Virtual power plant capacity optimal configuration method and system considering carbon tax

Abstract

The invention discloses a virtual power plant capacity optimization configuration method and system considering carbon tax, and belongs to the field of power markets. The invention discloses a virtual power plant capacity optimization configuration method considering carbon tax. The method comprises the following steps: analyzing virtual power plant capacity cost and operation income considering carbon tax and market transaction; constructing a virtual power plant internal resource capacity optimization configuration double-layer model based on the analysis of the virtual power plant capacity cost and the operation benefit; generating a spectral clustering typical scene based on an information gap decision theory considering source load uncertainty; and solving the virtual power plant internal resource capacity optimization configuration double-layer model by adopting a differential evolution algorithm to obtain a capacity optimization configuration scheme.

Description

Technical Field

[0001] The present invention belongs to the field of electric power market, and specifically relates to a virtual power plant capacity optimization configuration method and system considering carbon tax. Background Art

[0002] Under the new situation of power generation, distribution and consumption, virtual power plants (VPPs) that can aggregate traditional thermal power units, renewable energy power units, distributed energy storage, flexible loads and other resources distributed in different regions are increasingly becoming a typical "energy + Internet" technology configuration. Virtual power plants can tap the regulation potential of demand-side resources in a demand-responsive manner and improve user economics. In addition, the existence of carbon taxes will stimulate the enthusiasm of virtual power plants to reduce carbon emissions and promote the low-carbon economic development of virtual power plants. In order to better balance the low-carbon and economic development of virtual power plants, the study of optimization configuration methods has become particularly critical.

[0003] At present, most of the research on VPP focuses on dispatching and operation, and few studies consider the integration of VPP low-carbon planning and operation. Traditional research on virtual power plant planning has not considered the carbon emission responsibility assumed by virtual power plants during operation. In addition, the traditional carbon trading mechanism is not incentive enough, the mechanism is unclear, and it cannot reflect changes in the market environment. The introduction of a carbon tax mechanism, a carbon reduction method with unique advantages, can provide a new path for VPP low-carbon dispatching. Summary of the invention

[0004] In view of the deficiencies in the prior art, the purpose of the present invention is to provide a method and system for optimizing the capacity configuration of a virtual power plant taking carbon tax into consideration, thereby solving the problems in the prior art.

[0005] The purpose of the present invention can be achieved through the following technical solutions:

[0006] A virtual power plant capacity optimization configuration method considering carbon tax includes the following steps:

[0007] Analyze the capacity costs and operating benefits of virtual power plants considering carbon taxes and market transactions;

[0008] Based on the analysis of capacity cost and operational benefits of virtual power plants, a two-layer model for optimal allocation of internal resource capacity of virtual power plants is constructed;

[0009] Generate typical scenarios based on information gap decision theory-spectral clustering considering source-load uncertainty;

[0010] In the typical scenario, a differential evolution algorithm is used to solve the two-layer model of optimal configuration of internal resource capacity of the virtual power plant to obtain a capacity optimization configuration plan.

[0011] Furthermore, the virtual power plant capacity cost includes:

[0012] (1) Initial construction cost of virtual power plant equal to annual value:

[0013] C C =(N WT C MT +N MT C MT )c rf

[0014]

[0015] In the formula, C C is the initial construction cost of the virtual power plant in equivalent years, N WT , C WT are the number of wind turbines configured and the purchase cost of a single wind turbine, N MT , C MT are the number of micro gas turbines and the purchase cost of a single micro gas turbine, c rf is the coefficient factor for converting the initial construction cost into the initial construction cost of equivalent annual value, r is the discount rate, and m is the useful life of the system;

[0016] (2) Annual operation and maintenance costs of virtual power plants:

[0017]

[0018] In the formula, C M is the annual operation and maintenance cost of the virtual power plant, They are the annual operation and maintenance costs of a single wind turbine and a single micro gas turbine respectively;

[0019] (3) Micro gas turbine fuel cost:

[0020]

[0021] In the formula, C MT (s) is the fuel cost of the micro gas turbine under the sth scenario combination; is the output of a single micro gas turbine in period t under the sth scenario combination; a MT , b MT 、c MT is the fuel cost coefficient of the micro gas turbine; T is the total number of time periods;

[0022] (4) Energy storage regulation cost:

[0023]

[0024] In the formula, C ES (s) is the energy storage control cost of the virtual power plant under the sth scenario combination; c ESis the regulation cost per unit charge and discharge capacity, are the total charging power and total discharging power of all centralized energy storage stations in the virtual power plant in period t under the sth scenario combination;

[0025] (5) Carbon emission costs:

[0026]

[0027] In the formula, is the carbon emission cost in the tth period, τ is the unit carbon tax price; E is the length of the carbon emission interval; Δτ is the increase in carbon tax price; is the total carbon emissions in the tth period; κ TPG is the carbon emission intensity of the micro gas turbine; κ BUY The carbon emission conversion factor for purchasing electricity from the superior power grid.

[0028] Furthermore, the operating benefits of the virtual power plant include:

[0029] (1) Market transaction income:

[0030]

[0031] In the formula, R E (s) is the market transaction income; P t netsell (s), P t nebuy (s) and They are the market power sales, power sales price, market power purchase and power purchase price in time period t under scenario s respectively;

[0032] (2) Load management benefits:

[0033]

[0034]

[0035] In the formula, R LD (s) is the revenue from load electricity sales; R DR (s) compensation for demand response; is the internal electricity price during period t, P t LD (s) is the user load in time period t under scenario s, is the demand response compensation unit price during period t, P t DR (s) is the demand response quantity in period t.

[0036] Furthermore, the two-layer model includes an upper layer optimization model and a lower layer optimization model, and the expression of the upper layer optimization model is:

[0037]

[0038] N WT C WT +N MT C MT ≤C C,max

[0039] Where: B VPP The total revenue for the construction and operation of virtual power plants; R (s) is the optimal operating profit of the virtual power plant under the sth scenario combination; N ES N is the number of energy storage configurations; ES,max C is the total number of centralized energy storage allowed to be built in the region; C,max The investment ceiling.

[0040] Furthermore, the expression of the lower layer optimization model is:

[0041]

[0042] P t netsell (s)-P t nebuy (s) = N WT P t WT (s)+(P t ES,D (s)-P t ES,C (s))+N MT P t MT (s)-P t LD (s)

[0043] P MT,min ≤P t MT (s)≤P MT,max

[0044] 0≤P t ES,C (s)≤N ES P ES,Cmax

[0045] 0≤P t ES,D (s)≤N ES P ES,Dmax

[0046]

[0047] N ES Qmin ≤Q t (s)≤N ES Q max

[0048]

[0049]

[0050] In the formula, C MT (s), C Es (s), R E (s) are respectively the micro gas turbine fuel cost, centralized energy storage station regulation fee and market transaction income under the s-th scenario combination; is the predicted output value of the wind turbine in period t under the sth scenario combination; is the load forecast value for period t under the sth scenario combination; P MT，min , P MT，max are the lower and upper limits of the output of a single micro gas turbine respectively; P ES，Cmax , P ES，Dmax are the upper limits of charging and discharging power of a single energy storage, Q T (s) is the amount of energy stored in the t period under the s-th scenario combination; η C , η D are the charging and discharging efficiencies of energy storage, respectively; Δt is the duration of a period; Q min , Q max They are the lower and upper limits of the power of a single energy storage; For the configuration quantity N ES The minimum power of energy storage in period t under the condition; is the reserve capacity required for a single wind turbine during period t; It is the reserve capacity required for power load during period t.

[0051] Furthermore, the step of generating the typical scenario includes:

[0052] 1) Taking historical data of wind, light and load as input, a deterministic scheduling model of virtual power plant is established without considering uncertainty. The model is optimized with the minimum system power generation cost as the goal, the benchmark power generation cost within the scheduling planning cycle is determined, and the IGDT robust model and IGDT opportunity model are established respectively;

[0053] 2) Comprehensively process the source-load uncertainties obtained from the IGDT robust model and the IGDT opportunity model to fully quantify the source-load fluctuation range; for wind power and photovoltaic output, the robust uncertainty is used as the lower bound of their fluctuations, and the opportunity uncertainty is used as the upper bound of fluctuations; for load demand, the opportunity uncertainty is used as the lower bound of its fluctuations, and the robust uncertainty determines the upper bound of fluctuations, thus obtaining a comprehensive set of wind, light, and load uncertainties;

[0054] 3) Treat the uncertainty set as uniformly distributed and use Latin hypercube sampling to generate the original scene set;

[0055] 4) Preprocess the original scene data to obtain the input matrix P, extract the net load curve feature PNL and the net load ramp curve feature PNLC to construct the feature vector P = [PNL, PNLC], use the P data as nodes in the graph, build a graph based on the similarity between nodes, and further calculate the adjacency matrix E and the Laplace matrix I;

[0056] 5) Then, the Laplace matrix I is eigendecomposed, the eigenvectors corresponding to the first k smallest eigenvalues ​​are selected, and the eigenvectors are clustered using the K-means algorithm, and the results are mapped back to the original data space to obtain the final clustering results.

[0057] Furthermore, the IGDT robust model is:

[0058]

[0059] In the formula, α 1 is the uncertainty of the robust model system, β 1 is the robust deviation factor; F is the objective function of the deterministic model, F o To optimize the target benchmark value, (1+β 1 )F 0 is the expected cost of the robust model determined by the decision maker; α W1 , α S1 , α L1 are the wind, light, and load uncertainties in the robust model, respectively;

[0060] The IGDT opportunity model is:

[0061]

[0062] In the formula, α 2 is the uncertainty of the chance model system; β 2 is the opportunity deviation factor, (1-β 2 )F 0 is the expected cost of the opportunity model determined by the decision maker, α W2 , α S2 , α L2 are the wind, light and load uncertainties in the chance model, respectively.

[0063] Furthermore, the step of solving the two-layer model using the differential evolution algorithm includes:

[0064] Step 1: Initialize system parameters, including the population size N of the differential evolution algorithm p, scaling factor F, crossover probability C r , the maximum tolerable prediction error ε max ;

[0065] Step 2: Set the number of iterations l to 0 to generate the initial population Λ for differential evolution l ;

[0066] Step 3: Current population Λ l After differential mutation, crossover and selection, a new generation of population Λ is obtained l+1 , wherein the objective function value of the population individual is calculated by calling the lower layer optimization model;

[0067] Step 4: Select the new generation population Λ l+1 The best individual in Call the P2P market clearing model to calculate the actual objective function value of the individual

[0068] Step 5: Calculation The prediction error ε is used to determine whether ε≥ε max ; If satisfied, add the point to the sample point set and update the lower optimization model;

[0069] Step 6: Determine whether the iterative termination conditions of differential evolution are met. If so, output the optimal result Otherwise, l←l+1, jump to step 3.

[0070] A virtual power plant capacity optimization configuration system considering carbon tax, comprising:

[0071] Cost-benefit analysis module: Analyzes the capacity cost and operating benefits of virtual power plants considering carbon tax and market transactions;

[0072] Model building module: Based on the analysis of capacity cost and operational benefits of virtual power plants, a two-layer model for optimizing the internal resource capacity configuration of virtual power plants is constructed;

[0073] Scenario generation module: Generate typical scenarios based on information gap decision theory-spectral clustering considering source-load uncertainty;

[0074] And, model solving module: in the typical scenario, a differential evolution algorithm is used to solve the two-layer model of optimal configuration of internal resource capacity of the virtual power plant to obtain a capacity optimization configuration plan.

[0075] A computer storage medium stores a readable program, which, when running, can execute the above-mentioned virtual power plant capacity optimization configuration method considering carbon tax.

[0076] Beneficial effects of the present invention:

[0077] 1. The present invention adopts a carbon tax mechanism and a stepped carbon cost to address the low-carbon operation of virtual power plants. By introducing a carbon tax, carbon emissions are converted into carbon emission costs to quantify the carbon emission responsibility of virtual power plants, and a stepped carbon emission cost is constructed to fully tap the carbon emission reduction potential of virtual power plants. Compared with traditional carbon trading mechanisms, the carbon tax mechanism has a wider coverage and is fairer, which can promote virtual power plants to reasonably allocate the capacity of wind power, photovoltaics, energy storage and gas units in virtual power plants, thereby reducing system costs while meeting low-carbon operation needs and maximizing the interests of investors.

[0078] 2. The information gap decision theory-spectral clustering typical scenario generation method adopted by the invention provides a more accurate planning scenario for determining the operation mode of the virtual power plant system, and can effectively quantify the uncertainty of source and load. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0080] Figure 1 It is a flow chart of the virtual power plant capacity optimization configuration method of the present invention. DETAILED DESCRIPTION

[0081] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0082] Example 1

[0083] like Figure 1 As shown, a virtual power plant capacity optimization configuration method considering carbon tax includes the following steps:

[0084] S1, analysis of capacity cost and operating benefits of virtual power plants considering carbon tax and market transactions;

[0085] The internal resources of the virtual power plant include distributed wind turbines, micro gas turbines, centralized energy storage stations and power users. The output of distributed wind turbines has significant volatility and uncertainty. Micro gas turbines and centralized energy storage stations can provide flexible and adjustable resources for distributed wind turbines, thereby smoothing the volatility and uncertainty of the output of distributed wind turbines.

[0086] Virtual power plant capacity cost analysis includes:

[0087] (1) Initial construction cost of virtual power plant

[0088] In the virtual power plant in this embodiment, the centralized energy storage station is assumed to be purchased by the energy storage station company, so there is no need to consider the initial construction cost of the centralized energy storage station; therefore, the initial construction cost of the virtual power plant includes the purchase cost of the wind turbine and the micro gas turbine; the initial construction cost of the virtual power plant is converted into the initial construction cost of the same annual value, which can be expressed as:

[0089] C C =(N WT C MT +N MT C MT )c rf (1)

[0090]

[0091] In the formula, C C is the initial construction cost of the virtual power plant in equivalent years, N WT , C WT are the number of wind turbines configured and the purchase cost of a single wind turbine, N MT , C MT are the number of micro gas turbines and the purchase cost of a single micro gas turbine, c rf is the coefficient factor for converting the initial construction cost into the initial construction cost of equivalent annual value, r is the discount rate, and m is the useful life of the system.

[0092] (2) Annual operation and maintenance costs of virtual power plants

[0093] The distributed resources within the virtual power plant require regular inspection and maintenance, so the virtual power plant needs to bear the operation and maintenance costs; the operation and maintenance costs of the virtual power plant include the operation and maintenance costs of wind turbines and micro gas turbines, which can be expressed as follows based on an annual cycle:

[0094]

[0095] In the formula, C M is the annual operation and maintenance cost of the virtual power plant, They are the annual operation and maintenance costs of a single wind turbine and a single micro gas turbine respectively.

[0096] (3) Micro gas turbine fuel cost

[0097] Unlike conventional units with larger power generation capacity, micro gas turbines can be started and stopped quickly and at a lower cost. Therefore, this project ignores the start and stop process of micro gas turbines and uses the quadratic function model commonly used in existing studies to represent the relationship between micro gas turbine output and fuel cost. Assuming that the parameters of each micro gas turbine are the same, the fuel cost of the micro gas turbine can be expressed as:

[0098]

[0099] In the formula, C MT (s) is the fuel cost of the micro gas turbine under the sth scenario combination; is the output of a single micro gas turbine in period t under the sth scenario combination; a MT , b MT 、c MT is the fuel cost coefficient of the micro gas turbine; T is the total number of time periods.

[0100] (4) Energy storage regulation cost

[0101] The regulation fee paid by the virtual power plant is proportional to the charging and discharging power of the energy storage. The energy storage regulation cost is expressed as:

[0102]

[0103] In the formula, C ES (s) is the energy storage control cost of the virtual power plant under the sth scenario combination; c ES is the regulation cost per unit charge and discharge capacity, They are respectively the total charging power and total discharging power of all centralized energy storage stations within the virtual power plant in period t under the sth scenario combination.

[0104] (5) Carbon emission costs

[0105] By introducing carbon tax, carbon emissions are converted into carbon emission costs. Considering that wind power generation does not produce carbon emissions, the carbon emissions generated in VPP come from micro gas turbines. Assuming that the electricity generated by the upper power grid is also traditional thermal power, in order to increase the carbon emission reduction enthusiasm of VPP, a stepped carbon emission cost is constructed to fully tap the carbon emission reduction potential of VPP system operation. Therefore, the carbon emission cost in the tth period is for:

[0106]

[0107]

[0108] In the formula, τ is the unit carbon tax price; E is the length of the carbon emission interval; Δτ is the increase in carbon tax price; is the total carbon emissions in the tth period; κ TPGis the carbon emission intensity of the micro gas turbine; κ BUY The carbon emission conversion factor for purchasing electricity from the superior power grid.

[0109] The annual initial construction cost and annual operation and maintenance cost of the virtual power plant mentioned above are both based on an annual cycle, so the operating cost benefits of the virtual power plant that need to be studied in this embodiment are also modeled on an annual cycle. The virtual power plant optimizes operation on a daily cycle, but there are differences in the market electricity price, wind power output curve, and power user load curve on each day of the year, so it is necessary to consider multiple scenarios of market electricity prices, wind power output curves, and load curves. Consider all possible combinations of market electricity prices, wind power output curves, and load curves, and use s to represent the sth scenario combination of market electricity prices, wind power output curves, and load curves, and use S to represent the set of all possible scenario combinations.

[0110] Virtual power plant operating benefit analysis includes:

[0111] (1) Market transaction income

[0112] The capacity of the virtual power plant in this article is small and it is difficult to affect the electricity market price. Therefore, the virtual power plant in this article participates in the energy market transaction in the electricity market as a price acceptor, and the transaction volume in the market is equal to the declared volume. Therefore, the market transaction income obtained by the virtual power plant can be expressed as:

[0113]

[0114] In the formula, R E (s) is the market transaction income; P t netsell (s), P t nebuy (s) and They are the market electricity sales power, electricity sales price, market electricity purchase power and electricity purchase price in period t under scenario s respectively.

[0115] (2) Load management benefits

[0116] Load management benefits include internal electricity sales revenue and demand response compensation costs.

[0117]

[0118]

[0119] In the formula, R LD (s) is the revenue from load electricity sales; R DR (s) compensation for demand response; is the internal electricity price during period t, P t LD(s) is the user load in time period t under scenario s, is the demand response compensation unit price during period t, P t DR (s) is the demand response quantity in period t.

[0120] S2, based on the analysis of capacity cost and operating benefits of virtual power plants in S1, a two-layer model for optimal configuration of internal resource capacity of virtual power plants is constructed;

[0121] The cost-benefit of the internal resource capacity configuration of a virtual power plant mainly includes two parts. One part is the initial construction cost and annual operation and maintenance cost of equal annual value, which is directly related to the quantity of resource configuration and can be determined accordingly when the quantity of resource configuration is determined; the other part is the optimal operating benefit corresponding to the optimal operating strategy of the virtual power plant during its life cycle.

[0122] The two-layer model for optimizing the internal resource capacity configuration of the virtual power plant includes an upper optimization model and a lower optimization model;

[0123] (1) Upper-level optimization model

[0124] 1) Objective function

[0125] The goal of the upper-level optimization of the internal resource capacity configuration of the virtual power plant is to maximize the total benefits of the construction and operation of the virtual power plant, including the initial construction cost of equal annual value, the annual operation and maintenance cost, and the optimal operating income of the virtual power plant, which can be expressed as:

[0126]

[0127] Where: B VPP The total revenue for the construction and operation of virtual power plants; R (s) is the optimal operating profit of the virtual power plant under the sth scenario combination.

[0128] 2) Constraints

[0129] ① Constraints on the number of energy storage: The total number of energy storage construction in the region is limited, and the number of energy storage included in the virtual power plant should be less than or equal to the total number of centralized energy storage stations that have been built in the region:

[0130] N ES ≤N ES,max (12)

[0131] Where: N ES N is the number of energy storage configurations; ES,max The total number of centralized energy storage allowed to be built in the region.

[0132] ② Initial construction cost constraint: The initial construction funds of the virtual power plant are limited, and the initial construction cost should be less than or equal to the initial construction cost upper limit:

[0133] NWT C WT +N MT C MT ≤C C,max (13)

[0134] In the formula, C C,max The upper limit of construction investment.

[0135] (2) Lower-level optimization model

[0136] 1) Objective function

[0137] The goal of the lower-level optimization of the internal resource capacity configuration of the virtual power plant is to maximize the operating income of the virtual power plant, including the fuel cost of the micro gas turbine, the regulation cost of the centralized energy storage station, and the market transaction income, which can be expressed as:

[0138]

[0139] In the formula, C MT (s), C Es (s), R E (s) are the micro gas turbine fuel cost, centralized energy storage station regulation fee and market transaction income under the sth scenario combination.

[0140] 2) Constraints

[0141] ① Power balance constraints

[0142] The meaning of power balance constraint is that the market transaction power of virtual power plant should be equal to the difference between the power generated and consumed by internal resources of virtual power plant, where the power generated includes wind turbine output, micro gas turbine output and energy storage discharge, and the power consumed includes energy storage charging and power load. Assuming that the parameters of each wind turbine are exactly the same, it is approximately considered that the output of each wind turbine is equal. Without considering wind abandonment, the power balance constraint can be expressed as:

[0143]

[0144] Where: is the predicted output value of the wind turbine in period t under the sth scenario combination; is the load forecast value for period t under the sth scenario combination.

[0145] ②Micro gas turbine output constraints

[0146] The output of the micro gas turbine should meet the lower and upper limit constraints:

[0147] P MT,min ≤P t MT (s)≤P MT,max (16)

[0148] Where: P MT，min , P MT，max They are respectively the lower and upper limits of the output of a single micro gas turbine.

[0149] ③Energy storage constraints:

[0150] The battery charging and discharging process of the centralized energy storage station should meet the upper and lower limits of charging and discharging power, upper and lower limits of power, and minimum power constraints. The energy storage constraints can be expressed as:

[0151] 0≤P t ES,C (s)≤N ES P ES,Cmax (17)

[0152] 0≤P t ES,D (s)≤N ES P ES,Dmax (18)

[0153]

[0154] N ES Q min ≤Q t (s)≤N ES Q max (20)

[0155]

[0156] Where: P ES，Cmax , P ES，Dmax are the upper limits of charging and discharging power of a single energy storage, Q t (s) is the amount of energy stored in the t period under the s-th scenario combination; η C , η D are the charging and discharging efficiencies of energy storage, respectively; Δt is the duration of a period; Q min , Q max They are the lower and upper limits of the power of a single energy storage; For the configuration quantity N ES The minimum power of energy storage during period t under the condition.

[0157] ④ Reserve capacity constraints

[0158] There are errors in the prediction of wind turbine output and power load. To ensure the safe and stable operation of the virtual power plant, the virtual power plant needs to provide spare capacity for wind turbines and power loads. To ensure that the virtual power plant can still operate normally under various extreme conditions of wind turbine output and power load, the virtual power plant should have a sufficiently large spare capacity when configuring the capacity. It is approximately assumed that the spare capacity required by the wind turbine is proportional to the number of wind turbines configured, and the spare capacity required by the wind turbine and power load is considered separately. The spare capacity constraint can be expressed as:

[0159]

[0160] In the formula, is the reserve capacity required for a single wind turbine during period t; It is the reserve capacity required for power load during period t.

[0161] S3, generating typical scenarios based on information gap decision theory-spectral clustering considering source-load uncertainty;

[0162] The steps to generate a typical scenario include:

[0163] S31, generating an IGDT scheduling model and an original scenario considering source-load uncertainty;

[0164] S311, using historical data of wind, light and load as input, establish a deterministic scheduling model for a virtual power plant without considering uncertainty, optimize with the minimum system power generation cost as the goal, determine the benchmark power generation cost within the scheduling planning cycle, and establish an IGDT robust model and an IGDT opportunity model respectively;

[0165] S312, in order to ensure the robustness of the system while taking into account the economy, the uncertainty of the source and load obtained by the IGDT robust model and the IGDT opportunity model is comprehensively processed to fully quantify the source and load fluctuation range; for wind power and photovoltaic output, the robust uncertainty is used as the lower bound of its fluctuation, and the opportunity uncertainty is used as the upper bound of the fluctuation; for load demand, the opportunity uncertainty is used as the lower bound of its fluctuation, and the robust uncertainty determines the upper bound of the fluctuation, so that a comprehensive set of wind, light, and load uncertainties can be obtained;

[0166] S313, regards the uncertainty set as uniformly distributed, and uses Latin hypercube sampling to generate the original scene set, obtaining sufficient samples to represent different uncertainty situations with a small computational cost.

[0167] S32, clustering the new feature vector using a spectral clustering method;

[0168] S321, preprocessing the original scene data to obtain an input matrix P, extracting the net load curve feature PNL and the net load ramp curve feature PNLC to characterize the system regulation capability to meet different dispatching requirements under uncertain circumstances, and constructing a feature vector P = [PNL, PNLC], using the P data as nodes in the graph, constructing a graph based on the similarity between the nodes, and further calculating the adjacency matrix E and the Laplace matrix I;

[0169] S322, then perform eigendecomposition on the Laplace matrix I, select the eigenvectors corresponding to the first k smallest eigenvalues, cluster the eigenvectors using the K-means algorithm, and map the results back to the original data space to obtain the final clustering result.

[0170] Among them, the envelope constraint model is used to model the wind power, photovoltaic and load uncertainty sets respectively;

[0171] (1) Modeling of wind power uncertainty set:

[0172]

[0173] Where P w is the actual value of wind power output during the dispatch period, is the historical data of wind power output during the dispatch period, α w is the uncertainty of wind power output, which is a dimensionless value; this formula reflects the maximum range of wind power output deviation from the corresponding historical data.

[0174] (2) Modeling of photovoltaic uncertainty set:

[0175]

[0176] Where P s is the actual value of photovoltaic output during the dispatch period, is the historical data of photovoltaic output during the dispatch period, α S is the uncertainty of photovoltaic output, which is a dimensionless value; this formula reflects the maximum range of photovoltaic output deviation from the corresponding historical data.

[0177] (3) Modeling of load uncertainty set:

[0178]

[0179] Where P L is the actual value of load demand during the dispatch period; is the historical data of load demand during the dispatching period: Lis the uncertainty of the load demand, which is a dimensionless value; this formula reflects the maximum range of the load demand deviation from the corresponding historical data. For problems involving multiple uncertain variables, this embodiment uses the CRITIC weight method to calculate the overall uncertainty α of the system, so as to reflect the overall uncertainty level of the system after comprehensively considering various uncertain factors.

[0180]

[0181] Where α is the system uncertainty; λ W , S , L are the weights of wind power output, photovoltaic output and load demand uncertainty respectively.

[0182] IGDT robust model and IGDT opportunity model:

[0183] The IGDT robust model is essentially a max-max nested optimization problem. The sub-goal indicates that when wind, light, and load fluctuate within the uncertainty set, any decision-making scheme can ensure that the dispatch cost is no higher than (1+β 1 )F 0 ; The main goal is to find the maximum system uncertainty when the sub-goals are met. It can be expressed as:

[0184]

[0185] In the formula, α 1 is the uncertainty of the robust model system, β 1 is the robust deviation factor. The larger its value, the stronger the system's robustness and risk resistance. F is the objective function of the deterministic model. o To optimize the target benchmark value, that is, without considering the uncertainty of wind, light and load, the cost size is obtained by substituting historical data into the deterministic scheduling model; (1+β 1 )F 0 is the expected cost of the robust model determined by the decision maker; α W1 , α S1 , α L1 are the wind, light and load uncertainties in the robust model, respectively.

[0186] The IGDT opportunity model is essentially a min-min nested optimization problem. The sub-goal indicates that when wind, light, and load fluctuate within the uncertainty set, there is at least one decision-making scheme that makes the dispatch cost no higher than (1-β 2 )F 0 ; The main goal is to find the minimum system uncertainty when the sub-goals are met. The specific model is shown in the formula:

[0187]

[0188] In the formula, α 2 is the uncertainty of the chance model system; β 2 is the opportunity deviation factor. The larger its value, the greater the potential benefit of the model, but the risk faced by the system will also increase. (1-β 2 )F 0 is the expected cost of the opportunity model determined by the decision maker, α W2 , α S2 , α L2 are the wind, light and load uncertainties in the chance model, respectively.

[0189] S4, under the typical scenario generated by S3, uses the differential evolution algorithm to solve the two-layer model of optimal configuration of internal resource capacity of the virtual power plant constructed in S2, and obtains the resource capacity configuration plan of the virtual power plant.

[0190] Differential evolution is a heuristic random search algorithm based on population evolution for solving global optimization problems, including differential mutation, crossover, and selection. The specific process of the algorithm is shown in Table 1.

[0191] Table 1 Process of differential evolution algorithm

[0192]

[0193]

[0194] 1) Differential mutation

[0195] The mutant individual is formed by the basis vector and the difference vector, as shown in the following formula:

[0196]

[0197] Where: r1 , r2 , r3 is a random individual in the population Λ. r1 is called the basis vector, (λ r2 -λ r3 ) is called the difference vector. Since the basis vector is randomly selected, there is only one difference vector. This difference mutation is recorded as DE / rand / 1. Another variant of the difference mutation DE / best / 1 is shown as follows:

[0198]

[0199] Where: best is the best individual in population Λ.

[0200] 2) Cross

[0201] Crossover refers to exchanging some elements of the current individual and the variant individual, which can be divided into binomial crossover and exponential crossover. Assume that the dimension of the individual is n, and there is λ i =[λ i,1 , ..., λ i,j , ..., λ i,n ] T , i =[υ i,1 , ..., υ i ,j,...,υ i,n ] T , i =[ο i,1 , ...,ο i,j , ...,ο i,n ] T Taking binomial crossover as an example, the jth element of the test individual is determined according to the following formula:

[0202]

[0203] Where: rand(0,1) is a random number between [0,1].

[0204] 3) Select

[0205] The greedy criterion is used to select between the current individual and the test individual, that is, to compare λ i and i The function value of , individuals with better function values ​​enter the new generation population.

[0206] The specific process of adopting the algorithm is as follows.

[0207] Step 1: Initialize system parameters, including the population size N of the differential evolution algorithm p , scaling factor F, crossover probability C r , the maximum tolerable prediction error ε max ;

[0208] Step 2: Set the number of iterations l to 0 to generate the initial population Λ for differential evolution l ;

[0209] Step 3: Current population Λ l After differential mutation, crossover and selection, a new generation of population Λ is obtained l+1 , where the objective function value of the population individual is calculated by calling the lower-level optimization model (14), and constraints (16)-(23) need to be satisfied;

[0210] Step 4: Select the new generation population Λ l+1 The best individual in Call the P2P market clearing model to calculate the actual objective function value of the individual Assume that the optimal solution currently found by the algorithm is if and Satisfying the constraints of equations (12) and (13), then otherwise,

[0211] Step 5: Calculation The prediction error ε is used to determine whether ε≥ε max ; If satisfied, add the point to the sample point set and update the lower optimization model.

[0212] Step 6: Determine whether the iterative termination conditions of differential evolution are met. If so, output the optimal result Otherwise, l←l+1, jump to step 3.

[0213] Example 2

[0214] In this embodiment, a specific case is disclosed to illustrate the capacity optimization configuration method of the present invention;

[0215] The rated capacity, purchase cost and annual operation and maintenance cost of the internal resources of the virtual power plant are shown in Table 2. The service life of the virtual power plant system is 20 years, and the discount rate is 5%. The upper limit of the initial construction cost of the virtual power plant is 10 million yuan. The fuel cost coefficient of the micro gas turbine is: the quadratic coefficient is 0.0001 yuan·(kW2)-1, the linear coefficient is 0.5 yuan / kW, the constant coefficient is 0, and the lower and upper limits of the output of a single micro gas turbine are 20kW and 100kW respectively. The upper limit of the charging and discharging power of the energy storage is 12kW, the lower and upper limits of the power are 12 and 54kW·h respectively, and the charging and discharging efficiency is 95%. The spare capacity of a single wind power and photovoltaic unit is 20kW, and the spare capacity of the power load is 40kW. Considering the power transmission restrictions of the network, the upper limit of the market power purchase and sale is set to 5MW. The above model is used for calculation, the time step is 1h, and the calculation time is the whole year. In order to simplify the calculation, the typical day is used instead of the parameters of each day of the year for calculation.

[0216] Table 2 Related parameters of each power source of solar-diesel-storage

[0217] Resource Type Rated capacity Acquisition cost Annual operation and maintenance cost Wind turbines 200kW 4800 yuan / kW 150 yuan / kW Photovoltaic unit 200kW 3300 yuan / kW 60 yuan / kW Micro gas turbine 100kW 1800 yuan / kW 75 yuan / kW Energy Storage 60kW 1200 yuan / kW 10 yuan / kW

[0218] Impact of risk factors on capacity allocation strategies:

[0219] The calculation results of the capacity configuration of VPP and each part of the objective function under different risk preference coefficients are shown in Table 3. It can be seen that with the increase of the risk preference coefficient L, the total construction and operation cost of VPP increases, while the CVaR value gradually decreases, indicating that investors' risk aversion deepens and the planning strategy tends to be conservative.

[0220] Table 3 Capacity configuration results

[0221]

[0222]

[0223] When L increases, VPP mainly focuses on meeting load demand and reducing the risk brought by electricity price fluctuations. The power sales are greatly reduced, and the wind and light configurations are reduced accordingly. Although both energy storage and gas turbines can meet load demand when there is a shortage of renewable energy power, only energy storage can flexibly adjust power sales through charging and discharging according to electricity price fluctuations. Obviously, its risk is smaller than that of diesel generators, so the proportion of energy storage configuration is relatively increased.

[0224] Impact of resource forecast output on capacity configuration:

[0225] As shown in Table 4, when the preference coefficient is 0.2, comparing the results of scenario 2 with scenario 1, it can be seen that when the regional light intensity is slightly low, the virtual power plant will be suitable for increasing the number of photovoltaic units purchased, because the cost of configuring photovoltaic units can still be recovered in the market. However, as the light intensity decreases, configuring photovoltaic units will greatly reduce the profit margin. At this time, the goal of the virtual power plant is only to meet the regional load needs, so the number of gas turbines configured increases.

[0226] Table 4 Capacity configuration results in different scenarios

[0227]

[0228] The method of the present invention may be implemented in hardware, firmware, or as software or computer code that may be stored in a recording medium (such as a CDROM, RAM, floppy disk, hard disk, or magneto-optical disk), or as computer code that is originally stored in a remote recording medium or a non-temporary machine-readable medium downloaded over a network and will be stored in a local recording medium, so that the method described herein may be stored in such software processing on a recording medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware (such as an ASIC or FPGA). It is understood that a computer, processor, microprocessor controller, or programmable hardware includes a storage component (e.g., RAM, ROM, flash memory, etc.) that can store or receive software or computer code, and when the software or computer code is accessed and executed by a computer, processor, or hardware, the method described herein is implemented. In addition, when a general-purpose computer accesses the code for implementing the method shown herein, the execution of the code converts the general-purpose computer into a dedicated computer for executing the method shown herein.

[0229] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, and these changes and improvements all fall within the scope of the present invention to be protected.

Claims

1. A virtual power plant capacity optimization configuration method considering carbon tax, characterized in that: The following steps are involved: Analyze the capacity costs and operating benefits of virtual power plants considering carbon taxes and market transactions; Based on the analysis of capacity cost and operational benefits of virtual power plants, a two-layer model for optimal allocation of internal resource capacity of virtual power plants is constructed; Generate typical scenarios based on information gap decision theory-spectral clustering considering source-load uncertainty; In the typical scenario, a differential evolution algorithm is used to solve the two-layer model of optimal configuration of internal resource capacity of the virtual power plant to obtain a capacity optimization configuration plan.

2. A virtual power plant capacity optimization configuration method considering carbon tax according to claim 1, characterized in that: The virtual power plant capacity costs include: (1) Initial construction cost of virtual power plant equal to annual value: C C =(N WT C MT +N MT C MT )c rf In the formula, C C is the initial construction cost of the virtual power plant in equivalent years, N WT , C WT are the number of wind turbines configured and the purchase cost of a single wind turbine, N MT , C MT are the number of micro gas turbines and the purchase cost of a single micro gas turbine, c rf is the coefficient factor for converting the initial construction cost into the initial construction cost of equivalent annual value, r is the discount rate, and m is the useful life of the system; (2) Annual operation and maintenance costs of virtual power plants: In the formula, C M is the annual operation and maintenance cost of the virtual power plant, They are the annual operation and maintenance costs of a single wind turbine and a single micro gas turbine respectively; (3) Micro gas turbine fuel cost: In the formula, C MT (s) is the fuel cost of the micro gas turbine under the sth scenario combination; is the output of a single micro gas turbine in period t under the sth scenario combination; a MT 、b MT 、c MT is the fuel cost coefficient of the micro gas turbine; T is the total number of time periods; (4) Energy storage regulation cost: In the formula, C ES (s) is the energy storage control cost of the virtual power plant under the sth scenario combination; c ES is the regulation cost per unit charge and discharge capacity, are the total charging power and total discharging power of all centralized energy storage stations in the virtual power plant in period t under the sth scenario combination; (5) Carbon emission costs: In the formula, is the carbon emission cost in the tth period, τ is the unit carbon tax price; E is the length of the carbon emission interval; Δτ is the increase in carbon tax price; is the total carbon emissions in the tth period; κ TPG is the carbon emission intensity of the micro gas turbine; κ BUY The carbon emission conversion factor for purchasing electricity from the superior power grid.

3. A virtual power plant capacity optimization configuration method considering carbon tax according to claim 2, characterized in that: The virtual power plant operating benefits include: (1) Market transaction income: In the formula, R E (s) is the market transaction income; P t netsell (s), λ t netsell (s), P t nebuy (s) and λ t netbuy (s) are the market power sales, power sales price, market power purchase and power purchase price in time period t under scenario s; (2) Load management benefits: In the formula, R LD (s) is the revenue from load electricity sales; R DR (s) is the demand response compensation; λ t LD (s) is the internal electricity price in period t, P t LD (s) is the user load in time period t under scenario s, λ t DR (s) is the demand response compensation unit price during period t, P t DR (s) is the demand response quantity in period t.

4. The method for optimizing capacity configuration of virtual power plants considering carbon tax according to claim 3 is characterized in that: The two-layer model includes an upper layer optimization model and a lower layer optimization model. The expression of the upper layer optimization model is: N ES ≤N ES,max N WT C WT +N MT C MT ≤C C,max Where: B VPP The total revenue for the construction and operation of virtual power plants; R (s) is the optimal operating profit of the virtual power plant under the sth scenario combination; N ES N is the number of energy storage configurations; ES,max C is the total number of centralized energy storage allowed to be built in the region; C,max The investment ceiling.

5. A virtual power plant capacity optimization configuration method considering carbon tax according to claim 4, characterized in that: The expression of the lower layer optimization model is: In the formula, C MT (s), C Es (s), R E (s) are respectively the micro gas turbine fuel cost, centralized energy storage station regulation fee and market transaction income under the s-th scenario combination; is the predicted output value of the wind turbine in period t under the sth scenario combination; is the load forecast value for period t under the sth scenario combination; P MT，min , P MT，max are the lower and upper limits of the output of a single micro gas turbine respectively; P ES，Cmax , P ES，Dmax are the upper limits of charging and discharging power of a single energy storage, Q t (s) is the amount of energy stored in the t period under the s-th scenario combination; η C , η D are the charging and discharging efficiencies of energy storage, respectively; Δt is the duration of a period; Q min , Q max They are the lower and upper limits of the power of a single energy storage; For the configuration quantity N ES The minimum power of energy storage in period t under the condition; is the reserve capacity required for a single wind turbine during period t; It is the reserve capacity required for power load during period t.

6. The method for optimizing capacity configuration of virtual power plants considering carbon tax according to claim 1, characterized in that: The steps of generating the typical scenario include: 1) Taking historical data of wind, light and load as input, a deterministic scheduling model of virtual power plant is established without considering uncertainty. The model is optimized with the minimum system power generation cost as the goal, the benchmark power generation cost within the scheduling planning cycle is determined, and the IGDT robust model and IGDT opportunity model are established respectively; 2) Comprehensively process the source-load uncertainties obtained from the IGDT robust model and the IGDT opportunity model to fully quantify the source-load fluctuation range; for wind power and photovoltaic output, the robust uncertainty is used as the lower bound of their fluctuations, and the opportunity uncertainty is used as the upper bound of fluctuations; for load demand, the opportunity uncertainty is used as the lower bound of its fluctuations, and the robust uncertainty determines the upper bound of fluctuations, thus obtaining a comprehensive set of wind, light, and load uncertainties; 3) Treat the uncertainty set as uniformly distributed and use Latin hypercube sampling to generate the original scene set; 4) Preprocess the original scene data to obtain the input matrix P, extract the net load curve feature PNL and the net load ramp curve feature PNLC to construct the feature vector P = [PNL, PNLC], use the P data as nodes in the graph, build a graph based on the similarity between nodes, and further calculate the adjacency matrix E and the Laplace matrix I; 5) Then, the Laplace matrix I is eigendecomposed, the eigenvectors corresponding to the first k smallest eigenvalues ​​are selected, and the eigenvectors are clustered using the K-means algorithm, and the results are mapped back to the original data space to obtain the final clustering results.

7. A virtual power plant capacity optimization configuration method considering carbon tax according to claim 6, characterized in that: The IGDT robust model is: Where α1 is the uncertainty of the robust model system, β1 is the robust deviation factor; F is the objective function of the deterministic model, F o is the target benchmark value to be optimized, (1+β1)F0 is the expected cost of the robust model determined by the decision maker; α W1 , α S1 , α L1 are the wind, light, and load uncertainties in the robust model, respectively; The IGDT opportunity model is: In the formula, α2 is the uncertainty of the opportunity model system; β2 is the opportunity deviation factor, (1-β2)F0 is the expected cost of the opportunity model determined by the decision maker, and α W2 , α S2 , α L2 are the wind, light and load uncertainties in the chance model, respectively.

8. The method for optimizing capacity configuration of virtual power plants considering carbon tax according to claim 5, characterized in that: The steps of solving the two-layer model using the differential evolution algorithm include: Step 1: Initialize system parameters, including the population size N of the differential evolution algorithm p , scaling factor F, crossover probability C r , the maximum tolerable prediction error ε max ; Step 2: Set the number of iterations l to 0 to generate the initial population Λ for differential evolution l ; Step 3: Current population Λ l After differential mutation, crossover and selection, a new generation of population Λ is obtained l+1 , wherein the objective function value of the population individual is calculated by calling the lower layer optimization model; Step 4: Select the new generation population Λ l+1 The best individual in Call the P2P market clearing model to calculate the actual objective function value of the individual Step 5: Calculation The prediction error ε is used to determine whether ε≥ε max ; If satisfied, add the point to the sample point set and update the lower optimization model; Step 6: Determine whether the iterative termination conditions of differential evolution are met. If so, output the optimal result Otherwise, l←l+1, jump to step 3.

9. A virtual power plant capacity optimization configuration system considering carbon tax, characterized in that: include: Cost-benefit analysis module: Analyzes the capacity cost and operating benefits of virtual power plants considering carbon tax and market transactions; Model building module: Based on the analysis of capacity cost and operational benefits of virtual power plants, a two-layer model for optimizing the internal resource capacity configuration of virtual power plants is constructed; Scenario generation module: Generate typical scenarios based on information gap decision theory-spectral clustering considering source-load uncertainty; And, model solving module: in the typical scenario, a differential evolution algorithm is used to solve the two-layer model of optimal configuration of internal resource capacity of the virtual power plant to obtain a capacity optimization configuration plan.

10. A computer storage medium storing a readable program, characterized in that: When the program is running, it can execute the virtual power plant capacity optimization configuration method considering carbon tax as described in any one of claims 1-8.

Images