Virtual power plant market-oriented optimal dispatch method based on multiple uncertainty representation

Abstract

This invention relates to a market-based optimal scheduling method for virtual power plants (VPPs) based on multiple uncertainty representations. It employs a scenario-based method using Monte Carlo and Manhattan probability distances to model wind and solar power output and electricity prices, while simultaneously utilizing an interval method to model carbon prices and demand response. With the goal of minimizing VPP operating costs, it comprehensively considers electricity market trading, carbon trading, and incentive-based demand response mechanisms, establishing an optimal scheduling model for VPPs based on interval linear programming. The model is solved using a two-stage decomposition algorithm within strong interval linear programming. This invention effectively reflects the impact of multiple uncertainties on system scheduling results through the changing trends of indicators, improving economic efficiency while ensuring the safe operation of the VPP system. The interval form representation provides decision-makers with a certain degree of realistic choice space, offering guidance for the optimal scheduling of future VPP systems.

Description

Technical Field

[0001] This invention relates to an energy management technology, and more particularly to a market-based optimization scheduling method for virtual power plants based on multiple uncertainty characteristics. Background Technology

[0002] With the continuous transformation and upgrading of the energy industry, the development of my country's new power system faces the challenges of structural diversification and clean and low-carbon development. The large-scale grid connection of new energy sources on the supply and demand sides will have a certain impact on the safety and stability of the power grid operation due to their strong power output fluctuations and geographically dispersed locations.

[0003] Virtual Power Plants (VPPs) serve as a bridge between the power grid management center and distributed energy resources. Through precise control of flexible resources, electric vehicles, controllable loads, and energy storage devices, they achieve goals such as resource aggregation, peak shaving and valley filling, and carbon emission reduction. Regarding the establishment of optimal scheduling models for VPPs, current research both domestically and internationally has constructed VPP cluster architectures that include distributed renewable energy, energy storage, and controllable loads, establishing day-ahead and day-intraday two-stage optimal scheduling models. Additionally, some studies have incorporated carbon emissions into unit operation constraints, aiming to maximize VPP revenue in the electricity and carbon markets, thus constructing economic scheduling models for VPPs. However, these studies primarily focus on deterministic optimal scheduling models for VPPs, neglecting the uncertainties in factors such as renewable energy output, demand response, carbon trading prices, and electricity purchase and sale prices within VPPs. This leads to discrepancies between simulation results and actual conditions.

[0004] Currently, common methods for modeling the uncertainties of different scheduling resources include fuzzy logic, stochastic programming, scenario-based approaches, robust optimization, and interval methods. Some studies simulate electricity price uncertainty using a set of electricity price prediction scenarios, proposing an optimal bidding strategy for energy storage under uncertain electricity prices and establishing an electricity price-energy supply bidding model. Other studies consider uncertainty factors, introducing a VPP supply architecture that includes jointly optimizing multiple markets (e.g., energy, storage), systems (e.g., inertia, reactive power), and local networks (e.g., voltage support) to maximize revenue. However, these methods only consider the impact of individual uncertainties in the VPP model and do not characterize the multiple uncertainties in the VPP optimal scheduling model. Therefore, a comprehensive approach is needed to address the combined effects of multiple uncertainties, ensuring the economic operation of the VPP while maintaining its stability and security. Summary of the Invention

[0005] To address the challenges of virtual power plant (VPP) dispatching due to numerous uncertainties, a market-based optimization dispatching method for VPPs based on multiple uncertainty representations is proposed. Considering the VPP dispatching system under the influence of multiple uncertainties related to wind and solar power output, electricity price, carbon price, and demand response, and under energy market trading conditions, scenario-based and interval-based methods based on Monte Carlo and Manhattan probability distances are used to model the uncertainties of wind and solar power output, electricity price, carbon price, and demand response, respectively. An optimal VPP dispatching model based on interval linear programming is established, and the two-stage decomposition algorithm in strong interval linear programming is used to solve the model. Finally, simulation examples verify the rationality and applicability of the proposed optimization method and model.

[0006] The technical solution of this invention is: a market-based optimal scheduling method for virtual power plants based on multiple uncertainty representations, specifically including the following steps:

[0007] 1) Utilizing the fluctuation patterns of historical data in the energy system, a large number of corresponding scenarios S are generated using the Monte Carlo method. pre The probability p of the corresponding scene occurrence is then used to reduce the scene using a fast previous generation elimination technique based on Manhattan probability distance, and finally the scene information is generated.

[0008] 2) For uncertainties in energy systems that do not require precise probability distributions and have limited historical data, the interval method is used to solve for the upper and lower boundaries;

[0009] 3) The virtual power plant acts as the power grid management center, communicating with distributed energy resources, participating in power system dispatch, and connecting with the main grid for energy exchange. Based on meeting user energy demands and corresponding constraints, it aims to minimize the total cost of the objective function, thereby establishing a day-ahead dispatch model for the virtual power plant based on interval linear programming.

[0010] To address uncertainty, steps 2) and 3) are integrated, and the objective function of the established virtual power plant day-ahead scheduling model is optimized to obtain the optimized scheduling scheme.

[0011] Furthermore, the specific steps for implementing step 1) are as follows:

[0012] 1.1) For typical prediction data, the standard deviation s is multiplied by the standard normal distribution σ of the generated random numbers as the data fluctuation and superimposed on the original prediction value D. pre Above, the Monte Carlo method is used to generate m sets of typical prediction scenarios S with equal probability. pre :

[0013] S pre =D pre +sσ;

[0014] 1.2) The scene generated in step 1.1) is an equally probable scene with a probability of 1 / m. Therefore, the Manhattan distances between each pair of the m scenes are calculated to form a matrix d. ge ;

[0015] 1.3) Based on the Manhattan distance matrix d between scenes obtained in step 1.2), ge Calculate the sum of the probabilistic distances between each scene and the remaining scenes, and form the probabilistic distance matrix Y. ge ;

[0016] 1.4) Based on the probability distance matrix Y ge The scene with the smallest probability distance from the remaining scenes is selected and denoted as scene S. pre (a), where a is the index value of the scene, based on matrix d ge In the prediction scenario set S pre Let S be the scene with the smallest Manhattan distance to scene a. pre (b), where b is the index of the scene; Scene S pre (a) Merge into scene S pre (b) and scene S pre (a) Probability p a Overlay onto scene S pre (b) probability p b The new scene probabilities are obtained and applied to the prediction scene set S. pre Delete scene a;

[0017] 1.5) The number of scenes is continuously reduced to k using the method described in 1.4), ultimately resulting in k reduced scenes {S}. pre1 ,S per2 ,S pre3 ,…S prek} and the probabilities {p1, p2, p3, ... p} corresponding to each scene after reduction. k}

[0018] Furthermore, step 2) involves using the interval method to handle uncertainties in carbon prices and demand response:

[0019] 2.1) Regarding the carbon trading price k c The interval method model is as follows:

[0020]

[0021] In the formula, [α] ± The range of carbon price volatility coefficients is determined based on historical carbon price volatility data, with upper and lower limits α. min With α max ;k c0 The average carbon price over a certain period is obtained from historical data; [k c ]± This represents the price range for carbon trading.

[0022] 2.2) When the incentive price is x0, the critical point N is reached. At this point, users participate in demand response and reduce their load, and the overall system load does not increase. When the incentive price continues to increase to x1, the system saturates, and user response reaches its maximum value. When the incentive price is x, the upper and lower limit curves of demand response λ... min With λ max Satisfy the following expression:

[0023]

[0024]

[0025] 2.3) Load reduction ΔP due to user participation in demand response load for:

[0026]

[0027] In the formula, N load The number of users participating in the demand response; [λ] ± Let [λ] be the interval number of the demand response coefficient. ± ∈ [λ min ,λ max ];[W DR ] ± P represents the number of load ranges participating in demand response. load_i For the load of the i-th user;

[0028] 2.4) Number of total user load intervals after demand response [P] load_DR ] ± for:

[0029] [P load_DR ] ± =P load -[ΔP load ] ± P load This represents the total user load.

[0030] Furthermore, the interval method is applied to wind and solar power, electricity prices, carbon prices, and demand response in the energy system.

[0031] Furthermore, the objective function is solved as follows:

[0032] 3.1) Based on satisfying the constraints of the energy system, minimize the operating cost of the virtual power plant within a single scheduling cycle T, using [ ] as a unified approach. ± The objective function represents the interval number as follows:

[0033] min[Cvpp ] ± =min(C op,T +C th,T +C bs,T +[C ce,T ] ± +[C DR,T ] ± ), where, [C vpp ] ± C represents the total operating cost of the VPP system during time period T; op,T The operating and maintenance costs of wind power, solar power, and energy storage; C th,T For gas turbine operating costs; [C bs,T ] ± For the range of electricity purchase and sale costs; [C] ce,T ] ± For the carbon emission cost range; [C DR,T ] ± This represents the range of demand response costs.

[0034] 3.2) Establish an interval number linear programming model based on steps 1) and 2), and solve it using the two-stage decomposition algorithm in the strong interval linear programming method. Decompose the model into the optimal sub-model and the worst sub-model, substitute them into the objective function, and solve for their optimal values ​​to form the optimal value interval number. Use the YALMIP toolbox in MATLAB to model the model and call the CPLEX12.8.0 solver to solve it to obtain the optimal scheduling scheme.

[0035] The beneficial effects of this invention are as follows: This invention is based on a market-oriented optimization scheduling method for virtual power plants with multiple uncertainty characteristics. It combines the scenario method and the interval method to handle multiple uncertainties, comprehensively considers multiple uncertainty factors, establishes an appropriate VPP optimization scheduling model, and maintains system stability and security while ensuring the economic operation of the VPP. Attached Figure Description

[0036] Figure 1 This is a flowchart of the scene generation and reduction method based on the probabilistic distance between Monte Carlo and Manhattan in the present invention.

[0037] Figure 2 This is a demand response coefficient curve diagram in the method of the present invention;

[0038] Figure 3 This is a diagram of the VPP optimized scheduling system architecture in the method of this invention;

[0039] Figure 4 This is a flowchart of the interval linear programming solution in the method of the present invention;

[0040] Figure 5a This is a landscape output diagram after scene processing in an embodiment of the method of the present invention;

[0041] Figure 5b This is a graph showing the electricity purchase and sale price after scenario-based processing in an embodiment of the method of the present invention.

[0042] Figure 6a This is the scheduling output diagram for scenario 1 in the method embodiment of the present invention;

[0043] Figure 6b This is a diagram showing the electricity purchase and sale and the charging and discharging state of the battery in Scenario 1 of the method embodiment of the present invention;

[0044] Figure 7a This is a power purchase and sale diagram for scenario 1 in the method embodiment of the present invention;

[0045] Figure 7b This is a power purchase and sale diagram for scenario 2 in the method embodiment of the present invention;

[0046] Figure 7c This is a diagram of the energy storage charging and discharging power in scenario 1 of the method embodiment of the present invention;

[0047] Figure 7d This is a diagram of the energy storage charging and discharging power in scenario 2 of the method embodiment of the present invention;

[0048] Figure 8 This is a diagram illustrating the impact of carbon price uncertainty on unit output.

[0049] Figure 9 This diagram illustrates the impact of demand response uncertainty on user load in this invention. Detailed Implementation

[0050] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. These embodiments are based on the technical solution of the present invention and provide detailed implementation methods and specific operating procedures. However, the scope of protection of the present invention is not limited to the following embodiments.

[0051] Current energy dispatching methods are mostly focused on deterministic optimization dispatching models for Virtual Power Plants (VPPs), without fully considering the uncertainties of factors such as renewable energy output, demand response, carbon trading prices, and electricity purchase and sale prices in VPPs. This leads to discrepancies between dispatching simulation results and actual conditions, resulting in inaccuracies. Furthermore, some existing methods consider the impact of individual uncertainties in VPP models but fail to characterize the multiple uncertainties in VPP optimization dispatching models.

[0052] This invention aims to establish a universally applicable VPP (Virtual Power Utility) optimal scheduling model using appropriate uncertainty handling methods. In a VPP system, wind power output is affected by factors such as regional wind speed, season, and policies, while photovoltaic (PV) output is affected by factors such as regional solar irradiance and day length, exhibiting strong time-varying characteristics and significant fluctuations over short periods. Simultaneously, electricity and carbon prices fluctuate due to supply and demand dynamics and pricing relationships. Demand response, as a means of user-side load regulation, is easily influenced by factors such as user psychology, user participation, and user age. Under the combined influence of multiple uncertainties, the VPP coordination and control center needs to coordinate the output of various aggregation units in actual operation to maintain the safe and stable operation of the system.

[0053] The technical solution adopted in this invention is as follows:

[0054] 1. A scene generation and reduction method based on Monte Carlo and probabilistic distance is used to handle uncertainties in wind, solar and electricity prices.

[0055] This invention employs the Monte Carlo method to mine the random characteristics of uncertainty sources, using historical data as samples to ultimately generate scenarios that satisfy these characteristics; it uses the Manhattan Distance Method to measure geometric distance, indicating the proximity between data points. The Manhattan Distance (MD) is defined as:

[0056]

[0057] In the formula, n is the index of the data, and i and j are different data groups. Scene reduction methods based on probabilistic distance are currently widely used in typical scenarios of microgrids, active distribution networks, and integrated energy systems. To obtain scenes containing probabilistic information and reduce them, this invention proposes a scene generation and reduction method based on a combination of Monte Carlo and probabilistic distance, and applies it to a VPP system. For the large set of scenes generated by the Monte Carlo method, a fast prior elimination technique based on probabilistic distance is used for scene reduction. The basic idea is as follows:

[0058] 1) For typical prediction data, a certain standard deviation s is selected and multiplied by the standard normal distribution σ of the generated random numbers as the data fluctuation, which is then superimposed on the original prediction value D. pre Above, the Monte Carlo method is used to generate m sets of typical prediction scenarios S with equal probability. pre As shown in equation (2).

[0059] S pre =D pre +sσ (2)

[0060] 2) The scenarios generated in step 1) are equally probable scenarios with a probability of 1 / m. Therefore, the Manhattan distances between each pair of the m scenarios are calculated to form a matrix d.ge As shown in equation (3).

[0061]

[0062] In the formula, d ge Let i be an m×m symmetric matrix, and let i and j be the scene numbers generated at the i-th and j-th time, respectively.

[0063] 3) Based on the Manhattan distance matrix d between scenes obtained in step 2), ge Calculate the sum of the probabilistic distances between each scene and the remaining scenes, and form the probabilistic distance matrix Y. ge .

[0064]

[0065] 4) Based on the probability distance matrix Y ge The scene with the smallest probability distance from the remaining scenes is selected and denoted as scene S. pre (a), where a is the index value of the scene, based on matrix d ge In the prediction scenario set S pre Let S be the scene with the smallest Manhattan distance to scene a. pre (b), where b is the index of the scene; Scene S pre (a) Merge into scene S pre (b) and scene S pre (a) Probability p a Overlay onto scene S pre (b) probability p b The new scene probabilities are obtained from the above, as shown in equation (5), and then applied to the prediction scene set S. pre Delete scene a.

[0066]

[0067] 5) Using the method described in 4), the number of scenes is continuously reduced to k (k is any positive integer), ultimately resulting in k reduced scenes {S}. pre1 ,S per2 ,S pre3 ,…S prek} and the probabilities {p1, p2, p3, ... p} corresponding to each scene after reduction. k}

[0068] The flowchart for this scenario generation and reduction method is attached. Figure 1 As shown.

[0069] The uncertainties in wind power, solar power output, and electricity prices can all be used to generate corresponding scenario S based on the fluctuation patterns of historical data. preGiven the probability p of the corresponding scenario, compared to other methods for handling uncertainty, the scenario generation and reduction method can fully utilize historical data, avoid errors caused by tedious mathematical calculations and assumptions, and has stronger universality. Therefore, the set W1 generated by the scenario generation and reduction method based on the Monte Carlo and Manhattan probability distances described above to handle uncertainty is:

[0070]

[0071] In the formula, {S~xpre1 S~xpre2.....S~xprek} and {P~xpre1 P~xpre2...P~xprek} represent the generated and reduced k uncertainty scenarios for wind power, photovoltaics, and electricity prices, respectively, and their corresponding scenario probabilities. xpre This method ultimately generates scenario information corresponding to post-wind power, photovoltaics, and electricity prices.

[0072] (2) The interval method is used to deal with the uncertainty of carbon price and demand response.

[0073] For dealing with uncertainties in carbon prices and demand response, scenario-based approaches are not suitable due to a lack of substantial supporting data. In contrast, interval methods require less data and do not need precise probability distribution models. They only need to solve for the upper and lower boundaries to optimize the results of the objective function interval, highlighting the impact of uncertain parameters on the system, which aligns with the current research situation.

[0074] Carbon trading price k c Due to the influence of market supply and demand, government policies, and the actual profit needs of enterprises, the fluctuations will occur in real time. The interval method model is as follows:

[0075]

[0076] In the formula, [α] ± The range of carbon price volatility coefficients is determined based on historical carbon price volatility data, with upper and lower limits α. min With α max ;k c0 The average carbon price over a certain period is obtained from historical data; [k c ] ± This represents the price range for carbon trading.

[0077] This invention primarily considers IBDR (Incentive-Based Demand Response), with participating users mainly being interruptible loads, transferable loads, and reduceable loads. The user demand response coefficient is λ, and the incentive price is x. The demand response coefficient curve based on user psychology is shown in the attached figure. Figure 2 As shown. (Attached) Figure 2 The shaded area and the line y = λ1 represent the effective response portion of the user. When the incentive price is 0, although the user has a certain response space [0, λ0], in reality, the user uncertainty is relatively high, and most users do not respond or even increase the load. Afterwards, as the incentive price increases, the demand response coefficient continues to increase, and the user behavior gradually transitions from increasing the load to participating in the response to reduce the load. When the incentive price is x0, the critical point N is reached. At this time, all users will participate in the demand response and reduce the load, and the overall system load will not increase. When the incentive price continues to increase to x1, the system saturates. At this time, the user response reaches its maximum value, and the uncertainty of user participation in the response can be ignored.

[0078] When the incentive price is x, the upper and lower bound curves of the demand response are λ. min With λ max Satisfy the following expression:

[0079]

[0080]

[0081] User-involved load reduction ΔP load for:

[0082]

[0083] In the formula, N load [λ] represents the number of users participating in the demand response. ± The interval number of the demand response coefficient can be represented as [λ]. ± ∈[λ min ,λ max ], [W DR ] ± P represents the number of load ranges participating in demand response. load_i Let represent the load of the i-th user.

[0084] The upper and lower limits of the demand response coefficient are based on Figure 2 The upper and lower limits can be determined based on the incentive price. The range of load proportions participating in demand response is given data, and the participating load volume is the actual load volume of users participating in demand response. All the summation formulas in the entire text are matched during the solution process based on... Figure 4 The solution process mainly involves decomposing the model into an upper limit sub-model and a lower limit sub-model, which correspond to the maximum and minimum values ​​of each interval, respectively. Then, the upper and lower limit values ​​are obtained by solving the upper limit model and the lower limit model respectively.

[0085] Number of total user load intervals after demand response [P] load_DR ] ± for:

[0086] [P load_DR ] ± =P load -[ΔP load ] ± (11), P load This represents the total user load.

[0087] (3) Establishment of VPP optimal scheduling model with source load and storage considering uncertainty

[0088] The VPP system studied in this invention utilizes communication technology to aggregate its internal units into a whole, participate in power system dispatch, and connect with the main grid for energy exchange. While meeting user energy demands and corresponding constraints, it aims to minimize total costs (including wind and solar operation and maintenance costs, gas turbine costs, demand response costs, carbon emission costs, and electricity purchase and sale costs). The key lies in handling multiple uncertainties, integrating scenario generation and reduction methods with interval methods to establish a day-ahead optimization dispatch model for VPP based on interval linear programming. The optimized dispatch architecture of the VPP system is attached. Figure 3 As shown.

[0089] Optimization objective: To minimize the VPP operating cost within a single scheduling period T while satisfying system constraints, using [ ] as a unified standard. ± The interval number is represented. The cost function is shown in equation (12):

[0090] min[C vpp ] ± =min(C op,T +C th,T +C bs,T +[C ce,T ] ± +[C DR,T ] ± (12)

[0091] In the formula, [C vpp ] ± C represents the total operating cost of the VPP system during time period T; op,T The operating and maintenance costs of wind power, solar power, and energy storage; C th,T For gas turbine operating costs; [C bs,T ] ± For the range of electricity purchase and sale costs; [C] ce,T ] ± For the carbon emission cost range; [C DR,T ] ± This represents the range of demand response costs.

[0092] 1) Operation and maintenance costs C of wind power, photovoltaic power, and energy storage op,T

[0093]

[0094] In the formula, t represents the specific scheduling period; the specific relationship between the scheduling period T and the specific scheduling period t is T = N * t, where N is a positive integer; s w s sol s sto The operation and maintenance costs are divided into wind turbine operation and maintenance costs, photovoltaic operation and maintenance costs, and energy storage battery operation and maintenance costs; P w (t), P sol (t), P dis (t) represents the wind power output, photovoltaic power output, and battery discharge capacity at time t, respectively.

[0095] 2) Gas turbine operating cost C th,T

[0096] The operating cost of a gas turbine unit consists of fuel costs and start-up and shutdown costs, with fuel costs C fue,T The start-up and shutdown cost C is often expressed as a quadratic function related to the unit's output. ss,T The startup and shutdown status of the gas turbine during cycle T is related to the startup and shutdown costs. If the unit starts or stops at this time, it will be included in the startup and shutdown costs, as shown in equation (14):

[0097]

[0098] In the formula, a, b, and c are the coefficients of the quadratic, linear, and constant terms of the fuel cost, respectively, which are usually determined by mathematical fitting; P Gi (t) represents the output value of the i-th gas turbine in time period t; C Gsi With C Gti Let δ represent the start-up cost and shutdown cost of the i-th gas turbine, respectively; Gsi (t) and δ Gti (t) is a Boolean variable representing the start-up and shutdown status of the i-th gas turbine during time period t. During startup, δ... Gsi (t) is set to 1, otherwise it is set to 0. Similarly, δ is set when the machine stops. Gti (t) Set to 1 otherwise set to 0; N u This refers to the number of gas turbines.

[0099] 3) VPP electricity purchase and sale cost C bs,T

[0100]

[0101] In the formula, C buy (t) and C sell (t) represents the purchase price and sales price of electricity in the power grid during the t-th time period, respectively; P buy (t) and Psell (t) represents the power purchased and the power sold by VPP to the grid at time t, respectively. (Power purchased is defined as a positive number, and power sold is defined as a negative number).

[0102] 4) Carbon emission costs [C] ce,T ] ±

[0103] A baseline method is used to determine carbon trading allowances. If a VPP's actual carbon emissions in time period t are less than its allowance, it can sell the remaining carbon emission rights on the carbon trading market for profit; otherwise, it must purchase the corresponding carbon emission rights on the carbon trading market. As shown below:

[0104]

[0105]

[0106]

[0107] In the formula, E Gc,i (T), E Gb,i (T) represents the net CO2 emissions and carbon quota of the i-th gas turbine within the scheduling period T, respectively; γ Gi λ represents the carbon emissions per unit output of the i-th gas turbine; Gi The baseline carbon emission allowance per unit of electricity; [k c ] ± This represents the carbon price range.

[0108] 5) Demand response cost [C] DR,T ] ±

[0109] Demand response cost is expressed as:

[0110]

[0111] In the formula, P load_DRi (t) represents the load that the i-th user is willing to participate in demand response at time t; x i (t) represents the incentive price for the i-th user at time t.

[0112] Constraints:

[0113] 1) Gas turbine constraints

[0114] The unit output and ramping constraints are shown in equation (20).

[0115]

[0116] In the formula, δ Gri(t) is a Boolean variable representing the operating status of the i-th unit at time t; it is set to 1 if the unit is running and 0 if it is shut down. Gi_min P G These are the lower and upper limits of the output of the i-th gas turbine, respectively; P Ci Let be the ramp power of the i-th gas turbine.

[0117] The unit state constraints are shown in equation (21).

[0118]

[0119] 2) Power balance constraints

[0120] P net (t)+P w (t)+P sol (t)+P Gi (t)=[P load_DR (t)] ± +P bat (t) (22)

[0121] In the formula, P net (t) represents the exchange power between VPP and the main network at time t, with the specific meaning shown in equation (23); P bat (t) represents the battery charging power at time t, and the other parameters are the same as before.

[0122] 3) VPP power purchase and sale transaction constraints

[0123]

[0124] In the formula, δ bs (t) is a Boolean variable representing the VPP electricity purchase / sale flag at time t; it is set to 1 when electricity is purchased and 0 when electricity is sold; P power This is the limit for power exchange between the VPP and the main grid.

[0125] 4) Energy storage constraints

[0126] The battery charging and discharging constraints are shown in equation (24).

[0127]

[0128] In the formula, P bat (t), P cha (t), P dis (t) represents the battery output, charging capacity, and discharging capacity during time period t, respectively; δ cha (t), δ dis (t) is a Boolean variable representing the battery charging and discharging status at time t, and δ during charging. cha (t) is set to 1, otherwise it is 0. During discharge, δ dis(t) is set to 1, otherwise it is 0; P cs These are the limits for charging and discharging batteries.

[0129] The battery state constraints are shown in equation (25).

[0130]

[0131] In the formula, δ sta (t) is the battery resting flag at time t. It is set to 1 when the battery is neither discharging nor charging, and 0 in other cases.

[0132] The state of charge (SOC) of a battery at time t is defined as shown in equation (26).

[0133]

[0134] In the formula, E bat (t) represents the battery charge at time t, E max This is the maximum capacity of the battery.

[0135] The battery charging constraint is as shown in equation (27).

[0136] E bat (t)=E bat (t-1)+P cha (t)-P dis (t) (27)

[0137] Divide both sides of equation (27) by E. max We can obtain:

[0138]

[0139] There are also

[0140] SOC min ≤SOC(t)≤SOC max (29)

[0141] By combining equations (28) and (29) and summing both sides, we can obtain the battery SOC constraint condition, as shown in equation (30).

[0142]

[0143] 5) Incentive-based demand response constraints

[0144] The volatility of the demand response at the incentive price of x is F(x).

[0145]

[0146] The corresponding constraints are as follows:

[0147]

[0148] In equation (32), the constraints are, in order, demand response volatility, response capability, and response coefficient. Where, F min With F max These represent the minimum and maximum volatility coefficients, respectively (the volatility constraint here is to reflect that volatility cannot be infinitely small or infinitely large, but rather varies within a specific range); χ represents the maximum value of the VPP system demand response coefficient (the demand response coefficient constraint here is to reflect the range of coefficient variation, combined with...). Figure 2 (where χ takes λ1, the response cannot be unlimited).

[0149] An interval linear programming (ILP) model is established. For this model, a two-stage decomposition algorithm from the strong interval linear programming method is used to solve it. The model is decomposed into optimal and worst sub-models, and their optimal values ​​are obtained by substituting them into the objective function, forming the optimal value interval number. The model is modeled using the YALMIP toolbox in MATLAB and solved using the CPLEX 12.8.0 solver. The specific process is shown in the attached figure. Figure 4 As shown.

[0150] This invention takes a small VPP system in East China as the research object, and the VPP aggregation unit is shown in the attached figure. Figure 3 This includes gas turbines, wind power, photovoltaics, battery energy storage, and flexible user loads, etc. The main parameters of the example are as follows: scheduling cycle T = 24h, optimization time scale Δt = 1h; based on the historical fluctuation patterns of wind and solar power and electricity prices in the region during summer, the wind and solar power fluctuation amplitude s is taken as 20%, and the electricity price fluctuation amplitude is taken as 50% in the scenario generation and reduction method used. The wind and solar power output, flexible user loads, and electricity price curves after scenario generation and reduction are shown in the attached figure. Figure 5a , 5b As shown. Parameters related to gas turbines, battery energy storage, and demand response are shown in Tables 1a, 1b, and 1c. The average value k during the carbon trading price scheduling cycle. c0 Taking a price of 1.25 yuan / kg, the fluctuation coefficient [α] is... ± The carbon emission benchmark for a unit of electricity is set at ±20% and 0.76 kg / kWh; the power exchange limit between the VPP system and the main grid is set at 200 kW; and the operation and maintenance costs for wind power, photovoltaic power, and energy storage batteries are set at 0.52 yuan / kW, 0.72 yuan / kW, and 0.5 yuan / kW, respectively.

[0151] Table 1a

[0152]

[0153]

[0154] Table 1b

[0155]

[0156] Table 1c

[0157]

[0158] To compare and analyze the impact of multiple uncertainties on the VPP optimal scheduling results in this paper, six scenarios are set up for analysis based on different uncertainty handling methods and the impact generated during various uncertainty handling processes. The specific scenario settings are as follows:

[0159] Scenario 1: Wind, solar and electricity prices are processed using the scenario generation and reduction method, while carbon price and demand response modeling uses the midpoint of the interval numbers.

[0160] Scenario 2: Wind and solar power are processed using the scenario generation and reduction method. Electricity price is taken as time-of-use electricity price, and carbon price and demand response are taken as the midpoint of the interval number.

[0161] Scenario 3: Wind, solar and electricity prices are handled using the scenario generation and reduction method, carbon prices are handled using the interval method, and the demand response model takes the median value of the interval.

[0162] Scenario 4: Wind, solar and electricity prices are processed using the scenario generation and reduction method, carbon price is taken as the midpoint of the interval number, and demand response is processed using the interval method.

[0163] Scenario 5: Wind, solar and electricity prices are handled using the scenario generation and reduction method, while carbon prices and demand response are handled using the interval method.

[0164] Scenario 6: Wind and solar power, electricity prices, carbon prices, and demand response are all processed using the interval method.

[0165] The optimized scheduling results for Scenario 1 are attached. Figure 6a , 6b As shown. (Combined with the appendix) Figure 6a , 6b and appendix Figure 5aAs shown in section 5b, during the period from 1:00 to 4:00, wind and solar power resources are relatively scarce, nighttime user load demand is low, and the electricity purchase price is low. At this time, the VPP mainly meets user load demand through purchased power and battery discharge, and the gas turbine output is limited due to carbon cost control. During the period from 5:00 to 12:00, wind and solar power resources increase, with solar power resources reaching their maximum. User load and electricity purchase and sale prices increase significantly. Due to the higher electricity sale price, the VPP system sells excess electricity to the electricity market at 5:00, 8:00, and 9:00 to generate profit and reduce the overall system operating cost. At this time, the gas turbine purchases more electricity through carbon trading. The increased carbon emission rights are used to charge batteries between 6:00 and 7:00, which will then discharge between 8:00 and 9:00. At this time, the VPP system meets user load demand by increasing gas turbine output and utilizing additional wind and solar resources. Between 13:00 and 19:00, user load continues to increase, and the electricity purchase and sale price decreases. Therefore, the VPP system purchases electricity and charges batteries. Between 20:00 and 24:00, wind power resources reach their maximum, and user load also reaches its maximum. The gas turbine maintains minimum output to reduce carbon costs. At this time, the purchase price of electricity is higher, while the sale price is lower. The VPP system meets user demand during peak load periods by purchasing electricity and discharging batteries.

[0166] The VPP coordination and control center can flexibly choose whether to increase the output of gas turbines or conduct power purchase and sale transactions based on the user's load demand and the electricity purchase and sale price. The energy storage equipment stores the system's excess electricity during off-peak hours and supplies power to the load during peak hours, alleviating the problem of time-sharing between user load demand and wind and solar resource supply, and improving the capacity for renewable energy consumption.

[0167] In summary, VPP systems can combine electricity purchase and sale prices, load demand, and energy storage charging and discharging to meet load demand while simultaneously achieving low-price purchase and high-price sale of electricity, thus profiting in the electricity market. By rationally regulating its internal resources during the dispatch cycle, the stability of the entire system is ensured, and the economic efficiency of VPP operation is improved.

[0168] Impact of Electricity Price Uncertainty on Dispatch Results: To describe the impact of electricity price uncertainty on dispatch results, scenarios 1 and 2 are compared and analyzed. The VPP's electricity purchase and sale and charging / discharging power under scenarios 1 and 2 are shown in the attached figures. Figure 7a , 7b As shown in Figures 7c and 7d. (Refer to the attached diagram.) Figures 7a-7d Zhonghe Fu Figure 5a , 5bAs can be seen, in Scenario 2, the VPP system purchases less electricity and sells more electricity compared to Scenario 1. The VPP system will purchase electricity from the electricity market when the local purchase price is low and sell excess electricity to the electricity market when the selling price is high. In Scenario 1, the main electricity selling periods are 5:00, 8:00, and 9:00, while Scenario 2 has two additional electricity selling periods: 21:00 and 22:00. Therefore, considering the uncertainty of electricity prices, the VPP's electricity selling volume is reduced. When using the time-of-use pricing in Scenario 2, the VPP profits more from electricity purchase and sale. Furthermore, horizontally, as the purchase price increases, in order to save costs, the VPP will discharge its energy storage batteries to meet its own needs, reducing electricity purchases. The charging and discharging behavior of energy storage batteries is similar in Scenario 1 and Scenario 2. Overall, electricity price uncertainty mainly affects the VPP's electricity purchase and sale behavior. Considering the uncertainty of electricity prices, the VPP system will change its electricity purchase and sale behavior to cope with the uncertainty.

[0169] Impact of Carbon Price and Demand Response Uncertainty on Scheduling Results: The uncertainty in carbon price and demand response is mainly characterized using the interval method. Therefore, the optimized scheduling data consists of interval numbers. To analyze the impact of this uncertainty, the optimization results from scenarios 3 and 4 are used, with the carbon price interval number being [1, 1.5] and the demand response coefficient interval number being [0.25, 0.35]. The results are shown in the attached figure. Figure 8 With appendix Figure 9 As shown.

[0170] From the appendix Figure 8 It can be seen that considering the uncertainty of carbon prices can effectively limit the output of gas turbines in VPP systems, thereby achieving the effect of low carbon emissions and carbon reduction. Under the carbon mechanism design in this paper, once the gas turbine generates electricity, it is necessary to purchase corresponding carbon emission rights from the carbon market, which increases costs and also has an impact on the environment. As the carbon price increases, the output of the gas turbine and the carbon emissions decrease continuously. The carbon emissions of the gas turbine are eventually limited to 700kW and 532kg. The upper limit curves of both are limited to the minimum value when the carbon price is 1.11 yuan / kg, and the lower limit curves are limited to the minimum value when the carbon price is 1.19 yuan / kg. The upper and lower limit ranges are most sensitive when the carbon price is 1.08 yuan / kg, indicating that the carbon reduction effect of the entire VPP system is best when the carbon price range is [1.11, 1.25], and carbon emissions can be effectively reduced in a short period of time when the range is [1.08, 1.11]. Therefore, the consideration of the uncertainty of carbon prices mainly affects the scheduling decision of gas turbines.

[0171] From the appendix Figure 9It can be seen that after the introduction of the demand response mechanism, the load demand is significantly reduced, the peak-to-valley difference of the load curve is significantly reduced, and the volatility of the curve is significantly improved. This is beneficial to mitigating the impact of the uncertainty of new energy on the system. Scenario 4 uses the interval method to analyze the demand response, which can obtain the corresponding upper and lower limit curves of user load and determine the range of user participation in demand response. The upper and lower limits cover the curves of Scenario 1, indicating that after considering the uncertainty of demand response using the interval method, the system can be guaranteed to operate within a controllable range, increasing the system stability. Combining with the electricity price curve in Figure 5, it can be seen that the demand response range is affected by the real-time electricity price. The higher the real-time electricity price, the larger the response range; the lower the real-time electricity price, the smaller the response range. The demand response range is also larger during the two peak electricity price periods of 11:00-13:00 and 19:00-21:00. The number of load reduction intervals within the maximum response range is [239.07, 638.14]. Therefore, the uncertainty of demand response mainly regulates the user load of the VPP system, smoothing peaks and filling valleys to maintain system stability.

[0172] Economic Analysis of VPP Operation under Multiple Uncertainties: The operating costs of VPP under each scenario are shown in Appendix Table 2. From the cost analysis of scenarios 1 and 2, it can be seen that considering the uncertainty of electricity prices, the operating cost of VPP increases by RMB 1337.63, an increase of 2.4%. This indicates that considering the uncertainty of electricity prices sacrifices some economic efficiency but improves the stability of the entire VPP system operation, which is of practical significance. Comparing scenario 1 with scenarios 3 and 4, the cost change ranges are [-798.49, 2144.52] and [-1134.45, 4951.63], respectively, with change magnitudes of [-1.40%, 3.76%] and [-1.99%, 5.68%]. Therefore, considering the uncertainty of carbon prices and demand response will increase the width of the cost range, but the model will be more accurate. At the same time, the uncertainty of demand response has a greater impact on the cost of the VPP system. A cost comparison between scenarios 5 and 6 shows that the total cost ranges are RMB 8768.68 and RMB 10986.25, respectively. The lower and upper limits of the cost range for scenario 6 include the costs in scenario 5, indicating that the method proposed in this paper can improve the overall economy while ensuring the characterization of multiple uncertainties in scenario 5. This method is more advantageous than the scenario method used only in scenario 1 and the interval method used only in scenario 6.

[0173] Table 2

[0174]

[0175] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. A market-based optimal scheduling method for virtual power plants based on multiple uncertainty representations, characterized in that, Specifically, the steps include the following: 1) Utilize the fluctuation patterns of historical data in the energy system to generate a large number of corresponding scenarios using the Monte Carlo method. S pre and the probability of the corresponding scenario occurring p Then, a fast previous generation elimination technique based on Manhattan probability distance is used to reduce the scene and finally generate scene information; 2) For uncertainties in energy systems that do not require precise probability distributions and have limited historical data, the interval method is used to solve for the upper and lower boundaries; 3) The virtual power plant acts as the power grid management center, communicating with distributed energy resources, participating in power system dispatch, and connecting with the main grid for energy exchange. Based on meeting user energy demands and corresponding constraints, it aims to minimize the total cost of the objective function, thereby establishing a day-ahead dispatch model for the virtual power plant based on interval linear programming. For handling uncertainty, steps 2) and 3) are integrated, and the objective function of the established virtual power plant day-ahead scheduling model is optimized to obtain the optimized scheduling scheme. The specific steps for implementing step 1) are as follows: 1.1) For typical forecast data, select the standard deviation. s Compared with the standard normal distribution of generated random numbers σ Multiplication is used to account for data fluctuations and is then added to the original predicted value. D pre The above is generated using the Monte Carlo method. m A set of typical prediction scenarios with equal probability S pre : ； 1.2) The scene generated in step 1.1) has a probability of 1 / m Given equal probability scenarios, the following can be calculated: m A matrix is ​​formed by the pairwise Manhattan distances between each scene. d ge ; 1.3) Based on the Manhattan distance matrix between scenes obtained in step 1.2), d ge Calculate the sum of the probabilistic distances between each scene and the remaining scenes, and form a probabilistic distance matrix. Y ge ; 1.4) Based on the probability distance matrix Y ge The scene with the smallest probability distance from the remaining scenes is selected and denoted as scene. S pre ( a ), a The index value of the scene, based on the matrix. d ge In the prediction scenario set S pre Find the scene a The scene with the smallest Manhattan distance is denoted as scene. S pre ( b ), b For the scene index; Scene S pre ( a Merge into scene S pre ( b In the scene, S pre ( a Probability p a Overlay on scene S pre ( b The probability of ) p b The new scene probabilities are obtained and applied to the prediction scene set. S pre Delete scene a ; 1.5) The number of scenes is continuously reduced to the method described in 1.4). k One, ultimately resulting in the reduction k One scenario { S pre1 , S per2 , S pre3 ,…S prek } and the probability of each scene after reduction { p 1 , p 2 , p 3 ,…p k }; Step 2) The interval method for handling carbon price and demand response uncertainty: 2.1) Regarding carbon trading prices k c The interval method model is as follows: ， In the formula, [ α ] ± The range of carbon price volatility coefficients is determined based on historical carbon price volatility data, with upper and lower limits respectively. α min and α max ; k c0 The average carbon price over a certain period is obtained from historical data; k c ] ± This represents the price range for carbon trading. 2.2) When the incentive price is x At time 0, the critical point N is reached. At this point, users participate in demand response and reduce their load, and the overall system load will not increase. When the incentive price continues to increase to... x At time 1, the system saturates, and the user response reaches its maximum value; when the incentive price is... x At that time, the upper and lower limits of the demand response curve λ min and λ max Satisfy the following expression: ， ， 2.3) Load reduction due to user participation in demand response P load for: ， In the formula, N load The number of users participating in the demand response; λ ] ± The number of intervals for the demand response coefficient is represented as [ λ ] ± ∈[ λ min , λ max ];[ W DR ] ± The number of load ranges participating in demand response. P load_i For the first i The load of each user; 2.4) Number of total user load intervals after demand response [ P load_DR ] ± for: , Total user load; Solving the optimization objective function: 3.1) Based on satisfying the constraints of the energy system, minimize the operating cost of the virtual power plant within a single scheduling cycle T, using [ ] as a unified standard. ± Let the interval number be represented, and the objective function be as follows: ， In the formula, [ C vpp ] ± The total operating cost of the VPP system during time period T; C op,T The operating and maintenance costs of wind power, solar power, and energy storage; C th,T For gas turbine operating costs; C bs,T For VPP electricity purchase and sale costs; C ce,T ] ± The range of carbon emission costs; C DR,T ] ± This represents the range of demand response costs. 3.2) Establish an interval number linear programming model based on steps 1) and 2), and solve it using the two-stage decomposition algorithm in the strong interval linear programming method. Decompose the model into the optimal sub-model and the worst sub-model, substitute them into the objective function, and solve for their optimal values ​​to form the optimal value interval number. Use the YALMIP toolbox in MATLAB to model the model and call the CPLEX12.8.0 solver to solve it and obtain the optimal scheduling scheme.

Images