the source-grid-load-storage data and network structure of the active distribution network, and determine the overlapping areas in the active distribution network;

[0276] The re-division module is used to partition and adjust the overlapping nodes in the overlapping area of ​​the active power distribution network to obtain the area re-division result;

[0277] The fine-tuning module is used to fine-tune the edge nodes in each partition based on the regional redivision result and consider the balance of active and reactive power in the partition to obtain the final partition result of the active distribution network.

[0278] When the distribution network operates normally or the source-load fluctuation is small, only the fine-tuning module works:

[0279] The fine-tuning module is used to consider the balance of active and reactive power within the partition, and to fine-tune the edge nodes within each partition to obtain the final partition result of the active distribution network.

[0280] The preliminary partitioning module includes a local similarity unit, a node partitioning unit and a partition similarity unit;

[0281] The local similarity unit is used to determine the local connection similarity of nodes in the active power distribution network using the following formula:

[0282]

[0283] In the above formula, w ij is the local connection similarity between node i and node j, k i is the degree of node i, A ij is the weight of the edge connecting node i and node j, k is the common neighboring node of node i and node j, S ij is the set of common neighbor nodes of node i and node j, k x is the degree of node x;

[0284] The node partitioning unit is used to partition the distribution network nodes by a fuzzy hierarchical clustering method based on the local connection similarity of the nodes in the active distribution network;

[0285] The partition similarity unit is used to calculate the interval similarity between partitions using the following formula. If the interval similarity between two partitions is greater than the set threshold Then these two partitions are merged to complete the preliminary partitioning of the active distribution network;

[0286]

[0287] In the above formula, δ(C i , C j ) is partition C i and partition C j The similarity between i ) is partition C i The nodes included, N(C j ) is partition C j The nodes contained.

[0288] The subdivision module includes an electrical distance construction unit, a modularity unit and an independent subdivision merging unit;

[0289] The electrical distance construction unit is used to construct the electrical distance between the distribution network nodes based on the sensitivity method and the transmission impedance method, and the calculation method thereof includes:

[0290] The electrical distance between the distribution network nodes based on the sensitivity method is calculated using the following formula:

[0291] D ij =1-L ij / max{L ij};

[0292]

[0293] In the above formula, D ij is the electrical distance between nodes i and j based on the sensitivity method, L ij is the influence of other nodes in the distribution network on the sensitivity of node i and node j, N is the number of nodes in the distribution network, α im is the sensitivity between node i and node m, α jm is the sensitivity between node j and node m;

[0294] The electrical distance between the nodes of the distribution network based on the transmission impedance method is calculated using the following formula:

[0295]

[0296] In the above formula, is the output power of branch ij, is the injected power of branch ij, z ij is the impedance of branch ij, representing the equivalent transmission impedance of power from node j to node j, and |z ij | is the electrical distance between node i and node j, is the equivalent impedance of the downstream of the branch ij;

[0297] Combining the two electrical distances, the following formula is used to calculate the electrical distance of the distribution network:

[0298]

[0299] In the above formula, D′ ij is the electrical distance between node i and node j, D ij , Z ij The weight of D ij is the electrical distance between nodes i and j based on the sensitivity method, |Z ij| is the electrical distance between node i and node j based on the transmission impedance method.

[0300] The modularity unit is used to determine the modularity of each partition based on the electrical distance of the distribution network using the following formula:

[0301]

[0302] A ij =D′ ij ;

[0303]

[0304] In the above formula, Q v is the modularity of partition v, e vv is the ratio of the total edge weight within partition v to the total edge weight of the entire network, a v is the ratio of the total edge weight of all nodes in partition v to the total edge weight of all nodes in the network, m is the sum of the weights of all edges in the distribution network, A ij is the edge weight between node i and node j, δ(c i , v) is partition C i The similarity between partitions v, δ(c j , v) is partition C j The similarity between the partition v and D′ ij is the electrical distance between node i and node j, k i is the sum of the weights of all edges connected to node i;

[0305] The independent sub-partition merging unit is used to take the overlapping nodes in the distribution network overlapping area determined by the preliminary partitioning as independent sub-partitions, and the non-overlapping areas other than the overlapping areas in a partition and the partitions without overlapping areas as independent partitions, and merge the independent sub-partitions into the adjacent independent partitions based on the modularity index to obtain the regional redivision result, including:

[0306] For each independent sub-partition, perform the following operations:

[0307] If the modularity increase after merging the independent sub-partition e into the independent partition u is greater than the modularity increase after merging e into the independent partition v, then e is merged into the independent partition u;

[0308] If the modularity reduction after merging the independent sub-partition e into the independent partition u is less than the modularity reduction after merging e into the independent partition v, then merge e into the independent partition u;

[0309] Calculating the reactive power reserve of each partition after the operation is performed on each independent sub-partition, judging whether the reactive power reserve of each partition satisfies the reactive power reserve constraint, and merging the partition that does not satisfy the reactive power reserve constraint with its adjacent partition that satisfies the reactive power reserve constraint;

[0310] Calculate the reactive power reserve of each partition after merging, and judge again whether the reactive power reserve of each partition meets the reactive power reserve constraint. For the partition that still does not meet the reactive power reserve constraint, repeatedly merge the partition with its adjacent partition that meets the reactive power reserve constraint until the reactive power reserve of each partition meets the reactive power reserve constraint. At this time, the regional redivision result is obtained.

[0311] The fine-tuning module includes a central node optimization unit, an edge node partitioning unit and a final partitioning unit;

[0312] The central node optimization unit is used to optimize the comprehensive indicators of nodes in each partition of the distribution network based on the regional redivision result by using the improved sand cat swarm optimization algorithm, and output the optimal central nodes of all partitions by iteratively updating particles;

[0313] The comprehensive index of the node is calculated using the following formula:

[0314] F cen =η1OBS+η2CON;

[0315] In the above formula, F cen is the comprehensive index of the node, η1 and η2 are the weights of OBS and CON respectively, OBS is the observability index of the node, and CON is the controllability index of the node;

[0316] The edge node division unit is used to calculate the comprehensive index value of the optimal central node of each partition, and set the minimum node comprehensive index preset value F of each partition according to the comprehensive index value of the optimal central node of each partition. min , keep the nodes in each partition whose comprehensive index value is greater than the preset value unchanged, take other nodes in the partition as edge nodes, and use the improved sand cat group optimization algorithm to find the optimal adjacent partition of each edge node from the outside to the inside according to the network structure of the distribution network and the connection relationship between nodes, and divide the edge node into its optimal adjacent partition;

[0317] The minimum node comprehensive index preset value F of each partition min , calculated using the following formula:

[0318]

[0319] In the above formula, is the preset value of the minimum node comprehensive index of partition k, μ is the proportional coefficient, set μ = 0.8, is the comprehensive index value of the optimal central node of partition k;

[0320] The final partitioning unit is used to calculate the comprehensive balance index of the distribution network once after each operation of partitioning the edge nodes in the partition. After all the partitions of the distribution network are divided, the partition result with the largest comprehensive balance index is selected as the final partition result S of the active distribution network;

[0321] The comprehensive balance index of the distribution network is calculated using the following formula:

[0322]

[0323] In the above formula, is the total active and reactive balance index of the comprehensive distribution network, λ1 and λ2 are The weight of is the total active power balance index of the distribution network, It is the total reactive power balance index of the distribution network.

Claims

1. A method for dynamic partitioning of an active distribution network, characterized in that: When the network topology changes or extreme source load fluctuations occur, the method includes: S1. Based on the source, grid, load and storage data of the active distribution network and its network structure, the active distribution network is preliminarily partitioned to determine the overlapping areas in the active distribution network, including: S11. The local connection similarity of nodes in the active distribution network is determined using the following formula: In the above formula, w ij is the local connection similarity between node i and node j, k i is the degree of node i, A ij is the weight of the edge connecting node i and node j, x is the common neighboring node of node i and node j, S ij is the set of common neighbor nodes of node i and node j, k x is the degree of node x; S12, partitioning the distribution network nodes by fuzzy hierarchical clustering method based on the local connection similarity of the nodes in the active distribution network; S13, use the following formula to calculate the partition similarity between each partition, if the interval similarity between two partitions is greater than the set threshold Then these two partitions are merged to complete the preliminary partitioning of the active distribution network; In the above formula, δ(C i , C j ) is partition C i and partition C j The similarity between i ) is partition C i The nodes included, N(C j ) is partition C j Contained nodes; S2, partitioning and adjusting the overlapping nodes in the overlapping area of ​​the active distribution network to obtain a regional redivision result; S3. Based on the regional redivision results, considering the balance of active and reactive power in the partition, fine-tune the edge nodes in each partition to obtain the final partition result of the active distribution network; When the distribution network operates normally or the source-load fluctuation is small, the method is: Considering the balance of active and reactive power within the partition, the edge nodes in each partition are fine-tuned to obtain the final partition result of the active distribution network.

2. A method for dynamic partitioning of an active power distribution network according to claim 1, characterized in that: The S2 includes: S21. Based on the electrical distance of the distribution network, the modularity of each partition is determined using the following formula: A ij =D′ ij ; In the above formula, Q v is the modularity of partition v, e vv is the ratio of the total edge weight within partition v to the total edge weight of the entire network, a v is the ratio of the total edge weight of all nodes in partition v to the total edge weight of all nodes in the network; m is the sum of the weights of all edges in the distribution network, A ij is the edge weight between node i and node j, δ(c i , v) is partition C i and the similarity between partition v, δ(c j , v) is partition C j The similarity between the partition v and D′ ij is the electrical distance between node i and node j, k i is the sum of the weights of all edges connected to node i; S22. The overlapping nodes in the overlapping area of ​​the distribution network determined by the preliminary partitioning are taken as independent sub-partitions, and the non-overlapping areas and the partitions without overlapping areas in a partition are taken as independent partitions. Based on the modularity index, the independent sub-partitions are merged into the adjacent independent partitions to obtain the regional redivision results, including: S221. Perform the following operations on each independent sub-partition: If the modularity increase after merging the independent sub-partition e into the independent partition u is greater than the modularity increase after merging e into the independent partition v, then e is merged into the independent partition u; If the modularity reduction after merging the independent sub-partition e into the independent partition u is less than the modularity reduction after merging e into the independent partition v, then merge e into the independent partition u; S222, calculating the reactive power reserve of each partition after the operation described in step S221, judging whether the reactive power reserve of each partition satisfies the reactive power reserve constraint, and merging the partition that does not satisfy the reactive power reserve constraint with its adjacent partition that satisfies the reactive power reserve constraint; S223. Calculate the reactive power reserve of each partition after merging, and judge again whether the reactive power reserve of each partition meets the reactive power reserve constraint. For the partitions that still do not meet the reactive power reserve constraint, repeat step S222 until the reactive power reserve of each partition meets the reactive power reserve constraint. At this time, the regional redivision result is obtained.

3. A method for dynamic partitioning of an active power distribution network according to claim 2, characterized in that: In S21, the electrical distance of the distribution network is an electrical distance that characterizes the power flow relationship and spatial position relationship of the distribution network and is constructed based on the sensitivity method and the transmission impedance method. The calculation method includes: S211. Calculate the electrical distances between the nodes of the distribution network based on the sensitivity method and the transmission impedance method respectively: The electrical distance between the distribution network nodes based on the sensitivity method is calculated using the following formula: D ij =1-L ij / max{L ij }; In the above formula, D ij is the electrical distance between nodes i and j based on the sensitivity method, L ij is the influence of other nodes in the distribution network on the sensitivity of node i and node j, N is the number of nodes in the distribution network, α im is the sensitivity between node i and node m, α jm is the sensitivity between node j and node m; The electrical distance between the nodes of the distribution network based on the transmission impedance method is calculated using the following formula: In the above formula, is the output power of branch ij, is the injected power of branch ij, z ij is the impedance of branch ij, representing the equivalent transmission impedance of power from node i to node j. ij | is the electrical distance between node i and node j, is the equivalent impedance of the downstream of the branch ij; S212. Combining the two electrical distances, the electrical distance of the distribution network is calculated using the following formula: In the above formula, D′ ij is the electrical distance between node i and node j, D ij , Z ij The weight of D ij is the electrical distance between nodes i and j based on the sensitivity method, |Z ij | is the electrical distance between node i and node j based on the transmission impedance method.

4. The method for dynamic partitioning of an active power distribution network according to claim 1, characterized in that: The balance of active power and reactive power in the partition is considered, and the edge nodes in each partition are fine-tuned to obtain the final partition result of the active distribution network, including: S31. Based on the regional redivision results, the improved sand cat swarm optimization algorithm is used to optimize the comprehensive indicators of nodes in each partition of the distribution network, and the optimal central nodes of all partitions are output by iteratively updating particles; The comprehensive index of the node is calculated using the following formula: F cen =η1OBS+η2CON; In the above formula, F cen is the comprehensive index of the node, η1 and η2 are the weights of OBS and CON respectively, OBS is the observability index of the node, and CON is the controllability index of the node; S32, calculate the comprehensive index value of the optimal central node of each partition, and set the minimum node comprehensive index preset value F of each partition according to the comprehensive index value of the optimal central node of each partition min , keep the nodes in each partition whose comprehensive index value is greater than the preset value unchanged, take the other nodes in the partition as edge nodes, and use the improved sand cat group optimization algorithm to find the optimal adjacent partition of each edge node from the outside to the inside according to the network structure of the distribution network and the connection relationship between nodes, and divide the edge node into its optimal adjacent partition; S33, after each operation of dividing the edge nodes in the partition, the comprehensive balance index of the distribution network is calculated once, and after all the partitions of the distribution network are divided, the partition result with the largest comprehensive balance index is selected as the final partition result S of the active distribution network; The comprehensive balance index of the distribution network is calculated using the following formula: In the above formula, is the total active and reactive balance index of the comprehensive distribution network, λ1 and λ2 are The weight of is the total active power balance index of the distribution network, It is the total reactive power balance index of the distribution network.

5. A system for dynamic partitioning of an active distribution network, characterized in that: The system includes a preliminary partitioning module, a repartitioning module and a fine-tuning module; When the network topology changes or there are extreme source load fluctuations: The preliminary partitioning module is used to perform preliminary partitioning of the active distribution network based on the source-grid-load-storage data of the active distribution network and its network structure, and determine the overlapping areas in the active distribution network, including a local similarity unit, a node partitioning unit and a partition similarity unit; The local similarity unit is used to determine the local connection similarity of nodes in the active power distribution network using the following formula: In the above formula, w ij is the local connection similarity between node i and node j, k i is the degree of node i, A ij is the weight of the edge connecting node i and node j, x is the common neighboring node of node i and node j, S ij is the set of common neighbor nodes of node i and node j, k x is the degree of node x; The node partitioning unit is used to partition the distribution network nodes by a fuzzy hierarchical clustering method based on the local connection similarity of the nodes in the active distribution network; The partition similarity unit is used to calculate the interval similarity between partitions using the following formula. If the interval similarity between two partitions is greater than the set threshold Then these two partitions are merged to complete the preliminary partitioning of the active distribution network; In the above formula, δ(C i , C j ) is partition C i and partition C j The similarity between i ) is partition C i The nodes included, N(C j ) is partition C j Contained nodes; The re-division module is used to perform partition adjustment on the overlapping nodes in the overlapping area of ​​the active power distribution network to obtain the area re-division result; The fine-tuning module is used to fine-tune the edge nodes in each partition based on the regional redivision result and consider the balance of active and reactive power in the partition to obtain the final partition result of the active distribution network; When the distribution network operates normally or the source load fluctuation is small: The fine-tuning module is used to consider the balance of active and reactive power within the partition, and to fine-tune the edge nodes within each partition to obtain the final partition result of the active distribution network.

6. The system for dynamic partitioning of an active power distribution network according to claim 5, characterized in that: The repartitioning module includes a modularity unit and an independent sub-partition merging unit; The modularity unit is used to determine the modularity of each partition based on the electrical distance of the distribution network using the following formula: A ij =D′ ij ; In the above formula, Q v is the modularity of partition v, e vv is the ratio of the total edge weight within partition v to the total edge weight of the entire network, a v is the ratio of the total edge weight of all nodes in partition v to the total edge weight of all nodes in the network; m is the sum of the weights of all edges in the distribution network, A ij is the edge weight between node i and node j, δ(c i , v) is partition C i and the similarity between partition v, δ(c j , v) is partition C j The similarity between the partition v and D′ ij is the electrical distance between node i and node j; k i is the sum of the weights of all edges connected to node i; The independent sub-partition merging unit is used to take the overlapping nodes in the distribution network overlapping area determined by the preliminary partitioning as independent sub-partitions, and the non-overlapping areas other than the overlapping areas in a partition and the partitions without overlapping areas as independent partitions, and merge the independent sub-partitions into the adjacent independent partitions based on the modularity index to obtain the regional redivision result, including: For each independent sub-partition, perform the following operations: If the modularity increase after merging the independent sub-partition e into the independent partition u is greater than the modularity increase after merging e into the independent partition v, then e is merged into the independent partition u; If the modularity reduction after merging the independent sub-partition e into the independent partition u is less than the modularity reduction after merging e into the independent partition v, then merge e into the independent partition u; Calculating the reactive power reserve of each partition after the operation is performed on each independent sub-partition, judging whether the reactive power reserve of each partition satisfies the reactive power reserve constraint, and merging the partition that does not satisfy the reactive power reserve constraint with its adjacent partition that satisfies the reactive power reserve constraint; Calculate the reactive power reserve of each partition after merging, and judge again whether the reactive power reserve of each partition meets the reactive power reserve constraint. For the partition that still does not meet the reactive power reserve constraint, repeatedly merge the partition with its adjacent partition that meets the reactive power reserve constraint until the reactive power reserve of each partition meets the reactive power reserve constraint. At this time, the regional redivision result is obtained.

7. The system for dynamic partitioning of an active power distribution network according to claim 6, characterized in that: The subdivision module further comprises an electrical distance construction unit; The electrical distance construction unit is used to construct the electrical distance between the distribution network nodes based on the sensitivity method and the transmission impedance method, and the calculation method thereof includes: The electrical distance between the distribution network nodes based on the sensitivity method is calculated using the following formula: D ij =1-L ij / max{L ij }; In the above formula, D ij is the electrical distance between nodes i and j based on the sensitivity method, L ij is the influence of other nodes in the distribution network on the sensitivity of node i and node j, N is the number of nodes in the distribution network, α im is the sensitivity between node i and node m, α jm is the sensitivity between node j and node m; The electrical distance between the nodes of the distribution network based on the transmission impedance method is calculated using the following formula: In the above formula, is the output power of branch ij, is the injected power of branch ij, z ij is the impedance of branch ij, representing the equivalent transmission impedance of power from node i to node j. ij | is the electrical distance between node i and node j, is the equivalent impedance of the downstream of the branch ij; Combining the two electrical distances, the following formula is used to calculate the electrical distance of the distribution network: In the above formula, D′ ij is the electrical distance between node i and node j, D ij , Z ij The weight of D ij is the electrical distance between nodes i and j based on the sensitivity method, |Z ij | is the electrical distance between node i and node j based on the transmission impedance method.

8. The system for dynamic partitioning of an active power distribution network according to claim 5, characterized in that: The fine-tuning module includes a central node optimization unit, an edge node partitioning unit and a final partitioning unit; The central node optimization unit is used to optimize the comprehensive indicators of nodes in each partition of the distribution network based on the regional redivision result by using the improved sand cat swarm optimization algorithm, and output the optimal central nodes of all partitions by iteratively updating particles; The comprehensive index of the node is calculated using the following formula: F cen =η1OBS+η2CON; In the above formula, F cen is the comprehensive index of the node, η1 and η2 are the weights of OBS and CON respectively, OBS is the observability index of the node, and CON is the controllability index of the node; The edge node division unit is used to calculate the comprehensive index value of the optimal central node of each partition, and set the minimum node comprehensive index preset value F of each partition according to the comprehensive index value of the optimal central node of each partition. min , keep the nodes in each partition whose comprehensive index value is greater than the preset value unchanged, take the other nodes in the partition as edge nodes, and use the improved sand cat group optimization algorithm to find the optimal adjacent partition of each edge node from the outside to the inside according to the network structure of the distribution network and the connection relationship between nodes, and divide the edge node into its optimal adjacent partition; The final partitioning unit is used to calculate the comprehensive balance index of the distribution network once after each operation of partitioning the edge nodes in the partition. After all the partitions of the distribution network are divided, the partition result with the largest comprehensive balance index is selected as the final partition result S of the active distribution network; The comprehensive balance index of the distribution network is calculated using the following formula: In the above formula, is the total active and reactive balance index of the comprehensive distribution network, λ1 and λ2 are The weight of is the total active power balance index of the distribution network, It is the total reactive power balance index of the distribution network.

Images