Power distribution network service provider-multi-microgrid distribution robust optimization cooperative scheduling method considering multiple uncertainty

Abstract

The invention discloses a power distribution network service provider-multi-microgrid distribution robust optimization cooperative scheduling method considering multiple uncertainty. The method comprises the following steps: constructing Wasserstein fuzzy set constraints of purchase and sale electricity price of a superior power grid and renewable output; constructing a distribution robust model of the Wasserstein distance of the power distribution and sale service provider; constructing a multi-microgrid system operation model; constructing a P2P transaction network game model between a power distribution network service provider and multiple microgrids based on Wasserstein distribution robust optimization; and solving the P2P transaction network game model between the power distribution network service provider and the multiple microgrids by combining a KKT, a duality principle, a large M method and a stationary point method hybrid solving method. According to the method, the P2P network game model between the power distribution network service provider and the multiple microgrids is established, benefits of all parties are balanced, new energy in the microgrids is fully consumed, and active participation of the service provider and the microgrids in electric carbon energy transaction is effectively mined.

Description

Technical Field

[0001] The present invention relates to the technical field of coordinated dispatching between multiple entities in a multi-microgrid system, and in particular to a distribution network service provider-multi-microgrid distributed robust optimization coordinated dispatching method taking into account multiple uncertainties. Background Art

[0002] In order to cope with the adverse effects of the increasing penetration rate of new energy on the safe and stable operation of the system and the quality of power, microgrid technology has emerged and gradually become an important way to locally absorb distributed new energy and reduce carbon emissions. However, for a single microgrid, considering the uncertainty of new energy output and the fluctuation of electricity prices in the power market, it will face complex scheduling and management problems during operation. To this end, related studies have proposed a multi-microgrid system based on peer-to-peer (P2P) transactions. The microgrids are connected through interconnection lines and can conduct P2P energy transactions, which can effectively deal with the problems faced by a single microgrid during operation. However, there are differences in composition structure, equipment capacity, etc. between different microgrids. At the same time, they are affected by uncertainty, and their optimization management is more difficult than that of a single microgrid. Therefore, how to use appropriate methods to coordinate and manage multi-microgrid systems is a problem that needs to be studied.

[0003] In the operation process of multi-microgrid system, with the rapid development of intelligent building technology and distributed resources, microgrids gradually have certain decision-making and management capabilities. Multi-microgrid system gradually develops towards the direction of multi-subject decision-making, which leads to the increase of complexity of system scheduling and operation. Reference [1]: "Optimal Scheduling of Integrated Energy Multi-Microgrid System Based on Hierarchical Constrained Reinforcement Learning" (Dong Lei, Yang Zimin, Qiao Ji, etc. "Optimal Scheduling of Integrated Energy Multi-Microgrid System Based on Hierarchical Constrained Reinforcement Learning". Transactions of China Electrotechnical Society, 2024, 39(05): 1436-1453.) For the interconnected multi-microgrid system, an optimal scheduling method based on hierarchical constrained reinforcement learning is proposed, which realizes the optimal scheduling and operation of the multi-microgrid system. At the same time, through the hierarchical architecture, the communication pressure of the system can be reduced and the privacy of each subject can be effectively protected. Reference [2]: A Hierarchical Game Approach for Optimization of Regional Integrated Energy System Clusters Considering Bounded Rationality (Dong Lei, Li Mengting, Hu Junjie, et al. “A Hierarchical Game Approach for Optimization of Regional Integrated Energy System Clusters Considering Bounded Rationality”. CSEE Journal of Power and Energy Systems, 2024, 10(1): 302-313.) proposed a two-layer optimization method for the operator-multi-microgrid system, which achieved optimal management of different entities. It is worth noting that in these works, the optimization goal is mainly the economic benefit of the system, and reasonable restrictions on the system's carbon emissions are not considered.

[0004] In order to achieve low-carbon economic development, the carbon emission level of microgrids has gradually become an important indicator. On the one hand, in some studies, by adding carbon capture systems (CCS) and power to gas (P2G) equipment to microgrids, the carbon emissions in the microgrid system are effectively reduced through the coupled operation of CCS and P2G, and the recycling of energy is achieved. On the other hand, with the development of the carbon trading market, some studies have set up corresponding market management mechanisms to limit the carbon emissions of the system. However, only by setting up corresponding management mechanisms can the electricity-carbon coupling transactions between multiple microgrids be realized, but the interaction between different subjects is ignored. For multi-microgrid systems, the multi-subject characteristics of their operation are gradually highlighted. It is urgent to design a suitable operating mechanism to meet the coordination of the interests of multiple subjects, and game theory can just meet this current need. After searching the existing technical literature, it was found that

[0005] Reference [3]: "Peer-to-peer transactive mechanism for residential shared energy storage" (Wenyi Z, Boshen Z, Wei W, Laijun C, Shengwei M. "Peer-to-peer transactive mechanism for residential shared energy storage". Energy. 2022; 246: 123204.) Considering the strategic behavior of individual participants, a P2P network game model describing market equilibrium is proposed. The service price is determined by the supply and demand relationship, while taking into account the interests of shared energy storage operators and consumers to form a market equilibrium. Reference [4]: ​​"Peer-to-Peer transactive energy trading of multiple microgrids considering renewable energy uncertainty." (Yan X, Song M, Cao J, Gao C, Jing X, Xia S, Ban M. "Peer-to-Peer transactive energy trading of multiple microgrids considering renewable energy uncertainty". International Journal of Electrical Power & Energy Systems. 2023 Oct 1; 152: 109-235.) A multi-microgrid point-to-point energy trading strategy based on distributed robust optimization (DRO) was proposed. The renewable energy uncertainty fuzzy set was constructed based on the Wasserstein distance, and the duality theory was used to solve the optimization problem.Reference [5]: "Data-based distributionally robust stochastic optimal power flow—Part II: Case studies" (Guo Y, Baker K, Dall'Anese E, Hu Z, Summers TH. "Data-based distributionally robust stochastic optimal power flow—Part II: Case studies". IEEE Transactions on Power Systems. 2018 Oct 28; 34(2): 1493-503.) The DRO method based on Wasserstein distance constructs a probability distribution uncertainty set centered on the empirical distribution of wind power prediction error based on the Wasserstein sphere, and uses the CVaR theorem and affine rule to transform it into a linear model for solution.

[0006] In summary, the DRO method based on Wasserstein distance has better risk resistance performance, although it results in higher system operation costs. Therefore, in the P2P network game process, how to achieve utilization equilibrium of different decision-makers under uncertain scenarios is a problem worth studying. Summary of the invention

[0007] In order to solve the above technical problems, the present invention provides a distribution network service provider-multiple microgrids distributed blue-robust optimization collaborative scheduling method taking into account multiple uncertainties. The method establishes a P2P network game model between distribution network service providers and multiple microgrids, which is conducive to balancing the interests of all parties, fully absorbing the new energy in the microgrid, and effectively exploring the active participation of service providers and microgrids in electricity-carbon energy transactions.

[0008] The technical solution adopted by the present invention is:

[0009] The distributed robust optimization collaborative dispatching method of multiple microgrids and distribution network service providers considering multiple uncertainties includes the following steps:

[0010] Step 1: Construct Wasserstein fuzzy set constraints on the power purchase and sales price of the upper-level power grid and renewable power output;

[0011] Step 2: Construct a distributed robust model of Wasserstein distance for distribution and sales power service providers;

[0012] Step 3: Build a multi-microgrid system operation model;

[0013] Step 4: Construct a P2P transaction network game model between distribution network service providers and multiple microgrids based on Wasserstein distributed robust optimization;

[0014] Step 5: Combining the hybrid solution method of KKT, duality principle, big M method and stationary point method, the P2P transaction network game model between distribution network service providers and multiple microgrids is solved.

[0015] In step 1, the Wasserstein fuzzy set constraints of the upper-level power grid's electricity purchase and sales price and renewable power output include:

[0016] Distributed robust model based on Wasserstein distance, Wasserstein fuzzy set constraints of upstream grid electricity price and Wasserstein fuzzy set constraints of renewable power output;

[0017] 1.1: Distributed Bruno-Rodson based on Wasserstein distance:

[0018] The distributed robustness method based on Wasserstein distance is used to deal with uncertain variables. This method measures the distance between the true probability distribution and the empirical distribution through Wasserstein distance, thereby constructing an uncertain set of probability distributions. Wasserstein distance is defined as:

[0019]

[0020] In formula (1): P N As the empirical distribution, let N groups of sample data {θ1,θ2,...,θ N} is uniformly distributed as P N ,θ1,θ2,...,θ N They represent the data of the first randomly drawn sample, the data of the second randomly drawn sample, ..., the data of the Nth randomly drawn sample; P is the true probability distribution; θ and ξ are respectively the probability distributions that obey P N and P-distributed random variables; Φ is the joint probability distribution of θ and ξ; Ξ is the support set of the random variable; d W (P N ,P) represents the distribution P N and P is the Wasserstein distance between them.

[0021] Based on Wasserstein distance, a fuzzy set of uncertain variables is constructed, namely, the uncertainty set:

[0022] B ε (P N )={P:P(ξ∈Ξ)=1}∩{P:d w (PN ,P)≤ε} (2);

[0023] In formula (2): This fuzzy set is a P N The sphere of probability distribution space with ε as the center and ε as the radius, the random variable is contained in this sphere with a high confidence level; ε is the radius of the Wasserstein sphere, that is, the radius of the probability distribution space sphere. ε (P N ) indicates that P N is the fuzzy set with fuzzy center; Ξ represents the support set of random variables; P(ξ∈Ξ) represents the distribution of the support set of random variables.

[0024] The relationship between the Wasserstein sphere radius and the confidence level is as follows:

[0025]

[0026] In formula (3), ρ represents the confidence level; N represents N groups of sample data; H and η represent auxiliary variables at the significant level; φ m represents the mth sample; μ represents the sample mean; The exponential function representing the norm of the difference between the mth sample and the sample mean; Indicates the significant level.

[0027] 1.2: Wasserstein fuzzy set constraints of the upper grid electricity price:

[0028] Establish the Wasserstein fuzzy set constraints of the upper-level power grid electricity price forecast value:

[0029]

[0030] In formula (4) and formula (5): are the probability distributions of the samples of electricity purchase price and electricity sales price of the upper power grid respectively; are the bad probability distributions of the predicted values ​​of the power purchase price and power sales price of the upper power grid; κ b , κ s are the radii of the Wasserstein sphere of the upper-level power grid’s electricity purchase and sales prices; ρ b , s They are respectively the confidence levels of the electricity purchase and sales prices of the superior power grid. represents the Wasserstein distance between the probability distribution of the power purchase price of the upper power grid and the bad probability distribution of the power purchase price of the upper power grid; Φ b Represents θ b and b The joint probability distribution of b The random variable representing the probability distribution of the electricity purchase price of the upper power grid; ξb The random variable representing the bad probability distribution of the power purchase price of the upper power grid; θ s The random variable representing the probability distribution of the electricity price of the upper power grid; ξ s A random variable representing the bad probability distribution of the electricity price sold by the upper power grid;

[0031] represents the probability distribution of the Wasserstein distance between the probability distribution of the power purchase price of the upper power grid and the probability distribution of the bad power purchase price of the upper power grid;

[0032] The probability distribution representing the Wasserstein distance between the probability distribution of the upper-level power grid's electricity sales price and the probability distribution of the upper-level power grid's electricity sales price's badness.

[0033] 1.3: Wasserstein fuzzy set constraints for renewable power output:

[0034] Establish the Wasserstein fuzzy set constraints for renewable power output forecast:

[0035]

[0036] In formula (6): is the probability distribution of renewable output samples; The bad probability distribution of renewable power output forecast value; κ re is the radius of the renewable output Wasserstein sphere; ρ re The confidence level for renewable output; The Wasserstein distance between the probability distribution of renewable output samples and the bad probability distribution of renewable output prediction values;

[0037] θ re A random variable representing the probability distribution of renewable output samples; ξ re A random variable representing the bad probability distribution of the renewable power forecast value; ||θ re -ξ re ||1 represents the distance from the probability distribution of the renewable output sample to the bad probability distribution modulus of the renewable output prediction value; Φ re (dθ re ,dξ re ) represents the joint distribution of the probability of renewable output samples and the bad probability of renewable output forecast values; The probability distribution representing the Wasserstein distance between the probability distribution of renewable power samples and the poor probability distribution of renewable power prediction values.

[0038] In step 2, the distributed robust model of Wasserstein distance of the distribution and sales power service provider includes: the objective function of the distribution network service provider, the interaction model with the external power grid, and the carbon quota interaction model;

[0039] 2.1: Objective function of distribution network service provider:

[0040] The distribution network service provider fully considers the uncertainty caused by the fluctuation of power market purchase and sale prices and source load, and adopts the distributed blue stick model of Wasserstein distance to construct the objective function of the distribution network service provider as follows:

[0041]

[0042] In formula (7), C CIESO represents the objective function of the distribution network service provider; is the carbon price; The carbon emission quota purchased by the service provider from the external carbon market; e Clearing electricity prices for the internal market; P t A2M The equivalent interactive power from the service provider to the microgrid; a price for clearing carbon allowances in the internal market; The carbon emission quota sold by the service provider to the microgrid; t and T both represent the meaning of time, where t∈T.

[0043] It means that the distribution network service provider includes the income from buying carbon, selling electricity, and selling carbon; It means that the profit difference between the distribution network service provider's purchase and sale of electricity from the upper-level power grid is maximized.

[0044] If the optimization problem is a convex optimization problem, the objective function model of the distribution network service provider can be transformed into the following model:

[0045]

[0046] In formula (8): N price is the sample data set of electricity price of the upper power grid, |N price | is the total number of corresponding sample data; are the sample values ​​of the electricity purchase price and electricity sales price of the lth upper-level power grid respectively; are the deviation values ​​of the purchase price and sales price of electricity of the lth upper power grid respectively; are the dual multipliers of the Wasserstein sphere radius constraints of the upper power grid’s purchase and sales prices; l represents the lth upper power grid.

[0047] Use duality theory to process the above model and transform the min-max model into a min model:

[0048]

[0049] 2.2: Interaction model with external power grid:

[0050] The distribution network service provider interacts with the external power grid for electric energy. The interaction model is as follows:

[0051] P t buy -P t sell =P t A2M (10);

[0052]

[0053]

[0054] Where: P t buy , P t sell They are the amount of electricity purchased and sold by the service provider from the external power grid; The upper limit of the service provider's electricity purchase and sale from external networks; P t A2M Represents the equivalent interactive power from the service provider to the microgrid.

[0055] 2.3: Carbon quota interaction model:

[0056] The service provider purchases carbon emission quotas from the external carbon market and sells carbon emission quotas to the microgrid. The transaction model is as follows:

[0057]

[0058]

[0059] Where: Represents the carbon emission credits sold by the service provider to the microgrid.

[0060] In step 3, the multi-microgrid system operation model includes the objective function of the multi-microgrid, the interaction constraints with the service provider, the constraints of the gas unit, the operation constraints of the power storage equipment, the thermal power balance constraint model, the electric power balance constraints of the park service provider after considering the DRCC equivalent, and the carbon trading model;

[0061] 3.1: Objective function of multi-microgrid:

[0062] In a multi-microgrid system, each microgrid independently conducts energy transactions with the service provider. Considering that the internal unit models of each microgrid are consistent, the objective function of microgrid operator 1 is as follows:

[0063]

[0064] In formula (15), P t M2A,1 is the equivalent interactive power from microgrid operator 1 to the service provider; Carbon emission quota purchased by microgrid operator 1 from the service provider; p gas is the gas price; p ess is the operation and maintenance cost coefficient of electric energy storage; Indicates the gas consumption of the cogeneration unit; Indicates the amount of gas consumed by the gas boiler; P t ech Indicates the charging power of the electric energy storage. t edis Indicates the discharge power of the electrical energy storage.

[0065] It represents the difference between the price at which the microgrid sells electricity to the service provider and the price at which the microgrid buys carbon from the service provider.

[0066] It represents the sum of the gas cost of the microgrid and the operation and maintenance cost of the point energy storage.

[0067] 3..2: Interaction constraints with service providers:

[0068] P t buy,1 -P t sell,1 =P t M2A,1 (16);

[0069]

[0070]

[0071] Where: P t buy,1 , P t sell,1 They are the amount of electricity purchased and sold by microgrid operator 1 to the service provider; The upper limit of the power that the microgrid can purchase and sell from service providers.

[0072] 3.3: Constraints of gas units:

[0073] The combined heat and power unit burns natural gas to generate electricity and heat. Its power generation, thermal power and constraints are as follows:

[0074]

[0075]

[0076] Where: P tCHP,e , P t CHP,h They are the power generation and heating power of the CHP unit respectively; are the power generation and heating efficiencies of the CHP unit, respectively; The amount of gas consumed by the CHP unit; It is the upper limit of power generation of CHP unit.

[0077] Gas boilers burn natural gas to generate heat energy. Their heating power and constraints are as follows:

[0078]

[0079] Where: P t GB,h The heating power of the gas boiler; The heating efficiency of the gas boiler; Gas consumption for gas boilers; It is the upper limit of heating capacity of gas boiler.

[0080] 3.4: Operation constraints of power storage equipment:

[0081] Electric energy storage can achieve efficient utilization of energy storage resources. The state of charge and charging and discharging power constraints of electric energy storage are as follows:

[0082]

[0083]

[0084] μ ch +μ dis ≤1(30);

[0085] Where: Real-time storage capacity for electric energy storage; P1 edis Indicates the energy storage discharge power of microgrid 1; are the charging efficiency and discharging efficiency of electric energy storage respectively; P t ech , P t edis are the charging power and discharging power of the energy storage respectively; μ ch , μ dis They are the status bits for charging and discharging of the electric energy storage respectively; They are the upper and lower limits of the real-time storage capacity of the electric energy storage respectively; They are the upper limits of the charging power and discharging power of the energy storage respectively.

[0086] represents the electric energy storage capacity of microgrid 2 at the 24th hour; represents the electric energy storage capacity of microgrid 1 at the 23rd hour; represents the energy storage charging power of microgrid 2 at the 24th hour; represents the energy storage discharge power of microgrid 2 at the 24th hour; Indicates the storage capacity of the electric energy storage at the 24th hour; Indicates the storage capacity of the energy storage at hour 0.

[0087] 3.5: Thermal power balance constraint model:

[0088]

[0089] In formula (31): Forecasted thermal loads for microgrid operators.

[0090] 3.6: Considering the power balance constraints of the campus service provider after DRCC equivalence:

[0091]

[0092] In formula (32), P t load,1 represents the predicted electric load of microgrid operator 1; P W,C,bound represents the worst lower bound of renewable energy output of microgrid operator 1; P t buy,1 P represents the carbon cost of electricity purchased by microgrid operator 1. t sell,1 Represents the carbon revenue from selling electricity of microgrid operator 1.

[0093] 3.7: Carbon trading model:

[0094]

[0095]

[0096]

[0097] Where: is the actual carbon emissions of microgrid operator 1; is the correlation coefficient; Carbon emission quota for microgrid operator 1; is the correlation coefficient; Represents the carbon emission quota purchased by microgrid operator 1 from the service provider.

[0098] In step 4, the P2P transaction network game model between distribution network service providers and multiple microgrids based on Wasserstein distribution robust optimization includes equations (4) to (36).

[0099] A global electricity-carbon balance coupling model is established based on the P2P trading network game model between the distribution network service provider and multiple microgrids; the global electricity-carbon balance coupling model refers to the electricity and carbon quota balance model between the distribution network service provider and the three microgrids.

[0100] P t A2M =P t M2A,1 +P t M2A,2 +P t M2A,3 :p e (37);

[0101]

[0102] In the formula, p e clearing electricity prices for the internal market; The price of carbon quotas for clearing the internal market e is the clearing electricity price in the internal market, which are the dual variables of the above equation; P t A2M represents the interactive power between the service provider and the microgrid; P t M2A,1 represents the interactive power between microgrid 1 and the service provider; P t M2A,2 represents the interactive power between microgrid 2 and the service provider; P t M2A,3 Indicates the interactive power between microgrid 3 and the service provider; Indicates the interactive carbon quota between the service provider and the microgrid; It represents the interactive carbon quota between the service provider and microgrid 1; represents the interactive carbon quota between the service provider and microgrid 2; Represents the interactive carbon quota between the service provider and microgrid 3.

[0103] The step 5 comprises the following steps:

[0104] Step 5.1: For the P2P trading network game model, use the duality principle to construct an equivalent KKT system in parallel;

[0105] Step 5.2: The model shown in equation (39) is converted into a MILP problem through the big M method to solve it, so as to locate the equilibrium state of the multi-agent game in the park service provider-multi-microgrid system;

[0106] Step 5.3: To verify the existence and uniqueness of the equilibrium between multi-agent games, the KKT conditions are combined to calculate the fixed points for judgment. Finally, it is concluded that the Nash equilibrium solution exists and is unique, and can be solved directly.

[0107] In step 5, the solution steps are as follows:

[0108] According to the P2P trading network game model, the decision variable on the park service provider side is the power sold by the park service provider to the microgrid. t A2M The power purchased by the microgrid side P t M2A,1 , P t M2A,2 and P t M2A,3 Satisfy power balance constraints; Decision variables on the park service provider side The carbon emission quota sold by the park service provider to the microgrid Purchase carbon quotas with microgrid side and Meet the carbon quota balance constraint.

[0109] Therefore, the mutual constraints between the decision variables of multiple agents lead to mutual constraints in the strategy space. Therefore, the multi-agent non-cooperative game is modeled as a generalized Nash equilibrium problem (GNEP), which is defined as shown in the following formula, and the Nash equilibrium of the multi-agent game is solved by the stationary point optimization method:

[0110]

[0111] Where: C i 、x i , X i are the objective function, decision variables, and strategy space of participant i respectively; x -i is the determining variable of all participants except participant i; g i is the inequality constraint of participant i; h i is the equality constraint for participant i.

[0112] Firstly, the P2P trading network game model constructs an equivalent KKT system in parallel, and then converts it into a MILP problem through the big M method to solve it, so as to locate the equilibrium state of the multi-agent game in the park service provider-multi-microgrid system.

[0113] Since GNEP may have multiple equilibrium situations - no equilibrium and finite or infinite equilibrium, in order to verify the existence and uniqueness of equilibrium, the KKT conditions are directly combined to calculate the fixed point judgment.

[0114] The Nash equilibrium solution of the model shown in equation (39) of the present invention exists and is unique, so it can be solved directly.

[0115] In addition, the equilibrium solution can also be obtained according to the above process in case of an increase in the number of entities or changes in modeling.

[0116] 1) KKT system for:

[0117]

[0118] Where λ1 is the dual variable of the equality constraint; μ1 is the dual variable of the inequality constraint. represents the gradient of the solution of the KKT system; C i represents the i-th objective function; h i represents the i-th equality constraint; g i represents the i-th inequality constraint; ⊥ represents a constraint that satisfies both inequality and equality; i represents a variable.

[0119] The equilibrium point solution is as follows:

[0120]

[0121] Where: x A 、x M are the decision variables of microgrid and park service provider respectively; U VPP For the park service provider - the overall cost of multiple microgrid systems; x M1 represents the decision variable of microgrid 1; x M2 represents the decision variable of microgrid 2; x M3 represents the decision variables of microgrid 3; C CIESO represents the objective function of the distribution network service provider; C MGO,1 It is expressed as the objective function of microgrid 1. C MGO,2 Expressed as the objective function of microgrid 2; C MGO,3 It is expressed as the objective function of microgrid 3; x A represents the decision variable of the service provider; Indicates the KKT system of the service provider; represents the KKT system of microgrid 1; represents the KKT system of microgrid 2; Represents the KKT system of the microgrid 3.

[0122] The present invention provides a distribution network service provider-multi-microgrid distributed robust optimization collaborative scheduling method taking into account multiple uncertainties, and the technical effects are as follows:

[0123] 1) The method of the present invention promotes the low-carbon economic operation of microgrids through electricity-carbon coupling transactions between distribution network service providers and multiple microgrids.

[0124] 2) The method of the present invention establishes a P2P network game model between distribution network service providers and multiple microgrids, which is conducive to balancing the interests of all parties, fully absorbing the new energy in the microgrid, and effectively exploring the active participation of service providers and microgrids in electricity-carbon energy transactions.

[0125] 3) The method of the present invention takes into account the uncertainty factors of different subjects, establishes a distributed robust model, takes into account the advantages of SO and RO, and effectively avoids the unknown risks brought by the uncertainty factors in the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0126] Figure 1 Schematic diagram of the electrical load of the microgrid.

[0127] Figure 2 Schematic diagram of the heat load of the microgrid.

[0128] Figure 3 Schematic diagram of the worst lower bound of renewable energy for microgrids.

[0129] Figure 4 This is a schematic diagram of historical samples of electricity prices for service providers.

[0130] Figure 5 This is a schematic diagram of the historical samples of electricity prices sold by service providers.

[0131] Figure 6 This is a schematic diagram of carbon quota prices in the external carbon market.

[0132] Figure 7 The results of electricity trading between the service provider and three microgrids.

[0133] Figure 8 The carbon energy trading results between the service provider and three microgrids.

[0134] Fig. 9 The price results of electric carbon energy trading between the service provider and three microgrids.

[0135] Fig.10 Power optimization results for microgrid 1.

[0136] Fig.11 Thermal optimization results for microgrid 1.

[0137] Fig.12 Power optimization results for microgrid 2.

[0138] Fig.13 Thermal optimization results for microgrid 2.

[0139] Fig.14 3 Power optimization results for microgrid.

[0140] Fig.15 3 Thermal energy optimization results for microgrids.

[0141] Fig.16 Provide power optimization results for service providers.

[0142] Fig.17Schematic diagram of the worst lower bound of the renewable output of microgrid 1.

[0143] Fig.18 Schematic diagram of the worst lower bound of the renewable output of microgrid 2.

[0144] Fig.19 Schematic diagram of the worst lower bound of the renewable output of microgrid 3.

[0145] Fig. 20 Comparison chart of uncertainty optimization methods for microgrid 1. DETAILED DESCRIPTION

[0146] A distribution network service provider-multi-microgrid P2P network game optimization scheduling method based on Wasserstein distributed robust optimization is proposed. The distribution network service provider is the upper layer, considering the uncertainty of electricity prices in the power market and is responsible for formulating energy trading prices. The microgrid is the lower layer, considering the uncertainty of renewable energy output, accepting the energy price of the distribution network service provider and responding. The method based on Wasserstein distributed robust opportunity constraints is used to solve the uncertainty factors of the upper and lower layers, forming a P2P network game as a whole.

[0147] First, a P2P trading network game model between distribution network service providers and multiple microgrids is constructed. Secondly, the Wasserstein distributed robust optimization model is introduced to characterize the multiple uncertainties of new energy and market electricity prices. Then, a distribution network service provider-multi-microgrid P2P network game optimization scheduling model based on Wasserstein distributed robust optimization is proposed. Then, a hybrid solution method combining KKT, duality principle, big M method and stationary point method is proposed, and the Yalmip toolbox and Gurobi solver are called for solution. Finally, the effectiveness and feasibility of all methods of the present invention are verified through case analysis, which provides new ideas and methods for the coordinated scheduling between distribution network service providers and multiple microgrids.

[0148] The method for distributed robust optimization and collaborative scheduling of distribution network service providers and multiple microgrids taking into account multiple uncertainties is characterized by first constructing a Wasserstein fuzzy set constraint model of the upper-level power grid's electricity purchase and sales price and renewable output, a distributed robust model of the Wasserstein distance of the distribution and sales power service provider, and a multi-microgrid system operation model. The above-mentioned multiple subjects can be constructed as a point-to-point transaction network game model between distribution network service providers and multiple microgrids. For energy transactions between multiple subjects, a global power balance coupling constraint is constructed. The mutual constraints between the decision variables of multiple subjects lead to mutual constraints in the strategy space. Therefore, the multi-agent non-cooperative game is modeled as a generalized Nash equilibrium problem, and the Nash equilibrium of the multi-agent game is solved by the stationary point optimization method.

[0149] The solution steps are as follows:

[0150] Step 1: For the non-cooperative game structure model of power distribution service providers and multiple microgrids, an equivalent KKT system is constructed in parallel using the duality principle.

[0151] Step 2: The joint model is converted into a MILP problem through the big M method to solve it in order to locate the equilibrium state of the multi-agent game in the park service provider-multi-microgrid system.

[0152] Step 3: In order to verify the existence and uniqueness of the equilibrium between multi-agent games, the fixed points are calculated by combining the KKT conditions for judgment, and finally it is concluded that the Nash equilibrium solution of the model proposed by the present invention exists and is unique, and can be solved directly.

[0153] The Wasserstein fuzzy set constraint model for the power purchase and sale price of the upper power grid and renewable output is a distributed robust model based on Wasserstein distance, a Wasserstein fuzzy set constraint for the power price of the upper power grid, and a Wasserstein fuzzy set constraint for renewable output. The specific derivation process is as follows:

[0154] (1): Basic principle of the distributed robustness model based on Wasserstein distance:

[0155] The present invention uses a distributed robust method based on Wasserstein distance to process uncertain variables. The method measures the distance between the true probability distribution and the empirical distribution through Wasserstein distance, thereby constructing an uncertain set of probability distributions. Wasserstein distance is defined as:

[0156]

[0157] Where: P N For empirical distribution, the present invention transforms N groups of sample data {θ1, θ2, ..., θ N} is uniformly distributed as P N ; P is the true probability distribution; θ and ξ are respectively N and P distributed random variables; Φ is the joint probability distribution of θ and ξ; Ξ is the support set of the random variable.

[0158] Based on Wasserstein distance, a fuzzy set of uncertain variables is constructed, namely, the uncertainty set:

[0159] B ε (P N )={P:P(ξ∈Ξ)=1}∩{P:d w (P N ,P)≤ε} (2);

[0160] Where: ε is the radius of the Wasserstein ball, that is, the radius of the probability distribution space sphere. N A sphere in the probability distribution space with ε as the center and ε as the radius. The random variable is contained in this sphere with a high confidence level.

[0161] The relationship between the Wasserstein sphere radius and the confidence level is as follows:

[0162]

[0163] (2): Wasserstein fuzzy set constraints of the upper grid electricity price:

[0164] Based on the above theory, the Wasserstein fuzzy set constraints of the upper-level power grid electricity price forecast value can be established:

[0165]

[0166] Where: are the probability distributions of the samples of electricity purchase price and electricity sales price of the upper power grid respectively; are the bad probability distributions of the predicted values ​​of the power purchase price and power sales price of the upper power grid; κ b , κ s are the radii of the Wasserstein sphere of the upper-level power grid’s electricity purchase and sales prices; ρ b , s They are respectively the confidence levels of the electricity purchase and sales prices of the superior power grid.

[0167] (3): Wasserstein fuzzy set constraints on renewable power output

[0168] Similarly, based on the above theory, the Wasserstein fuzzy set constraints for renewable power output prediction can be established:

[0169]

[0170] Where: is the probability distribution of renewable output samples; The bad probability distribution of renewable power output forecast value; κ re is the radius of the renewable output Wasserstein sphere; ρ re is the confidence level for renewable output.

[0171] The distributed robust model of Wasserstein distance for the distribution and sales power service provider is the objective function of the distribution network service provider, the interaction model with the external power grid, and the carbon quota interaction model. The specific derivation process is as follows:

[0172] (1): The objective function of the distribution network service provider:

[0173] The distribution network service provider fully considers the uncertainty caused by the fluctuation of power market purchase and sale prices and source load, and adopts the distributed robust model of Wasserstein distance to construct it. The objective function of the distribution network service provider is as follows:

[0174]

[0175] In the formula, is the carbon price; The carbon emission quota purchased by the service provider from the external carbon market; e Clearing electricity prices for the internal market; P t A2M The equivalent interactive power from the service provider to the microgrid; a price for clearing carbon allowances in the internal market; Carbon emission quotas sold by service providers to microgrids.

[0176] If the optimization problem is convex, the distribution network service provider's objective function model can be transformed into the following model:

[0177]

[0178] Where: N price is the sample data set of electricity price of the upper power grid, |N price | is the total number of corresponding sample data; and are the sample value and deviation value of the electricity purchase price and electricity sales price of the lth upper-level power grid respectively; They are the dual multipliers of the Wasserstein sphere radius constraints of the upper-level power grid’s purchase price and sales price.

[0179] Use duality theory to process the above model and transform the min-max model into a min model:

[0180]

[0181] (2): Interaction model with external power grid:

[0182] The distribution network service provider interacts with the external power grid for electric energy. The interaction model is as follows:

[0183] P t buy -P t sell =P t A2M (10);

[0184]

[0185]

[0186] Where: P t buy , P t sell They are the amount of electricity purchased and sold by the service provider from the external power grid; It is the upper limit for service providers to purchase and sell electricity from external networks.

[0187] (3): Carbon quota interaction model:

[0188] The service provider purchases carbon emission quotas from the external carbon market and sells carbon emission quotas to the microgrid. The transaction model is as follows:

[0189]

[0190]

[0191] The operation model of the multi-microgrid system includes the objective function of the multi-microgrid, the interaction constraints with the service provider, the constraints of the gas unit, the operation constraints of the power storage equipment, the thermal power balance constraint model, the power balance constraints of the park service provider after considering the DRCC equivalent, the carbon trading model, the global power balance coupling constraints, and the proof of non-cooperative game and solving it with the stationary point method. The specific derivation process is as follows:

[0192] (1): Objective function of multi-microgrid:

[0193] In a multi-microgrid system, each microgrid independently conducts energy transactions with the service provider. The present invention takes into account the consistency of the internal unit models of each microgrid. Taking microgrid 1 as an example, the objective function of microgrid operator 1 is as follows:

[0194]

[0195] Where P t M2A,1 is the equivalent interactive power from microgrid operator 1 to the service provider; Carbon emission quota purchased by microgrid operator 1 from the service provider; p gas is the gas price; p ess is the energy storage operation and maintenance cost coefficient.

[0196] (2): Interaction constraints with service providers:

[0197] P t buy,1 -P t sell,1 =P t M2A,1 (16);

[0198]

[0199]

[0200] Where: P t buy,1 , P t sell,1 They are the amount of electricity purchased and sold by microgrid operator 1 to the service provider; The upper limit of the power that the microgrid can purchase and sell from service providers.

[0201] (3): Constraints of gas units:

[0202] The combined heat and power unit burns natural gas to generate electricity and heat. Its power generation, thermal power and constraints are as follows:

[0203]

[0204] Where: P t CHP,e , P t CHP,h Provides power generation and heating for the CHP unit; The power generation and heating efficiency of the CHP unit; The amount of gas consumed by the CHP unit; It is the upper limit of power generation of CHP unit.

[0205] Gas boilers burn natural gas to generate heat energy. Their heating power and constraints are as follows:

[0206]

[0207] Where: P t GB,h The heating power of the gas boiler; The heating efficiency of the gas boiler; Gas consumption for gas boilers; It is the upper limit of heating capacity of gas boiler.

[0208] (4): Operation constraints of power storage equipment:

[0209] Electric energy storage can achieve efficient utilization of energy storage resources. The state of charge and charging and discharging power constraints of electric energy storage are as follows:

[0210]

[0211] μ ch +μ dis ≤1(30);

[0212] Where: Real-time storage capacity for electric energy storage; are the charging efficiency and discharging efficiency of electric energy storage respectively; P t ech , P t edis are the charging power and discharging power of the energy storage respectively; μ ch , μ dis They are the status bits for charging and discharging of the electric energy storage respectively; They are the upper and lower limits of the real-time storage capacity of the electric energy storage respectively; They are the upper limits of the charging power and discharging power of the energy storage respectively.

[0213] (5): Thermal power balance constraint model:

[0214]

[0215] Where: Thermal load forecasts for microgrid operators

[0216] (6): Considering the power balance constraints of the park service provider after DRCC equivalent

[0217] P t load,1 -P t CHP,e -(P t buy,1 -P t sell,1 )-(P t edis -P t ech )≤P W,C,bound (32);

[0218] Where P t load,1 represents the predicted electric load of microgrid operator 1; P W,C,bound It represents the worst lower bound value of renewable energy output of microgrid operator 1.

[0219] (7): Carbon trading model

[0220]

[0221]

[0222] Where: is the actual carbon emissions of microgrid operator 1; is the correlation coefficient; Carbon emission quota for microgrid operator 1; is the correlation coefficient.

[0223] (8): Global electric-carbon balance coupling model:

[0224] P t A2M =P t M2A,1 +P t M2A,2 +P t M2A,3 :p e (37);

[0225]

[0226] In the formula, p e clearing electricity prices for the internal market; The price of carbon quotas for clearing the internal market e is the clearing electricity price in the internal market, and are the dual variables of the above formula.

[0227] The solution steps are as follows:

[0228] According to the operation decision optimization model of the park service provider-microgrid, the decision variable on the park service provider side is the power sold by the park service provider to the microgrid. t A2M The power purchased by the microgrid side P t M2A,1 , P t M2A,2 and P t M2A,3 Satisfy the power balance constraint; Decision variables on the park service provider side The carbon emission quota sold by the park service provider to the microgrid Q t A2M Purchase carbon quotas with microgrid side Q t M2A,1 , Q t M2A,2 and Q t M2A,3 The carbon quota balance constraint is satisfied. Therefore, the mutual constraints between the decision variables of multiple agents lead to mutual constraints in the strategy space. Therefore, the multi-agent non-cooperative game is modeled as a generalized Nash equilibrium problem (GNEP), which is defined as shown in the following formula, and the Nash equilibrium of the multi-agent game is solved by the stationary point optimization method:

[0229]

[0230] Where: C i 、x i , X i are the objective function, decision variables, and strategy space of participant i respectively; x -iis the determining variable of all participants except participant i; g i is the inequality constraint of participant i; h i is the equality constraint for participant i.

[0231] Firstly, the non-cooperative game structure model of the park service provider-multi-microgrid is constructed, and the equivalent KKT system is paralleled. Then, it is converted into a MILP problem through the big M method for solution, so as to locate the equilibrium state of the multi-agent game in the park service provider-multi-microgrid system.

[0232] Since GNEP may have multiple equilibrium situations - no equilibrium and finite or infinite equilibrium, in order to verify the existence and uniqueness of equilibrium, the KKT conditions are directly combined to calculate the fixed point judgment.

[0233] The following theorem guarantees that the solution of the proposed method is a Nash equilibrium. The proof of the theorem is given in

[20] Facchinei F, Kanzow C. Generalized Nash equilibrium problems. Annals of Operations Research, 2010, 175(1): 177-211.

[0234] Theorem 1: If all functions in GNEP are continuously differentiable, assuming that (x,λ) is the solution of the KKT system corresponding to GNEP, and GNEP is convex with respect to x, then x is a Nash equilibrium point of GNEP. The optimization model equations (1)-(38) are all continuously differentiable, and the constraint equation (39) can replace the following planning problem KKT condition:

[0235]

[0236] Because the clearing price p e and In equations (9) and (15), is only a parameter, so the decision model of each subject is linear with respect to its own decision variable and is convex. The theorem is applicable, and the obtained equilibrium point is the Nash equilibrium.

[0237] In addition, the single iteration method is used to check the equilibrium solution problem. The solutions of equation (43) are all Nash equilibrium. According to the definition of Nash equilibrium:

[0238]

[0239] That is, at the equilibrium point, no participant can gain greater benefits by changing their own strategies. The construction and testing steps are as follows:

[0240] 1) The optimal decisions of the park service provider and multi-microgrid obtained by the multi-agent equilibrium model are recorded as

[0241] 2) Fixed multi-microgrid strategy set Solve the source side model equation (7)-(14) and record its optimal decision as

[0242] 3) Repeat step 2) to obtain the optimal strategy for multiple microgrids

[0243] 4) Verify whether If both are satisfied, the solution of the stationary point optimization method is a Nash equilibrium. If any one of them is not satisfied, the solution is a non-equilibrium.

[0244] It can be seen from the above proof that the Nash equilibrium solution of the model proposed in the present invention exists and is unique, so it can be solved directly.

[0245] In addition, the equilibrium solution can also be obtained according to the above process in case of an increase in the number of entities or changes in modeling.

[0246] 1) KKT system for:

[0247]

[0248] Where: λ1 is the dual variable of the equality constraint; μ1 is the dual variable of the inequality constraint. The equilibrium point solution is as follows:

[0249]

[0250] Where: x A 、x M are the decision variables of microgrid and park service provider respectively; U VPP For the park service provider - the overall cost of the multi-microgrid system.

[0251] The energy transaction volume, energy transaction price and dispatch plan of each entity are analyzed and compared, which are the basic data and system structure description, energy transaction volume and energy transaction price, dispatch plan and operation cost of each entity and uncertainty analysis.

[0252] In the Matlab compilation environment, the present invention establishes a P2P network game model between the service provider and three microgrids through the Yalmip toolbox, and calls the Gurobi business solver to solve the energy transaction volume and energy transaction price problems between the service provider and the three microgrids.

[0253] 1): Basic data and system structure description

[0254] The present invention considers the energy transaction between the service provider and three microgrids, where the service provider is a pure electricity model and the three microgrids are cogeneration models. The gas purchase price of the three microgrids from the gas grid is 3.2 (unit: yuan / m 3). All three microgrids are equipped with renewable energy units. The source-load data of the three microgrids, the historical samples of the electricity prices purchased and sold by the service providers to the external power grid, and the prices of carbon quotas purchased from the external carbon market are as follows: Figure 1 to Figure 6 shown.

[0255] 2): Energy trading volume and energy trading price:

[0256] a: The electricity-carbon energy transaction volume between the service provider and the three microgrids is as follows: Figure 7 , Figure 8 Among them, positive values ​​indicate that the three microgrids purchase energy from service providers, and negative values ​​indicate that the three microgrids sell energy to service providers.

[0257] From the analysis of the power transaction volume between the service provider and the three microgrids, it can be seen that for microgrid 1, it sells power to the service provider during the period of 13:00-14:00. This is because during this period, the output of renewable energy in microgrid 1 is high, while the load demand is low. Therefore, it will sell excess power to the service provider to improve its own benefits. In the rest of the time, microgrid 1 mainly purchases power from the service provider to maintain its own power balance. For microgrid 2, since the power load demand of microgrid 2 is large, it purchases power from the service provider in a scheduling cycle to maintain its own power load balance. Among them, during the period of 10:00-14:00, the output of renewable energy in microgrid 2 increases, so the power purchased from the service provider decreases, and in the rest of the time, it needs to purchase power from the service provider to meet its own load demand. For microgrid 3, it purchases power from the service provider in a scheduling cycle to maintain its own power load balance. During the period of 8:00-16:00, as the output of renewable energy in the microgrid increases, the amount of electricity purchased by Microgrid 3 from the service provider decreases to improve its own benefits.

[0258] From the analysis of the carbon energy trading volume between the service provider and the three microgrids, it can be seen that the three microgrids all purchased carbon quotas from the service provider. For microgrids 1 and 2, the power load is at its peak during the period of 9:00-16:00, so it is necessary to increase the output of distributed power sources in the microgrid to maintain the power demand in the microgrid. Distributed power sources often produce a certain amount of carbon dioxide when generating electricity, so more carbon quotas are purchased from the service provider. For microgrid 3, since the power load of microgrid 3 is low and the output of new energy is high, the distributed power generation in microgrid 3 is relatively small, so the carbon quota purchased from the service provider is also low.

[0259] b: Electric carbon energy transaction price between the service provider and the three microgrids

[0260] The price of carbon energy transaction between the service provider and the three microgrids is as follows: Fig. 9As shown. It is worth noting that the electricity price and carbon price traded between the service provider and the three microgrids are unified. In the period of 1:00-8:00, in order to encourage the microgrids to purchase electricity and carbon quotas from the service provider, the prices of electricity and carbon quotas during this period are relatively low. During the peak period of electricity consumption from 10:00 to 15:00, each microgrid cannot meet its own energy demand and needs to purchase energy from the service provider. Therefore, during this period, the electricity price and carbon price are relatively high to maximize the benefits of the service provider itself.

[0261] c: Scheduling plan and operating costs of each entity:

[0262] ①. Microgrid operation and dispatching plan:

[0263] The power optimization results and thermal energy optimization results of the three microgrids are as follows: Figure 10-15 As shown. Take the operation and dispatching plan of microgrid 1 as an example for analysis. From the power optimization results of microgrid 1, it can be seen that the CHP unit of microgrid 1 has output in the period of 8:00-21:00; the energy storage is charged in the periods of 2:00, 5:00, 7:00, 15:00, 17:00-18:00, etc., and discharged in the period of 13:00-14:00, that is, it is charged when the power transaction price is low and discharged when it is high, thereby reducing its own operating costs; renewable energy has output throughout the day, and the output is the largest in the period of 9:00-15:00; microgrid 1 and service providers also have different degrees of power interaction throughout the day, among which the interactive power is at a higher level in the morning period of 5:00-7:00 to meet the steep increase in power load demand during this period.

[0264] It can be seen from the thermal energy optimization results of microgrid 1 that the thermal load of microgrid 1 is shared by the CHP unit and the GB unit. The heat output of the CHP unit of microgrid 1 is mainly concentrated in the period of 8:00-22:00, and the rest of the time, that is, at night, is mainly produced by the GB unit.

[0265] ②、Service provider operation scheduling plan:

[0266] The service provider's power optimization results are as follows: Fig.16 As shown. Fig.16 Analysis shows that service providers purchase electricity to varying degrees throughout the day, with higher purchases during periods of lower electricity prices such as 15:00-19:00 and 23:00-7:00. At the same time, service providers interact with other microgrids to varying degrees throughout the day, with more interactive power in the morning and afternoon. In addition, service providers have to supply electricity to each microgrid to meet their energy balance, so there is no surplus electricity to go online.

[0267] ③. Operating costs of service providers and three microgrids:

[0268] The operating costs of the service provider and the microgrid are shown in Table 1.

[0269] Table 1. Operating costs of service providers and microgrids

[0270]

[0271] From the analysis in Table 1, we can see that the sum of the untransacted costs, electricity transaction costs, and carbon transaction costs of each microgrid is the final cost of the P2P network game, which is 25197.05 yuan, 38443.59 yuan, and 22975.53 yuan respectively; the untransacted costs of the service provider's P2P network game are 46913.22 yuan, the electricity transaction costs are -42704.13 yuan, and the carbon transaction costs are -4209.09 yuan. Since the electricity and carbon benefits of the service provider and the three microgrids offset the untransacted costs of the P2P network game, the final cost of the P2P network game is 0.

[0272] ④. Uncertainty analysis:

[0273] Figure 17-Figure 19 is the optimization result of the distributed robust boundary of renewable energy in each microgrid. There is uncertainty in the renewable output of each microgrid. Therefore, the present invention adopts the DRCC method based on Wasserstein to obtain the worst lower bound of renewable output, where the number of historical samples of renewable output is 5 and the confidence level is 70%. Figure 17-Figure 19 It can be seen that the worst lower bound of renewable output is always between the maximum and minimum output of each period of the historical sample.

[0274] Taking microgrid 1 as an example, the comparative analysis of uncertainty optimization methods is as follows: Fig. 20 As shown. The sample performance of the proposed method is compared with that of the SO and RO methods by using Monte Carlo simulation of 5000 renewable output samples. For the SO model, the Sample Average Approximation (SAA) method is used, and the radius of the Wasserstein sphere is set to zero; for the RO model, it corresponds to the RO-SAA method. Fig. 20 It can be seen that the cost generated by the RO-SAA method is higher than that of the SAA and the distributed robust optimization method. This is because the RO-SAA method only uses the lower and upper bounds of uncertainty to make decisions, resulting in overly conservative solutions. Another observation is that the cost of the SAA and DRO methods decreases with the increase in the number of samples. This is because the more samples of uncertainty, the more accurate the probability distribution of uncertainty. In particular, it is observed that within the sample range, the cost of the DRO method is lower than that of the SAA method, which proves that the method proposed in this invention is superior to the SAA method. Table 2 shows the cost of service providers under different scenario scales of the upper-level power grid purchase and sales price.

[0275] Table 2 Costs of service providers under different scenarios and scales of electricity purchase and sales prices of the upper grid

[0276]

[0277] As can be seen from Table 2, as the scale of electricity price scenarios continues to increase, the operating costs of service providers have gradually stabilized, basically hovering around 632 yuan.

Claims

1. A distributed and optimized coordinated dispatching method for distribution network service providers and multiple microgrids taking into account multiple uncertainties, characterized by The following steps are involved: Step 1: Construct Wasserstein fuzzy set constraints between the power purchase and sales price of the upper-level power grid and renewable power output; Step 2: Construct a distributed blue stick model of Wasserstein distance for distribution and sales power service providers; Step 3: Build a multi-microgrid system operation model; Step 4: Construct a P2P transaction network game model between distribution network service providers and multiple microgrids based on Wasserstein distributed robust optimization; Step 5: Combining the hybrid solution method of KKT, duality principle, big M method and stationary point method, the P2P transaction network game model between distribution network service providers and multiple microgrids is solved.

2. The distribution network service provider-multi-microgrid distributed robust optimization collaborative scheduling method taking into account multiple uncertainties according to claim 1, characterized in that: In the step 1, the Wasserstein fuzzy set constraints of the power purchase and sale price of the upper power grid and the renewable output include: a distributed robust model based on the Wasserstein distance, a Wasserstein fuzzy set constraint of the power price of the upper power grid, and a Wasserstein fuzzy set constraint of the renewable output; 1.1: Distributed Bruno-Rodson based on Wasserstein distance: The distributed robustness method based on Wasserstein distance is used to deal with uncertain variables. This method measures the distance between the true probability distribution and the empirical distribution through the Wasserstein distance, thereby constructing an uncertain set of probability distributions. The Wasserstein distance is defined as: In formula (1): P N As the empirical distribution, let N groups of sample data {θ1,θ2,...,θ N } is uniformly distributed as P N ,θ1,θ2,...,θ N They represent the data of the first randomly drawn sample, the data of the second randomly drawn sample, ..., the data of the Nth randomly drawn sample; P is the true probability distribution; θ and ξ are respectively the probability distributions that obey P N and P-distributed random variables; Φ is the joint probability distribution of θ and ξ; Ξ is the support set of the random variable; d W (P N ,P) represents the distribution P N The Wasserstein distance between and P; Based on Wasserstein distance, a fuzzy set of uncertain variables is constructed, namely, the uncertainty set: B ε (P N )={P:P(ξ∈Ξ)=1}∩{P:d w (P N ,P)ε}(2); In formula (2): This fuzzy set is a P N The sphere of probability distribution space with ε as the center and ε as the radius, the random variable is contained in this sphere with a high confidence level; ε is the radius of the Wasserstein sphere, that is, the radius of the probability distribution space sphere; B ε (P N ) indicates that P N is the fuzzy set with fuzzy center; Ξ represents the support set of random variables; P(ξ∈Ξ) represents the distribution of the support set of random variables; The relationship between the Wasserstein sphere radius and the confidence level is as follows: In formula (3), ρ represents the confidence level; N represents N groups of sample data; H and η represent auxiliary variables at the significant level; φ m represents the mth sample; μ represents the sample mean; The exponential function representing the norm of the difference between the mth sample and the sample mean; Indicates the significance level; 1.2: Wasserstein fuzzy set constraints of the upper grid electricity price: Establish the Wasserstein fuzzy set constraints of the upper-level power grid electricity price forecast value: In formula (4) and formula (5): are the probability distributions of the samples of electricity purchase price and electricity sales price of the upper power grid respectively; are the bad probability distributions of the predicted values ​​of the power purchase price and power sales price of the upper power grid; κ b , κ s are the radii of the Wasserstein sphere of the upper-level power grid’s electricity purchase and sales prices; ρ b , s are the confidence levels of the power purchase and sales prices of the superior power grid; represents the Wasserstein distance between the probability distribution of the power purchase price of the upper power grid and the bad probability distribution of the power purchase price of the upper power grid; Φ b Represents θ b and b The joint probability distribution of b The random variable representing the probability distribution of the electricity purchase price of the upper power grid; ξ b The random variable representing the bad probability distribution of the power purchase price of the upper power grid; θ s The random variable representing the probability distribution of the electricity price of the upper power grid; ξ s A random variable representing the bad probability distribution of the electricity price sold by the upper power grid; represents the probability distribution of the Wasserstein distance between the probability distribution of the power purchase price of the upper power grid and the probability distribution of the bad power purchase price of the upper power grid; The probability distribution representing the Wasserstein distance between the probability distribution of the power selling price of the upper power grid and the probability distribution of the bad power selling price of the upper power grid; 1.3: Wasserstein fuzzy set constraints for renewable power output: Establish the Wasserstein fuzzy set constraints for renewable power output forecast: In formula (6): is the probability distribution of renewable output samples; The bad probability distribution of renewable power output forecast value; κ re is the radius of the renewable output Wasserstein sphere; ρ re The confidence level for renewable output; The Wasserstein distance between the probability distribution of renewable output samples and the bad probability distribution of renewable output prediction values; θ re A random variable representing the probability distribution of renewable output samples; ξ re A random variable representing a bad probability distribution of the predicted value of renewable output; ||θ re -ξ re ||1 represents the distance from the probability distribution of the renewable output sample to the bad probability distribution modulus of the renewable output prediction value; Φ re (dθ re ,dξ re ) represents the joint distribution of the probability of renewable output samples and the bad probability of renewable output forecast values; The probability distribution representing the Wasserstein distance between the probability distribution of renewable power samples and the poor probability distribution of renewable power prediction values.

3. The distribution network service provider-multi-microgrid distributed robust optimization collaborative scheduling method taking into account multiple uncertainties according to claim 2, characterized in that: In step 2, the distributed robust model of Wasserstein distance of the distribution and sales power service provider includes: the objective function of the distribution network service provider, the interaction model with the external power grid, and the carbon quota interaction model; 2.1: Objective function of distribution network service provider: The distribution network service provider fully considers the uncertainty caused by the fluctuation of power market purchase and sale prices and source load, and adopts the distributed blue stick model of Wasserstein distance to construct the objective function of the distribution network service provider as follows: In formula (7), C CIESO represents the objective function of the distribution network service provider; is the carbon price; The carbon emission quota purchased by the service provider from the external carbon market; e Clearing electricity prices for the internal market; P t A2M The equivalent interactive power from the service provider to the microgrid; a price for clearing carbon allowances in the internal market; The carbon emission quota sold by the service provider to the microgrid; t and T both represent the meaning of time, where t∈T; It means that the distribution network service provider includes the income from buying carbon, selling electricity, and selling carbon; It means that the profit difference between the distribution network service provider's purchase and sale of electricity from the upper-level power grid is maximized; If the optimization problem is a convex optimization problem, the objective function model of the distribution network service provider can be transformed into the following model: In formula (8): N price is the sample data set of electricity price of the upper power grid, |N price | is the total number of corresponding sample data; are the sample values ​​of the electricity purchase price and electricity sales price of the lth upper-level power grid respectively; are the deviation values ​​of the purchase price and sales price of electricity of the lth upper power grid respectively; are the dual multipliers of the Wasserstein sphere radius constraints of the power purchase price and power sales price of the upper power grid respectively; l represents the lth upper power grid; Use duality theory to process the above model and transform the min-max model into a min model: 2.2: Interaction model with external power grid: The distribution network service provider interacts with the external power grid for electric energy. The interaction model is as follows: P t buy -P t sell =P t A2M (10); Where: P t buy , P t sell They are the amount of electricity purchased and sold by the service provider from the external power grid; The upper limit of the service provider's electricity purchase and sale from external networks; P t A2M represents the equivalent interactive power from the service provider to the microgrid; 2.3: Carbon quota interaction model: The service provider purchases carbon emission quotas from the external carbon market and sells carbon emission quotas to the microgrid. The transaction model is as follows: Where: Represents the carbon emission credits sold by the service provider to the microgrid.

4. The distribution network service provider-multi-microgrid distributed robust optimization collaborative scheduling method taking into account multiple uncertainties according to claim 3 is characterized by: In step 3, the multi-microgrid system operation model includes the objective function of the multi-microgrid, the interaction constraints with the service provider, the constraints of the gas unit, the operation constraints of the power storage equipment, the thermal power balance constraint model, the electric power balance constraints of the park service provider after considering the DRCC equivalent, and the carbon trading model; 3.1: Objective function of multi-microgrid: In a multi-microgrid system, each microgrid independently conducts energy transactions with the service provider. Considering that the internal unit models of each microgrid are consistent, the objective function of microgrid operator 1 is as follows: In formula (15), P t M2A,1 is the equivalent interactive power from microgrid operator 1 to the service provider; Carbon emission quota purchased by microgrid operator 1 from the service provider; p gas is the gas price; p ess is the operation and maintenance cost coefficient of electric energy storage; Indicates the gas consumption of the cogeneration unit; Indicates the amount of gas consumed by the gas boiler; P t ech Indicates the charging power of the electric energy storage; P t edis Indicates the discharge power of the electric energy storage; It represents the difference between the price at which the microgrid sells electricity to the service provider and the price at which the microgrid buys carbon from that service provider; It represents the sum of the gas cost of the microgrid and the operation and maintenance cost of the point energy storage; 3.2: Interaction constraints with service providers: P t buy,1 -P t sell,1 =P t M2A,1 (16); Where: P t buy,1 , P t sell,1 They are the amount of electricity purchased and sold by microgrid operator 1 to the service provider; An upper limit on the amount of electricity that a microgrid can purchase and sell from a service provider; 3.3: Constraints of gas units: The combined heat and power unit burns natural gas to generate electricity and heat. Its power generation, thermal power and constraints are as follows: Where: P t CHP,e , P t CHP,h They are the power generation and heating power of the CHP unit respectively; are the power generation and heating efficiencies of the CHP unit, respectively; The amount of gas consumed by the CHP unit; It is the upper limit of power generation of CHP unit; Gas boilers burn natural gas to generate heat energy. Their heating power and constraints are as follows: Where: P t GB,h The heating power of the gas boiler; The heating efficiency of the gas boiler; Gas consumption for gas boilers; The upper limit of the heating capacity of the gas boiler; 3.4: Operation constraints of power storage equipment: Electric energy storage can achieve efficient utilization of energy storage resources. The state of charge and charging and discharging power constraints of electric energy storage are as follows: m ch +m dis ≤1(30); Where: Real-time storage capacity for electric energy storage; P1 edis Indicates the energy storage discharge power of microgrid 1; are the charging efficiency and discharging efficiency of electric energy storage respectively; P t ech , P t edis are the charging power and discharging power of the energy storage respectively; μ ch , μ dis They are the status bits for charging and discharging of the electric energy storage respectively; They are the upper and lower limits of the real-time storage capacity of the electric energy storage respectively; They are the upper limits of the charging power and discharging power of the electric energy storage, respectively; represents the electric energy storage capacity of microgrid 2 at the 24th hour; represents the electric energy storage capacity of microgrid 1 at the 23rd hour; represents the energy storage charging power of microgrid 2 at the 24th hour; represents the energy storage discharge power of microgrid 2 at the 24th hour; Indicates the storage capacity of the electric energy storage at the 24th hour; It represents the storage capacity of the electric energy storage at hour 0; 3.5: Thermal power balance constraint model: In formula (31): forecasted heat loads for microgrid operators; 3.6: Considering the power balance constraints of the campus service provider after DRCC equivalence: P t load,1 -P t CHP,e -(P t buy,1 -P t sell,1 )-(P t edis -P t ech )≤P W,C,bound (32); In formula (32), P t load,1 represents the predicted electric load of microgrid operator 1; P W,C,bound represents the worst lower bound of renewable energy output of microgrid operator 1; P t buy,1 represents the carbon cost of electricity purchased by microgrid operator 1; P t sell,1 represents the carbon revenue from electricity sales of microgrid operator 1; 3.7: Carbon trading model: Where: is the actual carbon emissions of microgrid operator 1; is the correlation coefficient; Carbon emission quota for microgrid operator 1; is the correlation coefficient; Represents the carbon emission quota purchased by microgrid operator 1 from the service provider.

5. The distribution network service provider-multi-microgrid distributed robust optimization collaborative scheduling method taking into account multiple uncertainties according to claim 4, characterized in that: In step 4, the P2P transaction network game model between distribution network service providers and multiple microgrids based on Wasserstein distribution robust optimization includes equations (4) to (36).

6. The distribution network service provider-multi-microgrid distributed robust optimization collaborative scheduling method considering multiple uncertainties according to claim 5, characterized in that: Based on the P2P trading network game model between the distribution network service provider and multiple microgrids, a global electricity-carbon balance coupling model is established, including equations (37) to (38): P t A2M =P t M2A,1 +P t M2A,2 +P t M2A,3 :p e (37); In the formula, p e clearing electricity prices for the internal market; The price of carbon quotas for clearing the internal market e is the clearing electricity price in the internal market, which are the dual variables of the above equation; P t A2M represents the interactive power between the service provider and the microgrid; P t M2A,1 represents the interactive power between microgrid 1 and the service provider; P t M2A,2 represents the interactive power between microgrid 2 and the service provider; P t M2A,3 Indicates the interactive power between microgrid 3 and the service provider; Indicates the interactive carbon quota between the service provider and the microgrid; It represents the interactive carbon quota between the service provider and microgrid 1; It represents the interactive carbon quota between the service provider and microgrid 2; Represents the interactive carbon quota between the service provider and microgrid 3.

7. The distribution network service provider-multi-microgrid distributed robust optimization collaborative scheduling method considering multiple uncertainties according to claim 6, characterized in that: The step 5 comprises the following steps: Step 5.1: For the P2P trading network game model, use the duality principle to construct an equivalent KKT system in parallel; Step 5.2: The model shown in equation (39) is converted into a MILP problem through the big M method to solve it, so as to locate the equilibrium state of the multi-agent game in the park service provider-multi-microgrid system; Step 5.3: To verify the existence and uniqueness of the equilibrium between multi-agent games, the KKT conditions are combined to calculate the fixed points for judgment, and finally it is concluded that the Nash equilibrium solution exists and is unique, and the solution is directly obtained.

8. The distribution network service provider-multi-microgrid distributed robust optimization collaborative scheduling method taking into account multiple uncertainties according to claim 7, characterized in that: In step 5, the solution steps are as follows: According to the P2P trading network game model, the decision variable on the park service provider side is the power sold by the park service provider to the microgrid. t A2M The power purchased by the microgrid side P t M2A,1 , P t M2A,2 and P t M2A,3 Meeting power balance constraints; Decision variables on the park service provider side: The carbon emission quota sold by the park service provider to the microgrid Purchase carbon quotas with microgrid side and Meeting carbon quota balance constraints; Therefore, the mutual constraints between the decision variables of multiple agents lead to mutual constraints in the strategy space. Therefore, the multi-agent non-cooperative game is modeled as a generalized Nash equilibrium problem (GNEP), which is defined as shown in the following formula, and the Nash equilibrium of the multi-agent game is solved by the stationary point optimization method: Where: C i 、x i , X i are the objective function, decision variables, and strategy space of participant i respectively; x -i is the determining variable of all participants except participant i; g i is the inequality constraint of participant i; h i is the equality constraint for participant i; Firstly, the P2P trading network game model constructs an equivalent KKT system and sets it in parallel, and then converts it into a MILP problem through the big M method to solve it, so as to locate the equilibrium state of the multi-agent game in the park service provider-multi-microgrid system; In order to verify the existence and uniqueness of equilibrium, the KKT conditions are directly combined to calculate the fixed point judgment; the Nash equilibrium solution of the model shown in formula (39) exists and is unique, so it can be solved directly.

9. The distribution network service provider-multi-microgrid distributed robust optimization collaborative scheduling method taking into account multiple uncertainties according to claim 8, characterized in that: Find equilibrium solutions for situations where the number of entities increases and modeling changes; 1) KKT system for: Where: λ1 is the dual variable of the equality constraint; μ1 is the dual variable of the inequality constraint; represents the gradient of the solution of the KKT system; C i represents the i-th objective function; h i represents the i-th equality constraint; g i represents the i-th inequality constraint; ⊥ represents constraints that satisfy both inequality and equality; i represents a variable; The equilibrium point solution is as follows: Where: x A 、x M are the decision variables of microgrid and park service provider respectively; U VPP For the park service provider - the overall cost of multiple microgrid systems; x M1 represents the decision variable of microgrid 1; x M2 represents the decision variable of microgrid 2; x M3 represents the decision variables of microgrid 3; C CIESO represents the objective function of the distribution network service provider; C MGO,1 It is expressed as the objective function of microgrid 1; C MGO,2 Expressed as the objective function of microgrid 2; C MGO,3 It is expressed as the objective function of microgrid 3; x A represents the decision variable of the service provider; Indicates the KKT system of the service provider; represents the KKT system of microgrid 1; represents the KKT system of microgrid 2; Represents the KKT system of the microgrid 3.

Images