Power distribution network reconstruction method under collaborative optimization of energy storage-containing source load uncertainty and demand response

Abstract

The invention discloses a power distribution network reconstruction method under collaborative optimization of energy storage-containing source load uncertainty and demand response. The method comprises the following steps: constructing topological structure information of a power distribution network; time sequence fluctuation characteristics and influences on power grid operation are considered, and a distributed power supply output and load fluctuation probability model is established; carrying out uncertainty processing on the distributed power supply and the load; constructing a power distribution network reconstruction optimization model considering demand response and source load characteristics; and introducing a second-order cone-convex relaxation technology and a linearization method into the constructed power distribution network reconstruction optimization model, and converting the original model into a form with a convex feasible region for solving. The method aims at overcoming the defects of a traditional method in the aspect of considering distributed power supply access and uncertainty of the distributed power supply access. By optimizing the configuration and scheduling of the energy storage resources, the flexibility and stability of the power distribution network can be effectively improved, so that the absorption capability of the power grid to renewable energy sources is improved; and the adaptability of the system to time sequence fluctuation is enhanced. And the defect of considering the structural change and load difference of the power distribution network in the traditional method is overcome.

Description

Technical Field

[0001] The present invention relates to the technical field of distribution network reconstruction optimization, and in particular to a distribution network reconstruction method under the coordinated optimization of source-load uncertainty and demand response including energy storage. Background Art

[0002] In the process of achieving the "carbon peak and carbon neutrality" goals of the new energy system, a distribution network with a stronger carrying capacity will help build a new power system that is clean, low-carbon, economical, efficient, and coordinated in supply and demand. Distribution network reconstruction is a strategy to optimize power grid operation, which aims to optimize by adjusting the network topology, operation mode, and control strategy. With the connection of various distributed power sources (DG) and distributed energy storage devices (EES), it is of great significance to consider their uncertainty and random fluctuation characteristics for the optimization of distribution network reconstruction.

[0003] At present, the research on the reconstruction of DG access distribution network mainly focuses on the solution algorithm of the reconstruction model and the processing of DG timing fluctuations. The reconstruction of the distribution network involves the combination and optimization of multiple interconnection switches. At present, the distribution network reconstruction algorithm is mainly divided into intelligent algorithms and heuristic algorithms. Among them, the intelligent algorithms mainly include: genetic algorithms, particle swarm optimization and ant colony algorithms, which have excellent performance in dealing with non-convex optimization problems. Heuristic algorithms can be used to optimize a large number of variables and complex constraints in the reconstruction of the distribution network. They can plan and allocate the capacity of each device in the distribution network. However, they may fall into the local optimal solution, and thus cannot guarantee the global optimal solution.

[0004] In the context of DG access to the distribution network system, the commonly used methods for source-load uncertainty problems include: robust optimization, interval number optimization, multi-scenario stochastic programming, and fuzzy programming. When studying the impact and optimization of DG access on new energy systems, most of them did not consider the scenarios of different types of DG working together and output uncertainty, but mainly used the traditional constant power model. Few studies have considered the time series fluctuation characteristics of DG and the uncertain factors under the load-side demand response, resulting in limitations in practical applications.

[0005] Existing technologies involving distributed generation access to distribution network reconstruction include:

[0006] The Chinese patent "Distribution network reconstruction method containing distributed power sources based on improved neural network algorithm" (application number: 202310782165.8) first establishes a mathematical model for distribution network reconstruction with the goal of minimizing active power loss and maximizing voltage stability index for the distribution network connected to distributed power sources, then improves the standard neural network algorithm, and finally reconstructs the distribution network containing distributed power sources based on the improved neural network algorithm.

[0007] The Chinese patent "A dynamic reconstruction strategy for a distribution network containing distributed power sources" (application number: 202110006434.2) first establishes a renewable distributed power model to be optimized, and then statically reconstructs the distribution network for multiple unit time periods within a day to form an optimal switch set, evaluates the effectiveness of the reconstruction of each time period, and determines the optimal time period division scheme based on this. Finally, the bacterial foraging algorithm is used to select the grid with the greatest reconstruction effectiveness as the optimal grid after the time period is merged.

[0008] The Chinese patent "Distribution network reconstruction method taking into account the time-varying characteristics of load and distributed power sources" (application number: 202410507932.9) first predicts the load based on historical data and machine learning algorithms, and models the distributed power sources; combines load prediction and distributed power response models, optimizes the topology of the distribution network, adjusts grid parameters and line configurations to improve system stability and efficiency; finally, designs online scheduling and control strategies to dynamically adjust system operating parameters according to the real-time load and distributed power status to achieve intelligent management of the distribution network.

[0009] The above patents in the prior art have made certain contributions in the field of distribution network reconstruction with distributed power generation access, but they have not been fully explored and deeply analyzed in terms of the impact of the timing fluctuation characteristics after the distributed power generation access to the distribution network and the power grid's ability to absorb clean energy. This may limit the comprehensive understanding of the dynamic characteristics of the distribution network operation and how to optimize the power grid to better adapt to the integration of renewable energy. Therefore, further research on these aspects will help improve the overall effect of distribution network reconstruction and the effective use of clean energy. Summary of the invention

[0010] In order to solve the above technical problems, the present invention proposes a distribution network reconstruction method under the coordinated optimization of source-load uncertainty and demand response with energy storage. The method aims to solve the shortcomings of traditional methods in considering the access of distributed power sources and their uncertainties, and better cope with the complex interactive relationship between sources and loads in the distribution network. By optimizing the configuration and scheduling of energy storage resources and making full use of the demand response mechanism, the flexibility and stability of the distribution network can be effectively improved, thereby improving the grid's ability to absorb renewable energy; and enhancing the system's adaptability to time series fluctuations. It makes up for the shortcomings of traditional methods in considering changes in the distribution network structure and load differences.

[0011] The technical solution adopted by the present invention is:

[0012] A distribution network reconstruction method under the coordinated optimization of source-load uncertainty and demand response with energy storage includes the following steps:

[0013] Step 1: Construct the initial distribution network topology information;

[0014] Step 2: Considering the characteristics of time series fluctuations and their impact on grid operation, a probability model of distributed generation output and load fluctuations is established;

[0015] Step 3: Deal with the uncertainty of distributed power sources and loads;

[0016] Step 4: Construct a distribution network reconstruction optimization model that takes into account demand response and source-load characteristics;

[0017] Step 5: Introduce the second-order cone convex relaxation technology and linearization method to the distribution network reconstruction optimization model constructed in step 4, and transform the original model into a form with a convex feasible domain for solution.

[0018] In step 1, in the distribution network topology information, the set formed by the nodes is called the node set E, and the set formed by the edges is called the branch set L. Therefore, the power grid structure can be described by an abstract graph G and an adjacency matrix A, that is:

[0019] G=(E,L)

[0020]

[0021] Where: The value of the elements of matrix A is 0 or 1. When node i and node j are connected by a line, it takes 1, otherwise it takes 0. ij Represents the connection relationship between nodes i and j in the adjacency matrix A, that is, it is used to indicate whether the node connection relationship in the graph is connected and its weight; (α ij ) n*n represents the connection relationship between nodes ij in the n-order adjacency matrix; n represents the total number of nodes. In step 2, a distributed power output and load fluctuation probability model is constructed, including:

[0022] (1) Wind power generation probability model:

[0023]

[0024] Where: f(P WT) represents the probability density function of wind power output; P WT Indicates the output power of the wind turbine; represents an exponential function with e as the base; v represents the actual wind speed in the area; k w represents the shape parameter, reflecting the distribution characteristics of wind speed in the area; k c is a scale parameter reflecting the average wind speed in the area.

[0025] k w , k c It can be calculated by the following formula using actual wind speed data:

[0026]

[0027] Where: δ, μ are the standard deviation and mean value of wind speed respectively; Γ is the gamma function.

[0028] (2) Photovoltaic power generation probability model:

[0029]

[0030] Where: f(P PV) represents the probability density function of photovoltaic output power; α, β are the shape parameters of Beta distribution; P PV is the actual light intensity; P PV,max is the maximum light intensity; Γ(α+β) represents the gamma function of the sum of shape parameters, which is used to describe the Beta distribution characteristics; Γ(α) represents the value of the gamma function at parameter α; Γ(β) represents the value of the gamma function at parameter β.

[0031] (3) Load probability model:

[0032]

[0033] Where f(P load ) represents the probability density function of active power; f(Q load ) represents the probability density function of reactive power; P load , Q load are active load and reactive load respectively; μ P , σ P are the mean and standard deviation of load active power respectively; μ Q , σ Q are the mean and standard deviation of the load reactive power respectively.

[0034] The step 3 comprises the following steps:

[0035] S3.1: Use Latin hypercube to simulate the random variables of solar irradiance, wind speed and load, and simulate and solve the random variables of wind and solar load;

[0036] (1) Sampling process:

[0037] Assume X k is any one of the k random variables, then X k The corresponding cumulative probability distribution function is: Y k =F k (X k );

[0038] Assuming the sampling scale is N, divide the vertical axis of the cumulative probability density curve into N equally spaced non-overlapping intervals, each interval length is 1 / N, and use the median Latin hypercube sampling method to obtain the nth sampling value y kn =(n-0.5) / N, its sampling value x kn for: n=1,2,…,N

[0039] In the formula, is the inverse probability distribution function.

[0040] According to the above steps, we can get the sample matrix X KN .

[0041] (2) Sorting process:

[0042] L KN is the sorting matrix of the sampling matrix elements, and the correlation coefficients between its rows are represented by the matrix ρ. It is easy to prove that the matrix ρ is a positive definite matrix, so the matrix ρ can be decomposed into:

[0043] ρ L =DD T

[0044] In the formula, ρ L The correlation coefficient matrix is ​​represented as matrix L; D is a non-singular lower triangular matrix.

[0045] In order to accurately reflect the row correlation of the sampling matrix, the root mean square ρ of the correlation coefficient matrix ρ is introduced msr , which is defined as:

[0046]

[0047] In the formula, ρ kj Represents the off-diagonal elements in the correlation matrix.

[0048] Since the matrix D is non-singular, there must be an inverse matrix D-- , combined with the initial sorting matrix L KN , we can get a matrix G with smaller row correlation:

[0049] G=D -1 L KN

[0050] Repeat the above steps until the row correlation is reduced to a preset value.

[0051] S3.2: Using the improved K-means clustering algorithm, a large number of wind power and photovoltaic output scenarios and their load fluctuation Latin hypercube simulation data are effectively clustered to ensure that the number of scenarios can be effectively reduced based on the wind power, photovoltaic and load distribution characteristics. Specifically, it includes:

[0052] 1. Determine the optimal number of clusters based on clustering effectiveness indicators:

[0053] (1) Distance between samples:

[0054] The distance between samples is usually calculated using the Euclidean distance. For two m-dimensional samples x i and x j , whose Euclidean distance is defined as:

[0055]

[0056] In the above formula: x ik 、x jk They represent the coordinate vectors of the i-th and j-th samples respectively; k means that the sample data is divided into k categories; m represents the feature dimension of the sample point.

[0057] (2) Class center point:

[0058] Class K i The center point c i is the mean of all samples in the class:

[0059]

[0060] In the above formula: Ni is K-like i The number of samples in; x represents class K i The feature vector of the i-th sample point.

[0061] (3) Intra-class dispersion:

[0062] The intra-class dispersion is the class radius r i It can be defined as the average distance from all samples in a class to the center point of its class:

[0063]

[0064] In the above formula: d(x,ci ) represents the sample x to the class center point c i distance.

[0065] (4) Distance between classes:

[0066] That is, the distance between the center points of two classes:

[0067]

[0068] In the above formula: c i ,c j Respectively represent K i ,K j The inter-class distance between ik 、c jk Represents class K i and class K j The center point of .

[0069] (5) DBI indicator:

[0070] That is, a measure to evaluate the quality of clustering, which is calculated based on the intra-class dispersion and the inter-class distance. For each class, the DBI indicator is defined as:

[0071]

[0072] In the above formula: r i ,r j K i ,K j The intra-class dispersion of d(c i ,c j ) is class K i and class K j The distance between the center points.

[0073] The smaller the DBI index is, the better the sample clustering effect is. The optimal number of clusters for K-means clustering is determined based on the DBI index, that is, the K value that minimizes the DBI index is found:

[0074]

[0075] In the above formula: K best Indicates the number of clusters that determine the minimum DBI index; DBI(K) represents the DBI index function calculated when the number of clusters K is given; Indicates the K value corresponding to the minimum value of the calculated DBI indicator.

[0076] 2. Use density clustering optimization to select cluster centers:

[0077] First, calculate the Euclidean distance between data points in the sample, that is:

[0078]

[0079] Where: x ik and x jk Respectively represent the values ​​of data points i and j at the kth eigenvalue.

[0080] Set the threshold distance d t , and calculate the local density ρ within the distance threshold i , the calculation formula is:

[0081]

[0082] Where: d ij is the Euclidean distance between data points i and j.

[0083] In step 4, the distribution network reconstruction optimization model includes:

[0084] 4.1. Objective function:

[0085] The model comprehensively considers the minimization of the comprehensive operating cost of the distribution network, namely:

[0086]

[0087] Where: C is the total cost in the distribution network reconstruction model; T is the duration of distribution network reconstruction; f 1,t is the network loss cost at time t; f 2,t is the cost of wind and solar power abandonment at time t; f 3,t is the penalty cost of the segment switch operation at time t.

[0088] 4.2 Constraints:

[0089] ①.Node power balance constraints:

[0090] The distribution network satisfies the active / reactive power balance constraint at node i at time t:

[0091]

[0092] Where: are the active and reactive power of node i at time t; V i,t ,V j,t are the voltage of node i at time t; B ij ,G ij are the susceptance and conductance between branches ij respectively; θ ij,t is the power angle difference of branch ij at time t; P ij,t , Q ij,t They represent the active power and reactive power flowing from node i to node j at time t respectively; α ij Indicates the switch status of line ij after reconstruction; It indicates that the power output and input of any intermediate system node maintain power balance; N represents the number of system branches.

[0093] ②.Safe operation constraints:

[0094]

[0095] Where: P ij ,Q ij are the active and reactive power transmitted between nodes i and j respectively; X ij ,R ij are the reactance and resistance between branches ij respectively; V i,min ,V i,max are the lower and upper limits of the voltage amplitude at node i respectively; I ij,max is the maximum current amplitude of line ij at time t.

[0096] ③.DG power constraints

[0097]

[0098] Where: P DG,max is the upper limit of the DG capacity connected to the i-node; is the DG power factor angle at node i; P i,DG represents the active power of DG at node i.

[0099] ④. Distribution network radiation and connectivity constraints:

[0100] During the distribution network reconstruction process, its topology should be guaranteed to be in a radial state, without ring networks and islands. Its constraints can be expressed as:

[0101]

[0102] Where: n is the number of system branches; β ij is the subordinate relationship between node i and node j; it is 1 when node i is the parent node of node j, otherwise it is 0; N i is the number of nodes in the network; β jt represents the state attribute of node i at time t; β 0j , β 1j They respectively represent the specific states of node j, namely, node i is the parent node of node j and node i is not the parent node of node j.

[0103] ⑤. Energy storage constraints:

[0104] Energy storage charging and discharging constraints can be defined as:

[0105]

[0106] Where: They represent the charging and discharging status of the energy storage at node i at time t respectively; They represent the charging and discharging power of the energy storage at node i at time t respectively; is the maximum charging and discharging power of the energy storage at node i at time t.

[0107] The external energy transfer characteristics of energy storage depend on the charge and discharge power and the state of charge (SOC). SOC is defined as:

[0108]

[0109] Where: is the SOC of the energy storage at node i at time t; η is the self-discharge rate of the energy storage; η charge ,η discharge are the charge and discharge conversion efficiency of energy storage respectively; △T is the scheduling time interval. represents the energy storage capacity of node i from time t to the next time t; They represent the minimum and maximum capacity limits of energy storage at node i respectively.

[0110] ⑥. Demand response constraints

[0111]

[0112] Where: i,t is the price elasticity coefficient of node i at time t; P i,t , are the loads of node i before and after DR at time t; ρ i,t , are the electricity prices before and after DR of node i at time t; They are the upper and lower limits of the electricity price after DR is implemented at node i at time t. peak represents the peak electricity price; ρ valley Indicates the valley electricity price; T peak Indicates the peak load moment; T valley valley load moment; Indicates any node in the value system.

[0113] In step 5, the original distribution network reconstruction optimization model is converted into a form with a convex feasible domain for solution, as follows:

[0114] Introduce intermediate variables to optimize the coupling relationship between various physical quantities:

[0115]

[0116] Where: I ij,t represents the effective value of the current flowing through branch ij at time t; V i,t represents the voltage of node i at time t; They respectively represent the quadratic relationship of the current at the node ij at time t after the introduction of the intermediate variable; They represent the quadratic relationship of the voltage at node ij at time t after the introduction of the intermediate variable; α ij,t , β ij,t They respectively represent the trigonometric function relationship of the voltage at the node ij at time t after the introduction of the intermediate variable; They represent the quadratic relationship of the voltage at node ij at time t after the introduction of the intermediate variable. ij,t Represents the common angle difference of branch ij at time t.

[0117] The present invention introduces two virtual node voltages are the voltages at the starting point i and the end point j of branch λ at time t. When branch λ is running, a λ =1, these two virtual voltages directly reflect the voltages of the actual nodes i and j, that is, When branch λ is disconnected, we have Its connection to the node voltage can be limited by the following equation:

[0118]

[0119] Where: α λ Represents the branch λ in the system; V imax represents the maximum voltage of node i at time t; V jmax Represents the maximum voltage of node j at time t.

[0120] The nonlinear objective function and nonconvex constraints in the original model are relaxed and transformed as follows:

[0121]

[0122] Where: C loss Represents the network loss cost coefficient; R ij represents the equivalent resistance of branch ij; △t represents each moment; f 2,t represents the cost of wind and solar power abandonment at time t; f 3,t represents the operation penalty cost of the segment switch at time t; P ij,t , Q ij,t They represent the active power and reactive power flowing from node i to node j at time t respectively; α ij Represents the switch state of line ij after reconstruction; G ij , B ij represents the conductance and susceptance of branch ij; β ij,t It represents the trigonometric function relationship of the voltage of node ij at time t after the introduction of the intermediate variable; represents the maximum current flowing through node ij at time t; ψ b Indicates the system branch set.

[0123] Step 6: According to the distribution network reconstruction optimization model solved in step 5, analyze and evaluate the optimization results, formulate corresponding demand response strategies, guide users to use electricity during off-peak hours through price incentives, and provide real-time electricity price signals to encourage users to increase electricity consumption when electricity prices are low, thereby improving the economic and stability of power grid operation. Efficient energy management and resource allocation of the optimized distribution network can be achieved, and the overall operating performance and renewable energy absorption capacity can be improved, thereby promoting the realization of sustainable development and energy transformation goals.

[0124] The present invention provides a distribution network reconstruction method under the coordinated optimization of source-load uncertainty and demand response with energy storage, and the technical effects are as follows:

[0125] 1) In step 1 of the present invention, when constructing the initial distribution network topology information, complex network theory is used to abstract the actual distribution network into a graph, so that the structure of the complex distribution network can be analyzed using relevant graph theory, which helps to reveal the internal structural characteristics of the complex distribution network.

[0126] The distribution network will Figure 1 The set of nodes is called the node set E, and the set of edges is called the branch set L. Therefore, the power grid structure can be described by an abstract graph G and an adjacency matrix A, that is:

[0127] G=(E,L)

[0128]

[0129] In the formula, the value of the elements of matrix A is 0 or 1. When node i and node j are connected by a line, it takes 1, otherwise it takes 0. nn Indicates the connection relationship between node n and node n in the adjacency matrix A, that is, it is used to indicate whether the connection relationship between the nodes in the graph is connected and its weight; (α ij ) n*n Represents the connection relationship between nodes ij in the n-order adjacency matrix; n represents the total number of nodes.

[0130] 2) Step 2 of the present invention models wind power generation, photovoltaic power generation and load fluctuations respectively by considering the time series fluctuation characteristics and the impact on the operation of the power grid, thereby reducing the uncertain impact in the actual operation of the power grid. By integrating the output model, the operation cost of the power grid can be effectively reduced and the overall economic benefits can be improved.

[0131] 3) Step 3 of the present invention uses Latin hypercube to simulate the random variables of solar irradiance, wind speed and load, and adopts multi-scenario random programming method to convert the uncertainty problem of wind and solar power generation into a deterministic problem for solution. Multi-scenario random programming characterizes the uncertainty of wind and solar load, mainly including two aspects: scenario generation and scenario reduction. Scenario generation is mainly completed through random sampling, while the main method of scenario reduction is through clustering algorithm. To this end, the present invention proposes an improved K-means clustering algorithm, which aims to effectively cluster a large number of Latin hypercube simulation data of wind power, photovoltaic output scenarios and their load fluctuations. This method can effectively reduce the number of scenarios on the basis of ensuring the distribution characteristics of wind power, photovoltaics and load.

[0132] 4) Step 4 of the present invention aims at optimizing the operating cost of the distribution network system, constructs a multi-period distribution network reconstruction model, and constructs a distribution network reconstruction optimization model with the system's power balance, node voltage, branch current, distributed power source power, energy storage and demand response, and distribution network topology characteristics as constraints. This model aims to optimize the balance of power supply and demand and improve the economy and reliability of the system by rationally allocating various resources in the distribution network, while taking into account the load changes in multiple time periods and the volatility of renewable energy. By introducing energy storage equipment and demand response strategies, the model can effectively regulate power supply and demand, ensure that users' electricity needs are met in each time period, reduce overall operating costs, and enhance the adaptability of the distribution network to future intelligence and sustainable development. The construction of this optimization model will provide important support for the efficient management and flexible operation of the distribution network.

[0133] 5) Step 5 of the present invention introduces second-order cone convex relaxation technology and linearization method to the distribution network reconstruction model, and transforms the original model into a form with a convex feasible domain for solution. The introduction of intermediate variables to optimize the coupling relationship between various physical quantities can effectively simplify the complexity of the model, improve the solution efficiency, and ensure that various constraints are met during the solution process. The second-order cone convex relaxation technology converts nonlinear constraints into linear constraints, thereby enhancing the solvability of the model, and through the introduction of intermediate variables, further optimizes the interconnection between power flow, voltage level and current distribution in the distribution network. This method can improve the quality of the overall solution and the stability of operation, provide accurate data support for subsequent optimization decisions, and ensure the successful implementation of distribution network reconstruction and the maximization of operating efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0134] The present invention will be further described below in conjunction with the accompanying drawings and embodiments:

[0135] Figure 1 It is a structural diagram of the improved IEEE1333 node system of the present invention.

[0136] Figure 2It is a flowchart of scene reduction based on improved k-means clustering.

[0137] Figure 3 This is the DBI index change diagram of the improved k-means clustering of the present invention.

[0138] FIG4( a ) is a typical scene diagram after source load reduction of the present invention (typical photovoltaic scene after reduction);

[0139] FIG4( b ) is a typical scene diagram after source load reduction of the present invention (typical scene of wind turbine after reduction);

[0140] FIG4( c ) is a diagram of a typical scenario after source load reduction of the present invention (a typical scenario of electric load output after reduction).

[0141] Figure 5 It is a load curve diagram before and after the implementation of DR in the present invention.

[0142] Figure 6 It is the voltage distribution diagram of the nodes in the distribution network reconstruction before and after the implementation of DR.

[0143] Figure 7 This is a diagram of voltage changes at distribution network reconstruction nodes before and after DR is implemented during the peak period of 13:00 according to the present invention.

[0144] Figure 8 This is a diagram of the voltage changes at the distribution network reconstruction nodes before and after the implementation of DR during the valley period of 05:00 of the present invention.

[0145] Fig. 9 It is a diagram showing the relationship between DG output and energy storage charge state of the present invention. DETAILED DESCRIPTION

[0146] Adopting the improved IEEE33 node system, such as Figure 1 The present invention provides a distribution network reconstruction method under the coordination and optimization of source-load uncertainty and demand response with energy storage, which specifically includes the following steps:

[0147] Step S1: constructing initial distribution network topology information according to distribution network parameters;

[0148] Step S2: Establish a probability model of distributed power output and load fluctuation, taking into account the characteristics of time series fluctuations and the impact on grid operation, to ensure that the model can dynamically reflect changes in the real environment;

[0149] Step S3: Uncertainty processing is performed on distributed power sources and loads. Classic wind, solar and load scenarios are generated using Latin hypercube. Multi-scenario random programming methods are used to convert the uncertainty problem of wind and solar power generation into a deterministic problem for solution. The simulation scenarios are reduced using an improved K-means clustering algorithm to construct a typical wind, solar and load joint time series scenario set.

[0150] Step S4: Taking the optimal operation cost of the distribution network system as the goal, a multi-period distribution network reconstruction model is constructed, and the power balance, node voltage, branch current, distributed generation power, energy storage and demand response of the system and the characteristics of the distribution network topology are used as constraints to construct a distribution network reconstruction optimization model;

[0151] Step S5: Introduce the second-order cone convex relaxation technology and linearization method to the distribution network reconstruction model, transform the original model into a form with a convex feasible domain for solution. Introduce intermediate variables to optimize the coupling relationship between various physical quantities;

[0152] Step S6: Analyze and evaluate the optimization results and formulate corresponding demand response strategies to ensure the effective implementation of the optimization plan in actual operation and improve the overall performance of the distribution network and its ability to absorb clean energy.

[0153] In step S2, constructing a distributed power generation output and load fluctuation probability model includes:

[0154] S201, wind power generation probability model: The random output model of the wind turbine generator set, the wind speed obeys the two-parameter Weibull distribution, and its corresponding probability density function is:

[0155]

[0156] Where: v is the actual wind speed in the area, k w is a shape parameter, reflecting the distribution characteristics of wind speed in the area; k c is the scale parameter;

[0157] Wind power generation power P WT The relationship between and wind speed ν can be expressed as a piecewise function:

[0158]

[0159] Where: ci is the cut-in wind speed; r is the rated wind speed; ν co is the cut-out wind speed; r is the rated output power of the fan.

[0160] S202, Photovoltaic power generation probability model: The random characteristics of solar irradiance in photovoltaic power generation are usually considered to obey the Beta distribution, and its corresponding probability density function is:

[0161]

[0162] Where: α, β are the shape parameters of Beta distribution; P pv is the actual light intensity; P pv，maxis the maximum light intensity.

[0163] Photovoltaic output power pv The relationship between and light intensity r is:

[0164]

[0165] Where: r s is the rated light intensity; P r,pv It is the rated output of photovoltaic power generation.

[0166] S203, load probability model: The load model can be described by normal distribution:

[0167]

[0168] Where: P load , Q load are active load and reactive load respectively; μ p , σ p are the mean and standard deviation of load active power respectively; μ Q , σ Q are the mean and standard deviation of the load reactive power respectively.

[0169] like Figure 2 As shown, in step S3, the present invention proposes an improved K-means clustering algorithm, and the specific operation process of the algorithm improvement includes:

[0170] S301. Determine the optimal number of clusters based on clustering effectiveness indicators:

[0171] (1) Distance between samples:

[0172] The distance between samples is usually calculated using the Euclidean distance. For two m-dimensional samples x i and x j , whose Euclidean distance is defined as:

[0173]

[0174] In the above formula: x ik 、x jk They represent the coordinate vectors of the i-th and j-th samples respectively; k means that the sample data is divided into k categories; m represents the feature dimension of the sample point.

[0175] (2) Class center point:

[0176] Class K i The center point c i is the mean of all samples in the class:

[0177]

[0178] In the above formula, Ni is a K-like i The number of samples in; x represents class K i The feature vector of the i-th sample point.

[0179] (3) Intra-class dispersion:

[0180] The intra-class dispersion is the class radius r i It can be defined as the average distance from all samples in a class to the center point of its class:

[0181]

[0182] In the above formula: d(x,c i ) represents the sample x to the class center point c i distance.

[0183] (4) Distance between classes:

[0184] That is, the distance between the center points of two classes:

[0185]

[0186] In the above formula: c i ,c j Respectively represent K i ,K j The inter-class distance between ik 、c jk Represents class K i and class K j The center point of .

[0187] (5) DBI indicator:

[0188] That is, a metric for evaluating the quality of clustering, which is calculated based on the intra-cluster dispersion and the inter-cluster distance.

[0189] For K classes, the DBI metric is defined as:

[0190]

[0191] In the above formula: r i ,r j K i ,K j The intra-class dispersion of d(c i ,c j ) is class K i and class K j The smaller the DBI index, the better the sample clustering effect; the optimal number of clusters for K-means clustering is determined based on the DBI index, that is, to find the K value that minimizes the DBI index:

[0192]

[0193] In the above formula: K best Indicates the number of clusters that determine the minimum DBI index; DBI(K) represents the DBI index function calculated when the number of clusters K is given; Indicates the K value corresponding to the minimum value of the calculated DBI indicator.

[0194] S302. Use density clustering optimization to select cluster centers:

[0195] First, calculate the Euclidean distance between data points in the sample, that is:

[0196]

[0197] Where: x ik and x jk Respectively represent the values ​​of data points i and j at the kth eigenvalue.

[0198] Set the threshold distance d t , and calculate the local density ρ within the distance threshold i , the calculation formula is:

[0199]

[0200] Where: d ij is the Euclidean distance between data points i and j.

[0201] The source load data is clustered as the original scene. Figure 3 As shown in the figure, the DBI index decreases with the increase of the number of clusters until the DBI index reaches the minimum when the number of clusters is 10. The optimal number of clusters for the improved K-means clustering is determined to be 10, which is the number of typical source-load output scenarios. The comparison before and after the improved K-means clustering algorithm is shown in Table 1.

[0202] Table 1 Comparison of K-means clustering before and after improvement

[0203]

[0204] From Table 1, we can see that by comparing the K-means clustering before and after the improvement, we can conclude that by combining effective clustering indicators such as DBI to optimize the selection of K, and introducing the idea of ​​density clustering to improve the selection of cluster centers, the effect of traditional K-means clustering can be effectively improved. Specifically, it is manifested in a smaller DBI indicator, better clustering effect, and faster convergence speed. It can better handle data distribution with uneven density and improve the adaptability and robustness of the clustering algorithm. These improvements make the clustering algorithm more advantageous when processing complex data sets.

[0205] Considering that the source-load prediction error follows a normal distribution, its mean is the predicted value and the standard deviation is 10% of the predicted value. During the simulation, the Latin hypercube sampling method was used to generate 1,000 photovoltaic, wind power and load scenarios. The initial scenarios were analyzed by feature classification using the improved k-means clustering method to achieve scenario reduction, and 10 typical scenarios were obtained. Figure 4(a) to Figure 4(c) shown.

[0206] In step S4, a multi-period distribution network reconstruction model is constructed. The model proposed in the present invention aims to achieve the maximum social benefit under the operating constraints with the goal of optimizing the operating cost of the distribution network system. The specific steps of the reconstruction model are as follows:

[0207] S401. Objective function:

[0208] 1). Network loss cost 1,t :

[0209]

[0210] Where: b is the set of all branches in the system; C loss is the network loss cost coefficient; I ij,t is the effective value of the current flowing through branch ij at time t; R ij is the equivalent resistance of branch ij; △t represents each moment;

[0211] 2). Operation cost f2:

[0212]

[0213] Where: f wind represents the system wind curtailment cost; f pv Indicates the system abandoned light cost; C wind ,C pv are the penalty coefficients for wind and solar abandonment, respectively; ψ wind ,ψ pv They represent the node sets connected to wind turbines and photovoltaics in the power distribution system respectively; They represent the output power of wind turbine and photovoltaic at node i at time t respectively; They represent the actual power of wind turbine and photovoltaic connected to the distribution network at node i at time t respectively.

[0214] 3). Total cost of sectional switch f 3,t :

[0215]

[0216] Where: C switch is the cost coefficient of the switching operation; a ij,0 ,a ijThey represent the switch states of branch ij before and after the reconstruction of the distribution network flow, respectively. When its value is 1, the switch of branch ij is closed, and when its value is 0, the switch of branch ij is open; ψ b Represents a collection of distribution network branches.

[0217] S402. Constraints:

[0218] ①. Node power balance constraint: The distribution network satisfies the active / reactive power balance constraint at node i at time t:

[0219]

[0220] Where: are the active and reactive power of node i at time t; V i,t ,V j,t are the voltage of node i at time t; B ij ,G ij are the susceptance and conductance between branches ij respectively; θ ij,t is the power angle difference of branch ij at time t; P ij,t , Q ij,t They represent the active power and reactive power flowing from node i to node j at time t respectively; α ij Represents the switch state of line ij after reconstruction; It indicates that the power output and input of any intermediate system node maintain power balance; N represents the number of system branches.

[0221] ②.Safe operation constraints:

[0222]

[0223] Where: P ij ,Q ij are the active and reactive power transmitted between nodes i and j respectively; X ij ,R ij are the reactance and resistance between branches ij respectively; V i,min ,V i,max are the lower and upper limits of the voltage amplitude at node i respectively; I ij,max is the maximum value of the current amplitude of line ij at time t.

[0224] ③.DG power constraints

[0225]

[0226] Where: P DG,max is the upper limit of the DG capacity connected to the i-node; is the DG power factor angle at node i; P i，DG represents the active power of DG at node i.

[0227] ④. Distribution network radiation and connectivity constraints:

[0228] During the distribution network reconstruction process, its topology should be guaranteed to be in a radial state, without ring networks and islands. Its constraints can be expressed as:

[0229]

[0230] Where: n is the number of system branches; β ij is the subordinate relationship between node i and node j; it is 1 when node i is the parent node of node j, otherwise it is 0; N i is the number of nodes in the network; β jt represents the state attribute of node i at time t; β 0j , β 1j They respectively represent the specific states of node j, namely, node i is the parent node of node j and node i is not the parent node of node j.

[0231] ⑤. Energy storage constraints:

[0232] Energy storage charging and discharging constraints can be defined as:

[0233]

[0234]

[0235] Where: They represent the charging and discharging status of the energy storage at node i at time t respectively; They represent the charging and discharging power of the energy storage at node i at time t respectively; is the maximum charging and discharging power of the energy storage at node i at time t.

[0236] The external energy transfer characteristics of energy storage depend on the charge and discharge power and the state of charge (SOC). SOC is defined as:

[0237]

[0238] Where: is the SOC of the energy storage at node i at time t; η is the self-discharge rate of the energy storage; η charge ,η discharge are the charge and discharge conversion efficiency of energy storage respectively; △T is the scheduling time interval. represents the energy storage capacity of node i from time t to the next time t; They represent the minimum and maximum capacity limits of energy storage at node i respectively.

[0239] ⑥. Demand response constraints

[0240]

[0241] Where: i,t is the price elasticity coefficient of node i at time t; P i,t , are the loads of node i before and after DR at time t; ρ i,t , are the electricity prices before and after DR of node i at time t; They are the upper and lower limits of the electricity price after DR is implemented at node i at time t. peak represents the peak electricity price; ρ valley Indicates the valley electricity price; T peak Indicates the peak load moment; T valley valley load moment; Indicates any node in the value system.

[0242] Preferably, in order to solve the reconstructed model in step S5, a second-order cone convex relaxation technique and a linearization method are introduced to transform the original model into a form with a convex feasible domain for solving. The second-order cone optimization model used is as follows:

[0243]

[0244] The present invention introduces two virtual node voltages are the voltages at the starting point i and the end point j of branch λ at time t. When branch λ is running, a λ =1, these two virtual voltages directly reflect the voltages of the actual nodes i and j, that is, When branch λ is disconnected, we have Its connection with the node voltage can be limited by the following formula.

[0245]

[0246] Preferably, in step S6, according to the evaluation results of the model solution in step S5, the performance indicators of the distribution network after reconstruction are obtained, including power loss, power supply reliability and economic operation cost, etc., and the effectiveness of the optimization scheme is confirmed through comparative analysis of these indicators. At this time, the peak and valley time-of-use electricity prices obtained by the model are shown in Table 2.

[0247] Table 2 Time-of-use electricity price after optimization and reconstruction

[0248]

[0249] From Table 2, it can be seen that the reconstructed power grid can adjust the user's electricity consumption behavior by adjusting the time-of-use electricity price (peak, valley, and flat electricity price), and encourage users to increase electricity consumption during periods with lower electricity prices (such as low-valley electricity prices at night) and reduce electricity consumption during peak hours, thereby balancing power supply and demand and reducing the pressure on the power grid. Moreover, the optimized time-of-use electricity price can smooth the load curve and reduce the load fluctuation of the power grid during peak hours. This not only helps to improve the stability of the power grid, but also reduces equipment failures or power grid overload problems caused by sudden load increases. It also helps the power grid to dispatch more renewable energy, especially wind power and solar power generation. The power generation characteristics of these energy sources are time-varying and volatile. The adjustment of time-of-use electricity prices can encourage users to increase electricity consumption during periods of large renewable energy generation (such as peak solar power generation during the day), thereby improving the utilization rate of renewable energy and reducing the use of traditional fossil energy.

[0250] In order to better study the distribution network reconstruction strategy proposed in this invention, the following scenarios are set for comparative analysis under the premise of considering the uncertainty of the source and load on both sides of the energy storage:

[0251] Scenario 1: Distribution network reconstruction without considering the uncertainty on both sides of the source and load and DR.

[0252] Scenario 2: Only the uncertainty on both the source and load sides is considered, and the distribution network reconstruction of DR is not considered.

[0253] Scenario 3: Distribution network reconstruction that only considers DR and ignores the uncertainties on both the source and load sides.

[0254] Scenario 4: Distribution network reconstruction considering both the uncertainty of the source and the load and DR.

[0255] The comparative analysis of distribution network reconstruction in different scenarios is shown in Table 3:

[0256] Table 3 Comparison of distribution network reconstruction under different scenarios

[0257]

[0258] The comparative analysis of distribution network reconstruction in different scenarios is shown in Table 3. It can be seen from Table 3 that the total system operation cost of the distribution network reconstruction strategy (scenario 4) proposed in the present invention is the lowest at 610.32 yuan per day.

[0259] Consider the load change curve of the demand response system after optimization and reconstruction. Figure 5As shown, it can be seen that the peak-to-valley difference of the load curve before and after the implementation of DR changes from 0.31 to 0.18. After the DR strategy of the present invention is adopted, the load volatility is reduced and smoother than before DR. Considering that the demand response strategy aims to adjust the power demand to balance the power supply and demand to avoid the pressure caused by peak load and potential power shortages. In the distribution network containing DG, by using the energy storage system combined with the demand response strategy, the distribution network reconstruction can effectively achieve load balance regulation.

[0260] Before and after considering demand response, the curves of the voltage changes of each node over time are as follows: Figure 6 As shown. The minimum node voltage after the distribution network reconstruction is 0.951pu when demand response is not considered, and it is increased to 0.964pu after demand response is considered. The node voltage of the reconstruction strategy mentioned in the present invention is generally higher than the node voltage when demand response is not considered. The phenomenon of increased node voltage helps to improve the reliability, efficiency and flexibility of the power system as a whole.

[0261] Demand response measures can regulate the voltage of the distribution network, and can comprehensively evaluate the impact of demand response strategies on the power quality of the distribution network at different load periods. At the two time nodes of 13:00 and 05:00, the comparative analysis of the voltage curves of the distribution network reconstruction nodes before and after the implementation of demand response shows different power quality performances. Figure 7 and Figure 8 As shown in Figure 2, it shows the positive impact of demand response on grid operation, especially during peak load periods, demand response can effectively reduce peak load, alleviate grid pressure, and improve power quality and node voltage stability.

[0262] Demand response measures further increase the flexibility of the distribution network by incentivizing users to reduce electricity consumption during peak hours or increase consumption during off-peak hours. Fig. 9 The diagram shows the relationship between the energy storage charge state and DG output after the demand response strategy is implemented. Since the implementation of the demand response measures proposed in the present invention effectively balances the supply and demand of the power system, the energy storage does not need to be adjusted on a large scale, and the DG absorption capacity is improved.

[0263] Through the above-mentioned example verification and analysis, the present invention discloses a distribution network reconstruction method under the coordination and optimization of source-load uncertainty and demand response with energy storage, so that after the distributed power source is connected to the distribution network, the flexibility and stability of the distribution network can be effectively improved. This method takes into account the volatility and uncertainty of renewable energy, and by combining with demand response, optimizes the configuration of power resources, realizes the dynamic balance between load regulation and energy supply, and introduces energy storage equipment during operation to balance the supply and demand relationship, reduce peak load, and improve the economy and reliability of the system. Ultimately, this reconstruction method not only improves the access capacity of clean energy, but also promotes the realization of sustainable development goals, and provides new ideas and solutions for the construction of smart distribution networks.

[0264] This comprehensive approach to the strategy proposed in the present invention can significantly improve the reliability and adaptability of the system, reduce the pressure of peak and valley loads, and ultimately achieve higher economy and environmental friendliness. The present invention provides a distribution network reconstruction method under the coordination and optimization of source-load uncertainty and demand response with energy storage, which makes up for the shortcomings of traditional distribution network reconstruction methods in dealing with renewable energy uncertainty and load fluctuations.

Claims

1. A distribution network reconstruction method under the coordinated optimization of source-load uncertainty and demand response with energy storage, characterized in that The following steps are involved: Step 1: Construct the initial distribution network topology information; Step 2: Considering the characteristics of time series fluctuations and their impact on grid operation, a probability model of distributed generation output and load fluctuations is established; Step 3: Deal with the uncertainty of distributed power sources and loads; Step 4: Construct a distribution network reconstruction optimization model that takes into account demand response and source-load characteristics; Step 5: Introduce the second-order cone convex relaxation technology and linearization method to the distribution network reconstruction optimization model constructed in step 4, and transform the original model into a form with a convex feasible domain for solution.

2. A distribution network reconstruction method under the coordinated optimization of source-load uncertainty and demand response with energy storage according to claim 1, characterized in that: In step 1, in the distribution network topology information, the set formed by the nodes is called the node set E, and the set formed by the edges is called the branch set L. Therefore, the power grid structure can be described by an abstract graph G and an adjacency matrix A, that is: G=(E,L) Where: the value of the elements of matrix A is 0 or 1. When node i and node j are connected by a line, it takes 1, otherwise it takes 0; a ij Represents the connection relationship between nodes i and j in the adjacency matrix A, that is, it is used to indicate whether the node connection relationship in the graph is connected and its weight; (α ij ) n*n Represents the connection relationship between nodes ij in the n-order adjacency matrix; n represents the total number of nodes.

3. According to claim 1, a distribution network reconstruction method under the coordinated optimization of source-load uncertainty and demand response with energy storage, characterized in that: In step 2, a distributed power output and load fluctuation probability model is constructed, including: (1) Wind power generation probability model: Where: f(P WT) represents the probability density function of wind power output; P WT Indicates the output power of the wind turbine; represents an exponential function with e as the base; v represents the actual wind speed in the area; k w represents the shape parameter, reflecting the distribution characteristics of wind speed in the area; k c is a scale parameter, reflecting the average wind speed in the area; k w , k c It can be calculated by the following formula using actual wind speed data: Where: δ, μ are the standard deviation and mean of wind speed respectively; Γ is the gamma function; (2) Photovoltaic power generation probability model: Where: f(P PV) represents the probability density function of photovoltaic output power; α, β are the shape parameters of Beta distribution; P PV is the actual light intensity; P PV,max is the maximum light intensity; Γ(α+β) represents the gamma function of the sum of shape parameters, which is used to describe the Beta distribution characteristics; Γ(α) represents the value of the gamma function at parameter α; Γ(β) represents the value of the gamma function at parameter β; (3) Load probability model: Where f(P load ) represents the probability density function of active power; f(Q load ) represents the probability density function of reactive power; P load , Q load are active load and reactive load respectively; μ P , σ P are the mean and standard deviation of load active power respectively; μ Q , σ Q are the mean and standard deviation of the load reactive power respectively.

4. A distribution network reconstruction method under the coordinated optimization of source-load uncertainty and demand response with energy storage according to claim 1, characterized in that: The step 3 comprises the following steps: S3.1: Use Latin hypercube to simulate the random variables of solar irradiance, wind speed and load, and simulate and solve the random variables of wind and solar load; S3.2: Use the improved K-means clustering algorithm to effectively cluster the Latin hypercube simulation data of wind power, photovoltaic output scenarios and their load fluctuations.

5. A distribution network reconstruction method under the coordinated optimization of source-load uncertainty and demand response with energy storage according to claim 4, characterized in that: S3.1 specifically includes: (1) Sampling process: Let X k is any one of the k random variables, then X k The corresponding cumulative probability distribution function is: Y k =F k (X k ); Assume that the sampling scale is N, divide the vertical axis of the cumulative probability density curve into N equally spaced non-overlapping intervals, each interval length is 1 / N, and use the median Latin hypercube sampling method to obtain the nth sampling value y kn =(n-0.5) / N, its sampling value x kn for: In the formula, is the inverse function of the probability distribution function; According to the above steps, we can get the sample matrix X KN ; (2) Sorting process: L KN is the sorting matrix of the sampling matrix elements, and the correlation coefficient between its rows is represented by the matrix ρ. It is easy to prove that the matrix ρ is a positive definite matrix, so the matrix ρ is decomposed into: r L =DD T In the formula, ρ L The correlation coefficient matrix represented as matrix L; D is a non-singular lower triangular matrix; In order to accurately reflect the row correlation of the sampling matrix, the root mean square ρ of the correlation coefficient matrix ρ is introduced msr , which is defined as: In the formula, ρ kj represents the off-diagonal elements in the correlation coefficient matrix; Since the matrix D is non-singular, there must be an inverse matrix D -- , combined with the initial sorting matrix L KN , we can get a matrix G with smaller row correlation: G=D -1 L KN Repeat the above steps until the row correlation is reduced to a preset value.

6. A distribution network reconstruction method under the coordinated optimization of source-load uncertainty and demand response with energy storage according to claim 4, characterized in that: S3.2 specifically includes: S3.2.

1. Determine the optimal number of clusters based on clustering effectiveness indicators: (1) Distance between samples: The distance between samples is usually calculated using the Euclidean distance. For two m-dimensional samples x i and x j , whose Euclidean distance is defined as: In the above formula: x ik 、x jk Respectively represent the i-th and j-th sample coordinate vectors; k means that the sample data is divided into k categories; m represents the feature dimension of the sample point; (2) Class center point: Class K i The center point c i is the mean of all samples in the class: In the above formula: Ni is K-like i The number of samples in; x represents class K i The feature vector of the i-th sample point; (3) Intra-class dispersion: The intra-class dispersion is the class radius r i It can be defined as the average distance from all samples in a class to the center point of its class: In the above formula: d(x,c i ) represents the sample x to the class center point c i distance; (4) Distance between classes: That is, the distance between the center points of two classes: In the above formula: c i ,c j Respectively represent K i ,K j The inter-class distance between ik 、c jk Represents class K i and class K j The center point; (5) DBI indicator: That is, a metric for evaluating clustering quality, which is calculated based on the intra-class dispersion and inter-class distance; for each class, the DBI indicator is defined as: In the above formula: r i ,r j K i ,K j The intra-class dispersion of d(c i ,c j ) is class K i and class K j The distance between the center points of The smaller the DBI index is, the better the sample clustering effect is. The optimal number of clusters for K-means clustering is determined based on the DBI index, that is, the K value that minimizes the DBI index is found: In the above formula: K best Indicates the number of clusters that determine the minimum DBI index; DBI(K) represents the DBI index function calculated when the number of clusters K is given; Indicates the K value corresponding to the minimum value of the calculated DBI indicator; S3.2.2, use density clustering optimization to select cluster centers: First, calculate the Euclidean distance between data points in the sample, that is: Where: x ik and x jk Respectively represent the values ​​of data points i and j at the kth eigenvalue; Set the threshold distance d t , and calculate the local density ρ within the distance threshold i , the calculation formula is: Where: d ij is the Euclidean distance between data points i and j.

7. A distribution network reconstruction method under the coordinated optimization of source-load uncertainty and demand response with energy storage according to claim 1, characterized in that: In step 4, the distribution network reconstruction optimization model comprehensively considers the minimization of the comprehensive operation cost of the distribution network, that is: Where: C is the total cost in the distribution network reconstruction model; T is the duration of distribution network reconstruction; f 1,t is the network loss cost at time t; f 2,t is the cost of wind and solar power abandonment at time t; f 3,t is the penalty cost of the segment switch operation at time t.

8. The method for reconfiguring a distribution network under the coordinated optimization of source-load uncertainty and demand response with energy storage according to claim 1, characterized in that: The constraints of the distribution network reconstruction optimization model include: ①.Node power balance constraints: The distribution network satisfies the active / reactive power balance constraint at node i at time t: Where: are the active and reactive power of node i at time t; V i,t ,V j,t are the voltage of node i at time t; B ij ,G ij are the susceptance and conductance between branches ij respectively; θ ij,t is the power angle difference of branch ij at time t; P ij,t , Q ij,t They represent the active power and reactive power flowing from node i to node j at time t respectively; α ij Indicates the switch status of line ij after reconstruction; Indicates that the power output and input of any intermediate system node maintain power balance; N represents the number of system branches; ②.Safe operation constraints: Where: P ij ,Q ij are the active and reactive power transmitted between nodes i and j respectively; X ij ,R ij are the reactance and resistance between branches ij respectively; V i,min ,V i,max are the lower and upper limits of the voltage amplitude at node i respectively; I ij,max is the maximum current amplitude of line ij at time t; ③.DG power constraints Where: P DG,max is the upper limit of the DG capacity connected to the i-node; is the DG power factor angle at node i; P i,DG represents the active power of DG at node i; ④. Distribution network radiation and connectivity constraints: During the distribution network reconstruction process, its topology should be guaranteed to be in a radial state without ring networks and islands. Its constraints can be expressed as: Where: n is the number of system branches; β ij is the subordinate relationship between node i and node j; it is 1 when node i is the parent node of node j, otherwise it is 0; N i is the number of nodes in the network; β jt represents the state attribute of node i at time t; β 0j , β 1j They represent the specific state of node j, namely, node i is the parent node of node j and node i is not the parent node of node j; ⑤. Energy storage constraints: Energy storage charging and discharging constraints can be defined as: Where: They represent the charging and discharging status of the energy storage at node i at time t respectively; They represent the charging and discharging power of the energy storage at node i at time t respectively; is the maximum charging and discharging power of the energy storage at node i at time t; The external energy transfer characteristics of energy storage depend on the charge and discharge power and the state of charge (SOC). SOC is defined as: Where: is the SOC of the energy storage at node i at time t; η is the self-discharge rate of the energy storage; η charge ,η discharge are the charge and discharge conversion efficiency of energy storage respectively; △T is the scheduling time interval; represents the energy storage capacity of node i from time t to the next time t; They represent the minimum and maximum capacity limits of energy storage at node i respectively; ⑥. Demand response constraints Where: i,t is the price elasticity coefficient of node i at time t; P i,t , are the loads before and after DR of node i at time t; ρ i,t , are the electricity prices before and after DR of node i at time t; are the upper and lower limits of the electricity price after DR is implemented at node i at time t; ρ peak represents the peak electricity price; ρ valley Indicates the valley electricity price; T peak Indicates the peak load moment; T valley valley load moment; Indicates any node in the value system.

9. A distribution network reconstruction method under the coordinated optimization of source-load uncertainty and demand response with energy storage according to claim 1, characterized in that: In step 5, the distribution network reconstruction optimization model is converted into a form with a convex feasible domain for solution, which specifically includes: Introduce intermediate variables to optimize the coupling relationship between various physical quantities: Where: I ij,t represents the effective value of the current flowing through branch ij at time t; V i,t represents the voltage of node i at time t; They respectively represent the quadratic relationship of the current at the node ij at time t after the introduction of the intermediate variable; They represent the quadratic relationship of the voltage at node ij at time t after the introduction of the intermediate variable; α ij,t , β ij,t They respectively represent the trigonometric function relationship of the voltage at the node ij at time t after the introduction of the intermediate variable; They represent the quadratic relationship of the voltage at node ij at time t after the introduction of the intermediate variable; θ ij,t represents the common angle difference of branch ij at time t; Introducing two virtual node voltages are the voltages at the starting point i and the end point j of branch λ at time t; when branch λ is running, a λ =1, these two virtual voltages directly reflect the voltages of the actual nodes i and j, that is, When branch λ is disconnected, we have Its connection to the node voltage is limited by the following equation: Where: α λ Represents the branch λ in the system; V imax represents the maximum voltage of node i at time t; V jmax represents the maximum voltage of node j at time t; The nonlinear objective function and nonconvex constraints in the original model are relaxed and transformed as follows: Where: C loss Represents the network loss cost coefficient; R ij represents the equivalent resistance of branch ij; △t represents each moment; f 2,t represents the cost of wind and solar power abandonment at time t; f 3,t represents the operation penalty cost of the segment switch at time t; P ij,t , Q ij,t They represent the active power and reactive power flowing from node i to node j at time t respectively; α ij Represents the switch state of line ij after reconstruction; G ij , B ij represents the conductance and susceptance of branch ij; β ij,t It represents the trigonometric function relationship of the voltage of node ij at time t after the introduction of the intermediate variable; represents the maximum current flowing through node ij at time t; ψ b Indicates the system branch set.

10. The method for reconfiguring a distribution network under the coordinated optimization of source-load uncertainty and demand response with energy storage according to claim 1, characterized in that: In step 6, according to the distribution network reconstruction optimization model solved in step 5, the optimization results are analyzed and evaluated, and a corresponding demand response strategy is formulated. Through price incentives, users are guided to use electricity during off-peak hours, and real-time electricity price signals are provided to encourage users to increase electricity consumption when electricity prices are low, thereby improving the economy and stability of power grid operation.

Images