Dynamic partitioning method of active distribution network based on reactive power flow tracking and branch cutting

Abstract

The present invention proposes a method for dynamic partitioning of active distribution network based on reactive power flow tracking and branch cutting. First, the VVS of each reactive source node to the remaining nodes is calculated by adding voltage reactive output perturbation to the reactive source; secondly, the VVS is fuzzified to obtain the membership matrix between the nodes, and the power source nodes are pre-partitioned according to the membership relationship between the reactive source nodes, and the dominant nodes and the number of partitions are preliminarily determined; then, the branch range to be cut is determined based on the node power flow direction and the voltage reactive power partitioning principle; the optimal branch cutting point is found with the regional comprehensive coupling degree as the objective function of BPSO, and the partitioning result is determined; finally, the partitioning result is subjected to reactive reserve verification to obtain the final partitioning result. The VVS index proposed by the present invention changes dynamically with the changes in DG output and load, and the proposed "reactive source node pre-partitioning - load node mapping partitioning" idea ensures the effectiveness of regional voltage reactive power control.

Description

Technical Field

[0001] The invention relates to the technical field of distribution network partitioning, and in particular to an active distribution network dynamic partitioning method based on reactive power flow tracking and branch cutting. Background Art

[0002] Building a new power system with new energy as the main body is an important way to help achieve the "dual carbon" goals. As an important carrier for the consumption and efficient utilization of distributed new energy (Distributed Generation, DG), the coordinated optimization of multiple types of adjustable resources in the active distribution network (ADN) is an important problem that needs to be solved urgently. A large number of DGs are connected to the distribution network, which enables the traditional distribution network to have a certain active adjustment capability. However, at the same time, the number of control variables increases and the convergence of centralized optimization decreases. Distributed control schemes have been proven to effectively cope with the shortcomings of traditional centralized voltage control methods and promote effective coordination between DG and other controllable devices. Therefore, in order to achieve effective distributed control, a reasonable network partitioning method is of great significance to achieve distributed optimization of voltage control.

[0003] ADN partitioning is to divide the entire power grid into several sub-areas. Its characteristics are that the relationship between internal nodes is very strong, and the relationship between nodes in different sub-areas is very weak. At present, most distribution network partitions are based on electrical distance. There are many definitions of electrical distance, of which two are more common. One is to define electrical distance using the topological information of the power grid itself, such as node admittance matrix, impedance matrix, etc., but this type of method cannot reflect the power flow state of the system and the advantages of reactive power in voltage control. The other type is to define electrical distance using the voltage / reactive sensitivity (VVS) between nodes. Among them, the most common VVS matrix calculation method is the Jacobian matrix of the Newton method. The eigenvector corresponding to its minimum eigenvalue is closely related to the reactive power distribution, and can be used as the basis for regional division in secondary voltage control. However, the ΔQ value of this method is very small (usually less than 10E-5), which cannot characterize the VVS of the remaining nodes when the ΔQ of the reactive source changes greatly. In addition, the VVS matrix is ​​an asymmetric matrix. The existing method uses mathematical methods to process it and obtain a symmetrical electrical distance matrix, but the electrical distance cannot fully characterize the sensitivity relationship between the two nodes.

[0004] At present, the voltage and reactive power partitioning of distribution networks can be divided into clustering partitioning, multi-objective optimization partitioning, partitioning based on complex network theory, etc. Most literatures consider the compactness of the topological structure within the region and the appropriate reactive power margin in the region when partitioning. However, the conventional clustering method cannot guarantee the rationality of the partition module when determining the number of partitions, so the modularity function is introduced. The introduction of the modularity function makes the partitioning more accurate and the number of partitions can be automatically determined. Although the above methods comprehensively consider the electrical characteristics and reactive power regulation capabilities between partition nodes, the changes in DG output and load levels have a certain impact on the voltage and reactive power partitioning results, especially the reactive power regulation capability of some DGs is closely related to their active output. Therefore, the partitioning should change dynamically with the time sequence output of DG and load and the change of network topology. Summary of the invention

[0005] In view of the problems existing in the above-mentioned background technology, the present invention proposes a dynamic partitioning method for active distribution network based on reactive power flow tracking and branch cutting. Firstly, a VVS calculation method that can reflect the time series change of "source-load" is proposed, and the VVS matrix between nodes is fuzzified to obtain the membership matrix between nodes; secondly, the power supply nodes are pre-partitioned according to the membership relationship between reactive source nodes, and the dominant nodes and the number of partitions are preliminarily determined; then, the range of branches to be cut is determined based on the reactive power flow direction of the branches, and the optimal branch cutting point is found with the regional comprehensive coupling degree as the objective function of BPSO to determine the partitioning result; finally, the partitioning result is checked for reactive reserve, and the partitions that do not meet the reactive reserve requirements are merged to obtain the final partitioning result.

[0006] The technical solution of the present invention is achieved in this way:

[0007] A method for dynamic partitioning of active distribution network based on reactive power flow tracking and branch cutting, the steps of which are as follows:

[0008] Step 1: Calculate the voltage / reactive sensitivity matrix of each reactive source node to the other nodes by adding voltage reactive output perturbation to the reactive source;

[0009] Step 2: Fuzzify the voltage / reactive sensitivity matrix to obtain the membership matrix between nodes; analyze the membership relationship between reactive source nodes, pre-partition the reactive source nodes, and preliminarily determine the dominant nodes and the number of partitions;

[0010] Step 3: According to the pre-partitioning results, determine the range of branches to be cut based on the node power flow direction and voltage and reactive power partitioning principles;

[0011] Step 4: Based on the range of the branch to be cut, the binary particle swarm algorithm is used to find the optimal cutting point of the branch, and the partition result is determined by the optimal cutting point of the branch;

[0012] Step 5: Perform reactive power reserve check on the partition results. When the reactive power reserve requirements are met, output the partition results. Otherwise, merge the partitions and output the final partition results.

[0013] Preferably, the voltage / reactive power sensitivity matrix is ​​calculated as follows:

[0014]

[0015] Where deltQ k is the reactive power output perturbation change vector of the kth reactive source node, S VQ is the voltage / reactive power sensitivity matrix, the element (S VQ ) ij It represents the voltage change of controlled node i when the reactive power output disturbance of reactive source node j changes, dV is the change of load node voltage, and dQ is the change of reactive power of power source node.

[0016] Preferably, the membership matrix is ​​calculated as follows:

[0017]

[0018] In the formula, G represents the number of reactive source nodes, μ ij It represents the membership degree of the i-th controlled node belonging to the j-th reactive source node.

[0019] Preferably, the method for pre-partitioning reactive source nodes is: calculating the membership between reactive source nodes through a membership matrix, partitioning the power supply nodes according to the membership values, and preliminarily determining the dominant node and the number of partitions.

[0020] Preferably, the voltage and reactive power zoning principle refers to the number of zones, the rationality of the zone size and the connectivity of the nodes within the zone.

[0021] Preferably, the method of finding the optimal cutting point of a branch using a binary particle swarm algorithm is:

[0022] S41, initialization: input the data of the distribution network and set the parameters of the binary particle swarm algorithm; wherein the data of the distribution network includes the topological information of the distribution network, the DG output and the load; the parameters of the binary particle swarm algorithm include the particle swarm size N, the inertia factor w and the basic values ​​of the learning factors c1 and c2, and the maximum number of iterations;

[0023] S42, determining the dimension of the particle: the dimension of the particle is the sum of all interconnected branches in the distribution network, and the search space of each dimension is the branch cutting range of the partition where it is located;

[0024] S43, establish an objective function, randomly generate M solutions within the branch cutting range, calculate their respective objective functions, and take the solution corresponding to the minimum value as the current optimal solution;

[0025] S44, initializing the state of each dimension of the particle according to the position update rule of the binary particle swarm algorithm, and randomly initializing the speed of the particle;

[0026] S45, updating the optimal solution of each particle and the global optimal solution;

[0027] S46, determine whether the maximum number of iterations has been reached, if so, the iteration ends and the result is output; otherwise, go to step S47;

[0028] S47, update the speed and position of the particle, increase the number of iterations by 1, and go to step S45.

[0029] Preferably, the objective function is:

[0030]

[0031] in, It represents the product of the average membership between load nodes in region k and the maximum membership between any two nodes in the system. It represents the product of the average membership degree between load nodes in region k and the minimum membership degree of controlled nodes that do not belong to this region. k is the number of controlled nodes in region k.

[0032] Preferably, the position update rule of the binary particle swarm algorithm is as follows: the number of branches of the distribution network is n, which is divided into m dimensions according to the branch cutting range, 1) at most one of the cutting branches appearing in only one partition is in a disconnected state; 2) when encoding the switch states of the common branches coexisting in two partitions, more than two switches are not allowed to be opened; 3) the branch switches directly connected to the power supply must be in a closed state.

[0033] Preferably, the speed and position of the particle are updated by:

[0034]

[0035]

[0036] In the formula, is the position of particle i in the kth dimension, is the flying speed of particle i in the kth dimension, is the flight speed of particle i in the k-1th dimension, ω is the inertia factor, r1 and r2 are both random numbers in the interval [0,1], is the optimal position of particle i in the kth dimension, is the optimal position of the k-th dimensional population, rand() is a random number in the interval [0,1]; S(v id ) is the sigmoid function, S(v id )=1 / (1+exp(-v id )).

[0037] Preferably, the method for performing reactive power reserve verification on the partition result is:

[0038] Calculate reactive power reserve index β k for:

[0039] β k =(1-Q Lk / Q Gk )×100%;

[0040] In the formula, Q Gk is the total reactive power reserve in area k, Q Lk Reactive power reserve generated by loads in the area.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] 1) The zoning scheme based on branch cutting and reactive power flow tracking ensures the regional reactive balance and system connectivity; the regional comprehensive coupling degree is used as the objective function of BPSO to find the optimal branch cutting point, ensuring strong coupling within the partition and weak coupling between partitions.

[0043] 2) The VVS index changes dynamically with the changes in DG output and load. The proposed idea of ​​"reactive source node pre-partitioning-load node mapping partitioning" ensures the effectiveness of regional voltage reactive power control.

[0044] 3) Based on the timing characteristics of "source-load" and dynamic partitioning of reactive power reserve, the flexibility of partition voltage and reactive power regulation and the real-time matching of operating state changes are ensured, thus ensuring stable operation of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only 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.

[0046] Figure 11 is a comparison diagram of the VVS curves obtained by the method of the present invention and the Newton method; wherein, (a) the VVS of reactive source 34 node to each node, (b) the VVS of reactive source 35 node to each node, (c) the VVS of reactive source 36 node to each node, (d) the VVS of reactive source 37 node to each node, and (e) the VVS of reactive source 38 node to each node.

[0047] Figure 2 It is a partition flow chart of the present invention.

[0048] Figure 3 The present invention optimizes the selection and division of variables; wherein, (a) the distribution network is radially designed, and (b) the generation forest formed by finding the cutting points is obtained. DETAILED DESCRIPTION

[0049] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0050] Most of the existing sensitivity analysis methods are based on the power flow equation, and its basic equation is the node power balance equation. The state equations for establishing the active power and reactive power balance equations of the PQ node and the active power balance equation of the PV node are:

[0051] F(X,U,α)=0 (1)

[0052] Where X is the state variable, including the voltage amplitude V of the load node L , phase angle δ L and the phase angle δ of the generator node g ; U is the control variable, including the active power P of the load node L and reactive power Q L , the active power P of the generator node g and voltage V g , the voltage of the equilibrium node V o and phase angle δ o etc.; α is the relevant parameters for power flow calculation, including constant quantities such as line admittance parameters G and B.

[0053] There are many sensitivity indicators of state variables, including dV which reflects the change of load node voltage with load power. Li / dP Li and dV Li / dQ Li ; There is an indicator dQ that reflects the change of generator reactive power with load powergi / dQ Lj and dQ g∑ / dQ Lj Ignoring the relationship between the control variables, we differentiate equation (1) and get the state variable to control variable U k The sensitivity expression is:

[0054]

[0055] From formula (2), we can get the sensitivity matrix expression:

[0056]

[0057] The complex relationship between the control variables is linearized to obtain the control vector U and the control variable U concerned in the sensitivity calculation. k The linear relationship expression between ΔU=AΔU k and the total differential expression dU / dU k =A.

[0058] Depending on the selected variables, different types of sensitivity indicators can be constructed, which can be applied to different scenarios in the power system to solve problems such as voltage stability assessment, weak node determination, and reactive compensation point selection. In the voltage-reactive control partition, the reactive power of the power source node and the voltage of the load node are selected to construct the voltage-reactive sensitivity index, which is used to study the impact of the reactive power output change of the reactive source on the load node voltage. According to formula (3), the calculation formula of this indicator is as follows:

[0059] dV=S VQ dQ (4)

[0060]

[0061] In the formula, S VQ is the sensitivity of the load node voltage V to the power node reactive power Q, S Vα is the sensitivity of the load node voltage V to the power flow related parameter α, dV is the change in the load node voltage, dQ is the change in the reactive power of the power node, (dV / dQ) ij is the reactive sensitivity of the voltage at node i to node j, i.e. dV i / dQ j .

[0062] The above sensitivity calculation ignores the relationship between the control variables, so there are the following defects:

[0063] (1) The dynamic characteristics of the load are not considered, and the sensitivity changes under different load levels cannot be reflected;

[0064] (2) The connectivity of the system cannot be guaranteed and can only be verified and merged afterwards;

[0065] (3) The transmission network X>>R, so the voltage is strongly correlated with reactive power and weakly correlated with active power. The effect of active power on voltage can be ignored. However, the distribution network has a large R / X ratio, and the effect of active power on voltage cannot be ignored.

[0066] (4) Important constraints that affect the accuracy of sensitivity calculations, such as generator reactive power exceeding the limit and active power economic dispatch, are ignored.

[0067] Therefore, it is necessary to analyze the impact of the above factors on the sensitivity index, which is of great significance for the evaluation of the rationality of the partition.

[0068] Therefore, the embodiment of the present invention provides a method for dynamic partitioning of active distribution network based on reactive power flow tracking and branch cutting, such as Figure 2 As shown in the figure, the VVS of each reactive source to the remaining nodes is calculated by adding voltage reactive output perturbation to the reactive source, and the membership between nodes is calculated at the same time. Based on the above calculation results, a reactive dynamic partitioning method based on reactive power flow tracking is designed to cut branches in the distribution network using the BPSO algorithm. First, the pre-partition is determined according to the membership of the power supply node, and the dominant node and the number of partitions are preliminarily determined; secondly, according to the pre-partitioning results, the range of branches to be cut is determined based on the system flow direction combined with the voltage reactive power partitioning principle; then, the optimal cutting point of the branch is determined based on BPSO; finally, the partitioning results are checked for reactive reserve, and the partitions that do not meet the reactive reserve requirements are merged to obtain the final partitioning results.

[0069] The specific steps are as follows:

[0070] Step 1: Calculate the voltage / reactive sensitivity matrix of each reactive source node to the other nodes by adding voltage reactive output perturbation to the reactive source;

[0071] Strictly speaking, in actual systems, the control variables are not completely independent, and their changes have certain correlations. Based on the relationship between the control variables, we can get the control vector U and the control variable U concerned in the sensitivity calculation. k The linear relationship expression between ΔU=AΔU k , total differential dU / dU k =A. State variable to control variable U k The sensitivity expression is:

[0072]

[0073] The sensitivity matrix is ​​obtained by considering the perturbation of reactive power source output in voltage and reactive power control partition:

[0074]

[0075] Where deltQ k is the reactive power output perturbation change vector of the kth reactive source node, S VQ is the voltage / reactive power sensitivity matrix, the element (S VQ ) ij It represents the voltage change of controlled node i when the reactive power output disturbance of reactive source node j changes, dV is the change of load node voltage, and dQ is the change of reactive power of power source node.

[0076] The selection of reactive source perturbation is related to the type of reactive source. When the reactive source node is a PQ type node, the perturbation of the capacitor bank is deltQ G The single group capacity can be taken, the deltQ of DG G The installed capacity S DG The reactive power output perturbation of DG is an empirical value, which can be determined according to the output time series change curve of DG (daily output prediction curve). When the output changes slowly, a lower ratio is selected, such as 1% to 2%. When the output changes rapidly, a higher ratio is selected, such as 5%. When the load time series curve in the area is increasing, a positive value is selected, otherwise a negative value is taken. When DG is a PV type node, the voltage of the PV node is used as the control variable, and different voltage disturbance deltV is selected. Q changes with deltV to obtain deltQ G , and then we can get ΔV / deltQ G .

[0077] Figure 1 The following is a comparison chart of the sensitivity curves obtained by Newton-Raphson power flow calculation and reactive perturbation power flow calculation. The following conclusions can be drawn from the analysis of these curves:

[0078] (1) The VVS index curve obtained by the VVS analysis method of the present invention is consistent with the physical concept and is consistent with the change trend of the VVS curve obtained by the Newton-Raphson power flow algorithm, which shows the correctness of the method of the present invention.

[0079] (2) The method of the present invention can reflect the VVS under different load levels, which is more reasonable than the sensitivity analysis of voltage and reactive power control characteristics obtained by using only the Newton-Raphson power flow algorithm.

[0080] (3) If the VVS index is obtained by adding reactive power perturbation, it will be smaller and can better reflect the sensitive characteristics between voltage and reactive power.

[0081] Step 2: Fuzzify the voltage / reactive sensitivity matrix to obtain the membership matrix between nodes; analyze the membership relationship between reactive source nodes and pre-partition the reactive source nodes;

[0082] In order to measure the closeness of the electrical connection between nodes, the VVS is fuzzy divided, and the membership function is established to solve the membership matrix between nodes. The main idea is to make the VVS objects divided into the same category have the greatest similarity, while different categories have differences, which can reflect the coupling degree between nodes more intuitively and vividly.

[0083] (1) Basic knowledge of fuzzy sets

[0084] The membership function represents the degree to which an object x belongs to the data set A, denoted by μ A (x), object x is all data objects contained in the space where set A is located, and the membership value range is [0,1], that is, 0<=μ A (x) <= 1. μ A (x)=1 means x∈A, that is, x completely belongs to A.

[0085] The membership function defined on the space X = {x} is equivalent to defining a fuzzy set A, that is, a fuzzy subset A defined on the domain X = {x}. For a finite number of objects x1, x2, ..., x n The fuzzy set A can be expressed as:

[0086] A={μ A (x i ),x i |x i ∈X} (8)

[0087] The degree of membership of each data point to a certain class is represented by a value in the interval [0,1].

[0088] (2) Membership matrix based on VVS

[0089] The data to be classified is the required sensitivity matrix S VQ Element (S VQ ) ij The i-th row is the control node, and the j-th column is the reactive node. Then the output membership matrix A is:

[0090]

[0091] In the formula, G represents the number of reactive source nodes, μ ij It represents the membership of the i-th controlled node belonging to the j-th reactive source node. When there is only one reactive source in a partition, the membership of node i to the reactive source is the membership of the partition; when there are multiple reactive sources in a partition, the sum of the memberships of node i to all reactive sources in the partition is the membership of the partition.

[0092] Since the voltage of the load node changes through the regulation of the reactive voltage of the power node, the voltage control ability of the reactive source radiates to the surrounding nodes starting from itself, and the control ability of the nodes gradually weakens from near to far. Therefore, the membership between the reactive source nodes is calculated by formula (9). The value of the membership is the value of one node belonging to another node, and its size is the value of the node relative to the other nodes. The larger the value of the membership between the two nodes, the stronger the coupling relationship. The power nodes with strong coupling relationships are divided into the same area, the power nodes are partitioned, and the dominant nodes and the number of partitions are preliminarily determined.

[0093] Step 3: According to the pre-division results, the range of branches to be cut is determined based on the node flow direction and voltage-reactive power division principle; the main factor affecting the voltage amplitude in steady-state operation is the strength of reactive power. Too much or too little reactive power will cause voltage collapse. First, according to the number of reactive source pre-divisions, the range of branches to be cut is determined based on the node flow direction combined with the number of divisions in the voltage-reactive power division principle, the reasonableness of the division scale, and the connectivity within the division.

[0094] Step 4: Based on the range of the branch to be cut, the binary particle swarm algorithm is used to find the optimal cutting point of the branch, and the partition result is determined by the optimal cutting point of the branch;

[0095] Kennedy and Eberhart proposed the BPSO algorithm to solve discrete variable optimization problems. The algorithm uses binary coding, assumes that the probability of the particle position is 1 or 0, and selects the sigmoid function to convert the particle speed to the interval [0,1]. The speed and position update rules of each iteration particle are as follows:

[0096]

[0097]

[0098] In the formula, is the position of particle i in the kth dimension, is the flying speed of particle i in the kth dimension, is the flight speed of the k-1th dimension particle i, ω is the inertia factor, which is generally 0.8 to 1.2, r1 and r2 are random numbers in the interval [0,1], is the best position found so far by particle i in the kth dimension, is the best position found so far by the k-th dimensional population, rand() is a random number in the interval [0,1]; S(v id ) is the sigmoid function, S(v id )=1 / (1+exp(-v id )). To prevent saturation, the velocity is clamped to the interval [-4.0,4.0].

[0099] All the switches in the distribution network are closed to form a radial spanning tree. The essence of distribution network partitioning is to find branch cutting points to generate a forest. Figure 3 (a) is a radial design of a distribution network with 11 nodes. If branch {2,3} is the optimal cutting point, Figure 3 (b) is the spanning forest formed based on the cutting points. Therefore, the states of all branches are used as optimization variables, with 0 representing a cut branch and 1 representing a connected branch. The state of a branch is used as an optimization variable, which corresponds to a one-dimensional component in the particle vector position. All the states that form a connected branch are combined to represent the position of the particle, and the dimension of the position vector is the sum of the states of all branches that form the partition.

[0100] Using the basic BPSO algorithm, the number of dimensional components of 0 in the position vector updated in each iteration is not fixed, and isolated nodes are prone to appear. Based on the division of optimization variables, the position update rule of the basic BPSO algorithm is improved as follows: Assuming that the distribution network is divided into n branches, it is divided into m subdimensions according to the branch cutting range. 1) Only one cutting branch that appears in one partition is in the disconnected state at most; 2) When encoding the common branch switch states in two partitions, it is not allowed to open more than two; 3) The branch switch directly connected to the power supply must be in the closed state.

[0101] The method of using binary particle swarm algorithm to find the optimal cutting point of the branch is:

[0102] S41, initialization: input the data of the distribution network and set the parameters of the binary particle swarm algorithm; wherein the data of the distribution network includes the topological information of the distribution network, the DG output and the load; the parameters of the binary particle swarm algorithm include the particle swarm size N, the inertia factor w and the basic values ​​of the learning factors c1 and c2, and the maximum number of iterations;

[0103] S42, determining the dimension of the particle: the dimension of the particle is the sum of all interconnected branches in the distribution network, and the search space of each dimension is the branch cutting range of the partition where it is located;

[0104] S43, establish an objective function, randomly generate M solutions within the branch cutting range, calculate their respective objective functions, and take the solution corresponding to the minimum value as the current optimal solution;

[0105] According to the output membership matrix, the coupling index within sub-region k is calculated:

[0106]

[0107] In the formula, l kis the number of controlled nodes in zone k. This formula represents the product of the average membership between load nodes in zone k and the maximum membership between any two nodes in the system. The larger the value of this formula, the stronger the coupling degree in the region.

[0108] The inter-regional coupling index can be expressed as:

[0109]

[0110] This formula represents the product of the average membership between the k load nodes in the region and the minimum membership of the controlled nodes that do not belong to this region; the smaller the value of this formula, the weaker the coupling between regions.

[0111] M solutions are randomly generated within the range of the control variable, and their respective objective functions are calculated. The objective function of the comprehensive intra-regional coupling index and inter-regional coupling index can be expressed as:

[0112]

[0113] Take the solution corresponding to the minimum value in F as the current optimal solution.

[0114] S44, initializing the state of each dimension of the particle according to the position update rule of the binary particle swarm algorithm, and randomly initializing the speed of the particle;

[0115] S45, updating the optimal solution of each particle and the global optimal solution;

[0116] S46, determine whether the maximum number of iterations has been reached, if so, the iteration ends and the result is output; otherwise, go to step S47;

[0117] S47, update the particle's velocity and position according to equations (10) and (11), increase the number of iterations by 1, and go to step S45.

[0118] Step 5: Perform reactive power reserve check on the partition results. When the reactive power reserve requirements are met, output the partition results. Otherwise, merge the partitions and output the final partition results.

[0119] An appropriate amount of reactive power margin should be left in the partition to ensure the stability of the voltage control partition. Assume that the total reactive power reserve of area k is Q Gk , the reactive power generated by the load in the area is Q Lk , then the reactive power reserve index β k for:

[0120] β k =(1-Q Lk / Q Gk )×100% (15)

[0121] When the system is operating normally, β kThe reactive power reserve must be greater than or equal to a certain margin, here 10%. If the reactive power reserve does not meet the requirements, the partition should be adjusted.

[0122] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for dynamic partitioning of active distribution network based on reactive power flow tracking and branch cutting, characterized in that: The steps are as follows: Step 1: Calculate the voltage-reactive sensitivity matrix of each reactive source node to the other nodes by adding the voltage-reactive output perturbation to the reactive source; The voltage-reactive power sensitivity matrix is ​​calculated as follows: Where deltQ k is the reactive power output perturbation change vector of the kth reactive source node, S VQ is the voltage / reactive power sensitivity matrix, the element (S VQ ) ij It represents the voltage change of controlled node i when the reactive power output disturbance of reactive source node j changes, dV is the change of load node voltage, and dQ is the change of reactive power of power source node; Step 2: Fuzzify the voltage-reactive sensitivity matrix to obtain the membership matrix between nodes; analyze the membership relationship between reactive source nodes, pre-partition the reactive source nodes, and preliminarily determine the dominant nodes and the number of partitions; The calculation method of the membership matrix is: In the formula, G represents the number of reactive source nodes, μ ij represents the membership degree of the i-th controlled node belonging to the j-th reactive source node; The method for pre-partitioning reactive source nodes is as follows: the membership between reactive source nodes is calculated through the membership matrix, the power source nodes are partitioned according to the membership values, and the dominant nodes and the number of partitions are preliminarily determined; Step 3: According to the pre-partitioning result, the range of the branch to be cut is determined based on the node power flow direction and the voltage and reactive power partitioning principle; the voltage and reactive power partitioning principle refers to the number of partitions, the rationality of the partition scale and the connectivity of the nodes in the partition; Step 4: Based on the range of the branch to be cut, the binary particle swarm algorithm is used to find the optimal cutting point of the branch, and the partition result is determined by the optimal cutting point of the branch; Step 5: Perform reactive power reserve check on the partition results. When the reactive power reserve requirements are met, output the partition results. Otherwise, merge the partitions and output the final partition results.

2. The method for dynamic partitioning of active distribution network based on reactive power flow tracking and branch cutting according to claim 1 is characterized in that: The method of using binary particle swarm algorithm to find the optimal cutting point of the branch is: S41, initialization: input the data of the distribution network and set the parameters of the binary particle swarm algorithm; wherein the data of the distribution network includes the topological information of the distribution network, the DG output and the load; the parameters of the binary particle swarm algorithm include the particle swarm size N, the inertia factor w and the basic values ​​of the learning factors c1 and c2, and the maximum number of iterations; S42, determining the dimension of the particle: the dimension of the particle is the sum of all interconnected branches in the distribution network, and the search space of each dimension is the branch cutting range of the partition where it is located; S43, establish an objective function, randomly generate M solutions within the branch cutting range, calculate their respective objective functions, and take the solution corresponding to the minimum value as the current optimal solution; S44, initializing the state of each dimension of the particle according to the position update rule of the binary particle swarm algorithm, and randomly initializing the speed of the particle; S45, updating the optimal solution of each particle and the global optimal solution; S46, determine whether the maximum number of iterations has been reached, if so, the iteration ends and the result is output; otherwise, go to step S47; S47, update the speed and position of the particle, increase the number of iterations by 1, and go to step S45.

3. The method for dynamic partitioning of active distribution network based on reactive power flow tracking and branch cutting according to claim 2 is characterized in that: The objective function is: in, It represents the product of the average membership between load nodes in region k and the maximum membership between any two nodes in the system. It represents the product of the average membership degree between load nodes in region k and the minimum membership degree of controlled nodes that do not belong to this region. k is the number of controlled nodes in region k.

4. The method for dynamic partitioning of active distribution network based on reactive power flow tracking and branch cutting according to claim 2 is characterized in that: The position update rule of the binary particle swarm algorithm is as follows: the number of branches in the distribution network is n, which is divided into m dimensions according to the branch cutting range, 1) at most one of the cutting branches that appears in only one partition is in a disconnected state; 2) when encoding the common branch switch states that exist in two partitions, more than two switches are not allowed to be opened; 3) the branch switch directly connected to the power supply must be in a closed state.

5. The method for dynamic partitioning of active distribution network based on reactive power flow tracking and branch cutting according to claim 2 is characterized in that: The updating method of the particle speed and position is: In the formula, is the position of particle i in the kth dimension, is the flying speed of particle i in the kth dimension, is the flight speed of particle i in the k-1th dimension, ω is the inertia factor, r1 and r2 are both random numbers in the interval [0,1], is the optimal position of particle i in the kth dimension, is the optimal position of the k-th dimensional population, rand() is a random number in the interval [0,1]; S(v id ) is the sigmoid function, S(v id )=1 / (1+exp(-v id )).

6. The method for dynamic partitioning of active distribution network based on reactive power flow tracking and branch cutting according to claim 1 is characterized in that: The method for performing reactive power reserve verification on the partition result is: Calculate reactive power reserve index β k for: β k =(1-Q Lk / Q Gk )×100%; In the formula, Q Gk is the total reactive power reserve in area k, Q Lk Reactive power reserve generated by loads in the area.

Images