A cable path planning method for medium voltage distribution network based on intelligent ant colony algorithm

Abstract

The present invention relates to a cable path planning method for a medium-voltage distribution network based on an intelligent ant colony algorithm, comprising the following steps: first, grid division of the distribution network based on GIS; then, quantifying the grid construction cost in the power supply area by combining the KDE and AHP methods; thirdly, improving the ant colony algorithm in terms of pheromone factor, update, and search direction, establishing a path planning model for a full-cable medium-voltage distribution network, and solving the minimum construction investment cost; finally, based on the path planning method of the improved ant colony algorithm, the interconnection lines of the ring network path are planned by a genetic algorithm to achieve the optimal operating cost target. The present invention carries out cable path planning for a medium-voltage distribution network based on an intelligent ant colony algorithm, and the path planning result is reasonable, which facilitates the formulation of a medium-voltage distribution network planning scheme. The method of the present invention is easy to use and has strong applicability.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid planning, and specifically refers to a cable path planning method for a medium-voltage distribution network based on an intelligent ant colony algorithm. Background Art

[0002] Since cables are easy to maintain, beautify the city, and improve the level of power supply safety, cable ring networks have become the primary goal of current urban distribution network planning and line reconstruction. With the improvement of geographic information systems, distribution network path planning has begun to be linked to the spatiotemporal characteristics of geographic information to optimize path selection, and cable network path planning is also developing in a refined direction. Based on geographic information systems, factors such as corridor construction, power balance, load distribution, and wiring mode need to be considered in the cable network planning process. Therefore, the distribution network cable path planning problem is a complex optimization problem with multi-objective, nonlinear, and uncertain characteristics.

[0003] Although traditional mathematical optimization algorithms guarantee the optimality of the solution in theory, as the scale and complexity of the problem increase, the variables and constraints in the model also increase significantly, making it difficult to search for the optimal solution in the feasible domain and difficult to obtain the optimal solution within the required time. How to achieve higher search speed and solution efficiency in order to obtain better optimization results within a reasonable time is an urgent problem to be solved. Summary of the invention

[0004] In view of the above-mentioned defects of the existing model, the present invention provides a medium-voltage distribution network cable path planning method based on intelligent ant colony algorithm.

[0005] The technical solution of the present invention:

[0006] A cable path planning method for a medium voltage distribution network based on an intelligent ant colony algorithm comprises the following steps:

[0007] First, the distribution network grid is divided based on the Geographic Information System (GIS);

[0008] Then, the grid construction cost in the power supply area is quantified by combining the nonparametric kernel density estimation method (KDE) and the analytic hierarchy process (AHP).

[0009] Thirdly, the ant colony algorithm is improved in terms of pheromone factor, update, and search direction, and a path planning model for the all-cable medium-voltage distribution network is established to solve the minimum construction investment cost.

[0010] Finally, based on the path planning method of the improved ant colony algorithm, the contact lines of the ring network path are planned through the genetic algorithm to achieve the optimal operation cost target.

[0011] Preferably, the GIS-based distribution network grid division includes the following steps:

[0012] (1) Determine the power supply scope of the medium voltage distribution network path planning and define the geographical boundaries in combination with GIS;

[0013] (2) Select appropriate unit grid pixels, calculate the number of grids in the grid map R×C, and establish a grid matrix represented by N R×C ;

[0014] (3) Based on the land property information in GIS, fill in the load type of each grid.

[0015] Preferably, the filling rules are as follows:

[0016] ① When the effective object area in a single grid accounts for more than 1 / 2, it needs to be filled, otherwise it will not be filled;

[0017] ② If a single grid fill range contains multiple objects, the object with the largest area will cover the entire grid.

[0018] Preferably, the steps of quantifying the grid construction cost in the power supply area in combination with the KDE method are as follows:

[0019] The number of grids of a certain type of land use within the power supply range is n, x1, x2, ..., x n is the load density of the corresponding grid, and the corresponding KDE model is as follows:

[0020]

[0021] In the formula, f h is the load density probability function, h is the bandwidth, and K is the kernel function of the nonparametric kernel density estimation; the bandwidth h reflects the flatness of the entire KDE curve: the larger h is, the smaller the proportion of sample data points in the final curve shape, and the flatter the KDE curve is; conversely, the KDE curve is steeper;

[0022] In order to ensure that the load density probability function f h The kernel function K(x) is a unimodal smooth nonlinear function that is symmetric about the y-axis and satisfies the following characteristics:

[0023]

[0024] Preferably, the Gaussian function is selected as the kernel function, and the formula is as follows:

[0025]

[0026] When h is selected as 0.5, the load density probability function f h It is expressed as:

[0027]

[0028] Combined load density function f h The load density value corresponding to its maximum value is the typical value x of the load density level. tp ,Right now:

[0029] x tp = argmaxf h (x)

[0030] Thus, the typical load density level set {x tp |x tp,i ,0<i≤m}, where m is the number of land types.

[0031] Preferably, the steps of quantifying the grid construction cost in the power supply area in combination with the AHP method are as follows:

[0032] Using the typical value set of load density levels for each type of land {x tp |x tp,i ,0<i≤m} constructs the judgment matrix A of the unit grid quantization cost m×m , the matrix elements are:

[0033] a ij =x tp,i / x tp,j

[0034] Judgment Matrix A m×m Satisfy the following characteristics:

[0035]

[0036] Since the judgment matrix A m×m The consistency test is met and the analytic hierarchy process is directly used to calculate the unit quantitative cost c of each type of land grid i Weight, such as:

[0037]

[0038] Combined with the weight c of a certain type of sample grid * , and the construction cost of the unit cable corridor E * , and obtain the unit grid cost C corresponding to each land use property i , as follows:

[0039]

[0040] For power supply range N R×C All grids in the grid are assigned similar values ​​to obtain the unit grid cost matrix C R×C Preferably, establishing a path planning model for a full-cable medium-voltage distribution network includes establishing an objective function and constraining the objective function;

[0041] (1) Establishing the objective function

[0042] The comprehensive investment of distribution network path planning includes the construction cost and operation cost of cable lines. The objective function is to minimize the comprehensive investment, as follows:

[0043] minf=f inv +f ope

[0044] Among them, f is the comprehensive annual investment cost as the objective function, f inv is the annual investment cost of cable line construction, f ope is the annual operating cost of the cable line.

[0045] 1)f inv It includes two parts: cable cost and cable channel construction cost, and is amortized over the economic useful life using the present value annualization method, as calculated as follows:

[0046]

[0047] In the above formula, nf is the number of feeder branches, k is the number of grids that the i-th feeder passes through; α i is the unit length cost of the i-th feeder, β i,j-1 and β i,j is the construction cost of the unit cable channel in the two adjacent grids of the corresponding feeder j-1 and j; r i,(j-1,j) is the distance between the j-1th and jth grids; r0 is the return on investment, and s is the economic service life of the grid;

[0048] Combined with the unit grid cost matrix C R×C Analysis shows that for the planned path L i Passed grid, element C r,c With β i,j There is a corresponding relationship that satisfies {β i,j}={C r,c}; If F is the corresponding function, it can be expressed as:

[0049] β i,j =F(C R×C )

[0050] For r i,(j-1,j) , can be calculated as follows:

[0051]

[0052] In the above formula, x i,j ,y i,j Respectively represent the horizontal and vertical coordinates of the jth grid through which the i-th feeder passes;

[0053] If d is the side length of the unit grid, then r i,(j-1,j) Satisfy the following formula:

[0054]

[0055] 2)f ope It includes network loss costs and operation and maintenance costs. Since the operation and maintenance costs of the distribution network are relatively fixed and have no effect on the model optimization, only network loss is considered here, and the calculation is as follows:

[0056]

[0057] In the above formula, λ is the comprehensive unit electricity price; τ max P is the maximum annual utilization hours; i , Q i are the active and reactive power of the ith feeder, U i is the terminal voltage of the branch, ρ i is the resistance per unit length of cable, r i,(j-1,j) is the distance between the j-1th and jth adjacent grids.

[0058] Preferably, the constraints include: power balance constraints, safe operation constraints, connectivity constraints, node outgoing line quantity constraints, taboo area constraints, and power supply range and feeder investment constraints;

[0059] 1) Power balance constraints

[0060]

[0061] Where P i and Q i are the active and reactive power of the node, U i is the voltage at node i, U j is the voltage at node j, G ij and B ij is the element in the corresponding admittance matrix, θ ij is the phase angle difference between nodes;

[0062] 2) Safety operation constraints

[0063]

[0064] As shown in the formula, the safe operation constraints mainly include line flow constraints and node voltage constraints; among them, Sl , S l,max and S l,min are the line flow and its upper and lower limits, U i , U i,max and U i,min is the node voltage and its upper and lower limits;

[0065] 3) Connectivity constraints

[0066] Search direction between every two adjacent grids on the distribution network planning path and It must remain consistent and cannot be turned back immediately. Constraints:

[0067]

[0068] In the formula, when the search direction of the i-th feeder passing through the j-th grid is k, D i,j,k =1, otherwise 0;

[0069] 4) Constraints on the number of outgoing lines of a node

[0070] The outgoing line number constraint means that each node must be connected, and there are requirements for the number of outgoing lines of the power node and the load node;

[0071]

[0072] Where n load-mid,i is the number of outgoing lines from non-last load nodes of the line, n load-end,i is the number of outgoing lines at the last load node of the line, n pow,i and n pow,lim To meet the load rate requirements of the power supply site, the number of outgoing lines and restrictions, and there are no isolated nodes in the distribution network;

[0073] 5) Taboo area constraints

[0074] Assume that the planned path between two nodes is l ij The grid set of the path is {e ij}, the grid set of typical obstacle areas such as water areas is Π, then {e ij} and Π do not have an intersection, that is, they satisfy the following formula:

[0075]

[0076] 6) Power supply scope and feeder investment constraints

[0077]

[0078] In the formula, and N lim are the number and limit of load nodes connected by the i-th feeder respectively; and inv lim are the construction investment and limit of the i-th feeder respectively.

[0079] Preferably, an improved ant colony algorithm based on pheromone concentration, search range and search direction, which comprehensively considers the nonlinear process of pheromone concentration, as well as object screening and direction reduction of ants in the entire search process, plans the contact lines of the ring network path through a genetic algorithm, and achieves the optimal operation cost target. Preferably,

[0080] (1) Improvement of pheromone impact factor α

[0081] The original ant colony algorithm uses a fixed value α to measure the proportion of pheromones. In order to improve the convergence speed in the initial stage and avoid rapid local convergence in the subsequent process, the following corrections are made;

[0082] α(t)=λ(1+e -εt )

[0083] Where λ and ε are real numbers and satisfy λ,ε∈(0,1]; compared with the original λ, α is larger at the beginning and the convergence speed is faster; as the number of iterations increases, α becomes smaller, which makes the convergence gradually slow down;

[0084] (2) Improvement of pheromone increment Δτ ij k

[0085] The original ant colony algorithm updates Δτ according to the path length of ant k ij k , ignoring the intensive effect of the optimal path on the solution in the current iteration process, the following improvements are made:

[0086]

[0087] Where Q is the initial pheromone concentration, L k is the path length of ant k in the iteration, L low is the path length of the worst ant in the iteration, L ave is the average path length of the ants in the iteration.

[0088] Preferably,

[0089] (3) Improving the search range and direction of ants

[0090] In the optimization process, the search range and direction of ants have a great influence on the convergence speed and complexity of the algorithm. kThe optimization direction can greatly improve the overall solution level of the model; R is the radius of the ant's search range with the current node as the center, 2α is the search direction based on the range on both sides of the line connecting the current node and the target node, and only the nodes in the above two areas can be selected into A k gather.

[0091] Preferably, based on the path planning results of the cable ring network and the "hand-in-hand" communication method, by constructing the inter-node communication matrix and encoding the communication lines, the fitness function objective is to minimize the operating cost, and the optimal communication solution is obtained through genetic operation. The main steps are as follows:

[0092] (1) Based on the specific situation of the cable ring network planning results, select the appropriate inter-station tie line set between each cable loop and determine the switch station location information of the corresponding tie line;

[0093] (2) It is stipulated that each loop has only one contact line, and a genetic algorithm is used to randomly generate a contact line traversal sequence, and then an initial population is constructed after forming multiple chromosomes to form a set of optional solutions;

[0094] (3) Open-loop operation is performed according to the selected contact line position, and the fitness function is calculated based on the different contact point configuration schemes of each chromosome and the line operation cost;

[0095] (4) Determine the termination condition of the iteration number. If it is met, record the optimal planning scheme; otherwise, perform genetic operations on each chromosome to obtain the next generation population, and return to (3) to continue optimization until the iteration converges.

[0096] Beneficial effects of the present invention:

[0097] The present invention carries out medium-voltage distribution network cable path planning based on intelligent ant colony algorithm, the path planning result is reasonable, and it is convenient to carry out the formulation of medium-voltage distribution network planning scheme. The method of the present invention is easy to use and has strong applicability.

[0098] Compared with the original algorithm, the improved algorithm of the present invention pays more attention to the better ants, which makes the algorithm converge faster; in the optimization process, the search range and direction of the ants have a great influence on the convergence speed and complexity of the algorithm, and the set of candidate nodes A is reduced. k The optimization direction can greatly improve the overall solution level of the model, greatly reduce the range of nodes to be selected, and improve the solution speed of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0099] Figure 1 Grid direction search method diagram;

[0100] Figure 2 Improved diagram of ant search range and direction. DETAILED DESCRIPTION

[0101] A cable path planning method for a medium voltage distribution network based on an intelligent ant colony algorithm comprises the following steps:

[0102] First, the distribution network grid is divided based on GIS;

[0103] Then, the KDE and AHP methods are combined to quantify the grid construction cost in the power supply area;

[0104] Thirdly, the ant colony algorithm is improved in terms of pheromone factor, update, and search direction, and a path planning model for the all-cable medium-voltage distribution network is established to solve the minimum construction investment cost.

[0105] Finally, based on the path planning method of the improved ant colony algorithm, the contact lines of the ring network path are planned through the genetic algorithm to achieve the optimal operation cost target.

[0106] 1. Grid division of distribution network based on GIS

[0107] (1) Determine the power supply scope of the medium voltage distribution network path planning and define the geographical boundaries in combination with GIS;

[0108] (2) Select appropriate unit grid pixels, calculate the number of grids in the grid map R×C, and establish a grid matrix represented by N R×C ;

[0109] (3) Combined with the land property information in GIS, fill in the load type of each grid. The rules are as follows:

[0110] ① When the effective object area in a single grid accounts for more than 1 / 2, it needs to be filled, otherwise it will not be filled.

[0111] ② If a single grid fill range contains multiple objects, the object with the largest area will cover the entire grid.

[0112] 2. Unit cost of planning paths within the grid

[0113] (1) Load density level estimation based on KDE

[0114] Distribution network path planning requires laying lines or building corridors. Distribution network path planning based on grid-based distribution networks needs to determine the line construction cost per unit grid based on factors such as load density level and land use characteristics.

[0115] The land use characteristics of cities can be divided into residential, public administration and services, commercial, industrial, road traffic, green space, etc. The load density of plots with different land use characteristics is often significantly different. Since the non-parametric kernel density estimation method (kernel density estimation, KDE) does not use prior knowledge about data distribution, nor does it make any assumptions about data distribution, but is an objective method to study the data distribution characteristics from the data sample itself, it is suitable for estimating the load density level of each land type.

[0116] Assume that the number of grids of a certain type of land use within the power supply range is n, x1, x2, ..., x n is the load density of the corresponding grid, and the corresponding KDE model is as follows:

[0117]

[0118] In the formula, f h is the load density probability function, h is the bandwidth, and K is the kernel function of the nonparametric kernel density estimation. The bandwidth h reflects the flatness of the entire KDE curve: the larger h is, the smaller the proportion of sample data points in the final curve shape, and the flatter the KDE curve is; conversely, the KDE curve is steeper.

[0119] In order to ensure that the load density probability function f h The kernel function K(x) is generally a single-peak smooth and nonlinear function symmetric about the y-axis, satisfying the following characteristics:

[0120]

[0121] The commonly used kernel functions are Uniform function, Epanechikov function, Gaussian function and Quartic function. This patent uses Gaussian function as the kernel function, and the formula is as follows:

[0122]

[0123] When h is selected as 0.5, the load density probability function f h It can be expressed as:

[0124]

[0125] Combined load density function f h The load density value corresponding to its maximum value is the typical value x of the load density level. tp ,Right now:

[0126] x tp = argmaxf h (x)

[0127] Other land types are similar, and thus the typical load density level set {x tp |x tp,i ,0<i≤m}, where m is the number of land types.

[0128] (2) Unit grid cost based on AHP and typical load density values

[0129] Generally speaking, the higher the load density level of the land, the higher the line construction cost. Therefore, this paper uses the typical value set of load density levels of various types of land {x tp |x tp,i ,0<i≤m} constructs the judgment matrix A of the unit grid quantization cost m×m , the matrix elements are:

[0130] a ij =x tp,i / x tp,j

[0131] Obviously, the judgment matrix A m×m Satisfy the following characteristics:

[0132]

[0133] Since the judgment matrix A m×m If the consistency test is met, the analytic hierarchy process (AHP) can be directly applied to calculate the unit quantitative cost c of each type of land grid i Weight, such as:

[0134]

[0135] Combined with the weight c of a certain type of sample grid * , and the construction cost of the unit cable corridor E * , the unit grid cost C corresponding to each land use property can be obtained i , as follows:

[0136]

[0137] For power supply range N R×C All grids in the grid are assigned similar values ​​to obtain the unit grid cost matrix C R×C .

[0138] 3. Path planning model for all-cable medium-voltage distribution network

[0139] (1) Objective function

[0140] The comprehensive investment of distribution network path planning mainly includes the construction cost and operation cost of cable lines. The objective function is to minimize the comprehensive investment, as follows:

[0141] minf=f inv +f ope

[0142] Among them, f is the comprehensive annual investment cost as the objective function, f inv is the annual investment cost of cable line construction, f ope is the annual operating cost of the cable line.

[0143] 1)f inv It includes two parts: cable cost and cable channel construction cost, and is amortized to the economic useful life using the "present value to year method", as follows:

[0144]

[0145] In the above formula, nf is the number of feeder branches, k is the number of grids that the i-th feeder passes through; α i is the unit length cost of the i-th feeder, β i,j-1 and β i,j is the construction cost of the unit cable channel in the two adjacent grids of the corresponding feeder j-1 and j; r i,(j-1,j) is the distance between the j-1th and jth grids; r0 is the rate of return on investment, and s is the economic service life of the grid.

[0146] Combined with the unit grid cost matrix C R×C Analysis shows that for the planned path L i Passed grid, element C r,c With β i,j There is a corresponding relationship that satisfies {β i,j}={C r,c}; If F is the corresponding function, it can be expressed as:

[0147] β i,j =F(C R×C )

[0148] For r i,(j-1,j) , can be calculated as follows:

[0149]

[0150] In the above formula, x i,j ,y i,j They respectively represent the horizontal and vertical coordinates of the j-th grid through which the ith feeder passes.

[0151] Since the grids on the path are adjacent to each other, there are Figure 1There are 8 search directions shown. If d is the side length of the unit grid, then r i,(j-1,j) Satisfy the following formula:

[0152]

[0153] 2)f ope It includes network loss costs and operation and maintenance costs. Since the operation and maintenance costs of the distribution network are relatively fixed and have almost no impact on the model optimization, this patent only considers network losses and is calculated as follows:

[0154]

[0155] In the above formula, λ is the comprehensive unit electricity price; τ max P is the maximum annual utilization hours; i , Q i are the active and reactive power of the ith feeder, U i is the terminal voltage of the branch, ρ i is the resistance per unit length of cable, r i,(j-1,j) is the distance between the j-1th and jth adjacent grids.

[0156] (2) Constraints

[0157] For a full cable ring network where the locations of power stations and load nodes in the distribution network are determined, the path planning mainly has the following constraints:

[0158] 1) Power balance constraints

[0159]

[0160] Where P i and Q i are the active and reactive power of the node, U i is the node voltage, G ij and B ij is the element in the corresponding admittance matrix, θ ij is the phase angle difference between nodes.

[0161] 2) Safety operation constraints

[0162]

[0163] As shown in the formula, the safe operation constraints mainly include line flow constraints and node voltage constraints. l , S l,max and S l,min are the line flow and its upper and lower limits, U i , U i,max and U i,m i n is the node voltage and its upper and lower limits.

[0164] 3) Connectivity constraints

[0165] Search direction between every two adjacent grids on the distribution network planning path and Need to be consistent and cannot turn back immediately, combined with Figure 1 The available constraints are:

[0166]

[0167] In the formula, when the search direction of the i-th feeder passing through the j-th grid is k, D i,j,k =1, otherwise 0.

[0168] 4) Constraints on the number of outgoing lines of a node

[0169] The outgoing line number constraint means that each node must be connected, and there are requirements for the number of outgoing lines of the power node and the load node.

[0170]

[0171] Where n load-mid,i is the number of outgoing lines from non-last load nodes of the line, n load-end,i is the number of outgoing lines at the last load node of the line, n pow,i and n pow,lim To meet the load rate requirements, the number and restrictions of outgoing lines of the power supply site are met, and there are no isolated nodes in the distribution network.

[0172] 5) Taboo area constraints

[0173] Assume that the planned path between two nodes is l ij The grid set of the path is {e ij}, the grid set of typical obstacle areas such as water areas is Π, then {e ij} and Π do not have an intersection, that is, they satisfy the following formula:

[0174]

[0175] 6) Power supply scope and feeder investment constraints

[0176]

[0177] In the formula, and N lim are the number and limit of load nodes connected by the i-th feeder respectively; and inv lim are the construction investment and limit of the i-th feeder respectively.

[0178] 4. Path planning method based on improved intelligent ant colony algorithm

[0179] In order to improve the efficiency of path planning in cable ring networks, this patent proposes an improved ant colony algorithm based on pheromone concentration, search range and search direction. The algorithm comprehensively considers the nonlinear process of pheromone concentration, as well as object screening and direction reduction of ants in the entire search process.

[0180] (1) Improvement of pheromone impact factor α

[0181] The original ant colony algorithm uses a fixed value α to measure the proportion of pheromones. In order to improve the convergence speed in the initial stage and avoid rapid local convergence in the subsequent process, the following corrections are made.

[0182] α(t)=λ(1+e -εt )

[0183] Where λ and ε are real numbers and satisfy λ, ε∈(0,1]. Compared with the original λ, α is larger at the beginning and the convergence speed is faster; as the number of iterations increases, α becomes smaller, which makes the convergence gradually slow down. Therefore, the whole process strengthens the characteristic of "fast first and slow later" in the optimization process.

[0184] (2) Improvement of pheromone increment Δτ ij k

[0185] The original ant colony algorithm updates Δτ according to the path length of ant k ij k , ignoring the intensive effect of the optimal path on the solution in the current iteration process, the following improvements are made:

[0186]

[0187] Where Q is the initial pheromone concentration, L k is the path length of ant k in the iteration, L low is the path length of the worst ant in the iteration, L ave is the average path length of ants in iteration. Compared with the original algorithm, the improved algorithm pays more attention to the better ants, which makes the algorithm converge faster.

[0188] (3) Improving the search range and direction of ants

[0189] In the optimization process, the search range and direction of ants have a great influence on the convergence speed and complexity of the algorithm. k And the optimization direction can greatly improve the overall solution level of the model. This patent has made the above two improvements, such as Figure 2 .

[0190] like Figure 2, R is the radius of the ant's search range with the current node as the center, 2α is the search direction based on the range on both sides of the line connecting the current node and the target node, and only the nodes in the above two areas can be selected into A k Therefore, after selecting appropriate R and α values, the range of candidate nodes is greatly narrowed and the solution speed of the model is improved.

[0191] 5. Ring network tie line planning method based on genetic algorithm

[0192] Based on the path planning results of the cable ring network and the "hand-in-hand" communication method, the communication matrix between nodes is constructed, and the communication lines are coded and processed. The fitness function objective is to minimize the operating cost. After genetic operation, the optimal communication plan is obtained. The main steps are as follows:

[0193] (1) Based on the specific situation of the cable ring network planning results, select the appropriate inter-station tie line set between each cable loop and determine the switch station location information of the corresponding tie line;

[0194] (2) It is stipulated that each loop has only one contact line, and a genetic algorithm is used to randomly generate a contact line traversal sequence, and then an initial population is constructed after forming multiple chromosomes to form a set of optional solutions;

[0195] (3) Open-loop operation is performed according to the selected contact line position, and the fitness function is calculated based on the different contact point configuration schemes of each chromosome and the line operation cost;

[0196] (4) Determine the termination condition of the iteration number. If it is met, record the optimal planning scheme; otherwise, perform genetic operations on each chromosome to obtain the next generation population, and return to (3) to continue optimization until the iteration converges.

Claims

1. A cable path planning method for a medium voltage distribution network based on an intelligent ant colony algorithm, characterized in that: The following steps are involved: First, the distribution network grid is divided based on the geographic information system; Then, the grid construction cost in the power supply area is quantified by combining the nonparametric kernel density estimation method and the analytic hierarchy process method; Thirdly, the ant colony algorithm is improved in terms of pheromone factor, update, and search direction, and a path planning model for the all-cable medium-voltage distribution network is established to solve the minimum construction investment cost. Finally, based on the improved ant colony algorithm path planning method, the contact lines of the ring network path are planned through genetic algorithm to achieve the optimal operation cost target; An improved ant colony algorithm based on pheromone concentration, search range and search direction. The algorithm comprehensively considers the nonlinear process of pheromone concentration, as well as the object screening and direction reduction of ants in the entire search process. The genetic algorithm is used to plan the contact lines of the ring network path to achieve the optimal operation cost target. (1) Improvement of pheromone impact factor α The original ant colony algorithm uses a fixed value α to measure the proportion of pheromones. In order to improve the convergence speed in the initial stage and avoid rapid local convergence in the subsequent process, the following corrections are made; α(t)=λ(1+e -εt ) Where λ and ε are real numbers and satisfy λ,ε∈(0,1]; compared with the original λ, α is larger at the beginning and the convergence speed is faster; as the number of iterations increases, α becomes smaller, which makes the convergence gradually slow down; (2) Improvement of pheromone increment Δτ ij k The original ant colony algorithm updates Δτ according to the path length of ant k ij k , ignoring the intensive effect of the optimal path on the solution in the current iteration process, the following improvements are made: Where Q is the initial pheromone concentration, L k is the path length of ant k in the iteration, L low is the path length of the worst ant in the iteration, L ave is the average path length of ants in iteration; (3) Improving the search range and direction of ants In the optimization process, the search range and direction of ants have a great influence on the convergence speed and complexity of the algorithm. k The optimization direction can greatly improve the overall solution level of the model; R is the radius of the ant's search range with the current node as the center, 2α is the search direction based on the range on both sides of the line connecting the current node and the target node, and only the nodes in the above two areas can be selected into A k gather.

2. A medium voltage distribution network cable path planning method based on intelligent ant colony algorithm according to claim 1, characterized in that: The GIS-based distribution network grid division includes the following steps: (1) Determine the power supply scope of the medium voltage distribution network path planning and define the geographical boundaries in combination with GIS; (2) Select appropriate unit grid pixels, calculate the number of grids in the grid map R×C, and establish a grid matrix represented by N R×C ; (3) Based on the land property information in GIS, fill in the load type of each grid.

3. A medium voltage distribution network cable path planning method based on intelligent ant colony algorithm according to claim 2, characterized in that: The filling rules are as follows: ① When the effective object area in a single grid accounts for more than 1 / 2, it needs to be filled, otherwise it will not be filled; ② If a single grid fill range contains multiple objects, the object with the largest area will cover the entire grid.

4. The method for cable path planning in a medium voltage distribution network based on an intelligent ant colony algorithm according to claim 1, characterized in that: The steps to quantify the grid construction cost in the power supply area by combining the KDE method are as follows: The number of grids of a certain type of land use within the power supply range is n, x1, x2, ..., x n is the load density of the corresponding grid, and the corresponding KDE model is as follows: In the formula, f h is the load density probability function, h is the bandwidth, and K is the kernel function of the nonparametric kernel density estimation; the bandwidth h reflects the flatness of the entire KDE curve: the larger h is, the smaller the proportion of sample data points in the final curve shape, and the flatter the KDE curve is; conversely, the KDE curve is steeper; In order to ensure that the load density probability function f h The kernel function K(x) is a unimodal smooth nonlinear function that is symmetric about the y-axis and satisfies the following characteristics: The Gaussian function is selected as the kernel function, and the formula is as follows: When h is selected as 0.5, the load density probability function f h It is expressed as: Combined load density function f h The load density value corresponding to its maximum value is the typical value x of the load density level. tp ,Right now: x tp =arg max f h (x) Thus, the typical load density level set {x tp |x tp,i ,0<i≤m}, where m is the number of land types.

5. The method for cable path planning in a medium voltage distribution network based on an intelligent ant colony algorithm according to claim 1, characterized in that: The steps to quantify the grid construction cost in the power supply area by combining the AHP method are as follows: Using the typical value set of load density levels for each type of land {x tp |x tp,i ,0<i≤m} constructs the judgment matrix A of the unit grid quantization cost m×m , the matrix elements are: a ij =x tp,i / x tp,j Judgment Matrix A m×m Satisfy the following characteristics: Since the judgment matrix A m×m The consistency test is met and the analytic hierarchy process is directly used to calculate the unit quantitative cost c of each type of land grid i Weight, such as: Combined with the weight c of a certain type of sample grid * , and the construction cost of the unit cable corridor E * , and obtain the unit grid cost C corresponding to each land use property i , as follows: For power supply range N R×C All grids in the grid are assigned similar values ​​to obtain the unit grid cost matrix C R×C .

6. A medium voltage distribution network cable path planning method based on intelligent ant colony algorithm according to claim 1, characterized in that: Establishing a path planning model for a fully cabled medium voltage distribution network includes establishing an objective function and constraining the objective function; (1) Establishing the objective function The comprehensive investment of distribution network path planning includes the construction cost and operation cost of cable lines. The objective function is to minimize the comprehensive investment, as follows: minf=f inv +f ope Among them, f is the comprehensive annual investment cost as the objective function, f inv is the annual investment cost of cable line construction, f ope is the annual operating cost of the cable line; 1)f inv It includes two parts: cable cost and cable channel construction cost, and is amortized over the economic useful life using the present value annualization method, as calculated as follows: In the above formula, nf is the number of feeder branches, k is the number of grids that the i-th feeder passes through; α i is the unit length cost of the ith feeder, β i,j-1 and β i,j is the construction cost of the unit cable channel in the two adjacent grids of the corresponding feeder j-1 and j; r i,(j-1,j) is the distance between the j-1th and jth grids; r0 is the return on investment, and s is the economic service life of the grid; Combined with the unit grid cost matrix C R×C Analysis shows that for the planned path L i Passed grid, element C r,c With β i,j There is a corresponding relationship that satisfies {β i,j }={C r,c }; If F is the corresponding function, it can be expressed as: b i,j =F(C R×C ) For r i,(j-1,j) , can be calculated as follows: In the above formula, x i,j ,y i,j Respectively represent the horizontal and vertical coordinates of the jth grid through which the i-th feeder passes; If d is the side length of the unit grid, then r i,(j-1,j) Satisfy the following formula: 2)f ope It includes network loss costs and operation and maintenance costs. Since the operation and maintenance costs of the distribution network are relatively fixed and have no effect on the model optimization, only network loss is considered here, and the calculation is as follows: In the above formula, λ is the comprehensive unit electricity price; τ max P is the maximum annual utilization hours; i , Q i are the active and reactive power of the ith feeder, U i is the terminal voltage of the branch, ρ i is the resistance per unit length of cable, r i,(j-1,j) is the distance between the j-1th and jth adjacent grids.

7. A medium voltage distribution network cable path planning method based on intelligent ant colony algorithm according to claim 6, characterized in that: The constraints include: power balance constraints, safe operation constraints, connectivity constraints, node outgoing line quantity constraints, taboo area constraints, power supply scope and feeder investment constraints; 1) Power balance constraints Where P i and Q i are the active and reactive power of the node, U i is the node voltage, G ij and B ij is the element in the corresponding admittance matrix, θ ij is the phase angle difference between nodes; 2) Safety operation constraints As shown in the formula, the safe operation constraints include line power flow constraints and node voltage constraints; where S l , S l,max and S l,min are the line flow and its upper and lower limits, U i , U i,max and U i,min is the node voltage and its upper and lower limits; 3) Connectivity constraints The search direction D between every two adjacent grids on the distribution network planning path i,j,k1 and D i,j-1,k2 It must be consistent and cannot be turned back immediately. Constraints: In the formula, when the search direction of the i-th feeder passing through the j-th grid is k, D i,j,k =1, otherwise 0; 4) Constraints on the number of outgoing lines of a node The outgoing line number constraint means that each node must be connected, and there are requirements for the number of outgoing lines of the power node and the load node; Where n load-mid,i is the number of outgoing lines from non-last load nodes of the line, n load-end,i is the number of outgoing lines at the last load node of the line, n pow,i and n pow,lim To meet the load rate requirements of the power supply site, the number of outgoing lines and restrictions, and there are no isolated nodes in the distribution network; 5) Taboo area constraints Assume that the planned path between two nodes is l ij The grid set of the path is {e ij }, the grid set of typical obstacle areas in the water area is Π, then {e ij } and Π do not have an intersection, that is, they satisfy the following formula: 6) Power supply scope and feeder investment constraints In the formula, and N lim are the number and limit of load nodes connected by the i-th feeder respectively; and inv lim are the construction investment and limit of the i-th feeder respectively.

8. The method for cable path planning in a medium voltage distribution network based on an intelligent ant colony algorithm according to claim 1, characterized in that: Based on the path planning results of the cable ring network and the hand-in-hand communication method, the communication matrix between nodes is constructed, and the communication lines are coded and processed. The fitness function objective is to minimize the operating cost. The optimal communication plan is obtained through genetic operation. The steps are as follows: (1) Based on the specific situation of the cable ring network planning results, select the appropriate inter-station tie line set between each cable loop and determine the switch station location information of the corresponding tie line; (2) It is stipulated that each loop has only one contact line, and a genetic algorithm is used to randomly generate a contact line traversal sequence, and then an initial population is constructed after forming multiple chromosomes to form a set of optional solutions; (3) Open-loop operation is performed according to the selected contact line position, and the fitness function is calculated based on the different contact point configuration schemes of each chromosome and the line operation cost; (4) Determine the termination condition of the iteration number. If it is met, record the optimal planning scheme; otherwise, perform genetic operations on each chromosome to obtain the next generation population, and return to (3) to continue optimization until the iteration converges.

Images