Virtual power plant planning-operation two-stage distribution robust optimization method based on scene driving

Abstract

The invention discloses a virtual power plant planning-operation two-stage distribution robust optimization method based on scene driving, and the method comprises the steps: S1, building a capacity configuration model and a master-slave game model based on a master-slave game framework of a power supply, a load and energy storage; s2, considering the correlation and uncertainty of fan and photovoltaic output to construct an uncertainty set of wind and light output probability distribution, and generating a plurality of typical scenes through a sampling and clustering algorithm; s3, performing comprehensive norm constraint on the confidence interval of the uncertain set under the driving of a typical scene, and establishing a distribution robust optimization model of the virtual power plant in two stages of planning and operation, wherein the distribution robust optimization model considers uncertainty; wherein the distributed robust optimization model comprises a capacity configuration model and a master-slave game model; and S4, solving the distribution robust optimization model containing the master-slave game model based on the modified Kriging model and a column and constraint generation algorithm. By utilizing the method, the influence of uncertainty on the planning rationality and the operation economy of the virtual power plant can be reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of distributed robust optimization of virtual power plants, and in particular, relates to a two-stage distributed robust optimization method for virtual power plant planning and operation based on scenario driving. Background Art

[0002] Under the new power system operation mechanism with electricity as the main body, the connection between heterogeneous energy sources such as electricity, gas, heat and cold is becoming increasingly close. However, the large-scale access to distributed resources such as wind power, photovoltaics, and cogeneration units has brought severe challenges to the dispatching and operation of the existing power system. Renewable energy generation is uncertain, its capacity is small, and the power output is highly volatile. Large-scale grid connection will seriously affect the stability of the power system.

[0003] With the continuous improvement of the penetration rate of renewable energy in the distribution network and the continuous increase in the proportion of electrification of user terminal energy consumption, the power system has gradually changed its operating mode of adjusting the output of generators according to load changes, and transformed to coordinated and optimized operation on the power supply side, load side and energy storage side. Therefore, distributed resources make the system's operating mode and energy coupling relationship more complex, which also greatly increases the difficulty of energy management within the system. How to tap the interactive ability and regulation ability of source, load and storage flexibility resources is a key issue in the development of new power systems.

[0004] Virtual power plants break the limitations of geographical location and resource types. Through communication, metering and other control methods, distributed resources scattered in adjacent areas are aggregated into an organic whole that can accept unified scheduling. As an entity integrated in a "virtual" form, virtual power plants can independently perform energy scheduling and market transactions, providing an effective solution for the rational consumption and centralized grid connection of distributed resources. In the three links of source, load and storage, the output of renewable energy on the source side and the energy consumption of users on the load side are uncertain, which affects the rationality and economy of the internal resource capacity configuration and optimal scheduling of virtual power plants.

[0005] At present, there are mainly two methods for dealing with the operational uncertainty of virtual power plants: stochastic optimization methods and robust optimization methods. The probability distribution of variables in stochastic optimization is fixed or estimated based on historical data. For uncertainty problems, the deterministic probability distribution is too optimistic and may have large errors, which will affect the economic efficiency of system operation. Robust optimization considers the worst scenario, but the actual operating scenario is often not so extreme, so its optimization results are too pessimistic, resulting in overly conservative decisions. In terms of solving optimization methods, intelligent optimization algorithms are currently mainly used, and the speed and accuracy of the solution still need to be further improved. In addition, the existing optimization methods focus on considering the impact of uncertainty on energy scheduling in the operation stage of virtual power plants, but lack effective quantitative evaluation of the rationality of capacity configuration in the planning stage.

[0006] In summary, in order to reduce the impact of uncertainty on the rationality of virtual power plant planning and the economic efficiency of operation, it is necessary to design a new optimization method that takes into account both economy and conservatism and a supporting efficient solution algorithm. Summary of the invention

[0007] In view of the problems existing in existing methods such as the inability to balance conservatism and economy, the lack of quantitative evaluation of planning rationality, and the need to improve solution performance, the purpose of the present invention is to provide a scenario-driven two-stage distributed robust optimization method for virtual power plant planning and operation, so as to effectively deal with uncertainty problems and reduce the impact of uncertainty on the rationality of virtual power plant planning and the economy of operation.

[0008] A scenario-driven virtual power plant planning-operation two-stage distributed robust optimization method, characterized by comprising the following steps:

[0009] S1. Based on the master-slave game framework of power source, load and energy storage, the capacity configuration and energy scheduling of virtual power plants are considered to build a capacity configuration model and a master-slave game model;

[0010] S2. Considering the correlation and uncertainty of wind turbine and photovoltaic output, the uncertainty set of wind and solar output probability distribution is constructed, and multiple typical scenarios are generated through sampling and clustering algorithms;

[0011] S3. Under the drive of typical scenarios, comprehensive norm constraints are imposed on the confidence intervals of the uncertain set, and a distributed robust optimization model of the virtual power plant considering uncertainty in the two stages of planning and operation is established; wherein the distributed robust optimization model includes the capacity configuration model and the master-slave game model established in step S1;

[0012] S4. Based on the modified Kriging model and the column and constraint generation algorithm, the distributed robust optimization model including the master-slave game model is solved.

[0013] The present invention fully considers the impact of uncertainty on the medium- and long-term capacity configuration and day-ahead optimal scheduling of internal resources in virtual power plants, and solves the problem of difficulty in balancing conservatism and economy in current uncertainty optimization. In addition, the present invention integrates the proxy model algorithm and the intelligent optimization algorithm to solve the distributed robust optimization model, which can effectively improve the solution efficiency compared with conventional solution methods.

[0014] The specific steps of S1 are:

[0015] S101. With the goal of minimizing the investment cost of the virtual power plant in the planning stage, a capacity configuration model is constructed for the distributed resources aggregated within it, where the distributed resources include wind turbines, photovoltaics, cogeneration units, biomass boilers, electric ammonia systems, and energy storage systems;

[0016] S102. Establish an energy-benefit interaction model with the virtual power plant as the leader and the user as the follower, i.e., a master-slave game model; the virtual power plant prioritizes the energy sales price to the user based on the load day forecast information, and the user adjusts the energy consumption through the demand response mechanism under the guidance of the virtual power plant price signal. The two interact with each other in sequence in terms of pricing strategy and energy consumption strategy. The virtual power plant is in a dominant position in the interaction process, forming a master-slave game with the user;

[0017] S103. In the master-slave game relationship, the upper virtual power plant conducts electric heat and ammonia transactions with the lower users with the goal of minimizing the day-ahead operating cost;

[0018] S104. In response to the price strategy issued by the upper-level virtual power plant in the day-ahead stage, the lower-level users adjust their energy consumption through demand response with the goal of maximizing comprehensive benefits.

[0019] In step S101, the investment cost of the virtual power plant in the planning stage is as follows:

[0020]

[0021] In the formula, C inv represents the daily average investment cost of VPP; i represents the type of equipment inside VPP; Ω DG is the set of internal devices of VPP, i∈Ω DG ; C DG,i N represents the investment cost of the i-th type of equipment in the VPP; DG,i is the number of installations of the i-th type of equipment; r inv represents the discount rate; TL i represents the economic useful life of the i-th type of equipment; inv,i is the unit investment cost of the i-th type of equipment.

[0022] In step S102, the objective function and constraints of the master-slave game model are as follows:

[0023] minC op (λ et ,λ ht ,λ at ,P et ,Q ht ,B at )

[0024]

[0025] In the formula, Ω vpp and Ω user represents the master-slave game strategy set of the virtual power plant and the user respectively; C op and C eu Respectively represent the objective functions of the virtual power plant and the user in the day-ahead stage; λet,t ,λ ht,t ,λ at,t Respectively represent the transaction prices of electricity, heat and ammonia between the virtual power plant and the users; P et,t ,Q ht,t ,B at,t They respectively represent the transaction volumes of electricity, heat and ammonia between the virtual power plant and users.

[0026] The specific steps of S2 are:

[0027] S201. Analyze the historical sample data of wind and solar power output of virtual power plants based on the kernel density estimation method, and fit the probability distribution model of wind and solar power output, so as to deeply explore the wind and solar power distribution characteristics in the area where the virtual power plant is located;

[0028] S202. Since wind power and photovoltaic power have complementary outputs in some periods and present a negative correlation, a Frank-Copula function that takes into account both non-negative and negative correlations between random variables is selected to describe the correlation between wind and photovoltaic power.

[0029] S203, using the Copularnd function to sample and generate a certain number of edge distribution function values ​​(u, v) that satisfy the Frank-Copula function relationship, and performing The inverse transformation is used to obtain the wind and solar power output scenario (x, y) in each period that satisfies the joint distribution function;

[0030] S204. To ensure the fitting accuracy of the Frank-Copula function, the K-means clustering algorithm is used to cluster the sampling results, generate multiple typical scenarios, and obtain the probability of occurrence of each typical scenario.

[0031] The specific steps of S3 are:

[0032] S301. Based on multiple typical scenarios reflecting the uncertainty and correlation of wind and solar power output generated in step S2, the probability distribution uncertainty set in the worst scenario is optimized, and the planning-operation two-stage decision-making scheme under the worst scenario is obtained; wherein, the distributed robust optimization model of the virtual power plant in the two stages of planning and operation includes a capacity configuration model and a master-slave game model, with the economic optimization of the two stages as the goal, and the objective function is as follows:

[0033]

[0034] In the formula, x and y s They represent the decision variables in the planning stage and the operation stage respectively; A is the coefficient matrix corresponding to the variables in the objective function of the planning stage; B and C are the coefficient matrices of the quadratic term and the linear term in the objective function of the operation stage respectively; d is the constant term coefficient of the objective function of the operation stage; X is the set of variables in the planning stage; Ω πis the uncertainty set of wind and solar power output probability distribution; s is the typical scene number; N s is the number of typical scenarios; π s represents the probability of scene s occurring; Y(x,π s ) is a set of variables in the running phase, representing x and y in scene s s The coupling relationship;

[0035] S302, the uncertainty set of wind and solar power output probability distribution and the confidence constraints satisfied by the probability distribution of each scenario are shown in the following formula:

[0036]

[0037]

[0038] In the formula, R + Represents a positive real number; π s,0 represents the initial probability of a typical scenario; θ1 and θ ∞ They correspond to the probability fluctuation deviations allowed under the 1-norm and ∞-norm constraints respectively; Pr(·) represents the probability distribution function; S is the number of initial scenarios.

[0039] The specific steps of S4 are:

[0040] S401, using the Kriging model to visualize the interactive relationship between the virtual power plant and the user in the master-slave game model, thereby simplifying the double-layer optimization problem into a single-layer optimization problem;

[0041] S402, in order to reduce the impact of the number of sample data on the model fitting accuracy, a Kriging model correction mechanism that considers dynamic update of sample data is designed;

[0042] S403, decompose the two-stage distributed robust optimization problem of planning and operation into a main problem and sub-problems, embed the master-slave game model into the main problem, and use the column and constraint generation algorithm to perform phased iterative solution.

[0043] Step S401 includes:

[0044] Firstly, Latin hypercube sampling is used to generate a certain number of initial sample points containing the electric thermal ammonia transaction prices per unit time between the virtual power plant and the user for each discrete scenario. The number of initial sample points in each group is selected to be 10 times the number of transaction price variables. The initial sample points are substituted into the lower-level user model to obtain the electric thermal ammonia transaction volume corresponding to each group of transaction prices, thereby constructing an initial sample data set containing transaction prices and transaction volumes.

[0045] Then, the Kriging model is used to predict the mapping relationship between transaction price and transaction volume in the sample data set, so as to fit and replace the lower model according to the mapping relationship between the upper and lower layers. The single-layer optimization model of each scenario obtained after fitting is as follows:

[0046]

[0047] Where H(·) represents the mapping relationship between the upper and lower layer interactions predicted by the Kriging model.

[0048] Step S402 includes:

[0049] First, the initial sample data set of each scene is constructed by uniform sampling and calling the lower-level model, and the initial Kriging model is established based on the mapping relationship within the data set;

[0050] Then, the main problem corresponding to the upper virtual power plant model is solved to obtain the optimal transaction price and transaction volume. However, the transaction volume at this time is based on the predicted value of the Kriging model rather than the actual value. Therefore, it is necessary to substitute the optimal transaction price into the lower user model to obtain the actual transaction volume, and add the optimal transaction price and its corresponding actual transaction volume as a high-quality data set to the initial sample data set to realize the correction of the Kriging model.

[0051] Finally, the sample data is dynamically updated through multiple iterations until the objective function value of the main problem converges.

[0052] In step S403, the main problem and sub-problems obtained by decomposition are expressed as follows:

[0053] Main question:

[0054]

[0055] Where l represents the number of iterations, l∈{1,2,3,…,N l}, where N l is the total number of iterations;

[0056] (2) Sub-problems;

[0057]

[0058] The max-min double-layer structure of the sub-problem includes an outer maximum problem and an inner minimum problem. The probability distribution of the worst scenario within the confidence interval is found based on the solution of the main problem. Since the outer discrete scenario probability π of the sub-problem s,l and the inner runtime variable y s,lThey are independent of each other, so the sub-problems are divided into two steps for solution, that is, first solve the inner minimum problem, and then solve the outer maximum problem. The inner minimum problem can be obtained from the solution of the main problem.

[0059] Compared with the prior art, the present invention has the following beneficial effects:

[0060] 1. The present invention fits the probability distribution model of wind and solar power output based on the kernel density estimation method, so as to deeply explore the wind and solar distribution characteristics of the area where the virtual power plant is located, and at the same time selects the Frank-Copula function that can take into account the non-negative and negative correlation between random variables to effectively characterize the correlation of wind and solar power output. In order to ensure the fitting accuracy of the Frank-Copula function, the present invention uses a clustering algorithm to cluster the sampling results and generate multiple typical scenarios that reflect uncertainty and correlation. Compared with the existing uncertainty sampling method, the present invention comprehensively considers the temporal and spatial correlation between uncertainty variables, and is therefore more scientific.

[0061] 2. Driven by multi-dimensional uncertainty scenarios, this invention uses a distributed robust optimization method to construct an uncertain set of wind and solar output probability distribution, and designs a two-stage distributed robust optimization model for planning and operation that takes into account virtual power plant capacity configuration and energy scheduling based on the source-load-storage master-slave game framework. This model fully quantifies the impact of the uncertainty of virtual power plant operation on medium- and long-term planning and day-ahead scheduling, which helps to formulate more economical capacity configuration and optimized scheduling plans.

[0062] 3. The present invention designs a dynamic update mechanism for sampling points, modifies the traditional Kriging model, and uses the modified Kriging model to achieve effective fitting and transformation of the master-slave game two-layer optimization model. The present invention decomposes the distributed robust optimization problem into a main problem and sub-problems, and nests the modified Kriging model into the column and constraint generation algorithm, and performs hierarchical decoupling optimization and phased iterative solution on the two-stage optimization problem containing the two-layer optimization model. Compared with traditional analytical methods and intelligent optimization algorithms, the present invention can effectively reduce the solution time and reduce the complexity of the problem by nesting the proxy model for solution. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 It is an overall framework diagram of the two-stage distributed robust optimization method for virtual power plant planning and operation in an embodiment of the present invention;

[0064] Figure 2 A flowchart of solving a distributed robust optimization problem based on a modified Kriging model and a column and constraint generation algorithm in an embodiment of the present invention;

[0065] Figure 3It is a convergence process diagram of solving the main problem and sub-problems in the modified Kriging model algorithm, the traditional Kriging model algorithm, the particle swarm algorithm, and the KKT condition embedding column and constraint generation algorithm in the embodiment of the present invention. DETAILED DESCRIPTION

[0066] The present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be pointed out that the embodiments described below are intended to facilitate the understanding of the present invention and do not have any limiting effect on the present invention.

[0067] like Figure 1 As shown, this embodiment provides a scenario-driven virtual power plant planning-operation two-stage distributed robust optimization method, including the steps of:

[0068] S1. Based on the master-slave game framework of the three parties of source, load and storage, a conventional optimization model for planning and operation considering the capacity configuration and energy scheduling of virtual power plants is established.

[0069] S2. Considering the correlation and uncertainty of wind turbine and photovoltaic output, the uncertainty set of wind and solar output probability distribution is constructed, and multiple typical scenarios are generated through sampling and clustering algorithms.

[0070] S3. Under scenario-driven conditions, comprehensive norm constraints are imposed on the confidence intervals of the uncertain set, and a two-stage distributed robust optimization model for virtual power plant planning and operation considering uncertainties is established.

[0071] S4. Based on the modified Kriging model and the column and constraint generation algorithm, the distributed robust optimization problem containing the master-slave game model is solved.

[0072] Furthermore, the detailed steps of step S1 include:

[0073] S101. With the goal of minimizing the mid- and long-term investment cost of virtual power plants, capacity allocation is carried out for distributed resources such as wind turbines, photovoltaics, cogeneration units, biomass boilers, electric ammonia systems, and energy storage systems aggregated within the virtual power plant.

[0074] The investment costs of a virtual power plant in the planning stage are as follows:

[0075]

[0076] In the formula, C inv represents the daily average investment cost of VPP; i represents the type of equipment inside VPP; Ω DG is the set of internal devices of VPP, i∈Ω DG ; C DG,i N represents the investment cost of the i-th type of equipment in the VPP; DG,i is the number of installations of the i-th type of equipment; r inv represents the discount rate; TLi represents the economic useful life of the i-th type of equipment; inv,i is the unit investment cost of the i-th type of equipment.

[0077] S102. Establish an energy-benefit interaction model with virtual power plants as leaders and users as followers. Virtual power plants prioritize energy sales prices to users based on load forecast information. Users adjust energy consumption through demand response mechanisms under the guidance of virtual power plant price signals. The two interact with each other in sequence in terms of pricing strategies and energy consumption strategies. Virtual power plants are in a dominant position in the interaction process, forming a master-slave game with users.

[0078] The objective function and constraints of the master-slave game model between the virtual power plant and the user are as follows:

[0079] minC op (λ et ,λ ht ,λ at ,P et ,Q ht ,B at )(2)

[0080]

[0081] In the formula, Ω vpp and Ω user represents the master-slave game strategy set of the virtual power plant and the user respectively; C op and C eu Respectively represent the objective functions of the virtual power plant and the user in the day-ahead stage; λ et,t ,λ ht,t ,λ at,t Respectively represent the transaction prices of electricity, heat and ammonia between the virtual power plant and the users; P et,t ,Q ht,t ,B at,t They respectively represent the transaction volumes of electricity, heat and ammonia between the virtual power plant and users.

[0082] S103. In the master-slave game relationship, the upper virtual power plant conducts electric heat and ammonia transactions with the lower users with the goal of minimizing the day-ahead operating cost.

[0083] The objective function C of the virtual power plant in the day-ahead operation phase op As shown in the following formula, it specifically includes the unit fuel cost C raw , Unit operation and maintenance cost C om , energy transaction cost with the parent network C grid , and the user's energy sales income C user and demand response cost C DR .

[0084] minC op=C raw +C om +C grid +C DR -C user (4)

[0085] The unit fuel cost is as follows, including the natural gas purchase cost of the cogeneration unit and the biomass raw material acquisition cost of the biomass boiler:

[0086]

[0087] Where, L g is the lower calorific value of natural gas; bio is the purchase price of biomass raw materials; P gt,t Indicates the electrical power output of the cogeneration unit; Q bfb,t Represents the thermal power output of the biomass boiler; ρ g is the density of biogas; η gt and η bfb is the operating efficiency of the cogeneration unit and biomass boiler; η g It indicates the volume of natural gas produced by fermentation of unit biomass raw materials;

[0088] The unit operation and maintenance costs are shown below, including daily inspections and fault recovery of wind turbines, photovoltaics, cogeneration units, biomass boilers, electric ammonia systems, and energy storage systems.

[0089]

[0090] Where P pv,t and P wt,t Represent the power output of photovoltaic and wind power respectively; P PA,t represents the electric power consumption of P2A; P ch,t ,Q ch,t ,B ch,t and P dis,t ,Q dis,t ,B dis,t Respectively represent the energy absorbed and released by the battery, heat storage device and ammonia storage device in the energy storage system per unit time; λ pv ,λ wt ,λ CHP ,λ PA ,λ bfb ,λ BSS ,λ HSS ,λ GSS They represent the unit power operation and maintenance costs of photovoltaics, wind turbines, cogeneration units, electric ammonia systems, biomass boilers, batteries, heat storage devices and ammonia storage devices respectively.

[0091] The energy transaction cost with the parent network is shown below, which includes heterogeneous energy transactions between the virtual power plant and the parent power network, thermal network and ammonia network.

[0092]

[0093] In the formula, λ eb,t and λ es,t Respectively represent the purchase and sale prices of electricity between the virtual power plant and the upper power grid; P eb,t and P es,t Respectively represent the amount of electricity purchased and sold by the virtual power plant and the upper-level power grid; hb and λ hs They represent the heat purchase and sale prices of the virtual power plant and the upper-level heat network respectively; Q hb,t and Q hs,t Respectively represent the heat purchase and sale between the virtual power plant and the upper-level heat network; ab and λ as Respectively represent the purchase and sale ammonia prices of the virtual power plant and the upper ammonia network; B ab,t and B as,t They respectively represent the amount of ammonia purchased and sold by the virtual power plant and the upper-level ammonia network.

[0094] The revenue from the virtual power plant selling electricity, heat, and ammonia to users is as follows:

[0095]

[0096] The demand response costs of virtual power plants and users are shown below, which specifically include the compensation costs for reducing electricity load, thermal load and ammonia load on the user side.

[0097]

[0098] In the formula, β re ,β rh ,β ra are the compensation prices for the reduction of electric thermal ammonia load due to demand response; P re,t ,Q rh,t ,B ra,t They are the reduction in electric thermal ammonia load due to demand response.

[0099] S104. In response to the price strategy issued by the upper-level virtual power plant in the day-ahead stage, the lower-level users adjust their energy consumption through demand response with the goal of maximizing comprehensive benefits.

[0100] Comprehensive benefits for users C eu Including user demand response benefits and energy purchase costs, its objective function is as follows:

[0101] maxC eu =C DR -C user (10)

[0102] Furthermore, the detailed steps of step S2 include:

[0103] S201. Analyze the historical sample data of wind and solar power output of virtual power plants based on the kernel density estimation method, fit the probability distribution model of wind and solar power output, and deeply explore the wind and solar power distribution characteristics in the area where the virtual power plant is located.

[0104] Let x1,x1,…,x n is a sample that is independent and identically distributed with the wind turbine output x, then the probability density function f that x obeys is h The kernel density estimation of (x) is as follows. The probability density function f of photovoltaic output y h The kernel density estimate of (y) can be obtained similarly.

[0105]

[0106] Where n is the number of samples; h is the window width; K(·) represents the kernel function, which has many forms. The formula of the Gaussian kernel function selected in this chapter is as follows:

[0107]

[0108] S202. Since wind power and photovoltaic power have complementary outputs in some periods and show a negative correlation, a Frank-Copula function that can take into account both non-negative and negative correlations between random variables is selected to describe the correlation between wind and photovoltaic power.

[0109] The introduced Frank-Copula function C(u,v) that reflects the relationship between the marginal distribution and joint distribution of wind and solar power output in each period is shown as follows:

[0110]

[0111] In the formula, F X (x) and F Y (y) represent the marginal distribution functions of wind power and photovoltaic power output respectively; F h (x,y) is the joint distribution function of wind and solar power output.

[0112] S203, using the Copularnd function to sample and generate a certain number of edge distribution function values ​​(u, v) that satisfy the Frank-Copula function relationship, and performing The inverse transformation of can obtain the wind and solar power output scenario (x, y) in each period that satisfies the joint distribution function.

[0113] S204. Since the number of initial scenes obtained after sampling is relatively large, in order to ensure the fitting accuracy of the Frank-Copula function, the K-means clustering algorithm is used to cluster the sampling results, generate multiple typical scenes, and obtain the probability of occurrence of each typical scene.

[0114] Furthermore, the detailed steps of step S3 include:

[0115] S301. Based on multiple typical scenarios reflecting the uncertainty and correlation of wind and solar power output generated in step S2, the probability distribution uncertainty set in the worst scenario is optimized, and a planning-operation two-stage decision plan under the worst scenario is obtained.

[0116] The distributed robust optimization model of the virtual power plant in the two stages of planning and operation includes the capacity configuration model established in step S101 and the master-slave game model established in step S102, with the economic optimization of the two stages as the goal. The objective function is as follows:

[0117]

[0118] In the formula, x and y s They represent the decision variables in the planning stage and the operation stage respectively; A is the coefficient matrix corresponding to the variables in the objective function of the planning stage; B and C are the coefficient matrices of the quadratic term and the linear term in the objective function of the operation stage respectively; d is the constant term coefficient of the objective function of the operation stage; X is the set of variables in the planning stage; Ω π is the uncertainty set of wind and solar power output probability distribution; s is the typical scene number; N s is the number of typical scenarios; π s represents the probability of scene s occurring; Y(x,π s ) is a set of variables in the running phase, which can be represented as x and y in scene s s coupling relationship.

[0119] S302: To ensure that the actual probability distribution of each scenario still converges to the theoretical probability distribution when the number of generated scenarios is large enough, the confidence intervals of the 1-norm and the ∞-norm are combined to constrain the fluctuation range of the probability distribution.

[0120] The uncertainty set of wind and solar power output probability distribution and the confidence constraints satisfied by the probability distribution of each scenario are shown in the following formula:

[0121]

[0122]

[0123] In the formula, R + Represents a positive real number; π s,0 represents the initial probability of a typical scenario; θ1 and θ∞ They correspond to the probability fluctuation deviations allowed under the 1-norm and ∞-norm constraints respectively; Pr(·) represents the probability distribution function; S is the number of initial scenarios.

[0124] Let α1 and α ∞ They represent the confidence level of the probability distribution of each scenario in the 1-norm and ∞-norm confidence intervals respectively. Let the quantiles of formula (16) be α1 and α ∞ , then the probability fluctuation deviation can be calculated as follows:

[0125]

[0126] Further, Figure 2 This is a flow chart for solving the distributed robust optimization problem based on the modified Kriging model and the column and constraint generation algorithm in step S4. A dynamic update mechanism of sampling points is designed on the basis of the traditional Kriging model for correction, and the modified Kriging model is embedded in the column and constraint generation algorithm to iteratively solve the two-stage distributed robust optimization model containing the master-slave game two-level optimization problem. The detailed steps include:

[0127] S401. The interactive relationship between the virtual power plant and the user in the master-slave game model is visualized through the Kriging model, thereby simplifying the double-layer optimization problem into a single-layer optimization problem.

[0128] First, Latin hypercube sampling is used to generate a certain number of initial sample points containing the unit-time electric thermal ammonia transaction prices of virtual power plants and users for each discrete scenario. The number of initial sample points in each group is selected as 10 times the number of transaction price variables. The initial sample points are substituted into the lower-level user model to obtain the electric thermal ammonia transaction volume corresponding to each group of transaction prices, thereby constructing an initial sample data set containing transaction prices and transaction volumes. Then, the Kriging model is used to predict the mapping relationship between transaction prices and transaction volumes in the sample data set, so as to fit and replace the lower-level model according to the mapping relationship between the upper and lower layers. The single-layer optimization model of each scenario obtained after fitting is shown in formula (18):

[0129]

[0130] Where H(·) represents the mapping relationship between the upper and lower layer interactions predicted by the Kriging model.

[0131] S402: In order to reduce the impact of the number of sample data on the model fitting accuracy, a Kriging model correction mechanism that considers the dynamic update of sample data is designed.

[0132] First, the initial sample data set of each scenario is constructed by uniform sampling and calling the lower-level model, and the initial Kriging model is established based on the mapping relationship within the data set. Then, the main problem corresponding to the upper-level virtual power plant model is solved to obtain the optimal transaction price and transaction volume. However, the transaction volume at this time is based on the predicted value of the Kriging model rather than the actual value. Therefore, it is necessary to substitute the optimal transaction price into the lower-level user model to solve the actual transaction volume, and add the optimal transaction price and its corresponding actual transaction volume as a high-quality data set to the initial sample data set to realize the correction of the Kriging model. Finally, the sample data is dynamically updated through multiple iterations until the objective function value of the main problem converges.

[0133] S403, decompose the two-stage distributed robust optimization problem of planning and operation into a main problem and sub-problems, embed the master-slave game model into the main problem, and use the column and constraint generation algorithm to perform phased iterative solution.

[0134] The main problem and sub-problems obtained by decomposition are expressed as follows:

[0135] (1) Main question:

[0136]

[0137] Where l represents the number of iterations, l∈{1,2,3,…,N l}, where N l is the total number of iterations.

[0138] (2) Sub-problems;

[0139]

[0140] The max-min double-layer structure of the sub-problem includes an outer maximum problem and an inner minimum problem. Based on the solution of the main problem, the probability distribution of the worst scenario within the confidence interval is found. s,l and the inner runtime variable y s,l They are independent of each other, so the sub-problem can be solved in two steps, that is, first solve the inner minimum problem, and then solve the outer maximum problem. The inner minimum problem can be obtained from the solution of the main problem.

[0141] The detailed solution steps of step S403 include:

[0142] S4031: Set the model lower bound L B is -∞, the upper bound of the model U B is +∞, the number of iterations l is 1, and the convergence threshold ε r For 10 -2 , the initial probability of each scene is π s,0 .

[0143] S4032: Solve the main problem and obtain the optimal solution And update the lower bound

[0144] S4033: Optimal variables based on the master problem Solve the sub-problem and get the probability of the worst scenario and variables And the objective function value of the subproblem And update the upper bound

[0145] S4034: If U is satisfied B -L B ≤ε r , then stop the iteration and output the optimal solution Otherwise, set l = l + 1 and return to step 2 to update the probability of each scene And add the variable y in the main problem s,l+1 and variable y s,l+1 Relevant constraints.

[0146] Further, in step S403, the modified Kriging model algorithm, the traditional Kriging model algorithm, the particle swarm algorithm, the KKT condition embedding sequence and the constraint generation algorithm are used to solve the main problem and the sub-problems. Figure 3 The figure is a comparison of the convergence performance of the method of the present invention and other methods. When the modified Kriging model algorithm designed by the present invention is used in conjunction with the column and constraint generation algorithm to solve the problem, the number of iterations is reduced by 37.50% and 61.54% respectively compared with the traditional Kriging model algorithm and KKT conditions. However, when the set maximum number of iterations is reached, the particle swarm algorithm still does not converge. Therefore, the method designed by the present invention can effectively improve the convergence speed of solving the distributed robust optimization problem.

[0147] The embodiments described above provide a detailed description of the technical solutions and beneficial effects of the present invention. It should be understood that the above are only specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, supplements and equivalent substitutions made within the scope of the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A scenario-driven two-stage distributed robust optimization method for virtual power plant planning and operation, characterized in that: The following steps are involved: S1. Based on the master-slave game framework of power source, load and energy storage, considering the capacity configuration and energy scheduling of virtual power plants, a capacity configuration model and a master-slave game model are constructed; S2. Considering the correlation and uncertainty of wind turbine and photovoltaic output, the uncertainty set of wind and solar output probability distribution is constructed, and multiple typical scenarios are generated through sampling and clustering algorithms; S3. Under the drive of typical scenarios, comprehensive norm constraints are imposed on the confidence intervals of the uncertain set, and a distributed robust optimization model of the virtual power plant considering uncertainty in the two stages of planning and operation is established; wherein the distributed robust optimization model includes the capacity configuration model and the master-slave game model established in step S1; S4. Based on the modified Kriging model and the column and constraint generation algorithm, the distributed robust optimization model including the master-slave game model is solved.

2. The scenario-driven two-stage distributed robust optimization method for virtual power plant planning and operation according to claim 1 is characterized in that: The specific steps of S1 are: S101. With the goal of minimizing the investment cost of the virtual power plant in the planning stage, a capacity configuration model is constructed for the distributed resources aggregated within it, where the distributed resources include wind turbines, photovoltaics, cogeneration units, biomass boilers, electric ammonia systems, and energy storage systems; S102. Establish an energy-benefit interaction model with the virtual power plant as the leader and the user as the follower, that is, a master-slave game model; the virtual power plant prioritizes the energy sales price to the user based on the load forecast information on the day before, and the user adjusts the energy consumption through the demand response mechanism under the guidance of the price signal of the virtual power plant. The two interact with each other in sequence in terms of pricing strategy and energy consumption strategy. The virtual power plant is in a dominant position in the interaction process, forming a master-slave game with the user; S103. In the master-slave game relationship, the upper virtual power plant conducts electric heat and ammonia transactions with the lower users with the goal of minimizing the day-ahead operating cost; S104. In response to the price strategy issued by the upper-level virtual power plant in the day-ahead stage, the lower-level users adjust their energy consumption through demand response with the goal of maximizing comprehensive benefits.

3. The scenario-driven two-stage distributed robust optimization method for virtual power plant planning and operation according to claim 2 is characterized in that: In step S101, the investment cost of the virtual power plant in the planning stage is as follows: In the formula, C inv represents the daily average investment cost of VPP; i represents the type of equipment inside VPP; Ω DG is the set of internal devices of VPP, i∈Ω DG ; C DG,i N represents the investment cost of the i-th type of equipment in the VPP; DG,i is the number of installations of the i-th type of equipment; r inv represents the discount rate; TL i represents the economic useful life of the i-th type of equipment; inv,i is the unit investment cost of the i-th type of equipment.

4. The scenario-driven two-stage distributed robust optimization method for virtual power plant planning and operation according to claim 3 is characterized in that: In step S102, the objective function and constraints of the master-slave game model are as follows: minC op (λ et ,λ ht ,λ at ,P et ,Q ht ,B at ) In the formula, Ω vpp and Ω user represents the master-slave game strategy set of the virtual power plant and the user respectively; C op and C eu They represent the objective functions of the virtual power plant and the user in the day-ahead stage respectively; et,t ,λ ht,t ,λ at,t Respectively represent the transaction prices of electricity, heat and ammonia between the virtual power plant and the user; P et,t ,Q ht,t ,B at,t They respectively represent the transaction volumes of electricity, heat and ammonia between the virtual power plant and users.

5. The scenario-driven two-stage distributed robust optimization method for virtual power plant planning and operation according to claim 4 is characterized in that: The specific steps of S2 are: S201. Analyze the historical sample data of wind and solar power output of virtual power plants based on the kernel density estimation method, and fit the probability distribution model of wind and solar power output, so as to deeply explore the wind and solar power distribution characteristics in the area where the virtual power plant is located; S202. Since wind power and photovoltaic power have complementary outputs in some periods and present a negative correlation, a Frank-Copula function that takes into account both non-negative and negative correlations between random variables is selected to describe the correlation between wind and photovoltaic power. S203, using the Copularnd function to sample and generate a certain number of edge distribution function values ​​(u, v) that satisfy the Frank-Copula function relationship, and performing The inverse transformation is used to obtain the wind and solar power output scenario (x, y) in each period that satisfies the joint distribution function; S204. To ensure the fitting accuracy of the Frank-Copula function, the K-means clustering algorithm is used to cluster the sampling results, generate multiple typical scenarios, and obtain the probability of occurrence of each typical scenario.

6. The scenario-driven two-stage distributed robust optimization method for virtual power plant planning and operation according to claim 5 is characterized in that: The specific steps of S3 are: S301. Based on multiple typical scenarios reflecting the uncertainty and correlation of wind and solar power output generated in step S2, the probability distribution uncertainty set in the worst scenario is optimized, and the planning-operation two-stage decision-making scheme under the worst scenario is obtained; wherein, the distributed robust optimization model of the virtual power plant in the two stages of planning and operation includes a capacity configuration model and a master-slave game model, with the economic optimization of the two stages as the goal, and the objective function is as follows: In the formula, x and y s They represent the decision variables in the planning stage and the operation stage respectively; A is the coefficient matrix corresponding to the variables in the objective function of the planning stage; B and C are the coefficient matrices of the quadratic term and the linear term in the objective function of the operation stage respectively; d is the constant term coefficient of the objective function of the operation stage; X is the set of variables in the planning stage; Ω π is the uncertainty set of wind and solar power output probability distribution; s is the typical scene number; N s is the number of typical scenarios; π s represents the probability of scene s occurring; Y(x,π s ) is a set of variables in the running phase, representing x and y in scene s s The coupling relationship; S302, the uncertainty set of wind and solar power output probability distribution and the confidence constraints satisfied by the probability distribution of each scenario are shown in the following formula: In the formula, R + Represents a positive real number; π s,0 represents the initial probability of a typical scenario; θ1 and θ ∞ They correspond to the probability fluctuation deviations allowed under 1-norm and ∞-norm constraints respectively; Pr(·) represents the probability distribution function; S is the number of initial scenarios.

7. The scenario-driven two-stage distributed robust optimization method for virtual power plant planning and operation according to claim 6 is characterized in that: The specific steps of S4 are: S401, using the Kriging model to visualize the interactive relationship between the virtual power plant and the user in the master-slave game model, thereby simplifying the double-layer optimization problem into a single-layer optimization problem; S402, in order to reduce the impact of the number of sample data on the model fitting accuracy, a Kriging model correction mechanism that considers dynamic update of sample data is designed; S403, decompose the two-stage distributed robust optimization problem of planning and operation into a main problem and sub-problems, embed the master-slave game model into the main problem, and use the column and constraint generation algorithm to perform phased iterative solution.

8. The scenario-driven two-stage distributed robust optimization method for virtual power plant planning and operation according to claim 7 is characterized in that: Step S401 includes: Firstly, Latin hypercube sampling is used to generate a certain number of initial sample points containing the electric thermal ammonia transaction prices per unit time between the virtual power plant and the user for each discrete scenario. The number of initial sample points in each group is selected to be 10 times the number of transaction price variables. The initial sample points are substituted into the lower-level user model to obtain the electric thermal ammonia transaction volume corresponding to each group of transaction prices, thereby constructing an initial sample data set containing transaction prices and transaction volumes. Then, the Kriging model is used to predict the mapping relationship between transaction price and transaction volume in the sample data set, so as to fit and replace the lower model according to the mapping relationship between the upper and lower layers. The single-layer optimization model of each scenario obtained after fitting is as follows: Where H(·) represents the mapping relationship between the upper and lower layer interactions predicted by the Kriging model.

9. The scenario-driven two-stage distributed robust optimization method for virtual power plant planning and operation according to claim 8 is characterized in that: Step S402 includes: First, the initial sample data set of each scene is constructed by uniform sampling and calling the lower-level model, and the initial Kriging model is established based on the mapping relationship within the data set; Then, the main problem corresponding to the upper virtual power plant model is solved to obtain the optimal transaction price and transaction volume. However, the transaction volume at this time is based on the predicted value of the Kriging model rather than the actual value. Therefore, it is necessary to substitute the optimal transaction price into the lower user model to obtain the actual transaction volume, and add the optimal transaction price and its corresponding actual transaction volume as a high-quality data set to the initial sample data set to realize the correction of the Kriging model. Finally, the sample data is dynamically updated through multiple iterations until the objective function value of the main problem converges.

10. The scenario-driven two-stage distributed robust optimization method for virtual power plant planning and operation according to claim 9 is characterized in that: In step S403, the main problem and sub-problems obtained by decomposition are expressed as follows: Main question: Where l represents the number of iterations, l∈{1,2,3,…,N l }, where N l is the total number of iterations; (2) Sub-problems; The max-min double-layer structure of the sub-problem includes an outer maximum problem and an inner minimum problem. The probability distribution of the worst scenario within the confidence interval is found based on the solution of the main problem. Since the outer discrete scenario probability π of the sub-problem s,l and the inner runtime variable y s,l They are independent of each other, so the sub-problems are divided into two steps for solution, that is, first solve the inner minimum problem, and then solve the outer maximum problem. The inner minimum problem can be obtained from the solution of the main problem.

Images