Power distribution network double-layer planning method and system based on cluster division

By using electrical distance and active and reactive power balance indicators defined by voltage sensitivity in the distribution network, combining adaptive genetic algorithms for cluster division, and establishing a photovoltaic energy storage dual-layer planning model, the complex voltage control in the distribution network is solved. The problem of insufficient performance and calculation accuracy has achieved stable and economic operation of the distribution network.

CN118917753AInactive Publication Date: 2024-11-08STATE GRID SHANDONG ELECTRIC POWER CO

Patent Information

Application Number
CN202411404504.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-10-10
Publication Date
2024-11-08
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

During large-scale distributed photovoltaic access, the voltage control is complex, the calculation accuracy is insufficient, and the global optimal solution is difficult to find, and the impact of reactive power and cluster size on planning is not fully considered.

Method used

Using the electrical distance definition based on voltage sensitivity, combining active and reactive power balance indicators, a comprehensive cluster division indicator for the distribution network is constructed, an adaptive genetic algorithm is used to calculate cluster division, and a photovoltaic energy storage dual-layer planning model is established to optimize the distribution The power grid is running.

Benefits of technology

More accurate and reasonable cluster division is achieved, network loss and voltage offset are reduced, stable operation of the distribution network is ensured, and economic benefits are maximized by optimizing photovoltaic and energy storage configurations.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118917753A_ABST
    Figure CN118917753A_ABST
Patent Text Reader

Abstract

The invention belongs to the technical field of power distribution network planning, and particularly relates to a power distribution network double-layer planning method and system based on cluster division, and the method comprises the steps: constructing a modularity index, an active power balance index and a reactive power balance index of a power distribution network; combining the three indexes to form a power distribution network comprehensive cluster division index; based on the power distribution network comprehensive cluster division index, calculating power distribution network cluster division by using an adaptive genetic algorithm; on the basis of power distribution network cluster division, a power distribution network photovoltaic energy storage double-layer planning model is constructed, the power distribution network photovoltaic energy storage double-layer planning model comprises an upper-layer model and a lower-layer model, the upper-layer model is used for determining photovoltaic and energy storage configuration, and the lower-layer model further optimizes operation scheduling of a power distribution network system; and solving the power distribution network photovoltaic energy storage double-layer planning model to obtain an optimal power distribution network double-layer planning scheme. According to the invention, the network loss and the voltage offset can be minimized, and stable operation of the power grid is ensured.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of distribution network planning, and in particular relates to a distribution network double-layer planning method and system based on cluster division. Background Art

[0002] With the rapid development of renewable energy, distributed photovoltaics (DPV) has become an important part of modern distribution networks. DPV not only helps to reduce greenhouse gas emissions, but also improves the economy and sustainability of distribution networks. After the large-scale access of distributed power sources to the distribution network, the proportion of renewable energy power generation in the distribution network has continued to increase, changing the unidirectional flow characteristics of the traditional distribution network, making the overvoltage problem of the distribution network more serious, affecting the safety, stability and economic operation of the distribution network. In addition, high-proportion distributed power sources are large in scale, scattered in location, large in number, and have uncertain output. These characteristics greatly increase the complexity of voltage control in the distribution network. In recent years, many studies have focused on how to effectively integrate DPV into the distribution network to achieve efficient, stable and reliable power supply.

[0003] The large-scale decentralized access of distributed power sources such as wind and solar to the distribution network has an unignorable impact on the operation control and economic dispatch of the power system, making the traditional centralized control mode face problems such as complex control and "dimensionality curse". Therefore, the optimization and development of distribution network zoning is particularly important. When large-scale distributed photovoltaics are connected to the distribution network, due to the wide range of distributed photovoltaic installation points, the variable dimension in the centralized planning model increases sharply, and the model will become too complex and difficult to solve. In order to solve the problem of centralized optimization, planning the distribution network in clusters provides a new idea for solving the problem. Cluster planning can not only decompose the complex centralized optimization problem into simple sub-problems of optimization within the cluster, but also consider the power matching degree between the photovoltaic output and the load connected to the cluster during the planning process, which is convenient for subsequent coordinated control in clusters and improves the absorption rate of distributed photovoltaics.

[0004] In terms of cluster division, existing research focuses on several key indicators: electrical distance, power balance, and system coupling. Electrical distance is usually defined based on the sensitivity of voltage amplitude to reactive power, while power balance focuses on the balance of active power within the cluster. System coupling indicators analyze the closeness of connections within a cluster and the sparseness of connections between clusters. These indicators provide important references for cluster planning, but many studies have not fully considered the impact of reactive power on planning and the impact of cluster size on planning results.

[0005] In terms of clustering algorithms, existing optimization methods include clustering algorithms, particle swarm optimization algorithms, community detection algorithms, and genetic algorithms. These algorithms have the problem of insufficient calculation accuracy when dealing with complex clustering indicators and may fall into local optimal solutions. Therefore, developing new clustering algorithms to solve complex distribution network planning problems and find the global optimal solution is an urgent challenge. Summary of the invention

[0006] In view of the above deficiencies in the prior art, the purpose of the present invention is to provide a two-layer planning method and system for a distribution network based on cluster division. The cluster division is more accurate and reasonable, and a two-layer planning model is established on the basis of the cluster division, which can minimize network losses and voltage deviations and ensure the stable operation of the power grid.

[0007] To achieve the above objectives, the present invention provides a two-layer planning method for a distribution network based on cluster division, comprising the following steps: S1. Construct the modularity index, active power balance index and reactive power balance index of the distribution network; S2. Combining the modularity index, active power balance index and reactive power balance index of the distribution network to form a comprehensive cluster division index of the distribution network; S3, based on the comprehensive cluster division index of the distribution network, using the adaptive genetic algorithm to calculate the distribution network cluster division; S4. Based on the distribution network cluster division, a two-layer planning model for photovoltaic energy storage in the distribution network is constructed. The two-layer planning model for photovoltaic energy storage in the distribution network includes an upper model and a lower model. The upper model is used to determine the photovoltaic and energy storage configuration, and the lower model further optimizes the operation and scheduling of the distribution network system. S5. Solve the photovoltaic energy storage double-layer planning model of the distribution network to obtain the optimal double-layer planning scheme of the distribution network.

[0008] As a preferred solution of the present invention, in S1, the construction process of the modularity index, the active power balance index and the reactive power balance index are respectively: S11, Modularity index: For nodes i and j in the distribution network, the electrical distance between them is defined by voltage sensitivity as: (1); (2); In the formula, , They represent the electrical distances of node i and node j based on active and reactive voltage sensitivity matrices respectively; , They represent the sensitivity coefficients of active and reactive power injected into node i to its own voltage amplitude; , They represent the sensitivity coefficients of active and reactive power injected into node j to its own voltage amplitude; , They represent the sensitivity coefficients of active and reactive power injected into node j to the voltage amplitude of node i respectively; , They represent the sensitivity coefficients of active and reactive power injected into node i to the voltage amplitude of node j respectively; The electrical distance between node i and node j Defined as: (3); In the formula, the weight coefficient , They respectively represent the recovery ratio of active power regulation and reactive power compensation to the voltage exceeding the limit when the voltage exceeds the limit; The modularity index Q is expressed as: (4); In the formula, k i , k j are the degrees of node i and node j respectively (in graph theory, the degree of a node refers to the number of edges connecting a node to other nodes. In the above formula, the degree k of node i i represents the number of edges connecting node i to other nodes); m is the sum of the weights of all nodes; is 01 variable, Indicates that node i and node j are in the same cluster, otherwise it is 0; is a function of the electrical distance between nodes, and its value is between [0, 1], expressed as: (5); S12, Active power balance index It is expressed as: (6); (7); In the formula, is the active power balance of cluster c; T is the time length (the value is 24, which means one day); is the net power of cluster c at time t; Nc is the number of cluster divisions; S13, reactive power balance index It is expressed as: (8); (9); In the formula, Q c is the reactive balance degree of cluster c; is the maximum value of reactive power supply within cluster c; is the reactive power demand value inside cluster c.

[0009] As a preferred solution of the present invention, in S2, combining equations (4), (7) and (9), the distribution network comprehensive cluster division index f is expressed as: (10); In the formula, , , They are the weights of modularity index, active power balance index, and reactive power balance index, respectively. They can be set according to requirements.

[0010] As a preferred solution of the present invention, in S3, the adaptive genetic algorithm simulates the inheritance, mutation and selection mechanisms in the biological evolution process. The solution is represented as an individual, i.e., a chromosome, which is composed of genes representing parameters. The initial population is randomly generated, and then selection, crossover and mutation are performed. After multiple generations of iterations, the global optimal solution is obtained. The solution idea is that when the fitness of each individual in the population tends to be consistent or tends to be locally optimal, the crossover probability P c and mutation probability P m Increase to jump out of the local optimum; when the population fitness is relatively dispersed, make P c and P m Reduce to facilitate the survival of superior individuals; Improve the adaptive genetic algorithm and adjust P through adaptive strategy c and P m The value of is expressed as: (11); (12); In the formula, , They are the maximum fitness value in the population and the average fitness value of the population in each generation; is the larger fitness value of the two individuals to be crossed; f fit is the fitness value of the individual to be mutated; g1, g2, g3, and g4 are all adaptive control parameters, which take fixed values ​​in the interval (0, 1).

[0011] As a preferred solution of the present invention, in S3, the method for calculating the distribution network cluster division based on the improved adaptive genetic algorithm is: S31. Initialize the population and define the encoding method of individuals: each individual represents a possible cluster division scheme, where each gene represents the cluster number of a node; when initializing the population, randomly generate a group of individuals, each of which has a length equal to the number of nodes, and encode them; S32, calculating the fitness value: the fitness value of each individual reflects the quality of its corresponding cluster division scheme. The fitness function is the comprehensive cluster division index f of the distribution network. The higher the fitness value, the better the cluster division scheme corresponding to the individual. S33, select M individuals according to the roulette rules, calculate and , randomly pair the individuals in the population into M / 2 pairs in total, and for each pair of individuals, calculate P according to formula (11) c , randomly generate R (0, 1), if R < P c Then the crossover operation is performed on the pair of individuals; S34. For all M individuals in the population, calculate P according to formula (12): m , randomly generate O(0,1), if O<P m Then perform a crossover operation on the individual; S35, calculate the fitness of the new individuals generated by crossover and mutation, and select the next generation parent individuals; add the newly generated offspring individuals to the population to form a new generation population; during the update process, retain several individuals with the highest fitness values ​​(the number of retained individuals is set according to demand) to ensure that the optimal solution is not discarded; S36, check termination conditions: determine whether the termination conditions are met. If the maximum number of iterations is reached or the fitness value reaches a preset threshold, the iteration is terminated and the optimal cluster division scheme is output; otherwise, return to S33 and continue iteration.

[0012] As a preferred solution of the present invention, in S4, the goal of the upper model is to minimize the economic cost of each cluster. Given the photovoltaic and load forecast values, the photovoltaic and energy storage of each cluster are sited and capacity determined, which can be expressed as: (13); (14); (15); (16); (17); Where C represents the set of all clusters; C1 represents the equivalent annual installation investment cost (i.e., the initial investment in equipment, including the installation cost of photovoltaic and energy storage systems, borne by the distribution network); and C2 represents the operating cost of the distribution network. represents the total economic cost of cluster c; represents the installation cost of the planned distribution network system in cluster c; r is the bank interest rate; is the photovoltaic life, in years; The middle is the life of the energy storage system, in years; , Respectively represent the investment cost per unit photovoltaic and energy storage capacity; , They represent the planned capacity of photovoltaic and energy storage at node j respectively; , is an intermediate variable. This means that node j is a node in cluster c.

[0013] As a preferred solution of the present invention, in S4, the goal of the lower model is to minimize the voltage deviation and operating cost of the distribution network system, which is expressed as: (18); (19); (20); (twenty one); (twenty two); (twenty three); (twenty four); In the formula, is the overall objective function of the underlying model; is the voltage deviation of the distribution network system; Run network loss for the cluster; represents the operation and maintenance cost of cluster c; represents the abandoned light cost of cluster c; represents the electricity purchase cost of cluster c; , Respectively represent the unit photovoltaic and energy storage unit operating costs; is the unit abandoned light cost; is the maximum photovoltaic output of node j; , They represent the actual active power of photovoltaic and energy storage at node j at time t respectively; is the collection of main network contact branches of cluster c, l represents The branch road in is the cluster c line set; is the main grid electricity price at time t; is the actual active power of the main network contact branch of cluster c at time t; is the actual voltage of node i at time t; N is the total number of nodes; is the line impedance between node i and node j; is the line current between node i and node j.

[0014] As a preferred solution of the present invention, the constraints of the upper model are: The node capacity constraint is expressed as: (25); In the formula, represents the maximum photovoltaic capacity allowed to be installed at node j; The cluster capacity penetration constraint is expressed as: (26); In the formula, is the cluster capacity penetration rate; is the maximum apparent power of the node load in cluster c; The constraints of the lower model are: The power flow constraint is expressed as: (27); (28); (29); In the formula, , are respectively the active and reactive power flowing in the line between node i and node j at time t; is the square of the voltage value of node i at time t; is the square value of the current in the line between node i and node j at time t; , They are the upper and lower voltage limits respectively; is the maximum current limit; The power balance constraint is expressed as: (30); In the formula, is the load power of node j at time t; is the distribution network loss; The charging and discharging power constraints of the energy storage system are expressed as: (31); (32); In the formula, The energy storage capacity installed for node j; , They are respectively the energy storage charging and discharging power at time t; , They are 01 variables for charging and discharging, 1 represents charging and discharging; , are the energy storage charging and discharging efficiency respectively; The state of charge constraint of the energy storage system is expressed as: (33); (34); (35); (36); In the formula, , are the charge states of the energy storage at time t and time t+1 respectively; , They are the minimum and maximum states of charge for energy storage, respectively.

[0015] As a preferred embodiment of the present invention, in the S5, the method for solving the two-level programming model of photovoltaic energy storage in the distribution network is to use the cplex solver to solve the upper model, and use the improved multi-objective optimization particle swarm algorithm (here, the well-known improved algorithm MOPSO can be used) to solve the lower model. When initializing the position and speed of the particles, the result of cluster division is used as the position constraint of the particles; after reaching the maximum number of iterations, the information entropy method is used to determine the weight, and the entropy weight method is used to select the optimal solution; the solution process of the lower model is specifically as follows: S51, parameter setting: Set the number of objective functions to two, representing two different optimization objectives (corresponding to and C2); determine the minimum and maximum values ​​of the objective function, define the number of initial populations, the dimension of decision variables, and the maximum number of iterations; set the upper and lower bounds of the inertia coefficient, learning factor, speed, and position; S52, during the initialization process, the initial position X and velocity V of the particle are generated. The generation of position and velocity follows the upper and lower bounds X of the decision space. max and X min ; Combined with the randomly generated values, each particle in the population is subjected to energy storage state constraint processing, the individual and group historical optimal positions and fitness are initialized, and an external archive set is set to store the initial state of the non-inferior solution; S53. During the iteration process, the fitness value of each particle is calculated, the historical optimal position and fitness of the individual are updated to reflect the current optimal solution, the historical optimal position and fitness of the group are updated, and the new solution is added to the external archive set. The external archive set is screened by dominance relationship and crowding distance to retain the optimal solution, and the size of the archive set is controlled. The solutions in the archive set are sorted using the entropy weight method to determine the optimal solution; S54, the speed and position of the particle are updated in each iteration according to the current inertia coefficient, learning factor and the historical optimal position of the individual and group (the speed and position update can be calculated by the known formula); when the position is updated, the particle position is ensured to be within the decision space to maintain an effective search space; S55, iterating according to the maximum number of iterations, updating the speed and position of the particle each time to gradually improve the quality of the solution, and calculating the final optimal solution by the entropy weight method after all iterations are completed; S56. When the algorithm ends, the entropy weight method is used to select the optimal solution from the external archive set, and finally the decision variables of the optimal solution and their corresponding fitness values ​​are output as the result of the algorithm.

[0016] A two-layer planning system for distribution network based on cluster division is used to implement the above method, including: An indicator construction module is used to construct the modularity index, active power balance index and reactive power balance index of the distribution network, and combine the modularity index, active power balance index and reactive power balance index of the distribution network to form a comprehensive cluster division index of the distribution network; The clustering module calculates the distribution network clustering based on the comprehensive distribution network clustering index and uses an adaptive genetic algorithm. The model building module builds a two-layer planning model for photovoltaic and energy storage in the distribution network based on the distribution network cluster division. The two-layer planning model for photovoltaic and energy storage in the distribution network includes an upper model and a lower model. The upper model is used to determine the photovoltaic and energy storage configuration, and the lower model further optimizes the operation and scheduling of the distribution network system. The model solving module solves the photovoltaic energy storage two-layer planning model of the distribution network to obtain the optimal two-layer planning scheme of the distribution network.

[0017] The algorithm and module involved in the present invention can be executed by an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. The above-mentioned algorithm calculation and module functions are realized by executing software by the processor.

[0018] The beneficial effects of the present invention are: The present invention is based on the definition method of electrical distance considering voltage sensitivity, considers the influence of active and reactive power on voltage, combines the active power balance index, the reactive power balance index and the modularity index considering electrical distance as the comprehensive performance index of the cluster, and uses an adaptive genetic algorithm with added crossover mutation operation to calculate the distribution network cluster division, so that the cluster division is more accurate and reasonable.

[0019] Based on the distribution network cluster division, the present invention establishes a two-layer planning model, in which the upper-layer economic optimization model focuses on the site selection and capacity configuration of photovoltaic and energy storage, and maximizes economic benefits by optimizing investment and operating costs; based on the photovoltaic and energy storage configuration determined by the upper-layer model, the lower-layer model further optimizes the operation and scheduling of the distribution network system, aiming to minimize network losses and voltage deviations and ensure the stable operation of the power grid; the lower-layer model includes detailed constraints such as flow constraints and energy storage system charging and discharging constraints to achieve stable and economical operation of the distribution network system. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a schematic diagram of the process of the method of the present invention; Figure 2 It is a schematic diagram of cluster division results in the verification process of the present invention; Figure 3 is a photovoltaic output curve diagram during the verification process of the present invention; Figure 4 It is a typical daily photovoltaic output curve diagram during the verification process of the present invention; Figure 5 It is a schematic diagram of the cluster optical storage configuration strategy in the verification process of the present invention; Figure 6 It is a schematic diagram of 24-hour load demand of the distribution network system during the verification process of the present invention; Figure 7 It is a schematic diagram of 24-hour dispatching output of energy storage during the verification process of the present invention; Figure 7 (a) is a schematic diagram of the dispatch output of the energy storage system 1; Figure 7 (b) is a schematic diagram of the dispatch output of the energy storage system 2; Figure 7 (c) is a schematic diagram of the dispatch output of the energy storage system 3; Figure 7 (d) is a schematic diagram of the dispatch output of the energy storage system 4; Figure 8 It is a comparison diagram between the centralized planning and the cluster partitioning planning in the verification process of the present invention; Fig. 9 It is a principle block diagram of the system of the present invention. DETAILED DESCRIPTION

[0021] The embodiments of the present invention are further described below in conjunction with the accompanying drawings: Example 1 like Figure 1 As shown, a two-layer planning method for distribution network based on cluster division includes the following steps: S1. Construct the modularity index, active power balance index and reactive power balance index of the distribution network; S2. Combining the modularity index, active power balance index and reactive power balance index of the distribution network to form a comprehensive cluster division index of the distribution network; S3, based on the comprehensive cluster division index of the distribution network, using the adaptive genetic algorithm to calculate the distribution network cluster division; S4. Based on the distribution network cluster division, a two-layer planning model for photovoltaic energy storage in the distribution network is constructed. The two-layer planning model for photovoltaic energy storage in the distribution network includes an upper model and a lower model. The upper model is used to determine the photovoltaic and energy storage configuration, and the lower model further optimizes the operation and scheduling of the distribution network system. S5. Solve the photovoltaic energy storage double-layer planning model of the distribution network to obtain the optimal double-layer planning scheme of the distribution network.

[0022] In S1, the construction processes of the modularity index, active power balance index and reactive power balance index are as follows: S11, Modularity index: For nodes i and j in the distribution network, the electrical distance between them is defined by voltage sensitivity as: (1); (2); In the formula, , They represent the electrical distances of node i and node j based on active and reactive voltage sensitivity matrices respectively; , They represent the sensitivity coefficients of active and reactive power injected into node i to its own voltage amplitude; , They represent the sensitivity coefficients of active and reactive power injected into node j to its own voltage amplitude; , They represent the sensitivity coefficients of active and reactive power injected into node j to the voltage amplitude of node i respectively; , They represent the sensitivity coefficients of active and reactive power injected into node i to the voltage amplitude of node j respectively; In this way, the electrical distance between nodes is obtained by comprehensively considering the influence of the nodes themselves and each other. The smaller the electrical distance, the closer the electrical connection between the two nodes. Given that the active voltage sensitivity and reactive voltage sensitivity between nodes are not proportional, the electrical distance between node i and node j is Defined as: (3); In the formula, the weight coefficient , They respectively represent the recovery ratio of active power regulation and reactive power compensation to the voltage exceeding the limit, which can be calculated and determined based on the most serious overvoltage scenario; Modularity index is often used to solve the partitioning problem in complex networks. If the modularity index is larger, it means that the node similarity within the cluster is higher and the node similarity with the external cluster is lower, that is, the divided cluster structure is stronger. The modularity index Q is expressed as: (4); In the formula, k i , k j are the degrees of node i and node j respectively; m is the sum of the weights of all nodes; is 01 variable, Indicates that node i and node j are in the same cluster, otherwise it is 0; is a function of the electrical distance between nodes, and its value is between [0, 1], expressed as: (5); The modularity index Q is a modularity index based on electrical distance. The larger its value is, the closer the electrical connection between nodes in the same cluster is, and the sparser the electrical connection between nodes in different clusters is.

[0023] In order to improve the local consumption capacity of distributed generation (DG) within the cluster and reduce the network loss during energy transmission, the active power transfer between clusters should be reduced, the cluster power backfeed should be reduced, and the power complementary characteristics of source, load and storage within the cluster should be enhanced as much as possible. This embodiment uses active power balance as the division indicator.

[0024] S12, Active power balance index It is expressed as: (6); (7); In the formula, is the active power balance of cluster c; T is the time length (the value is 24, which means one day); is the net power of cluster c at time t; Nc is the number of cluster divisions; Active power balance aims to balance the net power complementarity level between nodes in the network, help identify and optimize the complementary characteristics of power generation, load and energy storage systems within the cluster, and help maintain the voltage and frequency stability of the system by ensuring the balance of active power between nodes, avoiding large-scale voltage fluctuations and frequency deviations, so as to fully utilize the cluster's autonomous capabilities.

[0025] S13, reactive power balance index It is expressed as: (8); (9); In the formula, Q c is the reactive balance degree of cluster c; is the maximum value of reactive power supply within cluster c; is the reactive power demand value inside cluster c.

[0026] The reactive power balance index can effectively monitor and adjust the reactive power supply of each node in the distribution network system, ensure the reasonable distribution of reactive power, and help identify areas with insufficient or excessive reactive power supply in the distribution network system. Through reasonable resource allocation and optimization, it can ensure the efficient use of reactive power resources in the system and avoid resource waste and unnecessary losses.

[0027] In S2, combining equations (4), (7) and (9), the distribution network comprehensive cluster division index f is expressed as: (10); In the formula, , , They are the weights of the modularity index, active power balance index, and reactive power balance index respectively.

[0028] In S3, the adaptive genetic algorithm simulates the inheritance, mutation and selection mechanisms in the biological evolution process. The solution is represented as an individual, namely a chromosome, which is composed of genes representing parameters. The initial population is randomly generated, and then selection, crossover and mutation are performed. After multiple generations of iterations, the global optimal solution is obtained. The solution idea is to make the crossover probability P when the fitness of each individual in the population tends to be consistent or tends to the local optimum. c and mutation probability P m Increase to jump out of the local optimum; when the population fitness is relatively dispersed, make P c and P m Reduce to facilitate the survival of superior individuals; At the same time, for individuals whose fitness is higher than the average fitness of the group, a lower P is selected. c and P m , so that the solution can be protected into the next generation; for individuals with fitness values ​​below the average, a higher P is selected c and P m The adaptive genetic algorithm is improved by adjusting P through adaptive strategies. c and P m The value of is expressed as: (11); (12); In the formula, , They are the maximum fitness value in the population and the average fitness value of the population in each generation; is the larger fitness value of the two individuals to be crossed; f fit is the fitness value of the individual to be mutated; g1, g2, g3, and g4 are all adaptive control parameters, which take fixed values ​​in the interval (0, 1).

[0029] In S3, based on the improved adaptive genetic algorithm, the method for calculating the distribution network cluster division is: S31. Initialize the population and define the encoding method of individuals: each individual represents a possible cluster division scheme, where each gene represents the cluster number of a node; when initializing the population, randomly generate a group of individuals, each of which has a length equal to the number of nodes, and encode them; S32, calculating the fitness value: the fitness value of each individual reflects the quality of its corresponding cluster division scheme. The fitness function is the comprehensive cluster division index f of the distribution network. The higher the fitness value, the better the cluster division scheme corresponding to the individual. S33, select M individuals according to the roulette rules, calculate and , randomly pair the individuals in the population into M / 2 pairs in total, and for each pair of individuals, calculate P according to formula (11) c , randomly generate R (0, 1), if R < P c Then the crossover operation is performed on the pair of individuals; S34. For all M individuals in the population, calculate P according to formula (12): m , randomly generate O(0,1), if O<P m Then perform a crossover operation on the individual; S35, calculate the fitness of the new individuals generated by crossover and mutation, and select the next generation parent individuals; add the newly generated offspring individuals to the population to form a new generation population; during the updating process, retain several individuals with the highest fitness values ​​to ensure that the optimal solution is not discarded; S36, check termination conditions: determine whether the termination conditions are met. If the maximum number of iterations is reached or the fitness value reaches a preset threshold, the iteration is terminated and the optimal cluster division scheme is output; otherwise, return to S33 and continue iteration.

[0030] This embodiment establishes a two-level planning model based on the distribution network cluster division. In the upper-level economic optimization stage, the decision variables include the capacity of photovoltaic and energy storage. By optimizing these variables, the total investment cost and operating cost of photovoltaic and energy storage are minimized. The capacity decision output by the upper-level optimization will be passed as input to the lower-level operation optimization stage.

[0031] In the lower-level operation optimization stage, the power loss and voltage offset of the power grid are minimized by optimizing the power flow of each node. By solving the lower-level optimization problem, the stability of the power grid operation is ensured, and the optimization results are returned to the upper level to provide support for the economic benefits and stable operation of the overall system.

[0032] In S4, the goal of the upper model is to minimize the economic cost of each cluster. Given the photovoltaic and load forecast values, the photovoltaic and energy storage of each cluster are sited and capacity determined, taking into account the installation cost, operation and maintenance cost, aiming to minimize the economic cost through reasonable photovoltaic and energy storage configuration. This optimization process must consider not only the initial investment and long-term operating costs of the equipment, but also the effective use of power resources and the overall economy of the system. The upper model is expressed as: (13); (14); (15); (16); (17); In the formula, C represents the set of all clusters; C1 represents the equivalent annual installation investment cost, and C2 represents the distribution network operation cost; represents the total economic cost of cluster c; represents the installation cost of the planned distribution network system in cluster c; r is the bank interest rate; is the photovoltaic life, in years; The middle is the life of the energy storage system, in years; , Respectively represent the investment cost per unit photovoltaic and energy storage capacity; , They represent the planned capacity of photovoltaic and energy storage at node j respectively; , is an intermediate variable.

[0033] After the upper model determines the optimal configuration of photovoltaic and energy storage, the lower model further optimizes the operation and dispatching strategy of the distribution network system and calculates the operation cost based on these decision results. The main task of the lower model is to minimize the voltage deviation and operation cost of the distribution network system through a reasonable energy storage dispatching strategy. In S4, the goal of the lower model is to ensure the stable operation and economic benefits of the power grid, which is expressed as: (18); (19); (20); (twenty one); (twenty two); (twenty three); (twenty four); In the formula, is the overall objective function of the underlying model; is the voltage deviation of the distribution network system; Run network loss for the cluster; represents the operation and maintenance cost of cluster c; represents the abandoned light cost of cluster c; represents the electricity purchase cost of cluster c; , Respectively represent the unit photovoltaic and energy storage unit operating costs; is the unit abandoned light cost; is the maximum photovoltaic output of node j; , They represent the actual active power of photovoltaic and energy storage at node j at time t respectively; is the collection of main network contact branches of cluster c, l represents The branch road in is the cluster c line set; is the main grid electricity price at time t; is the actual active power of the main network contact branch of cluster c at time t; is the actual voltage of node i at time t; N is the total number of nodes; is the line impedance between node i and node j; is the line current between node i and node j.

[0034] The constraints of the upper model are: The node capacity constraint is expressed as: (25); In the formula, represents the maximum photovoltaic capacity allowed to be installed at node j; In order to ensure that DG is consumed within the cluster and reduce cross-cluster power exchange, the cluster is used as the basic unit during planning, and the total DG configuration capacity in each cluster should be determined by the total load in the cluster. The total DG configuration capacity in the cluster is limited by the cluster capacity penetration rate constraint. The cluster capacity penetration rate constraint is expressed as: (26); In the formula, is the cluster capacity penetration rate; is the maximum apparent power of the node load in cluster c; The constraints of the lower model are: The power flow constraint is expressed as: (27); (28); (29); In the formula, , are respectively the active and reactive power flowing in the line between node i and node j at time t; is the square of the voltage value of node i at time t; is the square value of the current in the line between node i and node j at time t; , They are the upper and lower voltage limits respectively; is the maximum current limit; The power balance constraint is expressed as: (30); In the formula, is the load power of node j at time t; is the distribution network loss; The charging and discharging power constraints of the energy storage system are expressed as: (31); (32); In the formula, The energy storage capacity installed for node j; , They are respectively the energy storage charging and discharging power at time t; , They are 01 variables for charging and discharging, 1 represents charging and discharging; , are the energy storage charging and discharging efficiency respectively; The state of charge constraint of the energy storage system is expressed as: (33); (34); (35); (36); In the formula, , are the charge states of the energy storage at time t and time t+1 respectively; , They are the minimum and maximum states of charge for energy storage, respectively.

[0035] In S5, the method for solving the two-level planning model of photovoltaic energy storage in the distribution network is to use the cplex solver to solve the upper model, and use the improved multi-objective optimization particle swarm algorithm to solve the lower model. When initializing the position and speed of the particles, the cluster division results are used as the position constraints of the particles; after reaching the maximum number of iterations, the information entropy method is used to determine the weights, and the entropy weight method is used to select the optimal solution; the specific solution process of the lower model is as follows: S51, parameter setting: set the number of objective functions to two, representing two different optimization objectives; determine the minimum and maximum values ​​of the objective function, define the number of initial populations, the dimension of decision variables, and the maximum number of iterations; set the upper and lower bounds of the inertia coefficient, learning factor, speed, and position; S52, during the initialization process, the initial position X and velocity V of the particle are generated. The generation of position and velocity follows the upper and lower bounds X of the decision space. max and X min ; Combined with the randomly generated values, each particle in the population is subjected to energy storage state constraint processing, the individual and group historical optimal positions and fitness are initialized, and an external archive set is set to store the initial state of the non-inferior solution; S53. During the iteration process, the fitness value of each particle is calculated, the historical optimal position and fitness of the individual are updated to reflect the current optimal solution, the historical optimal position and fitness of the group are updated, and the new solution is added to the external archive set. The external archive set is screened by dominance relationship and crowding distance to retain the optimal solution, and the size of the archive set is controlled. The solutions in the archive set are sorted using the entropy weight method to determine the optimal solution; S54, the speed and position of the particle are updated in each iteration according to the current inertia coefficient, learning factor and the historical optimal position of the individual and group; when updating the position, ensure that the particle position is within the decision space to maintain an effective search space; S55, iterating according to the maximum number of iterations, updating the speed and position of the particle each time to gradually improve the quality of the solution, and calculating the final optimal solution by the entropy weight method after all iterations are completed; S56. When the algorithm ends, the entropy weight method is used to select the optimal solution from the external archive set, and finally the decision variables of the optimal solution and their corresponding fitness values ​​are output as the result of the algorithm.

[0036] The verification process is: The IEEE33 node system is taken as the research object and the simulation analysis is carried out. The voltage level of the example system is 12.66kV and the reference power is 1MVA. The total active load of the node system is 4.04MW, and the reactive load is divided into clusters. The programming calculation is performed in the MATLAB2017a environment, and the population number is set to 100 and the maximum number of iterations is set to 200. The population is continuously evolved through selection, crossover, mutation and inversion (randomly selecting two positions of a chromosome and then reversing the gene order between the two positions) operations to obtain the best optimal individual and the corresponding fitness value.

[0037] Table 1 Comparison of indicators of clusters with different weight coefficients

[0038] As shown in Table 1, increasing the weight coefficient of the modularity index , the comprehensive index of cluster division is reduced, and the weight coefficient of active power balance index is increased Or the weight coefficient of reactive power balance index , the corresponding clustering index in the clustering result increases. Therefore, when the comprehensive clustering index f reaches the maximum value, the corresponding optimal weight coefficient is: =0.2, =0.5, = 0.3, and the weight combination is used to analyze the cluster division results. The corresponding cluster division results are as follows Figure 2 As shown, each dotted box represents a cluster, and there are 4 clusters in total.

[0039] The 24-hour light intensity data for 36 days is processed to obtain the photovoltaic power generation. Considering the impact of temperature on photovoltaic power generation efficiency, the minimum threshold of power generation and the invalid period (nighttime) are set, and the obtained data is clustered to obtain the following: Figure 3 The photovoltaic output characteristic curves under the four scenarios shown are each curve representing the output power of the photovoltaic system under different conditions.

[0040] Using the weighted summation of clustering probabilities, the weighted cluster centers are calculated to obtain a typical power generation curve reflecting the average daily variation, such as Figure 4 This model reflects the comprehensive effects under different weather and environmental conditions and serves as the standard performance model of the photovoltaic system in this verification example.

[0041] Based on the cluster division and photovoltaic output model, the upper layer takes the comprehensive economic cost of the system (that is, the node system, the same below) as the objective function, and the lower layer takes the system scheduling cost and voltage offset as the objective function. It takes into account the operation and maintenance costs, abandoned light costs and electricity purchase costs during system operation, aiming to minimize economic costs through reasonable resource allocation and energy storage scheduling.

[0042] Cplex is used to find the photovoltaic and energy storage configuration that minimizes costs and maximizes benefits. At the same time, multi-objective optimization is used to deal with multiple conflicting objectives that may exist, and the Pareto front is used to maintain the dynamic update of the best non-dominated solution. In addition, at the end of the iteration, the entropy weight method based on information entropy selects a final solution from the archive set to obtain a relatively optimal solution calculated based on the target weight and the objective function value.

[0043] Combining these two methods, we can effectively handle multi-objective optimization problems and ensure that good results are achieved on multiple evaluation criteria. The energy configuration strategy (installation location and capacity of photovoltaic and energy storage) is obtained as follows: Figure 5 shown.

[0044] Table 2 Cluster optical storage configuration strategy

[0045] As can be seen from Table 2, clusters 1 and 2 have large load demands, indicating that the load in the cluster is relatively high, and the photovoltaic and energy storage capacity is relatively large, which can balance the load through the energy storage system when the photovoltaic power supply is insufficient. Cluster 3 has moderate photovoltaic capacity and small energy storage capacity, and the load demand is well matched with the photovoltaic configuration. Cluster 4 has a small energy demand and a small corresponding photovoltaic storage capacity. Through analysis, it can be seen that the division plan may more accurately match the load demand and energy production, thereby optimizing investment and operating costs. The use of moderate photovoltaic and energy storage capacity can control costs while meeting demand.

[0046] Figure 6 It shows the power consumption and production at different times of the day, demonstrating the interactive role of photovoltaic systems and energy storage equipment in regulating power demand and supply. By properly configuring the energy storage system, it is possible to effectively balance the excess power generation during the day and the higher power demand at night. Figure 7 China Energy Storage's charging and discharging strategy effectively ensures that electricity can be supplemented by releasing stored energy when photovoltaic power generation is insufficient to meet demand, while storing energy when there is excess electricity.

[0047] Photovoltaics of different capacities are installed on the nodes. This can improve the reliability and stability of the system and prevent single-point failures from causing serious impacts on the overall system. Under centralized planning, the capacity is concentrated near some nodes, and the power transmission distance is long, which may lead to higher transmission losses. Through the above analysis, it can be seen that cluster planning has advantages in photovoltaic capacity distribution, such as more uniform distribution, higher flexibility, lower transmission loss, and optimized resource utilization. These advantages make cluster planning superior to centralized planning in terms of reliability, efficiency, and economy.

[0048] from Figure 8It can be seen that under cluster planning, the photovoltaic capacity is distributed more evenly on more nodes, and photovoltaics of different capacities are installed on nodes 5, 16, 22, and 29. It can improve the reliability and stability of the system and avoid single-point failures from causing serious impact on the overall system. Under centralized planning, the capacity is concentrated near some nodes, and the power transmission distance is long, which may lead to higher transmission losses. Through the above analysis, it can be seen that cluster planning has advantages in photovoltaic capacity distribution, such as more uniform distribution, higher flexibility, lower transmission loss, and optimization of resource utilization. These advantages make cluster planning superior to centralized planning in terms of reliability, efficiency, and economy.

[0049] As can be seen from Table 3, the total operating cost of the planning scheme based on cluster division is reduced by about 1.16%, the voltage deviation is reduced by about 5.88%, the cost of abandoned light is reduced by about 66.06%, the operation and maintenance cost is reduced by about 17.42%, the power purchase cost is slightly higher by about 1.64%, and the network loss cost is reduced by about 7.97%. It can be seen that the planning scheme based on cluster division is superior to the centralized planning scheme in multiple key cost indicators, improves the photovoltaic investment cost and output income efficiency, and significantly reduces the cost of abandoned light. It reduces the operation and maintenance cost and network loss cost, and improves the overall economy and efficiency of the power grid. It optimizes voltage stability and helps to improve the reliability of the power grid.

[0050] Table 3 Comparison of the cost of centralized planning and cluster division planning for node systems

[0051] Example 2 like Fig. 9 As shown, a two-layer planning system for a distribution network based on cluster division is used to implement the method in Example 1, including: An indicator construction module is used to construct the modularity index, active power balance index and reactive power balance index of the distribution network, and combine the modularity index, active power balance index and reactive power balance index of the distribution network to form a comprehensive cluster division index of the distribution network; The clustering module calculates the distribution network clustering based on the comprehensive distribution network clustering index and uses an adaptive genetic algorithm. The model building module builds a two-layer planning model for photovoltaic and energy storage in the distribution network based on the distribution network cluster division. The two-layer planning model for photovoltaic and energy storage in the distribution network includes an upper model and a lower model. The upper model is used to determine the photovoltaic and energy storage configuration, and the lower model further optimizes the operation and scheduling of the distribution network system. The model solving module solves the photovoltaic energy storage two-layer planning model of the distribution network to obtain the optimal two-layer planning scheme of the distribution network.

Claims

1. A two-layer planning method for distribution network based on cluster division, characterized in that The following steps are involved: S1. Construct the modularity index, active power balance index and reactive power balance index of the distribution network; S2. Combining the modularity index, active power balance index and reactive power balance index of the distribution network to form a comprehensive cluster division index of the distribution network; S3, based on the comprehensive cluster division index of the distribution network, using the adaptive genetic algorithm to calculate the distribution network cluster division; S4. Based on the distribution network cluster division, a two-layer planning model for photovoltaic energy storage in the distribution network is constructed. The two-layer planning model for photovoltaic energy storage in the distribution network includes an upper model and a lower model. The upper model is used to determine the photovoltaic and energy storage configuration, and the lower model further optimizes the operation and scheduling of the distribution network system. S5. Solve the photovoltaic energy storage double-layer planning model of the distribution network to obtain the optimal double-layer planning scheme of the distribution network.

2. A two-layer planning method for distribution network based on cluster division according to claim 1, characterized in that: In the above S1, the construction processes of the modularity index, the active power balance index and the reactive power balance index are respectively: S11, Modularity index: For nodes i and j in the distribution network, the electrical distance between them is defined by voltage sensitivity as: (1); (2); In the formula, , They represent the electrical distances of node i and node j based on active and reactive voltage sensitivity matrices respectively; , They represent the sensitivity coefficients of active and reactive power injected into node i to its own voltage amplitude; , They represent the sensitivity coefficients of active and reactive power injected into node j to its own voltage amplitude; , They represent the sensitivity coefficients of active and reactive power injected into node j to the voltage amplitude of node i respectively; , They represent the sensitivity coefficients of active and reactive power injected into node i to the voltage amplitude of node j respectively; The electrical distance between node i and node j Defined as: (3); In the formula, the weight coefficient , They respectively represent the recovery ratio of active power regulation and reactive power compensation to the voltage exceeding the limit when the voltage exceeds the limit; The modularity index Q is expressed as: (4); In the formula, k i , k j are the degrees of node i and node j respectively; m is the sum of the weights of all nodes; is 01 variable, Indicates that node i and node j are in the same cluster, otherwise it is 0; is a function of the electrical distance between nodes, and its value is between [0, 1], expressed as: (5); S12, Active power balance index It is expressed as: (6); (7); In the formula, is the active balance degree of cluster c; T is the time length; is the net power of cluster c at time t; Nc is the number of cluster divisions; S13, reactive power balance index It is expressed as: (8); (9); In the formula, Q c is the reactive balance degree of cluster c; is the maximum value of reactive power supply within cluster c; is the reactive power demand value inside cluster c.

3. A two-layer planning method for distribution network based on cluster division according to claim 2, characterized in that: In S2, combining equations (4), (7) and (9), the distribution network comprehensive cluster division index f is expressed as: (10); In the formula, , , They are the weights of the modularity index, active power balance index, and reactive power balance index respectively.

4. A two-layer planning method for distribution network based on cluster division according to claim 3, characterized in that: In S3, the adaptive genetic algorithm simulates the inheritance, mutation and selection mechanisms in the biological evolution process. The solution is represented as an individual, i.e., a chromosome, which is composed of genes representing parameters. The initial population is randomly generated, and then selection, crossover and mutation are performed. After multiple generations of iterations, the global optimal solution is obtained. The solution idea is to make the crossover probability P when the fitness of each individual in the population tends to be consistent or tends to the local optimum. c and mutation probability P m Increase to jump out of the local optimum; when the population fitness is relatively dispersed, make P c and P m Reduce to facilitate the survival of superior individuals; Improve the adaptive genetic algorithm and adjust P through adaptive strategy c and P m The value of is expressed as: (11); (12); In the formula, , They are the maximum fitness value in the population and the average fitness value of the population in each generation; is the larger fitness value of the two individuals to be crossed; f fit is the fitness value of the individual to be mutated; g1, g2, g3, and g4 are all adaptive control parameters, which take fixed values ​​in the interval (0, 1).

5. A two-layer planning method for distribution network based on cluster division according to claim 4, characterized in that: In S3, based on the improved adaptive genetic algorithm, the method for calculating the distribution network cluster division is: S31. Initialize the population and define the encoding method of individuals: each individual represents a possible cluster division scheme, where each gene represents the cluster number of a node; when initializing the population, randomly generate a group of individuals, each of which has a length equal to the number of nodes, and encode them; S32, calculating the fitness value: the fitness value of each individual reflects the quality of its corresponding cluster division scheme. The fitness function is the comprehensive cluster division index f of the distribution network. The higher the fitness value, the better the cluster division scheme corresponding to the individual. S33, select M individuals according to the roulette rules, calculate and , randomly pair the individuals in the population into M / 2 pairs in total, and for each pair of individuals, calculate P according to formula (11) c , randomly generate R (0, 1), if R < P c Then the crossover operation is performed on the pair of individuals; S34. For all M individuals in the population, calculate P according to formula (12): m , randomly generate O(0,1), if O<P m Then perform a crossover operation on the individual; S35, calculate the fitness of the new individuals generated by crossover and mutation, and select the next generation parent individuals; add the newly generated offspring individuals to the population to form a new generation population; during the updating process, retain several individuals with the highest fitness values ​​to ensure that the optimal solution is not discarded; S36, check the termination condition: determine whether the termination condition is met. If the maximum number of iterations is reached or the fitness value reaches a preset threshold, the iteration is terminated and the optimal cluster division scheme is output; Otherwise, return to S33 and continue iterating.

6. A two-layer planning method for distribution network based on cluster division according to claim 5, characterized in that: In S4, the goal of the upper model is to minimize the economic cost of each cluster. Given the photovoltaic and load forecast values, the photovoltaic and energy storage of each cluster are sited and capacity determined, which can be expressed as: (13); (14); (15); (16); (17); In the formula, C represents the set of all clusters; C1 represents the equivalent annual installation investment cost, and C2 represents the distribution network operation cost; represents the total economic cost of cluster c; represents the installation cost of the planned distribution network system in cluster c; r is the bank interest rate; is the photovoltaic life, in years; The middle is the life of the energy storage system, in years; , Respectively represent the investment cost per unit photovoltaic and energy storage capacity; , They represent the planned capacity of photovoltaic and energy storage at node j respectively; , is an intermediate variable.

7. A two-layer planning method for distribution network based on cluster division according to claim 6, characterized in that: In S4, the goal of the lower model is to minimize the voltage deviation and operating cost of the distribution network system, which can be expressed as: (18); (19); (20); (21); (22); (23); (24); In the formula, is the overall objective function of the underlying model; is the voltage deviation of the distribution network system; Run network loss for the cluster; represents the operation and maintenance cost of cluster c; represents the abandoned light cost of cluster c; represents the electricity purchase cost of cluster c; , Respectively represent the unit photovoltaic and energy storage unit operating costs; is the unit abandoned light cost; is the maximum photovoltaic output of node j; , They represent the actual active power of photovoltaic and energy storage at node j at time t respectively; is the collection of main network contact branches of cluster c, l represents The branch road in is the cluster c line set; is the main grid electricity price at time t; is the actual active power of the main network contact branch of cluster c at time t; is the actual voltage of node i at time t; N is the total number of nodes; is the line impedance between node i and node j; is the line current between node i and node j.

8. A two-layer planning method for distribution network based on cluster division according to claim 7, characterized in that: The constraints of the upper model are: The node capacity constraint is expressed as: (25); In the formula, represents the maximum photovoltaic capacity allowed to be installed at node j; The cluster capacity penetration constraint is expressed as: (26); In the formula, is the cluster capacity penetration rate; is the maximum apparent power of the node load in cluster c; The constraints of the lower model are: The power flow constraint is expressed as: (27); (28); (29); In the formula, , are respectively the active and reactive power flowing in the line between node i and node j at time t; is the square of the voltage value of node i at time t; is the square value of the current in the line between node i and node j at time t; , They are the upper and lower voltage limits respectively; is the maximum current limit; The power balance constraint is expressed as: (30); In the formula, is the load power of node j at time t; is the distribution network loss; The charging and discharging power constraints of the energy storage system are expressed as: (31); (32); In the formula, The energy storage capacity installed for node j; , They are respectively the energy storage charging and discharging power at time t; , They are 01 variables for charging and discharging, 1 represents charging and discharging; , are the energy storage charging and discharging efficiency respectively; The state of charge constraint of the energy storage system is expressed as: (33); (34); (35); (36); In the formula, , are the charge states of the energy storage at time t and time t+1 respectively; , They are the minimum and maximum states of charge for energy storage, respectively.

9. A two-layer planning method for distribution network based on cluster division according to claim 7, characterized in that: In the above S5, the method for solving the two-level planning model of photovoltaic energy storage in the distribution network is to use the cplex solver to solve the upper model, and use the improved multi-objective optimization particle swarm algorithm to solve the lower model. When initializing the position and speed of the particles, the cluster division result is used as the position constraint of the particles; after reaching the maximum number of iterations, the information entropy method is used to determine the weight, and the entropy weight method is used to select the optimal solution; the solution process of the lower model is specifically as follows: S51, parameter setting: set the number of objective functions to two, representing two different optimization objectives; determine the minimum and maximum values ​​of the objective function, define the number of initial populations, the dimension of decision variables, and the maximum number of iterations; set the upper and lower bounds of the inertia coefficient, learning factor, speed, and position; S52, during the initialization process, the initial position X and velocity V of the particle are generated. The generation of position and velocity follows the upper and lower bounds X of the decision space. max and X min ; Combined with the randomly generated values, each particle in the population is subjected to energy storage state constraint processing, the individual and group historical optimal positions and fitness are initialized, and an external archive set is set to store the initial state of the non-inferior solution; S53. During the iteration process, the fitness value of each particle is calculated, the historical optimal position and fitness of the individual are updated to reflect the current optimal solution, the historical optimal position and fitness of the group are updated, and the new solution is added to the external archive set. The external archive set is screened by dominance relationship and crowding distance to retain the optimal solution, and the size of the archive set is controlled. The solutions in the archive set are sorted using the entropy weight method to determine the optimal solution; S54, the speed and position of the particle are updated in each iteration according to the current inertia coefficient, learning factor and the historical optimal position of the individual and group; when updating the position, ensure that the particle position is within the decision space to maintain an effective search space; S55, iterating according to the maximum number of iterations, updating the speed and position of the particle each time to gradually improve the quality of the solution, and calculating the final optimal solution by the entropy weight method after all iterations are completed; S56. When the algorithm ends, the entropy weight method is used to select the optimal solution from the external archive set, and finally the decision variables of the optimal solution and their corresponding fitness values ​​are output as the result of the algorithm.

10. A two-layer planning system for distribution network based on cluster division, used to implement the method according to any one of claims 1 to 9, characterized in that: include: An indicator construction module is used to construct the modularity index, active power balance index and reactive power balance index of the distribution network, and combine the modularity index, active power balance index and reactive power balance index of the distribution network to form a comprehensive cluster division index of the distribution network; The clustering module calculates the distribution network clustering based on the comprehensive distribution network clustering index and uses an adaptive genetic algorithm. The model building module builds a two-layer planning model for photovoltaic and energy storage in the distribution network based on the distribution network cluster division. The two-layer planning model for photovoltaic and energy storage in the distribution network includes an upper model and a lower model. The upper model is used to determine the photovoltaic and energy storage configuration, and the lower model further optimizes the operation and scheduling of the distribution network system. The model solving module solves the photovoltaic energy storage two-layer planning model of the distribution network to obtain the optimal two-layer planning scheme of the distribution network.

Citation Information

Patent Citations

  • Method of determining locating and sizing planning of distributed power supplies in a power distribution network

    CN108764552A

  • Distributed photovoltaic cluster division method and system based on BAS-IGA algorithm

    CN114221350A

  • Distributed power generation cluster division method based on improved K-means algorithm

    CN116191544A

  • Side distributed resource partition self-adaptive control method

    CN116599123A

Cited By

  • Active power distribution network cluster dynamic division method and system based on deep reinforcement learning

    CN119154412A

  • Photovoltaic and energy storage optimal configuration method and device based on snowflake net

    CN119401526A

  • A photovoltaic and energy storage optimal configuration method and device based on a snowflake network

    CN119401526B

  • Distributed photovoltaic cluster division method considering frequency supporting capability

    CN119675104A

  • A distributed photovoltaic cluster partitioning method considering frequency support capability

    CN119675104B