Power distribution network partitioning method based on optimization of partition evaluation index and selection of dominant nodes

Abstract

The application relates to a power distribution network partitioning method based on optimization of partition evaluation indexes and selection of leading nodes. The partitioning method comprises the following steps: A, impedance and ground admittance of each branch of the power grid and active and reactive power loads injected in each node are counted; B, a node-to-node admittance matrix of the power grid is obtained according to the power grid topology; C, power flow calculation is carried out on the power grid according to the node-to-node admittance matrix and the power injection value of each node; D, node voltage variation sensitivity is solved, and the electrical distance between nodes is calculated; E, a power distribution network optimization partition evaluation index is selected; F, a leading node is selected; G, the final power grid partition is verified and determined. The method can simplify the calculation steps under the premise of ensuring the accuracy, is beneficial to realization of smaller network loss and node voltage deviation of the power distribution network system, enables the equipment such as the distributed power supply in the power grid to participate in the operation optimization of the power distribution network system more efficiently, and is suitable for voltage regulation of the power grid.

Description

Technical Field

[0001] The present invention relates to the technical field of smart grids, and in particular to a distribution network partitioning method based on optimized partition evaluation indicators and dominant node selection. Background Art

[0002] With the development of power networks, an increasing number of distributed generation (DG) and flexible loads are being connected to the grid. This large-scale integration of DG has changed the existing voltage distribution of the distribution system, significantly increasing the uncertainty of distribution network operation and the possibility of node voltage exceeding limits. Traditional distribution network voltage regulation methods are no longer effective. Traditional centralized voltage control methods are computationally intensive, inefficient, and lack high control accuracy. To address the difficulties that large-scale integration of DG and flexible loads poses to distribution network voltage control, consideration is being given to utilizing DG for regional autonomous voltage regulation, zoning the distribution network, and fully exploring the close connection between DG and distribution network voltage control. Summary of the Invention

[0003] The purpose of the present invention is to provide a distribution network partitioning method based on optimized partition evaluation index and dominant node selection to address the problems existing in the prior art.

[0004] To achieve the above-mentioned purpose, the technical solution of the present invention is: a distribution network partitioning method based on optimized partition evaluation index and dominant node selection, comprising the following steps:

[0005] A. Calculate the impedance and ground admittance of each branch of the power grid, as well as the active and reactive loads injected into each node;

[0006] B. Establish a mathematical model of the power grid based on the power grid topology and derive the admittance matrix between power grid nodes;

[0007] C. Based on the admittance matrix between grid nodes and the power injection value of each node, the power flow of the grid is calculated to obtain some variables required for sensitivity calculation;

[0008] D. Based on the results of the power flow calculation, solve the reactive voltage sensitivity based on HEM and calculate the electrical distance between nodes;

[0009] E. Select the reactive / active power reserve within the zone and the coupling degree between the inside and outside of the zone as the distribution network optimization zone indicators;

[0010] F. Select the dominant nodes of the distribution network based on reactive / active power controllability and voltage observability;

[0011] G. Construct and solve the partition objective function to determine the final grid partition result.

[0012] In one embodiment of the present invention, the grid node admittance matrix in step B is obtained based on the impedance between each branch and the connection between nodes. The expression of the grid node admittance matrix is:

[0013]

[0014] Among them, if i=j, then:

[0015]

[0016] If i≠j, then:

[0017] Y ji =-y ij (3)

[0018] In formulas (1)-(3), Y N is the node admittance matrix; Y ii is the self-admittance; Y ij|i≠j is the mutual admittance between nodes i and j; y ij is the admittance between nodes i and j; y i0 is the ground admittance of node i; y ji is the admittance between nodes j and i; N is the total number of nodes.

[0019] In one embodiment of the present invention, the power flow calculation process in step C is as follows:

[0020] Express the node power equation in polar coordinate form:

[0021]

[0022] In formula (4), ΔP and ΔQ are the increments of active power and reactive power of the node; H and L are coefficient matrices; Δθ is the increment of bus voltage phase angle; ΔV is the increment of node voltage amplitude; V is the node voltage amplitude;

[0023] The phase angle θ of the voltage across the line ij It is generally no more than 10° to 20°, so it can be considered that:

[0024]

[0025] And the admittance B corresponding to the reactive power of each node in the system LDi is much smaller than the imaginary part of the node's self-admittance, that is,

[0026]

[0027] Taking the above relationship into account, the elements in the coefficient matrix are expressed as:

[0028] H ij =V i Vj B ij (i,j=1,2,...,n-1) (7)

[0029] L ij =V i V j B ij (i,j=1,2,...,m) (8)

[0030] Therefore, the coefficient matrices H and L are written as:

[0031]

[0032]

[0033] Substituting (9) and (10) into (4), we get

[0034]

[0035] Then we get the simplified correction equation, which is expanded as follows:

[0036]

[0037]

[0038] The coefficient matrix elements in the two modified equations (12) and (13) are the imaginary parts of the system admittance matrix. Therefore, the coefficient matrix is ​​a symmetric matrix and remains unchanged during the iteration process. The node power increment expressed in polar coordinates is:

[0039]

[0040] Equations (12), (13) and (14) constitute the basic equations for power flow calculations.

[0041] In one embodiment of the present invention, the node voltage sensitivity solution process in step D is:

[0042] Based on the power flow calculation results, the active voltage sensitivity matrix M P and reactive voltage sensitivity matrix M Q From the following formula:

[0043]

[0044]

[0045] M Q / M P Element M Qij / M PijIt represents the change in voltage at node j when the reactive power injection at node i changes by one unit. The reactive power voltage sensitivity relationship between any two nodes in the N-node distribution network system is obtained as follows:

[0046]

[0047] Where V i and V j are the voltages at nodes i and j respectively; Q i is the reactive power of node i;

[0048] Formula (17) can be written as:

[0049]

[0050] Where, α ij The sensitivity of node i to the reactive voltage change of node j is derived from the above, and the sensitivity of node i to the active voltage change of node j is β ij , expressed as:

[0051]

[0052] The expression of the voltage increment at node i is obtained from equations (18) and (19), as follows:

[0053]

[0054] From the above derivation, we can get that α ij and β ij The two variables characterize the sensitivity of node voltage changes. Based on these two variables, an expression for the electrical distance between nodes is formed. The calculation formula for the electrical distance between two nodes in the distribution network system is defined as shown in formula (21):

[0055]

[0056] Where D is the electrical distance between nodes i and j. The greater the voltage sensitivity between nodes i and j, the smaller the electrical distance.

[0057] In one embodiment of the present invention, the distribution network optimization partition index selected in step E includes the reactive power reserve within the partition and the intra-regional coupling degree. The partition voltage control utilizes the distributed power sources in the region to flexibly adjust their active and reactive power to perform voltage regulation, and introduces the intra-regional voltage offset f 1k The evaluation index is calculated as follows:

[0058]

[0059]

[0060]

[0061] Where U Mi Indicates the node voltage rating of the region under the current operating state; and Respectively represent the voltage values ​​of nodes when all distributed generation nodes in the area operate at the upper limit of feasible reactive / active power and the lower limit of feasible reactive / active power;

[0062] In addition, to facilitate control within a region and avoid the impact of control within a region on other regions, strong coupling is required between nodes within the region and weak coupling between regions. The corresponding evaluation indicators are as follows:

[0063]

[0064]

[0065] Where, l kg is the number of distributed generation nodes in region k; l kl is the number of load nodes in region k; k g and k l are the sets of distributed generation nodes and load nodes in region k respectively; D is the electrical distance between any two nodes in the distribution network system. The numerator in formula (25) represents the average electrical distance between distributed generation nodes and load nodes in region k, and the denominator represents the maximum electrical distance between load nodes in the distribution network system. The numerator in formula (26) represents the average electrical distance between distributed generation nodes and load nodes in region k, and the denominator represents the minimum electrical distance between distributed generation nodes in region k and load nodes outside region k.

[0066] In one embodiment of the present invention, the method for selecting the dominant node in step F is as follows:

[0067] The selection of the dominant node needs to consider the observability index and the controllability index. The calculation of observability needs to consider the voltage sensitivity of other load nodes in the region to the dominant node. The observability index is defined as shown in formula (27):

[0068]

[0069] Where S k Represents the set of all load nodes in a certain area; l k is the number of load nodes in the region; α mk and β mk They represent the reactive voltage sensitivity and active voltage sensitivity of the dominant node to any load node in the area respectively;

[0070] Controllability mainly considers the sensitivity of the dominant node to the reactive / active power changes of the distributed power generation nodes in the region. The controllability index of the dominant node is defined as shown in (28):

[0071]

[0072] Where V m represents the dominant node voltage; P d and Q d Respectively represent the reactive power and active power of the distributed generation nodes in the region; l d represents the number of distributed power generation nodes in the area; δ1 and δ2 are weight coefficients; therefore, the selection formula for defining the dominant node is shown in formula (29):

[0073] ρ=max{γ1OBS+γ2CON} (29)

[0074] Where γ1 and γ2 represent the weights of observability and controllability respectively.

[0075] In one embodiment of the present invention, the specific process of power grid partitioning in step G is as follows:

[0076] G1. Establish a multi-objective optimization partition model. The objective function is shown in formula (30):

[0077]

[0078] Where, f i is the maximum value of the corresponding evaluation index; i is the weight of the evaluation index;

[0079] G2. Determine the evaluation level of the evaluation index according to the similarity coefficient method, calculate the single index measurement evaluation matrix, and then find the multi-index comprehensive measurement evaluation vector to obtain the weight of each evaluation index;

[0080] G3. Set constraints: a single node cannot form a region by itself and the number of nodes in a single region must not exceed 2 / 3 of the total number of nodes;

[0081] G4. Use the particle swarm algorithm with inertia weight to solve the above multi-objective partitioning model and output the partitioning results.

[0082] Compared with the prior art, the present invention has the following beneficial effects:

[0083] The grid partitioning method of the present invention enables distributed power sources to actively participate in the control of the distribution network, improves resource utilization, transforms the partitioning process into a multi-objective optimization solution process, and comprehensively considers the coupling relationship within and between regions and the reactive / active power reserve within the partition. In addition, the present invention can be readjusted according to changes in the operating parameters of the distributed power sources and the distribution network topology structure, and only requires modifying the input data, which has strong practical significance. BRIEF DESCRIPTION OF THE DRAWINGS

[0084] Figure 1 A flowchart of the partition solution process of the present invention;

[0085] Figure 2 Flowchart of the particle swarm algorithm used in solving the partitioning results of the present invention;

[0086] Figure 3 A topology diagram of a power grid used in an embodiment of the present invention;

[0087] Figure 4 This is a power grid partitioning result obtained by using the power grid partitioning method of the present invention based on the power grid topology diagram provided in the embodiment. DETAILED DESCRIPTION

[0088] The effects of the present invention are further described in detail below with reference to the accompanying drawings and embodiments. The specific embodiments described herein are intended only to explain the present invention and are not intended to limit the present invention. It should also be noted that, for ease of description, the accompanying drawings only illustrate portions relevant to the present invention, not all structures.

[0089] like Figure 1 As shown in FIG, a distribution network partitioning method based on multiple optimization partitioning evaluation indicators and dominant node selection includes the following steps:

[0090] (1) The admittance matrix between power grid nodes is obtained based on the impedance between each branch and the connection between nodes. The expression of the admittance matrix between power grid nodes is:

[0091]

[0092] Among them, if i≠j then:

[0093]

[0094] If i=j, then:

[0095] Y ji =-y ij (3)

[0096] In formulas (1)-(3), Y N is the node admittance matrix; Y ii is the self-admittance; Y ij|i≠jis the mutual admittance between nodes i and j; y ij is the admittance between nodes i and j; y i0 is the ground admittance of node i; y ji is the admittance between nodes j and i; N is the total number of nodes.

[0097] (2) Calculate the power flow of the power grid and obtain some variables required for sensitivity calculation:

[0098] Express the node power equation in polar coordinate form:

[0099]

[0100] In formula (4), ΔP and ΔQ are the increments of active power and reactive power of the node; H and L are coefficient matrices; Δθ is the increment of bus voltage phase angle; ΔV is the increment of node voltage amplitude; V is the node voltage amplitude;

[0101] In general, the phase angle θ of the voltage at both ends of the line is ij It is not large (no more than 10°~20°), so it can be considered that:

[0102]

[0103] And the admittance B corresponding to the reactive power of each node in the system LDi is much smaller than the imaginary part of the node's self-admittance, that is,

[0104]

[0105] Taking the above relationship into account, the elements in the coefficient matrix can be expressed as:

[0106] H ij =V i V j B ij (i,j=1,2,...,n-1) (7)

[0107] L ij =V i V j B ij (i,j=1,2,...,m) (8)

[0108] Therefore, the coefficient matrices H and L can be written as:

[0109]

[0110]

[0111] Substituting (9) and (10) into (4), we get

[0112]

[0113] Then we can get the simplified correction equation, which can be expanded into:

[0114]

[0115]

[0116] The coefficient matrix elements in the two modified equations (12) and (13) are the imaginary parts of the system admittance matrix. Therefore, the coefficient matrix is ​​a symmetric matrix and remains unchanged during the iteration process, which greatly reduces the computational workload. The node power increment expressed in polar coordinates is:

[0117]

[0118] Equations (12), (13) and (14) constitute the basic equations for power flow calculations.

[0119] (3) The process of solving the node voltage sensitivity is:

[0120] Based on the power flow calculation results, the active voltage sensitivity matrix M P and reactive voltage sensitivity matrix M Q It can be obtained from the following formula:

[0121]

[0122]

[0123] M Q / M P Element M Qij / M Pij It represents the change in voltage at node j when the reactive power injection at node i changes by one unit. The reactive power and voltage sensitivity relationship between any two nodes in the N-node distribution network system can be obtained as follows:

[0124]

[0125] Where, V i and V j are the voltages at nodes i and j respectively; Q i is the reactive power of node i.

[0126] Formula (17) can be written as:

[0127]

[0128] Where, α ijThe sensitivity of node i to the reactive voltage change of node j can be obtained by the above derivation: ij , which can be expressed as:

[0129]

[0130] The expression of the voltage increment at node i can be obtained from equations (18) and (19), as follows:

[0131]

[0132] From the above derivation, we can get that α ij and β ij The two variables characterize the sensitivity of node voltage changes. Based on these two variables, an expression for the electrical distance between nodes can be constructed. The calculation formula for the electrical distance between two nodes in the distribution network system is defined as shown in Equation (21):

[0133]

[0134] Where D is the electrical distance between nodes i and j. The greater the voltage sensitivity between nodes i and j, the smaller the electrical distance.

[0135] (IV) The indicators for selecting optimized distribution network zones include reactive power reserve within the zone and the degree of coupling between the inside and outside of the zone. The zone voltage control can flexibly adjust the active and reactive power of the distributed power sources within the zone to regulate the voltage. It requires sufficient reactive and active adjustable capacity in the zone. The reactive and active reserve f within the zone is introduced. 1k The evaluation index is calculated as follows:

[0136]

[0137]

[0138]

[0139] Where U Mi Indicates the node voltage rating of the region under the current operating state; and They respectively represent the voltage values ​​of the nodes when all distributed power generation nodes in the area operate at the upper limit of feasible reactive / active power and the lower limit of feasible reactive / active power.

[0140] In addition, in order to facilitate control within a region and avoid the influence of control within a region on other regions, strong coupling is required between nodes within a region and weak coupling between regions. The corresponding evaluation indicators are as follows:

[0141]

[0142]

[0143] where, l kg is the number of distributed generation nodes in region k; l kg is the number of load nodes in region k; k g and k l are the set of distributed generation nodes and load nodes in region k, respectively; D is the electrical distance between any two nodes in the distribution system, the numerator in equation (25) represents the average electrical distance between the distributed generation nodes and load nodes in region k, and the denominator represents the maximum electrical distance between the load nodes in the distribution system, equation (26) represents the average electrical distance between the distributed generation nodes and load nodes in region k, and the denominator represents the minimum electrical distance between the distributed generation nodes in region k and the load nodes outside region k.

[0144] (Five), the dominant node selection method is as follows:

[0145] The observability index and the controllability index need to be considered in the selection of the dominant node. The calculation of the observability needs to consider the voltage sensitivity of the other load nodes in the region to the voltage of the dominant node. The observability index is defined as shown in equation (27):

[0146]

[0147] where, S k represents the set of all load nodes in a region; a mk and β mk represent the reactive voltage sensitivity and active voltage sensitivity of the dominant node to any load node in the region, respectively.

[0148] The controllability mainly considers the sensitivity of the dominant node to the change in reactive power / active power of the distributed generation nodes in the region. The controllability index of the dominant node is defined as shown in equation (28):

[0149]

[0150] where, V m represents the voltage of the dominant node; P d and Q d represent the reactive power and active power of the distributed generation nodes in the region, respectively; l d represents the number of distributed generation nodes in the region; δ1 and δ2 are weight coefficients. Therefore, the selection formula of the dominant node is defined as shown in equation (29).

[0151] ρ = max{γ1OBS + γ2CON} (29)

[0152] where, γ1 and γ2 represent the respective weights of the observability and controllability.

[0153] (6) The specific process of determining the final grid partition is as follows:

[0154] 1. Establish a multi-objective optimization partition model. The objective function is shown in formula (30):

[0155]

[0156] Where, f i is the maximum value of the corresponding evaluation index; i is the weight of the evaluation index.

[0157] 2. According to the similarity coefficient method, determine the evaluation level of the evaluation index, calculate the single index measurement evaluation matrix, and then find the multi-index comprehensive measurement evaluation vector to obtain the weight of each evaluation index.

[0158] 3. Set constraints: A single node cannot form a region by itself and the number of nodes in a single region cannot exceed 2 / 3 of the total number of nodes.

[0159] 4. Use Figure 2 The particle swarm algorithm with inertia weights is used to solve the above multi-objective partitioning model and output the partitioning results.

[0160] Example

[0161] The embodiment is as follows Figure 3 The mathematical model of the modified IEEE33 system grid topology shown in the figure has a voltage level of 10kV. The distributed power generation installation locations are nodes 3, 6, 8, 11, 14, 16, 18, 20, 22, 25, 29, and 32, respectively. The particle swarm algorithm with inertia weight is used for optimization calculation. In the particle swarm algorithm, the value of the jth element of the particle during the optimization process represents the region number to which node j-1 belongs, and the particle swarm size is taken as 100. In the calculation of inertia weight, the maximum weight ω max =0.9, minimum weight ω min =0.4, maximum number of iterations k max = 100. After each iteration of particle updating, the node numbers contained in each region are found to form a new partition. The evaluation index values ​​are calculated, and the weight coefficients δ1, δ2, γ1, and γ2 are all set to 0.5. The weights λ1, λ2, and λ3 calculated using the similarity coefficient method are 0.33, 0.41, and 0.26, respectively. Then, the dominant node is selected, and the termination condition is determined or the loop limit is reached. After the iteration is completed, the optimal partition result and the dominant node selection result are output.

[0162] The partitioning results and leading node selection results of the partitioning scheme in this embodiment are as follows Figure 4As shown in the figure, after zoning optimization, the network loss was reduced to 69.64kW, a 25% reduction; the maximum voltage deviation was 4.49%, a 12% reduction. The zoning and leading node selection method described in this invention can achieve more ideal regional local control effects, achieving smaller network losses and smaller node voltage deviations in the distribution network system.

[0163] The above are preferred embodiments of the present invention. Any changes made according to the technical solution of the present invention, as long as the resulting functions and effects do not exceed the scope of the technical solution of the present invention, shall fall within the scope of protection of the present invention.

Claims

1. A distribution network partitioning method based on optimized partition evaluation index and dominant node selection, characterized in that: The steps include: A. Calculate the impedance and ground admittance of each branch of the power grid, as well as the active and reactive loads injected into each node; B. Establish a mathematical model of the power grid based on the power grid topology and derive the admittance matrix between power grid nodes; C. Based on the admittance matrix between grid nodes and the power injection value of each node, the power flow of the grid is calculated to obtain some variables required for sensitivity calculation; D. Based on the results of the power flow calculation, solve the reactive voltage sensitivity based on HEM and calculate the electrical distance between nodes; E. Select the reactive / active power reserve within the zone and the coupling degree between the inside and outside of the zone as the distribution network optimization zone indicators; F. Select the dominant nodes of the distribution network based on reactive / active power controllability and voltage observability; G. Construct and solve the partition objective function to determine the final grid partition result; The node voltage sensitivity solution process in step D is: Based on the power flow calculation results, the active voltage sensitivity matrix M P and reactive voltage sensitivity matrix M Q From the following formula: ΔP and ΔQ are the increments of active power and reactive power of the node; ΔV is the increment of node voltage amplitude; V is the node voltage amplitude; M Q / M P Element M Qij / M Pij It represents the change in voltage at node j when the reactive power injection at node i changes by one unit. The reactive power voltage sensitivity relationship between any two nodes in the N-node distribution network system is obtained as follows: Where V i and V j are the voltages at nodes i and j respectively; Q i is the reactive power of node i; Formula (17) can be written as: Where, α ij The sensitivity of node i to the reactive voltage change of node j is derived from the above, β ij , expressed as: The expression of the voltage increment at node i is obtained from equations (18) and (19), as follows: ΔV i =ΔV i Q +ΔV i P (20) From the above derivation, we can get that according to α ij and β ij These two variables constitute the expression of the electrical distance between nodes. The calculation formula for the electrical distance between two nodes in the distribution network system is defined as shown in formula (21): Where D ij is the electrical distance between nodes i and j. The greater the voltage sensitivity between nodes i and j, the smaller the electrical distance. The distribution network optimization partitioning indicators selected in step E include the reactive power reserve within the partition and the coupling degree between the inside and outside of the region. The partition voltage control uses the distributed power sources in the region to flexibly adjust their active and reactive power to regulate the voltage, and introduces the voltage offset f within the partition. 1k The evaluation index is calculated as follows: Where U Mi Indicates the node voltage rating of the region under the current operating state; and Respectively represent the voltage values ​​of nodes when all distributed generation nodes in the area operate at the upper limit of feasible reactive / active power and the lower limit of feasible reactive / active power; In order to facilitate control within a region and avoid the impact of control within a region on other regions, strong coupling is required between nodes within the region and weak coupling between regions. The corresponding evaluation indicators are as follows: Where, l kg is the number of distributed generation nodes in region k; l kl is the number of load nodes in region k; k g and k l are the sets of distributed generation nodes and load nodes in region k respectively; D is the electrical distance between any two nodes in the distribution network system. The numerator in formula (25) represents the average electrical distance between distributed generation nodes and load nodes in region k, and the denominator represents the maximum electrical distance between load nodes in the distribution network system. The numerator in formula (26) represents the average electrical distance between distributed generation nodes and load nodes in region k, and the denominator represents the minimum electrical distance between distributed generation nodes in region k and load nodes outside region k.

2. The distribution network partitioning method based on optimized partition evaluation index and dominant node selection according to claim 1 is characterized in that: The grid node admittance matrix in step B is obtained based on the impedance between each branch and the connection between nodes. The expression of the grid node admittance matrix is: Among them, if i=j, then: If i≠j, then: AND ji =-y ij (3) In formulas (1)-(3), Y N is the node admittance matrix; Y ii is the self-admittance; Y ij|i≠j is the mutual admittance between nodes i and j; y ij is the admittance between nodes i and j; y i0 is the ground admittance of node i; y ji is the admittance between nodes j and i; N is the total number of nodes.

3. The distribution network partitioning method based on optimized partition evaluation index and dominant node selection according to claim 1 is characterized in that: The method for selecting the dominant node in step F is as follows: The selection of the dominant node needs to consider the observability index and the controllability index. The calculation of observability needs to consider the voltage sensitivity of other load nodes in the region to the dominant node. The observability index is defined as shown in formula (27): Where S k Represents the set of all load nodes in a certain area; l k is the number of load nodes in the region; α mk and β mk They represent the reactive voltage sensitivity and active voltage sensitivity of the dominant node to any load node in the area respectively; Controllability mainly considers the sensitivity of the dominant node to the reactive / active power changes of the distributed power generation nodes in the region. The controllability index of the dominant node is defined as shown in (28): Where V m represents the dominant node voltage; P d and Q d Respectively represent the reactive power and active power of the distributed generation nodes in the region; l d represents the number of distributed power generation nodes in the area; δ1 and δ2 are weight coefficients; therefore, the selection formula for defining the dominant node is shown in formula (29): ρ=max{γ1OBS+γ2CON} (29) Where γ1 and γ2 represent the weights of observability and controllability respectively.

4. The distribution network partitioning method based on optimized partition evaluation index and dominant node selection according to claim 3 is characterized in that: The specific process of power grid partitioning in step G is as follows: G1. Establish a multi-objective optimization partition model. The objective function is shown in formula (30): Where, f i is the maximum value of the corresponding evaluation index; i is the weight of the evaluation index; G2. Determine the evaluation level of the evaluation index according to the similarity coefficient method, calculate the single index measurement evaluation matrix, and then find the multi-index comprehensive measurement evaluation vector to obtain the weight of each evaluation index; G3. Set constraints: a single node cannot form a region by itself and the number of nodes in a single region must not exceed 2 / 3 of the total number of nodes; G4. Use the particle swarm algorithm with inertia weight to solve the above multi-objective optimization partition model and output the partition results.

Images