Distribution and micro-grid collaborative planning method considering interaction power randomness description

Abstract

The invention provides a distribution and micro-grid collaborative planning method considering the randomness description of interaction power, which comprises the following steps: firstly, considering the randomness description of the interaction power between distribution and micro-grids, extracting feature information from a long-time sequence through a fuzzy C-means clustering algorithm by utilizing historical data of the interaction power of each micro-grid and the distribution network, and calculating the distribution and micro-grid collaborative planning method; a typical scene is formed through interception, reduction and combination, so that the randomness depiction of the interaction power is carried out, and an interaction power scene basis is provided for subsequent power distribution network collaborative planning; secondly, performing distribution and micro-grid collaborative planning, establishing a distribution-micro-grid double-layer collaborative planning model, and planning the capacities of photovoltaic devices, energy storage devices and reactive devices connected to a distribution network and the access positions of micro-grids by taking the minimum annual comprehensive cost as a target in the upper-layer planning; the lower-layer planning takes the operation cost of the power distribution network and the minimum voltage offset as targets, adopts an approximate ideal sorting method based on an entropy weight method to process multiple targets, performs collaborative optimization on photovoltaic, energy storage and reactive access node positions and capacities of the power distribution network, and adopts a particle swarm algorithm of double-layer self-adaptive parameters to solve. The method has important theoretical value and engineering application significance for improving the economic, safe and stable operation of the power distribution network, improving the new energy consumption rate and promoting the construction of a novel power system.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of power system planning and optimization, and particularly relates to a micro-grid coordinated planning method considering randomness of interactive power. BACKGROUND

[0002] At present, the coordinated planning of distribution network and micro-grid mainly focuses on load management and energy scheduling of power demand, but ignores the complex power interaction and randomness between the micro-grid and the distribution network. With large-scale access of renewable energy and continuous development of micro-grid system, how to accurately depict the randomness of power interaction and carry out coordinated planning is still a challenge in the current technology, and the existing coordinated planning method cannot meet the demand of efficient and safe operation of future power system. SUMMARY

[0003] To solve the problems in the prior art, the application provides a micro-grid coordinated planning method considering randomness of interactive power and a system, which can comprehensively consider randomness of interactive power and micro-grid coordinated planning, establish a distribution network-micro-grid double-layer coordinated planning model, adapt to distribution network systems with different new energy penetration rates and multiple micro-grid planning access, support multiple power source planning, and more comprehensively and accurately plan distribution network systems with multiple micro-grid access. The implementation of the application can provide important technical support for improving active and reactive power planning of distribution network with multiple micro-grid planning access, and provide a scientific basis for new power system construction and development of distribution network and micro-grid.

[0004] The technical solution for achieving the object of the application is a micro-grid coordinated planning method considering randomness of interactive power, comprising the following steps.

[0005] Step 1: considering randomness of interactive power, using the fuzzy C-means clustering algorithm, extracting feature information from long time series by using historical data of interactive power of each micro-grid and the distribution network, and forming typical scenarios by intercepting, reducing and merging to depict randomness of interactive power;

[0006] Step 2: establishing a double-layer coordinated planning model of distribution network-micro-grid, planning the capacity of photovoltaic, energy storage and reactive device connected to the distribution network and the access position of each micro-grid in the upper layer planning with the minimum annual comprehensive cost as the target, and solving by using the particle swarm algorithm;

[0007] The lower layer planning has the minimum distribution network operation cost and voltage deviation as the target, and cooperatively optimizes the access node position and capacity of photovoltaic, energy storage and reactive device connected to the distribution network, and selects the best as the optimal solution from the saved external file set by using the entropy weight-based optimal solution distance method for solving the model of lower layer multi-objective planning;

[0008] Step 3: Based on the established two-layer collaborative planning model of distribution network-microgrid, the generated interactive power scenario is used to perform iterative solution using a particle swarm algorithm with two-layer adaptive parameters to obtain the planning results of distribution network-microgrid and output the optimal planning scheme.

[0009] Compared with the prior art, the significant advantages of this invention are: 1. This invention fully considers the impact of the randomness of interactive power on planning: Traditional distribution network planning is mostly based on deterministic interactive power, without fully considering the impact of the randomness of interactive power.

[0010] 2. This invention establishes a two-layer collaborative planning model for distribution microgrids, which takes into account the planning of microgrid access locations. By planning the active and reactive power sources of the distribution network and the location planning of the microgrids, it effectively improves the economic, safe and stable operation of the distribution network and the renewable energy consumption rate.

[0011] The present invention will now be described in further detail with reference to the accompanying drawings. Attached Figure Description

[0012] Figure 1 This is a flowchart of the collaborative planning of distribution microgrids, which takes into account the randomness of interactive power in this invention.

[0013] Figure 2 This is a system diagram of the power distribution network example in this invention.

[0014] Figure 3 This is a graph showing the power generation curve of the multi-microgrid interaction in this invention.

[0015] Figure 4 This invention provides a diagram showing the locations of photovoltaic, energy storage, and reactive power connections in the power distribution network, as well as the locations of microgrid connections. Detailed Implementation

[0016] A method and system for collaborative planning of distribution microgrids considering the stochasticity of interactive power is proposed. First, the method considers the stochasticity of interactive power between distribution microgrids. Using a fuzzy C-means clustering algorithm, historical data on the interactive power between each microgrid and the distribution network is used to extract feature information from long-term series data. Typical scenarios are formed through truncation, reduction, and merging, thus characterizing the stochasticity of interactive power and providing a foundation for subsequent distribution network collaborative planning. Second, collaborative planning of distribution microgrids is performed, establishing a two-layer collaborative planning model. The upper-layer planning aims to minimize the annual comprehensive cost, planning the capacity of photovoltaic, energy storage, and reactive power devices connected to the distribution network, as well as the connection location of each microgrid. The lower-layer planning aims to minimize the operating cost and voltage deviation of the distribution network. An entropy weight-based approximation ideal ranking method is used to process multiple objectives, collaboratively optimizing the connection node location and capacity of photovoltaic, energy storage, and reactive power devices in the distribution network. A two-layer adaptive parameter particle swarm optimization algorithm is used to solve the problem.

[0017] Cooperative planning methods for distribution microgrids that consider the stochasticity of interaction power, such as Figure 1 As shown, the specific steps include:

[0018] Step 1: Considering the stochastic characterization of interactive power, the fuzzy C-means clustering algorithm is used to extract feature information from the long-term series using historical data of interactive power between each microgrid and the distribution network. Typical scenarios are formed by truncation, reduction, and merging, thereby characterizing the stochasticity of interactive power.

[0019] In step 1, the characterization of interactive power during distribution network collaborative planning should comprehensively consider the differences in the characteristics of each microgrid. For example, residential loads in residential microgrids fluctuate greatly, with distinct morning and evening peaks, exhibiting a bimodal curve; commercial loads in commercial microgrids are concentrated during business hours, with significant peaks, exhibiting a unimodal curve; intermittent industrial microgrids only operate during the day with relatively small fluctuations, while production ceases at night, resulting in low load levels. Therefore, directly using historical load data for each microgrid as a single dataset lacks representativeness. Thus, a fuzzy C-means clustering algorithm is employed to cluster the interactive power between each microgrid and the distribution network, aiming to obtain different typical load curves and generate different typical operating scenarios. Based on this, collaborative planning of the distribution network is carried out.

[0020] Currently, the selection of typical scenarios mainly relies on clustering algorithms to simplify and compress long-term series scenarios. Among them, the fuzzy C-means algorithm is widely used due to its good performance. This algorithm determines the category of the sample by optimizing the objective function and calculating the membership degree of each sample point relative to all cluster centers, thus achieving automatic data classification. Unlike traditional hard clustering methods, the fuzzy C-means algorithm allows a data point to belong to multiple categories simultaneously, thus showing stronger adaptability when dealing with datasets with random characteristics.

[0021] The specific steps of the fuzzy C-means algorithm are as follows:

[0022] Step 1.1: Obtain historical sample data of the interaction power between each microgrid and the distribution network, and establish an interaction power sample data matrix.

[0023] Step 1.2: Initialize the cluster center and membership matrix, and update the cluster center and membership matrix according to the following formula.

[0024]

[0025]

[0026] In the formula: μ(t)p is the p-th cluster center at the t-th iteration; N X x is the total number of samples; qLet τ(t)pq be the q-th sample; τ(t)pq is the membership degree of the q-th sample to the p-th cluster at the t-th iteration; N P δ represents the number of clusters; ||·|| denotes the Euclidean norm; δ is the fuzzy coefficient.

[0027] Step 1.3: Determine if the iteration meets the convergence condition. If it converges, output the clustering result; otherwise, repeat step 1.2. The iteration convergence condition is as follows:

[0028]

[0029]

[0030] In the formula: f(t) FCM is the objective function of the t-th iteration; ε is the convergence error.

[0031] Step 1.4: Based on the obtained clustering results, calculate the comprehensive clustering effectiveness index D corresponding to different numbers of clusters. comp D comp The maximum number of clusters is the optimal number of clusters. (D) comp The calculation formula is as follows:

[0032]

[0033]

[0034]

[0035]

[0036] In the formula: μ(0) p is the p-th initial cluster center; D FP D P′ D L Three clustering validity functions

[78] ;D comp D is a comprehensive index of clustering effectiveness. comp The maximum value corresponding to N P This is the optimal number of clusters.

[0037] Step 2: Establish a two-layer collaborative planning model for distribution and microgrids. The upper-layer planning aims to minimize the annual comprehensive cost and plans the capacity of photovoltaic, energy storage, and reactive power devices connected to the distribution network, as well as the connection location of each microgrid. The particle swarm optimization algorithm is used to solve the problem.

[0038] The lower-level planning aims to minimize the operating cost and voltage deviation of the distribution network. It coordinates the optimization of the location and capacity of the access nodes for photovoltaic, energy storage and reactive power in the distribution network. For solving the model of the lower-level multi-objective planning, the best solution is selected from the saved external archives by using the superior solution distance method based on the entropy weight method.

[0039] To address the distribution network-microgrid collaborative planning problem involving the planning and site selection of multiple microgrids, a two-layer collaborative planning model is established. In the upper-layer planning, the planning objectives are the annual investment cost of photovoltaic (PV), energy storage, and reactive power in the distribution network, as well as the system operation and maintenance cost. The decision variables are the PV, energy storage, and reactive power capacity allocated to the distribution network, and the microgrid access location; the constraints include investment capacity constraints for each power source, and the particle swarm optimization algorithm is used for solving the problem.

[0040] In the lower-level planning, the planning objectives are the distribution network operating cost (including system network losses, renewable energy curtailment costs, generation costs, and the operating costs of planned energy storage) and the system voltage deviation. The decision variables are the location and capacity of photovoltaic (PV), energy storage, and reactive power connections within the distribution network. Constraints include distribution network power flow constraints, PV, energy storage, and reactive power planning capacity constraints, system safety operation constraints, and individual generator unit operation constraints. A multi-objective particle swarm optimization algorithm is used for solving the problem.

[0041] In a bilevel programming model, both the upper and lower levels have their own objective functions, decision variables, and constraints. However, the optimization processes of the upper and lower levels are interdependent, requiring information exchange between levels through parameter passing. The upper-level programming model passes decision variables as parameters to the lower-level programming model, which serve as initial conditions and constraints for the optimization of the lower-level decision variables. Based on this, the lower-level programming model performs optimization and solves the problem, then feeds back the time-series output of each planned power source as parameters to the constraints and objective function of the upper-level programming model. This information exchange completes the iterative optimization solution.

[0042] Specifically, for the distribution network-microgrid collaborative planning problem involving the planning and site selection of multiple microgrids, a two-layer collaborative planning model is established. The upper-layer planning model is as follows:

[0043] The upper-level objective function is:

[0044]

[0045] In the formula, C1 represents the annual comprehensive cost of the distribution network, including the equivalent annual investment cost C of photovoltaic, energy storage, and reactive power devices. I System operation and maintenance costs C O The specific formulas for calculating each cost are as follows:

[0046] 1) Equivalent annual investment cost

[0047]

[0048] In the formula, C I The equivalent annual investment cost of the distribution network; y PV y ESS y SVC, respectively, represent the service life of photovoltaic, energy storage, and reactive power devices; cI PV, cI ESS, and cI SVC represent the unit capacity investment cost of photovoltaic, energy storage, and reactive power devices, respectively; PPV nEESS n and QSVC n represent the planned installation capacity of photovoltaic, energy storage, and reactive power devices in the distribution network, respectively.

[0049] 2) Operation and maintenance costs

[0050]

[0051] In the formula, KO PV, KO ESS, and KO SVC are the unit cost coefficients for the planned operation and maintenance of photovoltaic, energy storage, and reactive power, respectively; PPV n,t, PESS c,n,t, PESS dis,n,t, and QSVC n,t are the photovoltaic and energy storage charging and discharging power and reactive power generated by the reactive power devices planned to be installed in the distribution network at time t, respectively; where QSVC n,t>0 indicates the generation of inductive reactive power, and QSVC n,t<0 indicates the generation of capacitive reactive power.

[0052] The constraints include installed capacity constraints for photovoltaic, energy storage, and reactive power in the distribution network. The specific constraints are as follows:

[0053] Maximum planned installed capacity constraints for photovoltaic, energy storage and reactive power devices in power distribution networks

[0054]

[0055] In the formula, PPV n,max, EESS n,max, and QSVC n,max represent the maximum installed capacity of photovoltaic, energy storage, and reactive power devices that are allowed to be planned and installed in the distribution network, respectively.

[0056] The lower-level planning model is as follows:

[0057] The lower-level model uses each node of the distribution network as the basic unit and is responsible for solving the location and capacity determination problem of photovoltaic, energy storage, and reactive power devices in the distribution network, based on the upper-level model. Based on the generated interactive power, and with the goal of minimizing distribution network operating costs and voltage deviation, it optimizes the access location and capacity of photovoltaic, energy storage, and reactive power devices.

[0058] The objective function is as follows:

[0059] The distribution network takes the minimum operating cost C1 and voltage deviation C2 as its objective function. C1 includes the system generation cost CADNg, the wind and solar curtailment cost CADNcut, and the network loss cost CADNloss.

[0060]

[0061]

[0062]

[0063]

[0064] In the formula, T is the operation and scheduling cycle, generally taken as T=24h; KADN g is the generator power generation cost coefficient; PADN g,t is the generator power generation at time t; i is the distribution network node; KADN pv is the curtailment penalty coefficient; Φ pv Φ wt Let P be the set of photovoltaic access nodes i; PV,i,t P represents the photovoltaic output power of node i at time t. pv,i,t These represent the actual power injected into node i by the photovoltaic system at time t; KADN loss is the network loss cost coefficient; I ij,t Let r be the amplitude of the line current (i, j) at time t; ij Here is the resistance value of line (i, j); Ω line It is the collection of all lines in the distribution network.

[0065] The voltage offset C2 is the sum of the absolute values ​​of the deviations between the per-unit voltage values ​​of each node and the standard voltage value of 1 throughout the entire operating period, and can be expressed as:

[0066]

[0067] In the formula, Ω node Let U be the set of all nodes in the distribution network; U* i,t is the per-unit voltage value of node i at time t.

[0068] The constraints are as follows:

[0069] The lower-level planning model includes constraints on the total capacity of photovoltaic, energy storage, and reactive power connected to each node of the distribution network, distribution network power flow constraints, system operation security constraints, and operating constraints of each generating unit. Specifically, it is represented as follows:

[0070] 1) Total capacity constraints of photovoltaic, energy storage and reactive power connected to each node of the distribution network

[0071] The photovoltaic, energy storage, and reactive power capacity connected to each node in the distribution network is constrained by the upper-level decision variables (the planned total capacity of photovoltaic, energy storage, and reactive power).

[0072]

[0073] In the formula, Φ PV,n Φ ESS,n Φ SVC,n , i, , and QSVC, are the sets of planned access nodes i for photovoltaic, energy storage and reactive power devices in the distribution network; PPV i, EESS i and QSVC i are the capacities of photovoltaic, energy storage and reactive power devices at planned access node i, respectively.

[0074] 2) Distribution network power flow constraints

[0075]

[0076] 3) System operation safety constraints

[0077]

[0078] In the formula, Umin and Umax are the minimum and maximum node voltages allowed for system operation, respectively; Iij,max is the maximum current allowed for line (i, j) operation.

[0079] 4) Operating constraints of each unit

[0080] Constraints generally consider the operational constraints of each power source within the distribution network. The operational constraints for photovoltaic and wind power are relatively simple, generally only requiring that their output power does not exceed their maximum allowable power, and will not be elaborated upon here. The operational constraints for energy storage generally consider power coupling constraints, maximum charge / discharge power constraints, remaining battery capacity constraints, and capacity balance constraints, which can be specifically expressed as follows:

[0081]

[0082] In the formula, E es,t Let u be the amount of energy stored at time t; es,c,t u es,dis,t Let E represent the charge / discharge state of the stored energy at time t, respectively, and be a 0-1 variable; es,0 The initial energy level before the start of the energy storage dispatch cycle; P c,max P dis,max These are the maximum charging and discharging power of the energy storage, respectively; E es,min E es,max These are the minimum and maximum allowable electrical capacities for energy storage, respectively.

[0083] Specifically, the lower-level model is a multi-objective problem. Transforming it into a single-objective problem through weighted optimization makes it difficult to obtain reasonable weight coefficients, and normalization issues must be considered for different dimensions. Therefore, an adaptive parameter particle swarm optimization algorithm is used to solve the multi-objective programming model. Regarding the selection of the optimal individual non-dominated solution and the optimal swarm non-dominated solution in multi-objective particle swarm optimization, C1 and C2 are used as two indicators for selecting the optimal individual non-dominated solution and the optimal swarm non-dominated solution in multi-objective particle swarm optimization, and the following selection strategy is adopted: Optimal individual non-dominated solution (p best Selection strategy:

[0084] If the updated individual Pareto dominates its p best Then the updated individual is used as the new p. best If neither can dominate the other, then one of them is randomly selected as p. bestOtherwise p best It remains unchanged.

[0085] Optimal population non-dominated solution (g best Selection Strategy

[0086] The selection is made using the crowding distance sorting method. The crowding degree of each particle (the Euclidean distance to the two surrounding particles) is calculated, and any one of the top 10% with the smallest distance is selected.

[0087] Global optimal solution selection strategy

[0088] The best solution in the saved external archive set is selected as the global optimum using the best solution distance method (Topsis) based on the entropy weight method (EWM). The optimal population non-dominated solutions under all iterations are used as evaluation samples, and the C1 and C2 values ​​of each evaluation sample are used as evaluation indicators. The specific principle is as follows:

[0089] Assume there are n samples to be evaluated and m evaluation indicators, forming the original indicator data matrix.

[0090]

[0091] 1) First, the indicators are positive-oriented. Indicators are categorized into benefit-type indicators, cost-type indicators, intermediate-type indicators, and range-type indicators. Therefore, all indicators need to be converted into positive-oriented indicators, i.e., all converted into benefit-type indicators. After positive-oriented conversion, standardization is performed to eliminate the influence of different indicator dimensions.

[0092]

[0093] 2) Assign weights to each evaluation indicator using the entropy weight method. The entropy weight method is one of the most commonly used objective weighting methods. This method uses the entropy value of the data, i.e., the amount of information, to calculate the weights. It is suitable for situations where there are fluctuations between data, and these fluctuations are considered as information. The entropy weight method draws on the definitions of chemical entropy and information entropy. By defining the entropy value of each indicator, it quantifies and synthesizes the information of each unit to be evaluated in the assessment, resulting in a relatively objective weight for each indicator. In data, the greater the dispersion, the greater the impact of that indicator on the overall evaluation. In information theory, the greater the amount of information, the lower the uncertainty, and the lower the entropy; conversely, the lower the amount of information, the greater the uncertainty, and the greater the entropy. Specifically, the probability matrix is ​​first calculated from the positively standardized matrix:

[0094]

[0095] Then calculate the information entropy:

[0096]

[0097] Calculate the information benefit value and normalize it:

[0098]

[0099] This yields the weights of each indicator. The matrix is ​​then rewritten as follows:

[0100]

[0101] 3) Calculate the score. When there are multiple indicators, define the distance between the i-th evaluation object and the maximum value as:

[0102]

[0103] The distance between the i-th evaluation object and the minimum value is:

[0104]

[0105] The score for the i-th evaluation object is:

[0106]

[0107] This yields the final score for each evaluated object, which is then used as the basis for selecting the globally optimal solution.

[0108] Step 3: Based on the established two-layer collaborative planning model of distribution network-microgrid, the generated interactive power scenario is used to perform iterative solution using a particle swarm algorithm with two-layer adaptive parameters to obtain the planning results of distribution network-microgrid and output the optimal planning scheme.

[0109] The two-layer collaborative planning model for distribution microgrids includes an upper-layer planning model and a lower-layer planning model;

[0110] In the upper-level planning model, the planning objective is the annual comprehensive cost of the distribution network; the decision variables are the total capacity of energy storage and reactive power allocated to the distribution network and the energy storage capacity of the microgrid; the constraints include the investment capacity constraints of each power source. In the lower-level planning model, the planning objectives are the operating costs of the distribution network and the operating costs of the microgrid; the decision variables are the access location and capacity of each energy storage and reactive power source in the distribution network, and the interaction power between the microgrid and the distribution network; the constraints include the power flow constraints of the distribution network, the planned capacity constraints of energy storage and reactive power, the operating constraints of the microgrid, the system safety operation constraints, and the operating constraints of each generating unit.

[0111] Specifically, by solving the upper planning layer, the optimal planning schemes for the location and capacity determination of photovoltaic, energy storage, and reactive power, as well as the connection locations of each microgrid, can be obtained. Solving the lower operation layer yields the output power of photovoltaic, energy storage, and reactive power for each time period, and outputs the optimal planning scheme.

[0112] The solution process for the upper planning layer is as follows:

[0113] 1) Initialize the upper-level population.

[0114] 2) Based on the time-series data of the load, the lower-level optimization is initiated to calculate the scheduling scheme based on the upper-level planning. This scheme is the optimal solution obtained by the lower level from the non-dominated solutions in the external archive through the superior-inferior solution distance method based on the entropy weight method. Thus, the operation and maintenance cost of the upper level can be obtained, and its comprehensive cost can be obtained.

[0115] 3) Calculate the fitness of the upper-layer population based on the comprehensive cost of the upper layer obtained in step 2.

[0116] 4) Update the upper population.

[0117] 5) Repeat steps 2) to 4) until the termination condition is met, i.e., the difference between the comprehensive cost of the upper layer and the previous and next iterations is less than 100 yuan or the maximum number of iterations is reached. Then output the optimal solution of the upper layer.

[0118] The solution process for the lower-level runtime layer is as follows:

[0119] 1) Input the planning results obtained from the upper layer, and initialize the lower layer population based on the upper layer input.

[0120] 2) Calculate the power flow according to the time sequence, and calculate the lower-level operating cost and voltage offset based on the power flow results to obtain the fitness of the population.

[0121] 3) Update the lower-level population and its external files using the multi-objective particle swarm optimization algorithm.

[0122] 4) Repeat steps 2) and 3) until the termination condition is met, that is, the difference between the operating cost and voltage offset of the best non-dominant solution in the external archive obtained by the superior-inferior solution distance method is less than 100 yuan and 0.1 respectively after 8 consecutive iterations, or the maximum number of iterations is reached. Then output the external archive, that is, the lower-level Pareto optimal solution set.

[0123] 5) Finally, the optimal solution in the lower-level Pareto optimal solution set is obtained by using the Topsis method based on the entropy weight method.

[0124] The distribution-microgrid collaborative planning module establishes a two-layer collaborative planning model for the distribution network and microgrid based on the stochastic characterization of interactive power, and performs iterative solutions to obtain the planning results for the distribution network and microgrid.

[0125] To verify the effectiveness of the proposed distribution microgrid collaborative planning method that considers the stochastic characterization of interactive power, the second embodiment of the present invention employs the following... Figure 2 The following is an analysis of the power distribution network system configuration example. The three microgrids are configured as follows: one residential microgrid with a load capacity of 1000KW, one intermittent industrial microgrid with a load capacity of 1500KW, and one commercial microgrid with a load capacity of 1200KW.

[0126] The improved distribution network system includes 2 photovoltaic nodes, 3 wind power nodes, 2 energy storage nodes, 1 reactive power compensation device node, and 3 nodes to be connected to the microgrid.

[0127] Based on the constituent elements within each of the three microgrids and using the proposed method for characterizing the stochasticity of interactive power, the optimal number of scenarios is four, with probabilities of [0.334, 0.334, 0.139, 0.193]. The interactive power curve for scenario 1 is shown below. Figure 3 As shown.

[0128] Based on the proposed site selection method for multi-microgrid planning and access to the distribution network, the particle swarm optimization algorithm is used for optimization. Considering the existence of a set of nodes to be planned at the locations of each microgrid when it accesses the distribution network, the candidate nodes for residential microgrids are set as [1 2 3 4 18 19 20 21], the candidate nodes for intermittent production microgrids are set as [5 6 22 23 24 25 26 27], and the candidate nodes for commercial microgrids are set as [10 11 12 13 14 15 16 17].

[0129] The final planning and site selection results are as follows: Figure 4 As shown. The planning result is:

[0130] The photovoltaic installation location is 5, with 5 units installed; the energy storage installation location is 28, with 8 units installed; and the reactive power installation location is 14, with 1 unit installed.

[0131] Microgrid connection location planning: Microgrid 1 is connected at node 6, microgrid 2 is connected at node 3, and microgrid 3 is connected at node 15.

[0132] Table 1 shows a comparison of the operation of the distribution network before and after the collaborative planning.

[0133] Table 1 Comparison of Operation Before and After the Coordinated Planning of Distribution Microgrids

[0134] Planning scheme Equivalent annual investment cost (ten thousand yuan) Investment cost Total operation cost Generator power generation cost Network loss cost Wind and light curtailment cost Voltage deviation Unplanned 1394.15 0 38196 25532 10281 2383 123.52 Planned 1091.03 46.0564 28631 20856 7228 547 78.67

[0135] It can be seen that the operating cost of the distribution network has been significantly reduced after the planning, the power generation cost and network loss cost have decreased significantly, the renewable energy absorption rate has been improved, voltage fluctuations have been greatly reduced, and the economic security and stable operation of the distribution network system have been improved.

[0136] This method is applicable to distribution network planning scenarios where multiple microgrids are planned for future integration. It has significant theoretical value and engineering application significance for improving the economic, safe and stable operation of distribution networks, increasing the absorption rate of new energy sources, and promoting the construction of new power systems.

[0137] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A microgrid collaborative planning method considering interaction power randomness characterization, characterized in that, Comprise: Step 1, considering the randomness of the interaction power, the fuzzy C-means clustering algorithm is used to extract feature information from long time series by using the historical data of the interaction power of each micro-grid and the distribution network, and the typical scenarios are formed by intercepting, reducing and merging, so as to describe the randomness of the interaction power; Step 2, a double-layer collaborative planning model of distribution-micro-grid is established, the upper layer planning takes the minimum annual comprehensive cost as the target, plans the capacity of photovoltaic, energy storage and reactive power device connected to the distribution network and the access position of each micro-grid, and solves by using particle swarm algorithm; The lower layer planning takes the minimum distribution network operation cost and voltage deviation as the target, and cooperatively optimizes the access position and capacity of the distribution network photovoltaic, energy storage and reactive power, and solves the lower layer multi-objective planning model by using the entropy weight-based optimal solution distance method to select the best as the optimal solution from the saved external file set; Step 3, according to the established double-layer collaborative planning model of distribution-micro-grid, the generated interaction power scenario is used, and the double-layer adaptive parameter particle swarm algorithm is used for iterative solution to obtain the distribution network-micro-grid planning result and output the optimal planning scheme.

2. The microgrid co-planning method considering interaction power randomness characterization according to claim 1, characterized in that, In step 1, the specific steps of forming the typical scenario by intercepting, reducing and merging through the fuzzy C-means clustering algorithm and using the historical data of the interaction power of each micro-grid and the distribution network from long time series are as follows: Step 1.1: Obtain the historical sample data of the interaction power of each micro-grid and the distribution network, and establish the interaction power sample data matrix; Step 1.2: Initialize the clustering center and membership matrix, and update the clustering center and membership according to the following formula; ； ； where μ(t) p is the pth cluster center at the tth iteration; N X is the total number of samples; x q is the qth sample; τ(t) pq is the membership of the qth sample to the pth cluster at the tth iteration; N P is the number of clusters; ||·|| denotes the Euclidean norm; δ is the fuzziness coefficient; Step 1.3: Determine whether the iteration meets the convergence condition, if the convergence is met, output the clustering result, otherwise repeat step 1.2; Step 1.4: According to the obtained clustering results, the integrated clustering validity index D corresponding to different cluster numbers is calculated comp , D comp The maximum cluster number is the optimal cluster number.

3. The microgrid co-planning method considering interaction power randomness characterization of claim 2, wherein, The iteration convergence condition in step 1.3 is as follows: ； ； In the formula, Jtis the objective function for the tth iteration; ε is the convergence error.

4. The microgrid co-planning method considering interaction power randomness characterization of claim 2, wherein, The integrated clustering validity index D comp The calculation formula is as follows: ； ； ； ； In the formula: μ(0) p is the pth initial clustering center; D FP , D P′ , D L are three cluster validity functions; D comp D is the validity index of clustering comp The maximum value of N P That is, the optimal number of clusters.

5. The microgrid co-planning method considering interaction power randomness characterization of claim 1, wherein, In the upper layer planning of the double-layer collaborative planning model of distribution-micro-grid, the planning target is the annual investment cost and system operation and maintenance cost of the distribution network photovoltaic, energy storage and reactive power, and the decision variable is the photovoltaic, energy storage and reactive power capacity configured to the distribution network and the micro-grid access position; The constraint conditions include the investment capacity constraints of each power supply, and the particle swarm algorithm is used for solution; In the lower layer planning, the planning target is the distribution network operation cost and system voltage deviation, the decision variable is the access position and access capacity of the photovoltaic, energy storage and reactive power in the distribution network, and the constraint conditions include the distribution network power flow constraint, the photovoltaic, energy storage and reactive power planning capacity constraint, the system safe operation constraint and the unit operation constraint. A multi-objective particle swarm algorithm is used for solution; The decision variable of the upper layer planning is transmitted to the lower layer planning as a parameter and one of the initial conditions and constraints of the lower layer decision variable optimization; The lower layer planning model is optimized and solved based on this, and the time sequence output of each planning power supply is obtained as a parameter fed back to the constraint condition and objective function of the upper layer planning, and information interaction is carried out to complete the iterative optimization and solution.

6. The microgrid co-planning method considering interaction power randomness characterization of claim 1, wherein, The upper layer objective function of the upper layer planning is: ； In the formula, C1 is the annual comprehensive cost of the power distribution network, including the equivalent annual investment cost of photovoltaic, energy storage and reactive power devices , system operation and maintenance cost C O ; the specific calculation formula of each cost is as follows: Equivalent annual investment cost: ； where C I is the equivalent annual investment cost of the distribution network; y PV , y ESS , y SVC are the service life of photovoltaic, energy storage and reactive power devices, respectively; cIPV, cIESS, cISVC are the unit capacity investment cost of photovoltaic, energy storage and reactive power devices, respectively; PPVn, EESSn, QSVCn are the installed photovoltaic, energy storage and reactive power capacity of the distribution network planning, respectively. Operation and maintenance cost: ； Where, KO PV, KO ESS, KO SVC are the unit cost coefficients of the planned photovoltaic, energy storage and reactive power operation and maintenance, respectively; PPV n,t, PESS c,n,t, PESS dis,n,t, QSVC n,t are the installed photovoltaic, energy storage charging and discharging power, and the reactive power generated by the reactive power device of the distribution network at time t; wherein QSVC n,t > 0 indicates the generation of inductive reactive power, and QSVC n,t < 0 indicates the generation of capacitive reactive power; The constraint conditions include the installed capacity constraints of photovoltaic, energy storage and reactive power of the distribution network, and are specifically represented as follows: ； Where, PPV n,max, EESS n,max, QSVC n,max are the maximum installed capacities of photovoltaic, energy storage and reactive power devices allowed to be installed in the distribution network.

7. The microgrid co-planning method considering interaction power randomness characterization of claim 6, wherein, The lower target function of the lower layer planning is as follows: The distribution network takes the minimum operation cost C1 and voltage deviation C2 as the target function, and C1 includes the system generation cost CADN g, wind and light curtailment cost CADN cut, and network loss cost CADN loss: ； ； ； ； Where, T is the operation scheduling period, and KADN g is the generation cost coefficient of the generator; PADN g,t is the power generation of the generator at time t; i is the distribution network node; KADN pv is the light abandonment penalty coefficient; Φ pv , Φ wt are the set of photovoltaic access nodes i respectively; P PV,i,t is the photovoltaic output power of access node i at time t; P pv,i,t are the actual power of photovoltaic injected into node i at time t; KADN loss is the network loss cost coefficient; I ij,t is the amplitude of the current of line (i, j) at time t; A ij is the resistance value of line (i, j); Ω line is the set of all lines of the distribution network; The voltage deviation C2 is the sum of the absolute values of the deviations of the voltage per unit of each node from the standard voltage value 1 in the entire operation period, and is represented as follows: ； In the formula, Ω node is a set of all nodes of the power distribution network; U* i,t is the voltage per unit of node i at time t; The constraint conditions specifically include: The photovoltaic, energy storage and reactive power capacities connected to each node in the distribution network are constrained by the upper decision variables: ； where Φ PV,n , Φ ESS,n , and Φ SVC,n are the sets of planning access nodes i for photovoltaic, energy storage, and reactive devices in the distribution network, respectively; PPV i, EESS i, and QSVC i are the capacities of photovoltaic, energy storage, and reactive devices at planning access node i, respectively. The power flow constraint of the distribution network: ； The system operation safety constraint: ； Where, U min and U max are the minimum and maximum node voltages allowed in system operation, respectively; and I ij,max is the maximum current allowed in the operation of line (i, j); The operation constraints of each unit: ； In the formula, E es,t is the energy storage amount at time t; u es,c,t , u es,dis,t are the charge and discharge states of the energy storage at time t, which are 0-1 variables; E es,0 is the initial energy storage amount before the start of the energy storage scheduling period; P c,max , P dis,max are the maximum charge and discharge powers of the energy storage; E es,min , E es,max are the minimum and maximum capacitances allowed by the energy storage, respectively.

8. The microgrid co-planning method considering interaction power randomness characterization of claim 7, wherein, According to the established bi-level collaborative planning model of the distribution network-microgrid, the generated interactive power scenarios are used to solve the planning results of the distribution network-microgrid by using the bi-level adaptive parameter particle swarm algorithm, and the specific method for outputting the optimal planning scheme is as follows: The optimal planning scheme of photovoltaic, energy storage and reactive power site selection and capacity determination and the connection position of each microgrid is obtained through the solution of the upper layer planning, and the output power of photovoltaic, energy storage and reactive power in each period is obtained through the solution of the lower layer planning, and the optimal planning scheme is output; The solution process of the upper layer planning is as follows: 1) Initialize the upper layer population; 2) According to the time sequence data of the load, start the lower layer optimization, calculate the scheduling scheme based on the upper layer planning, and obtain the optimal solution from the external archive non-inferior solution by using the entropy weight method-based optimal and inferior solution distance method, and then the operation and maintenance cost of the upper layer planning is obtained, and the comprehensive cost of the upper layer planning is obtained; 3) Calculate the fitness of the upper layer population according to the comprehensive cost of the upper layer planning; 4) Update the upper layer population; 5) Repeat 2)~4) until the termination condition is met, and then output the optimal solution of the upper layer; The solution process of the lower layer planning is as follows: 1) Input the planning results obtained by the upper layer, and initialize the lower layer population according to the upper layer input; 2) Calculate the power flow according to the time sequence, calculate the operation cost and voltage deviation according to the power flow results, and obtain the fitness of the population; 3) The method of updating population by multi-objective particle swarm optimization algorithm is used to update the lower layer population and its external archive; 4) Repeat 2), 3) until the termination condition is met, and output the external archive, i.e. the lower layer Pareto optimal solution set; 5) Finally, the optimal solution in the lower layer Pareto optimal solution set is obtained by Topsis method based on entropy weight method.

Images