A method for uncertainty absorption and load balancing of flexible interconnected microgrid groups

Abstract

A method for uncertainty absorption and load balancing of a flexible interconnected microgrid group includes S1, quantifying the uncertainty sources of the microgrid group and determining a two-stage energy storage control strategy; S2, constructing a microgrid group optimization control model based on distributed robust opportunity constraints with the goal of minimizing operating cost, the model includes line transmission power opportunity constraints, node voltage opportunity constraints, energy storage charging and discharging power opportunity constraints, and energy storage capacity safety opportunity constraints, as well as energy storage capacity constraints, network flow constraints, VSC capacity constraints, and AC regional power balance constraints; S3, linearizing and reconstructing the uncertainty penalty cost using the Wasserstein Metric definition and duality theorem, and converting the opportunity constraints into deterministic constraints; S4, linearizing and reconstructing other nonlinear parts of the model objective function and constraint conditions; S5, solving the microgrid group optimization control model and optimizing the control of the flexible interconnected microgrid group.

Description

Technical Field

[0001] The present invention relates to the field of microgrid control technology, and in particular to a method for uncertainty absorption and load balancing of a flexible interconnected microgrid group. Background Art

[0002] As the power industry advances its green transformation, distribution networks are undergoing profound changes. The large-scale integration of distributed renewable energy and the growing demand for DC loads present unprecedented challenges. Specifically, the uneven deployment of renewable energy leads to unbalanced load distribution, which in turn causes voltage over-limit issues. Furthermore, the uncertainty caused by errors in renewable energy output forecasts exacerbates voltage fluctuations, seriously threatening the safe and stable operation of distribution networks.

[0003] However, with the rapid development of power electronics technology, multiple low-voltage distribution network substations have achieved flexible interconnection through voltage source converters (VSCs). The resulting flexible interconnected microgrid architecture based on a chain-type segmented topology has the advantages of temporal and spatial complementarity between sources and loads, complementarity between uncertain generation sources and energy storage, and coordinated scheduling of multiple resources, providing an effective way to solve the above problems.

[0004] The emergence of flexible, interconnected low-voltage AC / DC microgrids has driven the evolution of distribution networks toward a multi-tiered architecture. Through autonomous optimization, lower-tier microgrids can effectively reduce the control scale and coordination complexity of upper-tier distribution networks. However, with the integration of large numbers of wind power, photovoltaic power, and energy storage systems into the DC busbars of these flexible microgrids, and their energy exchange with the upper-tier distribution network via VSCs, system operation and control face increased uncertainty and complexity. Traditional system control methods fail to account for source-load uncertainty. While they can achieve an "optimal" control solution, their neglect of the actual fluctuations in energy storage output and the power purchase and sales at each substation significantly reduces the feasibility of the solution, increasing the difficulty of system control and making the "optimal" solution often difficult to implement. Current approaches to address multi-source uncertainty in systems primarily include stochastic optimization and robust optimization. The former offers a clear model structure and simple derivation, but requires the known probability distribution of uncertainties, which is often difficult to accurately obtain in practice, and the solution efficiency decreases significantly as the number of scenarios increases. The latter, based on worst-case modeling, can ensure safe and stable system operation under any scenario, but this approach is overly conservative, resulting in low system economics. Therefore, it is urgent to study an autonomous control method for flexible interconnected microgrid groups that takes uncertainty into consideration, so as to achieve coordinated optimization of microgrid group resources and ensure the economic, safe and stable operation of the microgrid group. Summary of the Invention

[0005] In order to take into account both the economy and control reliability of the system, the present invention provides a method for uncertainty absorption and load balancing of a flexible interconnected microgrid group. This method introduces a distributed robust chance-constrained optimization method driven by historical data. While improving the solution efficiency, it effectively reduces the problem of overly conservative robust optimization and achieves improved operating performance in multiple scenarios.

[0006] In order to solve the above technical problems, the present invention adopts the following technical method: a method for uncertainty absorption and load balancing of a flexible interconnected microgrid group, comprising:

[0007] S1: quantify the uncertainty of wind power, photovoltaic power, and DC loads in the microgrid group and determine a two-stage control strategy for energy storage day-ahead scheduling and real-time adjustment;

[0008] S2, with the goal of minimizing the sum of the upper-level power purchase and sales costs, energy storage operation costs, load-sharing penalty costs, and uncertainty penalty costs, constructs a microgrid group optimization control model based on distributed robust opportunity constraints. The constraints of this model include opportunity constraints and other constraints. The opportunity constraints include line transmission power opportunity constraints, node voltage opportunity constraints, energy storage charging and discharging power opportunity constraints, and energy storage capacity security opportunity constraints. The other constraints include energy storage capacity constraints, network flow constraints, VSC capacity constraints, and AC regional power balance constraints.

[0009] S3, uses the Wasserstein Metric definition and duality theorem to linearize and reconstruct the uncertainty penalty cost in the model objective function, and transforms the model's chance constraints into deterministic constraints;

[0010] S4, linearize and reconstruct other nonlinear parts of the model objective function and constraints;

[0011] S5, obtain the network topology, equipment and distribution network parameter information of the microgrid group, collect several groups of historical wind power and photovoltaic output fluctuation data sample points at each moment, organize the equivalent linearized microgrid group optimization control model and solve it, obtain the control parameters, and optimize the control of the flexible interconnected microgrid group.

[0012] Furthermore, in S3, when the uncertainty penalty function is linearly reconstructed using the Wasserstein Metric definition and the duality theorem, a series of historical sample data sets of random variables are first collected as historical uncertainty fluctuation data sets of the uncertainty source. Then, before obtaining the distribution fuzzy set of the random variable, the radius of the Wasserstein sphere is determined according to the fuzzy set radius size formula. Finally, based on the Wasserstein Metric definition, the uncertainty penalty function is transformed into a dual problem, and the supremum approximation is introduced to simplify it.

[0013] Furthermore, in S3, when the Wasserstein Metric definition and duality theorem are used to transform the opportunity constraints of the microgrid group optimization control model into deterministic constraints, the mean and variance of the historical sample data set of random variables at each moment are first calculated, and all sample points are normalized to the standard coordinate system. Then, the problem of minimizing the side length of the hypercube is solved to obtain the envelope under a certain confidence level in the standard coordinate system. t The side length of the hypercube of the sample point at the moment is determined, and the corresponding hypercube is determined. Then the hypercube is denormalized to the original coordinate system to obtain the envelope under a certain confidence level in the original coordinate system. t The parallel polyhedron of the sample points at each moment is constructed, and the corresponding box uncertainty set is determined. Finally, the duality theorem is used to transform the original chance constraint problem into a robust problem under the box uncertainty set, and then:

[0014] Substituting the network power flow constraint into the line transmission power opportunity constraint, the line transmission power deterministic constraint is equivalently converted.

[0015] Substitute the network power flow constraint into the node voltage opportunity constraint, and the node voltage deterministic constraint is equivalently converted;

[0016] Substituting the energy storage output into the energy storage charging and discharging power opportunity constraint, the energy storage output deterministic constraint is obtained through equivalent transformation.

[0017] Substituting the energy storage capacity constraint into the energy storage capacity safety opportunity constraint, the energy storage capacity deterministic constraint is obtained through equivalent transformation.

[0018] Furthermore, the control parameters include the day-ahead output benchmark and real-time adjustment factor of each energy storage in the microgrid group, as well as the power purchase and sales amount of each substation.

[0019] Furthermore, S1 quantifies the uncertainty of wind power, photovoltaic power, and DC loads of the microgrid group as follows:

[0020] (1)

[0021] (2)

[0022] (3)

[0023] Where, express The actual output column vector of each uncertainty source at each moment; Represents the uncertainty source-node correlation matrix, including the photovoltaic-node correlation matrix , Wind power-node correlation matrix , DC load-node correlation matrix ; express The predicted values ​​of the sources of uncertainty at each moment, including Photovoltaic forecast values ​​at each moment , wind power forecast values , DC load forecast value ; is the prediction error column vector of each uncertainty source, representing the uncertainty;

[0024] In S1, the energy storage output of the microgrid group is divided into day-ahead scheduling and real-time adjustment. The output of energy storage at each node is as follows:

[0025] (4)

[0026] Where, express Real-time output of each energy storage at all times; represents the energy storage-node association matrix, It is the benchmark for energy storage’s day-ahead output; To adjust the factor matrix in real time, its specific definition is:

[0027] (5)

[0028] Where, Middle Row, No. Elements of a column , indicating the The energy storage device The proportion of uncertainty reduction of each uncertainty source, is the number of energy storage devices, is the number of uncertainty sources, The elements in the table satisfy the following conditions:

[0029] (6)

[0030] (7)

[0031] Formula (6) indicates that the proportion of each energy storage to alleviate each uncertainty is less than 100%; Formula (7) indicates Moment Uncertainty The total amount of compensation by all stored energy is less than or equal to the uncertainty itself.

[0032] Furthermore, the microgrid group optimization control model based on distributed robust opportunity constraints constructed in S2 is:

[0033] (8)

[0034] Where, is the optimization variable; They are the upper level electricity purchase and sales costs, energy storage operation costs, load balancing penalty costs, and uncertainty penalty costs;

[0035] (9)

[0036] Where, is the set of all scheduling times, is the numbered set of all nodes in the microgrid; and are electricity purchase cost and electricity selling price respectively; For the Power interaction between each node and the upper distribution network; Operator satisfy ;

[0037] (10)

[0038] Where, It is the numbered set of all energy storage in the microgrid; Output operating costs for energy storage units; Provide active power for each energy storage;

[0039] (11)

[0040] Where, is the load-sharing penalty coefficient; 、 They are respectively 、 Transformer capacity of each node; For the The amount of power interaction between each node and the upper distribution network; is the number of nodes in the microgrid group;

[0041] (12)

[0042] (13)

[0043] (14)

[0044] Where, is the uncertainty penalty coefficient; A numbered set representing all sources of uncertainty; Indicates the The source of uncertainty is The uncertainty distribution fuzzy set of the moment; Represents the maximum expected value under all possible probabilities; is a column vector The elements, , its physical meaning is that all energy storage in the microgrid group The proportion of uncertainty generated by each uncertainty source absorbed; represents a column vector of all 1s with a dimension of n×1;

[0045] The opportunity constraints of the microgrid group optimization control model constructed in S2 based on distributed robust opportunity constraints include:

[0046] 1) Line transmission power opportunity constraints

[0047] (15)

[0048] Where, Represents the set of all probability distributions that an uncertain quantity may satisfy; Indicates the minimum probability that the formula in the brackets is true; Indicates The DC line transmission power vector of the microgrid group at time t is, 、 are the minimum and maximum transmission power vectors of the DC line respectively; Indicates confidence;

[0049] 2) Node voltage opportunity constraint

[0050] (16)

[0051] Where, Indicates The DC bus voltage vector of the microgrid group at time , 、 are the maximum and minimum values ​​of the DC bus voltage respectively;

[0052] 3) Energy storage charging and discharging power opportunity constraints

[0053] (17)

[0054] Where, Indicates The output column vector of the energy storage at each moment, which stipulates that the energy storage discharge is in the positive direction; 、 are the maximum and minimum charging and discharging power of energy storage respectively;

[0055] 4) Energy storage capacity security opportunity constraints

[0056] (18)

[0057] Where, Indicates The energy storage capacity vector of each microgrid group at the moment; 、 They are the lower and upper limits of energy storage capacity respectively.

[0058] Other constraints of the microgrid group optimization control model constructed in S2 based on distributed robust opportunity constraints include:

[0059] 1) Energy storage capacity constraints

[0060] (19)

[0061] Where, Charging efficiency for energy storage; is the energy storage self-discharge coefficient; Indicates the daily scheduling interval, set to 1;

[0062] 2) Network flow constraints

[0063] (20)

[0064] (twenty one)

[0065] (twenty two)

[0066] Where, represents the node injection power; represents the active power injected by the VSC, is the augmented node-line association matrix, which is obtained from the microgrid topology diagram, i.e., the substation connection relationship; is the augmented line resistance matrix; represents the DC bus voltage column vector of the microgrid group; is the voltage reference;

[0067] 3) VSC capacity constraints

[0068] (twenty three)

[0069] Where, 、 Respectively expressed in Moment Active power injected and reactive power output by VSC at each node; For the The capacity of the VSC on each node;

[0070] 4) AC area power balance constraints

[0071] (twenty four)

[0072] Where, represents the column vector of the upper-level purchase and sale power, is the AC load column vector of the AC node.

[0073] In S3, the Wasserstein Metric definition and the duality theorem are used to linearize and reconstruct the uncertainty penalty function as follows:

[0074] 1) Get the radius of the Wasserstein sphere

[0075] Collect a series of random variables The historical sample data set, that is, the historical uncertainty fluctuation data set of the uncertainty source, has a sample size of M ; After getting the random variable Before the distribution of fuzzy sets, the fuzzy set radius size It can be determined by the following formula:

[0076] (35)

[0077] Where, Indicates the diameter of the data support set; To include random variables Confidence of all distributions of ;

[0078] 2) Problem transformation

[0079] Based on the Wasserstein Metric definition, the original objective function Transformed into a dual problem:

[0080] (36)

[0081] Where, is the dual multiplier; is an auxiliary variable, is the sample number set; Indicates uncertainty The feasible set of

[0082] The constraints contained in formula (36) face the problem of scalability, that is, as the amount of sample data gradually increases, the scale of the solution gradually expands; therefore, the supremum approximation is introduced, and formula (36) is further simplified to:

[0083] (37)

[0084] (25)

[0085] Where, 、 Indicates Moment The maximum and minimum values ​​of the historical data of each uncertainty source; is the radius of the fuzzy set; represents the number of samples; Indicates the uncertainty at each moment Middle A historical sample dataset of components;

[0086] In S3, the Wasserstein Metric definition and the duality theorem are used to transform the opportunity constraints of the microgrid group optimization control model into deterministic constraints;

[0087] 1) Normalize all historical sample points to the standard coordinate system

[0088] Calculate the random variables at each moment Historical sample data set The mean ,variance , and all sample points Normalize to the standard coordinate system:

[0089] (38)

[0090] Where, Indicates the coordinates of the sample point in the standard coordinate system;

[0091] 2) Get the side length of the hypercube in the standard coordinate system

[0092] Solve the problem of minimizing the side length of the hypercube and obtain a certain confidence level in the standard coordinate system Lower envelope Hypercube of sample points at time Side length :

[0093] (39)

[0094] Where, is the dual variable, is the sample point in the standard coordinate system The radius of the fuzzy set is obtained according to formula (35); is the set of coordinates of the vertices of the hypercube, which is further described as:

[0095] (40)

[0096] Where, The dimension size is A column vector of all 1s;

[0097] 3) Denormalize the hypercube to the original coordinate system

[0098] Denormalize the obtained hypercube to the original coordinate system and obtain a certain confidence level in the original coordinate system Lower envelope Parallel polyhedron of sample points at the moment , its vertex can be obtained by substituting formula (40) into the following formula:

[0099] (41)

[0100] is the set containing the coordinates of the vertices of the parallelepiped, which is further described as the following box uncertainty set:

[0101] (42)

[0102] This equation (42) describes the parallel polyhedron Internal space, is a constant matrix, whose Rank The element of the column is 1, Rank The element of the column is -1, The rest of the elements are 0; is a constant column vector, Row and The rows are Middle The absolute values ​​of the coordinates of the vertices, ;

[0103] 4) Proof of Proposition

[0104] Set optimization variables , all chance-constrained problems in the model can be described in a unified form:

[0105] (43)

[0106] The above formula always holds true, that is, the maximum value on the left side of the inequality is less than or equal to 0:

[0107] (44)

[0108] To solve the problem of finding the maximum value on the left side of the inequality (44), the duality theorem is used to further transform (44) into:

[0109] (45)

[0110] Therefore, formula (43) can be equivalently transformed into the following proposition:

[0111] (46)

[0112] 5) Transforming Chance Constraints into Deterministic Constraints

[0113] Substituting Equations (20) and (21) into the chance constraint Equation (15), we can obtain the following robust problem:

[0114] (47)

[0115] Based on the above-proven proposition, Equation (47) can be equivalently transformed into the following deterministic constraint on line transmission capacity:

[0116] (26)

[0117] Where, 、 is the dual variable related to the power capacity constraint;

[0118] Similarly, by substituting Equations (20) and (22) into Equation (16), based on the above-proven proposition, the node voltage deterministic constraint is equivalently transformed to:

[0119] (27)

[0120] Where, 、 is the dual variable related to the voltage constraint;

[0121] Substituting Equation (4) into Equation (17), based on the above-proven proposition, the energy storage output deterministic constraint is equivalently transformed to:

[0122] (28)

[0123] Where, 、 is the dual variable related to the energy storage power constraint;

[0124] Substituting Equation (19) into Equation (18), based on the above-proven proposition, the energy storage capacity deterministic constraint is equivalently transformed to:

[0125] (29)

[0126] Where, is the energy storage self-discharge coefficient; is the initial capacity of energy storage; is the dual variable related to the energy storage capacity constraint.

[0127] Preferably, the confidence level of the chance constraint It is 95%.

[0128] Furthermore, in said S4, auxiliary variables are introduced to linearly reconstruct the upper-level electricity purchase and sales cost function and the energy storage operation cost function;

[0129] 1) Introducing auxiliary variables The upper level electricity purchase and sales cost function is linearized and reconstructed as:

[0130] (30)

[0131] 2) Introducing auxiliary variables The energy storage operation cost function is linearized and reconstructed as:

[0132] (31).

[0133] Furthermore, in S4, the linearization of the converter capacity constraint contains quadratic terms, and the circular feasible interval of the VSC capacity constraint is approximately simplified to an inscribed regular dodecagonal feasible interval, which is expressed as a standard straight line equation of each side length, as shown in the following formula:

[0134] (32)

[0135] Where, 、 and The first The coefficient of the line equation for the side.

[0136] Furthermore, in S5, the equivalent linearized microgrid group optimization control model obtained is:

[0137] (33)

[0138] (34)

[0139] Where, A set of optimization variables, including: , as well as the auxiliary variables and dual multipliers introduced in the equivalent transformation process.

[0140] Compared with traditional control methods, the uncertainty absorption and load balancing method of the flexible interconnected microgrid group provided by the present invention introduces a distributed robust opportunity constraint optimization method driven by historical data. While improving the solution efficiency, it effectively reduces the problem of overly conservative robust optimization, realizes the load balancing within the microgrid group, efficient regulation of energy storage and VSC, and complete absorption of wind power and photovoltaics. In addition, it reduces the adverse effects of fluctuations in renewable energy power generation on the microgrid group and the upper distribution network, and realizes the economic, safe and stable operation of the flexible interconnected microgrid group. Not only that, the present invention also optimizes the distributed robust opportunity constraint method of the main grid by utilizing the duality principle, so that the uncertainty source in the system of the present invention can be of any dimension, which greatly simplifies the difficulty of solving the system model. BRIEF DESCRIPTION OF THE DRAWINGS

[0141] Figure 1 This is a flow chart of the uncertainty absorption and load balancing method of the flexible interconnected microgrid group involved in the present invention;

[0142] Figure 2 This is a basic topological diagram of a flexible interconnected microgrid group in an embodiment of the present invention;

[0143] Figure 3 This is a normalized output curve diagram of wind power, photovoltaic power, AC and DC loads in the embodiment of the present invention;

[0144] Figure 4 Schematic diagram of the total uncertainty remaining in the microgrid group system in an embodiment of the present invention;

[0145] Figure 5 Schematic diagram of uncertainty mitigation ratio of each uncertainty source at different times based on the RO method in an embodiment of the present invention;

[0146] Figure 6 Schematic diagram of uncertainty mitigation ratio of each uncertainty source at different times based on the DRCO method in an embodiment of the present invention;

[0147] Figure 7 It is a graph of the upper purchase and sale fluctuation curve of the microgrid group system area 2 in the embodiment of the present invention;

[0148] Figure 8 Schematic diagram of DC bus voltage fluctuation in microgrid group system area 2 under the WU method in an embodiment of the present invention;

[0149] Figure 9 Schematic diagram of DC bus voltage fluctuation in microgrid group system area 2 under the RO method in an embodiment of the present invention;

[0150] Figure 10 Schematic diagram of DC bus voltage fluctuation in microgrid group system area 2 under the DRCO method in an embodiment of the present invention;

[0151] Figure 11 This is a comparison diagram of the total operating cost distribution under the three methods in the embodiment of the present invention;

[0152] Figure 12 Schematic diagram of the output of energy storage in Area 3 under the RO and DRCO methods in an embodiment of the present invention;

[0153] Figure 13 Schematic diagram of the load conditions of each substation in a microgrid group under different scenarios in an embodiment of the present invention. DETAILED DESCRIPTION

[0154] In order to facilitate understanding by those skilled in the art, the present invention will be further described below with reference to embodiments and drawings. The contents mentioned in the embodiments are not intended to limit the present invention.

[0155] The process of a flexible interconnected microgrid group uncertainty absorption and load balancing method is as follows: Figure 1 The specific steps are as follows:

[0156] S1, proposes a two-stage real-time adjustment strategy for energy storage.

[0157] In a low-voltage flexible interconnected AC / DC microgrid, wind power, photovoltaic power, DC loads, etc. are centrally connected to the DC busbars of each substation. The uncertainty caused by the output prediction error will have an adverse impact on the distribution network.

[0158] 1. Define the output mode of uncertainty source

[0159] To quantify the uncertainty of wind power, photovoltaic power, and DC loads in a microgrid cluster and develop strategies to reduce uncertainty, the output of these uncertainty sources at each node is specified as follows:

[0160] (1)

[0161] (2)

[0162] (3)

[0163] Where, express The actual output column vector of each uncertainty source at each moment; Represents the uncertainty source-node correlation matrix, including the photovoltaic-node correlation matrix , Wind power-node correlation matrix , DC load-node correlation matrix ,These matrices are known by the location information parameters of each wind power, photovoltaic and load access microgrid node; express The predicted values ​​of the sources of uncertainty at each moment, including Photovoltaic forecast values ​​at each moment , wind power forecast values , DC load forecast value ,These values ​​are obtained from the output and normalized output curves of wind power, photovoltaic, and DC loads; is the prediction error column vector of each uncertainty source, representing the uncertainty;

[0164] 2. Energy storage control strategy

[0165] The energy storage output of a microgrid group is divided into day-ahead scheduling and real-time adjustment: day-ahead scheduling is to ensure the safe, stable and economical operation of the microgrid group system, and real-time adjustment is to suppress the adverse effects of uncertainty. The output of energy storage at each node can be defined as:

[0166] (4)

[0167] Where, express Real-time output of each energy storage at all times; represents the energy storage-node association matrix, which is obtained from the location information parameters of the energy storage connected to the microgrid node; It is the benchmark for energy storage’s day-ahead output; To adjust the factor matrix in real time, its specific definition is:

[0168] (5)

[0169] Where, Middle Row, No. Elements of a column , indicating the The energy storage device The proportion of uncertainty reduction of each uncertainty source, is the number of energy storage devices, is the number of uncertainty sources, The elements in the table satisfy the following conditions:

[0170] (6)

[0171] (7)

[0172] Formula (6) indicates that the proportion of each energy storage to mitigate each uncertainty is less than 100%, that is, overcompensation of uncertainty is not allowed; Formula (7) indicates Moment Uncertainty The total amount of compensation by all stored energy is less than or equal to the uncertainty itself.

[0173] S2, construct a microgrid group optimization control model based on distributed robust opportunity constraints.

[0174] In order to achieve economical, safe and stable operation of microgrid clusters, balance the load of each substation and reduce the adverse effects of uncertainty on microgrid clusters, an optimization and control model of microgrid clusters based on distributed robust chance constraint method is constructed.

[0175] 1. Objective Function

[0176] The objective function of the microgrid group optimization control model is:

[0177] (8)

[0178] Where, is the optimization variable; They are the upper level electricity purchase and sales costs, energy storage operation costs, load balancing penalty costs, and uncertainty penalty costs;

[0179] 1) Cost of electricity purchase and sales from higher authorities

[0180] When the microgrid group has excess energy, it can sell it to the distribution network at a lower price; when the energy is insufficient, it can purchase electricity from the upper distribution network; therefore, the cost of purchasing and selling electricity is for:

[0181] (9)

[0182] Where, is the set of all scheduling times, is the numbered set of all nodes in the microgrid; and They are respectively the electricity purchase cost (i.e. time-of-use electricity price) and the electricity selling price (i.e. the price at which the microgrid group sells electricity to the distribution network); For the Power interaction between each node and the upper distribution network; Operator satisfy ;

[0183] 2) Energy storage operating costs

[0184] In a microgrid group, energy storage operation requires a certain cost :

[0185] (10)

[0186] Where, It is the numbered set of all energy storage in the microgrid; Output operating costs for energy storage units; Provide active power for each energy storage;

[0187] 3) Load-sharing penalty costs

[0188] In order to alleviate the operating pressure of transformers in each microgrid group area, the load penalty cost of each node in the microgrid group is set as:

[0189] (11)

[0190] Where, is the load-sharing penalty coefficient, The numerical setting of is to obtain a good load balancing effect during the debugging simulation process. Take it as 1. If you need a better load balancing effect, you can gradually achieve greater; 、 They are respectively 、 Transformer capacity of each node / station; For the The amount of power interaction between each node and the upper distribution network; is the number of nodes / stations in the microgrid group;

[0191] 4) Uncertainty penalty cost

[0192] In order to alleviate the adverse effects of uncertainty on each substation, the uncertainty penalty cost is defined for:

[0193] (12)

[0194] (13)

[0195] (14)

[0196] Where, is the uncertainty penalty coefficient, It is a very large number. In order to reduce the remaining uncertainty in the system, a better value for optimization results obtained during the debugging simulation process is 10^6. A numbered set representing all sources of uncertainty; Indicates the The source of uncertainty is The uncertainty distribution fuzzy set of the moment; Represents the maximum expected value under all possible probabilities; is a column vector The elements, , its physical meaning is that all energy storage in the microgrid group The proportion of uncertainty generated by each uncertainty source absorbed; Represents a column vector of all 1s with a dimension of n×1;

[0197] 2. Construct opportunity constraints

[0198] The flexible interconnected microgrid referred to in this invention is formed by interconnecting multiple substations through VSCs, forming a low-voltage AC / DC distribution network. New energy equipment such as wind power, photovoltaics, and energy storage are directly integrated into the DC bus. Therefore, prediction errors from uncertain sources such as wind power, photovoltaics, and DC loads can cause fluctuations in DC bus voltage, tributary line transmission power, and energy storage output. The risk of these fluctuations exceeding certain limits needs to be limited. Therefore, network security flow opportunity constraints and energy storage safe operation opportunity constraints are established for the microgrid cluster.

[0199] 1) Line transmission power opportunity constraints

[0200] In order to fully utilize energy storage resources and limit line transmission power to a safe range, the line transmission power at each moment must meet the safe transmission constraint under a certain confidence level, namely:

[0201] (15)

[0202] Where, Represents the set of all probability distributions that an uncertain quantity may satisfy; Indicates the minimum probability that the formula in the brackets is true; Indicates The DC line transmission power vector of the microgrid group at time t is, 、 are the minimum and maximum transmission power vectors of the DC line respectively; Indicates confidence;

[0203] 2) Node voltage opportunity constraint

[0204] The voltage of each node at each moment must meet the safety constraints under a certain confidence level, namely:

[0205] (16)

[0206] Where, Indicates The DC bus voltage vector of the microgrid group at time , 、 are the maximum and minimum values ​​of the DC bus voltage respectively;

[0207] 3) Energy storage charging and discharging power opportunity constraints

[0208] Under a certain confidence level, the charging and discharging power of each energy storage at each moment must meet the energy storage safe operation constraints, namely:

[0209] (17)

[0210] Where, Indicates The output column vector of the energy storage at each moment, which stipulates that the energy storage discharge is in the positive direction; 、 are the maximum and minimum charging and discharging power of energy storage respectively;

[0211] 4) Energy storage capacity security opportunity constraints

[0212] Under a certain confidence level, the capacity of each energy storage at each moment must meet the energy storage safe operation constraints, namely:

[0213] (18)

[0214] Where, Indicates The energy storage capacity vector of each microgrid group at the moment; 、 They are the lower and upper limits of energy storage capacity respectively.

[0215] 3. Construct other constraints

[0216] 1) Energy storage capacity constraints

[0217] The capacity safety constraint of energy storage capacity at each moment is:

[0218] (19)

[0219] Where, Charging efficiency for energy storage; is the energy storage self-discharge coefficient; Indicates the daily scheduling interval, set to 1;

[0220] 2) Network flow constraints

[0221] The relationship between node injection power, network line transmission power, and network node voltage is expressed as follows:

[0222] (20)

[0223] (twenty one)

[0224] (twenty two)

[0225] Where, represents the node injection power; represents the active power injected by the VSC; is the augmented node-line association matrix, which is obtained from the microgrid topology diagram, i.e., the substation connection relationship; is the augmented line resistance matrix, which is obtained by multiplying the unit resistance of the DC line by the distance between the substation and the line; represents the DC bus voltage column vector of the microgrid group; is the voltage reference;

[0226] 3) VSC capacity constraints

[0227] The VSC operation meets the capacity safety constraints:

[0228] (twenty three)

[0229] Where, 、 Respectively expressed in Moment Active power injected and reactive power output by VSC at each node; For the The capacity of the VSC on each node;

[0230] 4) AC area power balance constraints

[0231] The AC area node is the power interaction node between the microgrid and the distribution network, meeting the power balance:

[0232] (twenty four)

[0233] Where, represents the column vector of the power purchased and sold by the upper level, is the AC load column vector of the AC node.

[0234] S3 uses the Wasserstein Metric definition and duality theorem to linearly reconstruct the uncertainty penalty cost in the model objective function and transform the model's chance constraints into deterministic constraints.

[0235] 1. Reconstruction of uncertainty penalty function based on distributed robustness method.

[0236] Among the distributed robustness problems, Wasserstein's Metrics has attracted widespread attention due to its simplicity and intuitiveness. Starting from its basic definition, this paper uses the duality theorem to gradually simplify the high-dimensional and difficult-to-solve equation (12) into a linearly solvable problem.

[0237] 1) Get the radius of the Wasserstein sphere

[0238] Collect a series of random variables The historical sample data set, that is, the historical uncertainty fluctuation data set of the uncertainty source, has a sample size of M ; After getting the random variable Before the distribution of fuzzy sets, the fuzzy set radius size It can be determined by the following formula:

[0239] (35)

[0240] Where, Indicates the diameter of the data support set; To include random variables Confidence of all distributions of ;

[0241] 2) Problem transformation

[0242] Based on the Wasserstein Metric definition, the original objective function Transformed into a dual problem:

[0243] (36)

[0244] Where, is the dual multiplier; is an auxiliary variable, is the sample number set; Indicates uncertainty The feasible set of

[0245] The constraints contained in formula (36) face the problem of scalability, that is, as the amount of sample data gradually increases, the scale of the solution gradually expands; therefore, the supremum approximation is introduced, and formula (36) is further simplified to:

[0246] (37)

[0247] (25)

[0248] Where, 、 Indicates Moment The maximum and minimum values ​​of the historical data of each uncertainty source; is the radius of the fuzzy set, determined by formula (35); represents the number of samples; Indicates the uncertainty at each moment Middle A historical sample data set of components, which is generated based on the Gaussian distribution function;

[0249] 2. Reconstruct probability constraints based on distributed robust chance constraints.

[0250] The transformation and implementation idea of ​​the chance constraint proposed in this invention is: as the data-driven historical data set contains a small number of extreme scenarios or erroneous noise points, a distribution robust method based on Wasserstein Metric is adopted to obtain a box uncertainty set of sample points with a certain confidence level, and the original chance constraint problem is transformed into a robust problem under this set.

[0251] 1) Normalize all historical sample points to the standard coordinate system

[0252] Calculate the random variables at each moment Historical sample data set The mean ,variance , and all sample points Normalize to the standard coordinate system:

[0253] (38)

[0254] Where, Indicates the coordinates of the sample point in the standard coordinate system;

[0255] 2) Get the side length of the hypercube in the standard coordinate system

[0256] Solve the problem of minimizing the side length of the hypercube and obtain a certain confidence level in the standard coordinate system Lower envelope Hypercube of sample points at time Side length :

[0257] (39)

[0258] Where, is the dual variable, is the sample point in the standard coordinate system The radius of the fuzzy set is obtained according to formula (35); is the set of coordinates of the vertices of the hypercube, which is further described as:

[0259] (40)

[0260] Where, The dimension size is A column vector of all 1s;

[0261] 3) Denormalize the hypercube to the original coordinate system

[0262] Denormalize the obtained hypercube to the original coordinate system and obtain a certain confidence level in the original coordinate system Lower envelope Parallel polyhedron of sample points at the moment , its vertex can be obtained by substituting formula (40) into the following formula:

[0263] (41)

[0264] is the set containing the coordinates of the vertices of the parallelepiped, which is further described as the following box uncertainty set:

[0265] (42)

[0266] This equation (42) describes the parallel polyhedron Internal space, is a constant matrix, whose Rank The element of the column is 1, Rank The element of the column is -1, The rest of the elements are 0; is a constant column vector, Row and The rows are Middle The absolute values ​​of the coordinates of the vertices, ;

[0267] 4) Proof of Proposition

[0268] Set variable , all chance-constrained problems in the model can be described in a unified form:

[0269] (43)

[0270] The above formula always holds true, that is, the maximum value on the left side of the inequality is less than or equal to 0:

[0271] (44)

[0272] To solve the problem of finding the maximum value on the left side of the inequality (44), the duality theorem is used to further transform (44) into:

[0273] (45)

[0274] Therefore, formula (43) can be equivalently transformed into the following proposition:

[0275] (46)

[0276] 5) Transforming Chance Constraints into Deterministic Constraints

[0277] Substituting Equations (20) and (21) into the chance constraint Equation (15), we can obtain the following robust problem:

[0278] (47)

[0279] Based on the above-proven proposition, Equation (47) can be equivalently transformed into the following deterministic constraint on line transmission capacity:

[0280] (26)

[0281] Where, 、 is the dual variable related to the power capacity constraint;

[0282] Similarly, by substituting Equations (20) and (22) into Equation (16), based on the above-proven proposition, the node voltage deterministic constraint is equivalently transformed to:

[0283] (27)

[0284] Where, 、 is the dual variable related to the voltage constraint;

[0285] Substituting Equation (4) into Equation (17), based on the above-proven proposition, the energy storage output deterministic constraint is equivalently transformed to:

[0286] (28)

[0287] Where, 、 is the dual variable related to the energy storage power constraint;

[0288] Substituting Equation (19) into Equation (18), based on the above-proven proposition, the energy storage capacity deterministic constraint is equivalently transformed to:

[0289] (29)

[0290] Where, is the energy storage self-discharge coefficient; is the initial capacity of energy storage; is the dual variable related to the energy storage capacity constraint.

[0291] It is worth mentioning here that the traditional control method can only transform the uncertainty source of the system into one-dimensional or two-dimensional, while the present invention optimizes the distributed robust chance constraint method of the main network through the above-mentioned duality principle, so that the uncertainty source in the system of the present invention can be of any dimension.

[0292] Preferably, the confidence level of the aforementioned chance constraint It is 95%.

[0293] S4, linearize and reconstruct other nonlinear parts of the model objective function and constraints.

[0294] 1. Reconstruction of the objective function of total electricity purchase and sales cost

[0295] 1) Reconstruction of power purchase and sales costs at higher levels

[0296] Original electricity purchase and sales cost For nonlinearity, by introducing auxiliary variables , can be transformed back into:

[0297] (30)

[0298] 2) Reconstructing energy storage operating costs

[0299] Original energy storage operating cost For nonlinearity, by introducing auxiliary variables , can be transformed back into:

[0300] (31).

[0301] 2. Linearizing constraints with quadratic terms

[0302] The circular feasible interval of the VSC capacity constraint can be approximately simplified to the inscribed regular dodecagonal feasible interval, that is, constraint (23) can be simplified to the standard straight line equation form of each side length:

[0303] (32)

[0304] Where, 、 and The first The coefficients of the straight line equation of the edge, the currently recognized coefficients are as follows Table 1:

[0305] Table 1 Coefficients of linear equations

[0306] ;

[0307] S5, organize and solve the equivalent linearized microgrid group optimization control model.

[0308] Based on the above derivation and transformation, the optimization model of flexible interconnected microgrid group based on distributed robust opportunity constraint is:

[0309] (33)

[0310] (34)

[0311] Where, A set of optimization variables, including: And the auxiliary variables and dual multipliers introduced in the equivalent transformation process.

[0312] Obtain the network topology, equipment and distribution network parameter information of the microgrid group, and collect 100 sets of historical data sample points of wind power and photovoltaic power output fluctuation data at each moment. Figure 2 As shown, the basic topology of a flexible interconnected microgrid cluster is a chain-type DC bus structure, divided into a low-voltage AC region and a low-voltage DC region. The former enables energy exchange with the upstream distribution network, while the latter connects to a large number of distributed renewable energy sources to ensure safe, economical and stable operation within the microgrid cluster. The microgrid cluster is connected to three energy storage devices, each with a maximum discharge capacity of 50kW, a maximum charging capacity of -50kW, and a maximum capacity of 200kWh. The upper and lower capacity limits are set at 0.2 and 0.9 of the maximum capacity, respectively. The initial capacity is 0.4 of the maximum capacity, the self-discharge coefficient is 0.998, and the charging efficiency is 100%. A 40kW DC load is connected to the DC bus of all nodes. Figure 3 The normalized output curves for wind power, photovoltaic power, and AC / DC loads are shown. Node 1's DC bus is connected to a 90kW photovoltaic system and a 60kW wind turbine; node 2 is connected to a 300kW photovoltaic system; node 3 is connected to a 150kW photovoltaic system and a 150kW wind turbine; and node 4 is connected to a 120kW photovoltaic system. The unit resistance of the DC line is set to 0.0754Ω / km. The distances between substations 1-4 are 2.5km, 1.2km, and 1.6km, respectively. The VSCs connected to substations 1-4 have capacities of 100kVA, 500kVA, 500kVA, and 100kVA, respectively. The distribution transformer capacities of each substation are the same as the VSC capacity. The connected AC loads are 90kW, 420kW, 420kW, and 60kW, respectively. The maximum transmission capacity of each line is set to 300kW. The upper and lower node voltage limits are set to 1.05 and 0.95, respectively; the reference node voltage is set to 1. The time-of-use electricity price is set as follows: peak electricity price is 0.14$ / kWh (18:00-22:00), off-peak electricity price is 0.06$ / kWh (11:00-15:00), and normal electricity price is set at 0.10$ / kWh; the microgrid group sells electricity to the distribution network at 0.04$ / kWh, and the operating cost of energy storage is set at 0.015$ / kWh.

[0313] To verify the effectiveness of the proposed method, 1000 sets of scenario data were generated based on the Gaussian distribution function with a mean of 0 and a variance of 10% of the uncertainty source output. In the MATLAB environment on the PC side, the equivalent linearized microgrid group optimization control model obtained by YALMIP was called by the CPLEX solver to solve the control parameters of the flexible interconnected microgrid system, which mainly include the day-ahead output benchmark and real-time adjustment factor of each energy storage in the microgrid group, as well as the upper-level purchase and sales power of each substation, that is, the optimization variables 、 、 , based on these key control parameters, the flexible interconnected microgrid group is optimized and controlled, thereby achieving load balancing within the microgrid group, efficient control of energy storage and VSC, and complete absorption of wind power and photovoltaic power.

[0314] In order to verify the superiority of the proposed method (hereinafter referred to as DRCO), the simulation results of this method are compared with those of two other traditional benchmark methods, namely, the robust optimization method (hereinafter referred to as RO) and the uncertainty-free optimization method (hereinafter referred to as WU). Figure 4 The total uncertainty remaining in the microgrid system is shown. WU does not consider the uncertainty absorption, so Figure 4 The residual uncertainty shown in the WU method is the original total uncertainty generated by each uncertainty source in the system. Due to limited energy storage resources, both the RO method and the proposed method can absorb some uncertainty, but the proposed method is more effective in absorbing uncertainty than the RO method. This is because the energy storage in the RO method is primarily used to ensure the safe and stable operation of the microgrid group system, such as maintaining the DC bus within a safe range, leaving little residual capacity to absorb uncertainty. In contrast, the proposed method relaxes the constraints on parameters such as voltage, allowing BESs to absorb more uncertainty.

[0315] Figure 5 and Figure 6 The uncertainty mitigation of each uncertainty source at each moment in the RO and DRCO methods is further illustrated. Over time, the RO method's ability to absorb uncertainty diminishes as energy storage capacity is exhausted. In contrast, DRCO maintains high uncertainty absorption performance during periods of high uncertainty, such as when wind power (at night) and photovoltaic power (at midday) generation are high. This ensures that the remaining uncertainty in the microgrid cluster system remains relatively low. Figure 7 The fluctuation of energy interaction between substation 2 and the upper distribution network is further illustrated. It is obvious that the DRCO method provides the best performance in smoothing the fluctuation of purchased / sold electricity compared with other methods.

[0316] Figure 8 、 Figure 9 、 Figure 10 The DC bus voltage fluctuations in Substation 2 are shown for the WU, RO, and DRCO methods, respectively. The WU method exhibits a high percentage of voltage exceeding the upper limit. RO ensures that the voltage remains within the safe range in all cases, but the voltage distribution is primarily around 1 p.u., making the voltage optimization overly robust and conservative, limiting the potential for energy storage reuse. In contrast, DRCO only rarely exceeds the upper limit, improving the operational flexibility of energy storage while ensuring system voltage safety with a high probability.

[0317] Table 2 shows the average total cost and components of a microgrid cluster system under the WU, RO, and DRCO methods. Affected by internal system uncertainty, the distribution of purchase and sale costs reflects the characteristics of residual uncertainty, and the real-time output of energy storage is also closely related to uncertainty. Therefore, this paper uses the 95% confidence level value-at-risk (VaR) of the real-time output of energy storage in all scenarios as a metric for the real-time regulation cost of energy storage. Because residual uncertainty within the system can adversely impact the upper-level distribution network, the system must internally absorb this uncertainty as much as possible. However, given limited internal energy storage resources, all three methods require a certain penalty cost to the distribution network. As shown in Table 2, although the WU method has lower average purchase and sale costs than the RO and DRCO methods, its higher day-ahead energy storage output cost and uncertainty penalty cost ultimately result in the highest total cost. Furthermore, excluding the real-time energy storage regulation cost, the RO method has higher overall costs than the DRCO method. This is because the RO method uses worst-case optimization to ensure absolute system safety and stability. Therefore, under the RO method, the operation of energy storage is mainly based on the priority goal of ensuring system safety, but to a certain extent, it sacrifices the economic benefits of the microgrid group.

[0318] It's worth noting that Table 2 shows the actual operating costs incurred by optimizing the system using the DRCO, WU, and RO methods. These actual operating costs are not entirely consistent with the cost terms in the objective function of the proposed model, primarily because they serve different purposes. The costs in the objective function are intended to optimize decision-making and are not equivalent to actual operating costs. Specifically, the present invention incorporates energy storage costs, electricity purchase and sales costs, and load-sharing penalty costs into the objective function to guide the model in making more reasonable scheduling decisions in the face of uncertainty. For example, in real life, the power grid does not charge users for load-sharing penalty costs. The inclusion of these terms in the objective function is intended to guide power distribution and capacity sharing among substations. In the converted equivalent model, these uncertainties are eliminated. However, in the simulation, 1,000 uncertainties were randomly generated using a Gaussian function to simulate actual operational uncertainty. Therefore, the "mean" (a statistical quantity) is added to the electricity purchase cost. Therefore, these cost terms are not intended to directly reflect the economic expenditures of real-world operation, but rather to serve the model solution process. However, the simulation phase of this embodiment introduces actual uncertainty samples to evaluate the performance of the scheduling strategy. Although the present invention equivalently handles uncertainty through the distributed robust chance constraint method when solving the model, in order to test the effectiveness of this scheduling strategy under real uncertainty conditions, this embodiment also generates 1000 sets of sample scenarios using Gaussian distribution to simulate the operation process. Finally, after obtaining the energy storage day-ahead output benchmark, real-time adjustment factor, and power purchase amount based on model optimization, the data shown in Table 2 is calculated based on the 1000 generated uncertainties and the actual operating costs of the distribution network:

[0319] Table 2 Total operating costs within the microgrid cluster

[0320] ;

[0321] Figure 11 The overall operating cost distribution of the microgrid system under the three methods is further demonstrated. Figure 11 It can be observed that the expected cost of the DRCO method is significantly lower than that of the RO and WU methods. In addition, since the DRCO method minimizes the residual uncertainty within the system, its total cost distribution is more concentrated.

[0322] In order to further reveal the difference in output cost of energy storage systems under RO and DRCO methods, Figure 12 The day-ahead and real-time output regulation curves of the energy storage connected to the DC area of ​​​​the station area 3 are given. Figure 12As can be clearly seen in the figure, under the RO method, to ensure system safety and stability (especially voltage safety), energy storage charging and discharging are primarily scheduled during off-peak and off-valley electricity price periods. However, due to the large amount of energy storage capacity occupied during the day-ahead period, its ability to adjust to the uncertainty of real-time renewable energy output is significantly reduced. Therefore, the intraday regulation range of energy storage under the RO method is generally lower than that of the DRCO method. In contrast, the DRCO method, by appropriately relaxing constraints, not only effectively mitigates the impact of uncertainty but also employs an arbitrage strategy of charging during the low-price period (11:00) and discharging during the high-price period (21:00), thereby reducing the purchase and sale costs of electricity from the upper distribution network and optimizing the overall system revenue. Therefore, while allowing for slight deviations in the DC bus voltage within the microgrid cluster, the DRCO method fully exploits the multifunctional reuse potential of the energy storage system, ensuring the economic efficiency and stability of the microgrid cluster. Overall, the DRCO method significantly reduces system conservatism (RO method) and instability (WU method), effectively suppressing the transmission of uncertainty to the upstream distribution network.

[0323] Figure 13 The load rate distribution of each node in the microgrid system is shown when considering (scenario A, i.e. the method of the present invention) and not considering (scenario B) the balanced load target. Figure 13 As shown, the method of the present invention effectively achieves power balance and capacity sharing among different substations.

[0324] The above embodiments are preferred implementation schemes of the present invention. In addition, the present invention can also be implemented in other ways. Any obvious replacement without departing from the concept of the present technical solution is within the scope of protection of the present invention.

[0325] In order to make it easier for ordinary technicians in this field to understand the improvements of the present invention over the prior art, some drawings and descriptions of the present invention have been simplified, and for the sake of clarity, some other elements are omitted in this application document. Ordinary technicians in this field should realize that these omitted elements may also constitute the content of the present invention.

Claims

1. A method for uncertainty absorption and load balancing of a flexible interconnected microgrid group, characterized in that: include: S1: quantify the uncertainty of wind power, photovoltaic power, and DC loads in the microgrid group and determine a two-stage control strategy for energy storage day-ahead scheduling and real-time adjustment; S2, with the goal of minimizing the sum of the upper-level power purchase and sales costs, energy storage operation costs, load-sharing penalty costs, and uncertainty penalty costs, constructs a microgrid group optimization control model based on distributed robust opportunity constraints. The constraints of this model include opportunity constraints and other constraints. The opportunity constraints include line transmission power opportunity constraints, node voltage opportunity constraints, energy storage charging and discharging power opportunity constraints, and energy storage capacity security opportunity constraints. The other constraints include energy storage capacity constraints, network flow constraints, VSC capacity constraints, and AC regional power balance constraints. S3, uses the Wasserstein Metric definition and duality theorem to linearize and reconstruct the uncertainty penalty cost in the model objective function, and transforms the model's chance constraints into deterministic constraints; S4, linearize and reconstruct other nonlinear parts of the model objective function and constraints; S5, obtain the network topology, equipment and distribution network parameter information of the microgrid group, collect several groups of historical wind power and photovoltaic output fluctuation data sample points at each moment, organize the equivalent linearized microgrid group optimization control model and solve it, obtain the control parameters, and optimize the control of the flexible interconnected microgrid group.

2. The method for uncertainty elimination and load balancing of a flexible interconnected microgrid group according to claim 1 is characterized in that: In S3, when linearizing and reconstructing the uncertainty penalty function using the Wasserstein Metric definition and the duality theorem, we first collect a series of historical sample data sets of random variables as historical uncertainty fluctuation data sets of the uncertainty source. Then, before obtaining the distribution fuzzy set of the random variable, the radius of the Wasserstein sphere is determined according to the fuzzy set radius size formula. Finally, based on the Wasserstein Metric definition, the uncertainty penalty function is transformed into a dual problem, and the supremum approximation is introduced to simplify it.

3. The method for uncertainty elimination and load balancing of a flexible interconnected microgrid group according to claim 2 is characterized in that: In S3, when the Wasserstein Metric definition and duality theorem are used to transform the opportunity constraints of the microgrid group optimization control model into deterministic constraints, the mean and variance of the historical sample data set of random variables at each moment are first calculated, and all sample points are normalized to the standard coordinate system. Then, the problem of minimizing the side length of the hypercube is solved to obtain the envelope under a certain confidence level in the standard coordinate system. t The side length of the hypercube of the sample point at the moment is determined, and the corresponding hypercube is determined. Then the hypercube is denormalized to the original coordinate system to obtain the envelope under a certain confidence level in the original coordinate system. t The parallel polyhedron of the sample points at each moment is constructed, and the corresponding box uncertainty set is determined. Finally, the duality theorem is used to transform the original chance constraint problem into a robust problem under the box uncertainty set, and then: Substituting the network power flow constraint into the line transmission power opportunity constraint, the line transmission power deterministic constraint is equivalently converted. Substitute the network power flow constraint into the node voltage opportunity constraint, and the node voltage deterministic constraint is equivalently converted; Substituting the energy storage output into the energy storage charging and discharging power opportunity constraint, the energy storage output deterministic constraint is obtained through equivalent transformation. Substituting the energy storage capacity constraint into the energy storage capacity safety opportunity constraint, the energy storage capacity deterministic constraint is obtained through equivalent transformation.

4. The method for uncertainty elimination and load balancing of a flexible interconnected microgrid group according to claim 3 is characterized by: The control parameters include the day-ahead output benchmark and real-time adjustment factor of each energy storage in the microgrid group, as well as the power purchase and sales volume of each substation.

5. The method for uncertainty elimination and load balancing of a flexible interconnected microgrid group according to claim 4 is characterized in that: The S1 quantifies the uncertainty of wind power, photovoltaic power, and DC loads of the microgrid group as follows: （1） （2） （3） Where, express The actual output column vector of each uncertainty source at each moment; Represents the uncertainty source-node correlation matrix, including the photovoltaic-node correlation matrix , Wind power-node correlation matrix , DC load-node correlation matrix ; express The predicted values ​​of the sources of uncertainty at each moment, including Photovoltaic forecast values ​​at each moment , wind power forecast values , DC load forecast value ; is the prediction error column vector of each uncertainty source, representing the uncertainty; In S1, the energy storage output of the microgrid group is divided into day-ahead scheduling and real-time adjustment. The output of energy storage at each node is as follows: （4） Where, express Real-time output of each energy storage at all times; represents the energy storage-node association matrix, It is the benchmark for energy storage’s day-ahead output; To adjust the factor matrix in real time, its specific definition is: （5） Where, Middle Row, No. Elements of a column , indicating the The energy storage device The proportion of uncertainty reduction of each uncertainty source, is the number of energy storage devices, is the number of uncertainty sources, The elements in the table satisfy the following conditions: （6） （7） Formula (6) indicates that the proportion of each energy storage to alleviate each uncertainty is less than 100%; Formula (7) indicates Moment Uncertainty The total amount of compensation by all stored energy is less than or equal to the uncertainty itself.

6. The method for uncertainty elimination and load balancing of a flexible interconnected microgrid group according to claim 5, characterized in that: The microgrid group optimization control model constructed by S2 based on distributed robust opportunity constraints is: （8） Where, is the optimization variable; They are the upper level electricity purchase and sales costs, energy storage operation costs, load balancing penalty costs, and uncertainty penalty costs; （9） Where, is the set of all scheduling times, is the numbered set of all nodes in the microgrid; and are electricity purchase cost and electricity selling price respectively; For the Power interaction between each node and the upper distribution network; Operator satisfy ; （10） Where, It is the numbered set of all energy storage in the microgrid; Output operating costs for energy storage units; Provide active power for each energy storage; （11） Where, is the load-sharing penalty coefficient; 、 They are respectively 、 Transformer capacity of each node; For the The amount of power interaction between each node and the upper distribution network; is the number of nodes in the microgrid group; （12） （13） （14） Where, is the uncertainty penalty coefficient; A numbered set representing all sources of uncertainty; Indicates the The source of uncertainty is The uncertainty distribution fuzzy set of the moment; Represents the maximum expected value under all possible probabilities; is a column vector The elements, , its physical meaning is that all energy storage in the microgrid group The proportion of uncertainty generated by each uncertainty source absorbed; represents a column vector of all 1s with a dimension of n×1; The opportunity constraints of the microgrid group optimization control model constructed in S2 based on distributed robust opportunity constraints include: 1) Line transmission power opportunity constraints （15） Where, Represents the set of all probability distributions that an uncertain quantity may satisfy; Indicates the minimum probability that the formula in the brackets is true; Indicates The DC line transmission power vector of the microgrid group at time t is, 、 are the minimum and maximum transmission power vectors of the DC line respectively; Indicates confidence; 2) Node voltage opportunity constraint （16） Where, Indicates The DC bus voltage vector of the microgrid group at time , 、 are the maximum and minimum values ​​of the DC bus voltage respectively; 3) Energy storage charging and discharging power opportunity constraints （17） Where, Indicates The output column vector of the energy storage at each moment, which stipulates that the energy storage discharge is in the positive direction; 、 are the maximum and minimum charging and discharging power of energy storage respectively; 4) Energy storage capacity security opportunity constraints （18） Where, Indicates The energy storage capacity vector of each microgrid group at the moment; 、 are the lower and upper limits of energy storage capacity respectively; Other constraints of the microgrid group optimization control model constructed in S2 based on distributed robust opportunity constraints include: 1) Energy storage capacity constraints （19） Where, Charging efficiency for energy storage; is the energy storage self-discharge coefficient; Indicates the daily scheduling interval, set to 1; 2) Network flow constraints （20） （21） （22） Where, represents the node injection power; represents the active power injected by the VSC, is the augmented node-link association matrix, is the augmented line resistance matrix; represents the DC bus voltage column vector of the microgrid group; is the voltage reference; 3) VSC capacity constraints （23） Where, 、 Respectively expressed in Moment Active power injected and reactive power output by VSC at each node; For the The capacity of the VSC on each node; 4) AC area power balance constraints （24） Where, represents the column vector of the power purchased and sold by the upper level, is the AC load column vector of the AC node; In S3, the Wasserstein Metric definition and the duality theorem are used to linearize and reconstruct the uncertainty penalty function as follows: （25） Where, is the dual multiplier; is the radius of the fuzzy set; 、 Indicates Moment The maximum and minimum values ​​of the historical data of each uncertainty source; represents the number of samples; Indicates the uncertainty at each moment Middle A historical sample dataset of components; In S3, the Wasserstein Metric definition and the duality theorem are used to transform the opportunity constraints of the microgrid group optimization control model into deterministic constraints, as follows: Substituting Equations (20) and (21) into the opportunity constraint Equation (15), the equivalent transformation results in the deterministic constraint of line transmission power: （26） Where, 、 is the dual variable related to the power capacity constraint; Substituting Equation (20) and Equation (22) into Equation (16), the equivalent transformation results in the node voltage deterministic constraint: （27） Where, 、 is the dual variable related to the voltage constraint; Substituting Equation (4) into Equation (17), the equivalent transformation results in the energy storage output deterministic constraint: （28） Where, 、 is the dual variable related to the energy storage power constraint; Substituting Equation (19) into Equation (18), the energy storage capacity deterministic constraint is equivalently transformed to: （29） Where, is the energy storage self-discharge coefficient; is the initial capacity of energy storage; is the dual variable related to the energy storage capacity constraint; A parallel polyhedron The constant matrix of Rank The element of the column is 1, Rank The element of the column is -1, The rest of the elements are 0; A parallel polyhedron The constant column vector of Row and The rows are Middle The absolute values ​​of the coordinates of the vertices, .

7. The method for uncertainty elimination and load balancing of a flexible interconnected microgrid group according to claim 6, characterized in that: The confidence level of the chance constraint It is 95%.

8. The method for uncertainty elimination and load balancing of a flexible interconnected microgrid group according to claim 7, characterized in that: In said S4, auxiliary variables are introduced to linearly reconstruct the upper-level electricity purchase and sales cost function and the energy storage operation cost function; 1) Introducing auxiliary variables The upper level electricity purchase and sales cost function is linearized and reconstructed as: （30） 2) Introducing auxiliary variables The energy storage operation cost function is linearized and reconstructed as: （31）。 9. The method for uncertainty elimination and load balancing of a flexible interconnected microgrid group according to claim 8, characterized in that: In S4, the VSC capacity constraint containing quadratic terms is linearized, and the circular feasible interval of the VSC capacity constraint is approximately simplified to an inscribed regular dodecagonal feasible interval, which is expressed as the standard straight line equation form of each side length, as shown in the following formula: （32） Where, 、 and The first The coefficient of the line equation for the side.

10. The method for uncertainty elimination and load balancing of a flexible interconnected microgrid group according to claim 9, characterized in that: In S5, the equivalent linearized microgrid group optimization control model obtained is: （33） （34） Where, A set of optimization variables, including: And the auxiliary variables and dual multipliers introduced in the equivalent transformation process.

Images