Virtual power plant two-stage demand response optimization method based on adaptive control algorithm

Abstract

The invention discloses a virtual power plant two-stage demand response optimization method based on an adaptive control algorithm, and belongs to the technical field of resource scheduling, and the method comprises the steps: S1, obtaining the dynamic response excitation of a demand response main body based on a similarity index and the total response amount of the demand response main body; s2, in different stages, based on dynamic response excitation and constraint conditions, using an adaptive control algorithm to construct a corresponding optimization objective function, and based on the corresponding optimization objective function, constructing an improved quasi-linear demand response model; s3, on the basis of the improved quasi-linear demand response model, adopting a Shapley value distribution method considering risk preference to obtain the income of a demand response subject; and S4, predicting the generation power of the new energy based on the conditional value-at-risk, obtaining an accuracy factor based on the generation power, and optimizing the demand response based on the accuracy factor and the income. The problems that the uncertainty and volatility of large-scale renewable energy integration are difficult to flexibly deal with and the limitation of small-scale monomer capacity of user-side distributed resources is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of resource scheduling, and in particular to a two-stage demand response optimization method for a virtual power plant based on an adaptive control algorithm. Background Art

[0002] With the access of large-scale renewable energy, challenges have been brought to the operation of the power system. The challenges mainly come from the uncertainty and volatility of renewable energy such as wind and solar energy, which makes it more difficult to balance the supply and demand of the power system. In order to meet these challenges, in addition to the flexibility transformation of thermal power units on the power generation side, it is also necessary to tap the flexible response capabilities of distributed resources on the user side. Demand response, as an important part of the flexibility regulation of the power system, is an important way for virtual power plants to interact with the power grid. Improving quasi-linear demand response guides users to adjust their electricity consumption behavior through price signals or incentive mechanisms, thereby reducing the load on the power grid and enhancing the regulation capability of the system. However, the traditional quasi-linear demand response has some limitations, such as the patent number CN117424202A, which is a new energy power system dispatching method considering quasi-linear demand response, including: based on the historical predicted output and historical actual output of new energy, dividing the prediction box and obtaining the uncertainty set; based on the uncertainty set, constructing the uncertainty scenario and generating the severe scenario; constructing the system load quasi-line model under the severe scenario, and solving the system load quasi-line model to obtain the system load quasi-line; based on the historical node load data, obtaining the predicted response interval of each node; constructing the node load quasi-line model, and solving the node load quasi-line model to obtain the optimal load response and node load quasi-line; improving the power supply capacity of the system and the new energy consumption rate. However, the above scheme has the problem that it is difficult to flexibly cope with the uncertainty and volatility of large-scale renewable energy integration and is limited by the small-scale single-unit capacity of user-side distributed resources. Summary of the invention

[0003] In view of the problem that the existing technology is difficult to flexibly cope with the uncertainty and volatility of large-scale renewable energy integration and is limited by the small-scale single-unit capacity of distributed resources on the user side, the present invention provides a two-stage demand response optimization method for a virtual power plant based on an adaptive control algorithm. Through quasi-linear incentives and constraints, an adaptive control algorithm is used to construct a quasi-linear demand response model, which can dynamically adjust the demand response strategy of the two stages according to real-time monitoring demand, thereby flexibly coping with the uncertainty and volatility of large-scale renewable energy integration. At the same time, the introduction of constraints also effectively integrates the distributed resources on the user side, breaking through the limitation of small-scale single-unit capacity, solving the problem that it is difficult to flexibly cope with the uncertainty and volatility of large-scale renewable energy integration and is limited by the small-scale single-unit capacity of distributed resources on the user side, and improving the applicability and response speed of the power grid system.

[0004] In order to solve the above technical problems, the present invention provides a two-stage demand response optimization method for a virtual power plant based on an adaptive control algorithm, comprising the following steps: S1: Obtain the dynamic response incentive of the demand response subject based on the similarity index between the demand response subject and the load criterion and the total response amount of the demand response subject; S2: In the day-ahead stage, based on the dynamic response incentive and the first constraint, an adaptive control algorithm is used to construct the day-ahead optimization objective function. In the intraday stage, based on the dynamic response incentive and the second constraint, an adaptive control algorithm is used to construct the intraday optimization objective function. An improved quasi-linear demand response model is constructed based on the day-ahead optimization objective function and the intraday optimization objective function. S3: Based on the improved quasi-linear demand response model, the Shapley value allocation method considering risk preference is adopted to obtain the benefits of the demand response subject; S4: Predict the power generation of renewable energy based on conditional value at risk, obtain the accuracy factor based on the power generation, and optimize the demand response based on the accuracy factor and the benefit.

[0005] After adopting the above technical solution, the present invention has the following advantages: By dynamically responding to incentives and constraints, an improved quasi-linear demand response model is constructed using an adaptive control algorithm. The two-stage demand response strategy can be dynamically adjusted according to real-time monitoring demand, thereby flexibly responding to the uncertainty and volatility of large-scale renewable energy integration. Through continuous optimization before and during the day, the balance between power supply and demand is ensured, and the adaptability and response speed of the power market are improved. By introducing constraints to construct an improved quasi-linear demand response model, the integration of distributed resources on the user side is achieved, breaking through the limitation of small-scale single-unit capacity and significantly improving the regulation potential of the power system; The importance of different types of load resources is accurately assessed by the Shapley value allocation method that takes risk preference into consideration, which makes it easier to formulate corresponding strategies to incentivize users based on their importance, thereby increasing users' active responsiveness to response needs and thus improving the response quality of the power system and user participation. It solves the problem of being difficult to flexibly respond to the uncertainty and volatility of large-scale renewable energy integration and being limited by the small-scale single capacity of distributed resources on the user side.

[0006] Preferably, in S1, the similarity index χ=1-νe -λd , ν is the adjustment factor, e is the base of the natural logarithm, λ is the attenuation coefficient, and d is the Euclidean distance between the demand response subject and the load directrix.

[0007] Preferably, in S1, the dynamic response excitation M award =δb χP D , where δ b is the excitation coefficient, P D is the total response amount of the demand response entity.

[0008] Preferably, in S2, the day-ahead optimization objective function is min C=C buy +C ES +C TL -M before -M sale +Q t , where C is the cost of the virtual power plant in the day-ahead stage, C buy is the electricity purchase cost of the virtual power plant in the day-ahead stage, C ES is the operation and maintenance cost of energy storage in the virtual power plant in the day-ahead phase, C TL The compensation cost for industrial users to transfer load in the day-ahead phase, M before is the dynamic response incentive of the demand response subject in the day-ahead stage, M sale is the revenue from electricity sales of the virtual power plant in the day-ahead phase, Q t It represents the economic losses that may be caused by the uncertainty of wind and solar power output of the virtual power plant at time t in the day-ahead phase.

[0009] Preferably, in S2, the intraday optimization objective function is min B=ΔP AC K AC +ΔP EV K EV +ΔPb grid K grid +ΔPb W K W +ΔPb V K V +ΔP tr K tr +ΔP ES K ES , where B is the adjustment cost of the virtual power plant during the intraday stage, ΔP AC is the total amount of adjustment of the air conditioner in the virtual power plant compared with the day-ahead dispatch plan, K AC is the penalty adjustment coefficient of the air conditioner, ΔP EV is the total amount of electric vehicles in the virtual power plant adjusted compared to the day-ahead dispatch plan, K EV is the penalty adjustment coefficient of the electric vehicle, ΔPb grid K is the total amount of grid interaction power in the virtual power plant compared to the day-ahead dispatch plan, grid is the penalty adjustment coefficient of the grid interaction power, ΔPb W K is the total amount of wind power adjusted in the virtual power plant compared to the day-ahead dispatch plan,W is the penalty adjustment coefficient of wind power, ΔPb V is the total amount of adjustment of PV in the virtual power plant compared to the day-ahead dispatch plan, K V is the penalty adjustment coefficient of the photovoltaic power generation, ΔP tr K is the total amount of industrial transfer load in the virtual power plant compared to the day-ahead dispatch plan, tr is the penalty adjustment coefficient for the industrial load transfer, ΔP ES K is the total amount of energy storage in the virtual power plant compared to the day-ahead dispatch plan, ES is the penalty adjustment coefficient for the energy storage.

[0010] Preferably, S3 includes: S31: Based on the improved quasi-linear demand response model, the optimal target value is obtained using the historical data of the virtual power plant; S32: Obtain the risk preference corresponding to the demand response subject based on the state variable of the demand response subject, and obtain the benefit of the demand response subject based on the optimal target value and the risk preference corresponding to the demand response subject.

[0011] Preferably, in S32, the expression for obtaining the risk preference corresponding to the demand response subject based on the state variable of the demand response subject is: In the formula, ψ z is the fair revenue distribution that participant z deserves in the demand response subject, S represents the number of participants in the subset S of the demand response subject, n is the total number of flexible loads aggregated by the virtual power plant, l is the risk adjustment coefficient based on the improved load criterion, and M Z represents the risk-weighted marginal contribution of the participant z.

[0012] Preferably, in S32, the expression for obtaining the revenue of the demand response subject based on the optimal target value and the risk preference corresponding to the demand response subject is R x =(1-φ x )R, where R x is the benefit of participant x in the demand response entity, R represents the optimal target value, φ x is the profit proportion of the participant x in the virtual power plant.

[0013] Preferably, in S4, the expression for obtaining the accuracy factor based on the generated power is: In the formula, F represents new energy, is the accuracy factor of the new energy at time t, are the predicted renewable energy power generation power in the day-ahead stage and the intraday stage respectively, and μ and η are the parameters of the accuracy factor.

[0014] Beneficial effects of this program: By constructing an improved quasi-linear demand response model through dynamic response incentives and constraints, the two-stage demand response strategy can be dynamically adjusted according to real-time monitoring demand, thereby flexibly responding to the uncertainty and volatility of large-scale renewable energy integration. Through continuous optimization on the day before and within the day, the balance between power supply and demand is ensured, and the adaptability and response speed of the power market are improved. By constructing an improved quasi-linear demand response model based on two-stage constraint conditions, the distributed resources on the user side are integrated, the limitation of small-scale single capacity is broken, and the regulation potential of the power system is significantly improved. The importance of different types of load resources is accurately assessed by the Shapley value allocation method that takes risk attitude into consideration, which makes it easier to formulate corresponding strategies to incentivize users based on their importance, thereby increasing users' active responsiveness to response demands, and thus improving the response quality of the power system and user participation. It solves the problem of being difficult to flexibly respond to the uncertainty and volatility of large-scale renewable energy integration and being limited by the small-scale single capacity of distributed resources on the user side.

[0015] The present invention also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the steps of the virtual power plant two-stage demand response optimization method based on an adaptive control algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Other features, objects and advantages of the present invention will become more apparent by reading the detailed description of non-limiting embodiments made with reference to the following drawings. The drawings are only for the purpose of illustrating preferred embodiments and are not to be considered as limiting the present invention. Also, the same reference symbols are used throughout the drawings to represent the same parts.

[0017] Figure 1 It is a flow chart of a two-stage demand response optimization method of a virtual power plant based on an adaptive control algorithm of the present invention; Figure 2 It is an overall framework diagram of the virtual power plant two-stage demand response optimization method based on the adaptive control algorithm of the present invention; Figure 3 This is a graph of industrial load forecasting data in a two-stage demand response optimization method for a virtual power plant based on an adaptive control algorithm of the present invention; Figure 4 This is a wind power prediction data diagram in the virtual power plant two-stage demand response optimization method based on the adaptive control algorithm of the present invention; Figure 5 This is a photovoltaic prediction data diagram in the virtual power plant two-stage demand response optimization method based on the adaptive control algorithm of the present invention; Figure 6 It is the optimal dispatching diagram of the day-ahead flexible load of the virtual power plant in the virtual power plant two-stage demand response optimization method based on the adaptive control algorithm of the present invention; Figure 7 Whether to participate in improving the quasi-linear load per unit load curve diagram in the virtual power plant two-stage demand response optimization method based on the adaptive control algorithm of the present invention; Figure 8 It is the optimal dispatching diagram of the intra-day flexible load of the virtual power plant in the two-stage demand response optimization method of the virtual power plant based on the adaptive control algorithm of the present invention; Fig. 9 It is a day-ahead and day-intraday flexible load adjustment diagram of a virtual power plant in a two-stage demand response optimization method of a virtual power plant based on an adaptive control algorithm of the present invention; Fig.10 It is a normalized load curve diagram of the day-ahead and intra-day loads in the two-stage demand response optimization method of a virtual power plant based on an adaptive control algorithm of the present invention; Fig.11 This is a Shapley method profit distribution diagram that takes risk attitude into consideration in the two-stage demand response optimization method of a virtual power plant based on an adaptive control algorithm of the present invention. DETAILED DESCRIPTION

[0018] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific implementation method described herein is only an optimal embodiment of the present invention, which is only used to explain the present invention and does not limit the scope of protection of the present invention. All other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0019] Before discussing the exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flow charts. Although the flow charts describe the operations (or steps) as sequential processes, many of the operations (or steps) therein can be implemented in parallel, concurrently, or simultaneously. In addition, the order of the operations can be rearranged. The process can be terminated when its operation is completed, but can also have additional steps not included in the drawings; the process can correspond to a method, function, procedure, subroutine, subprogram, etc.

[0020] Embodiment 1: like Figure 1 As shown, a two-stage demand response optimization method for a virtual power plant based on an adaptive control algorithm includes the following steps: S1: obtaining a dynamic response incentive of a demand response subject based on a similarity index between the demand response subject and the load criterion and a total response amount of the demand response subject.

[0021] In S1, the similarity index χ=1-νe -λd , ν is the adjustment factor, e is the base of the natural logarithm, λ is the attenuation coefficient, and d is the Euclidean distance between the demand response subject and the load directrix.

[0022] In S1, the dynamic response excitation M award =δ b χP D , where δ b is the excitation coefficient, P D is the total response amount of the demand response entity.

[0023] In this embodiment, ν is used to strengthen the effect of the similarity between the load curve and the criterion on the incentive, and λ is used to adjust the effect of the similarity between the load curve and the criterion on the incentive. The configuration parameters and related variables of the dynamic response incentive include the similarity index between the demand response market entity and the load criterion and the incentive obtained by each demand response market entity after participating in the criterion-type demand response. The power grid company provides market participants with a reference standard for electricity adjustment by publishing the load criterion to achieve the stability and peak load regulation of the power supply. Market entities can obtain economic incentives by matching the load criterion to optimize the electricity consumption mode. The Euclidean distance can be used to measure the closeness between the standardized load curve and the load criterion. The smaller the Euclidean distance, the more consistent the load curve of the market entity is with the load criterion set by the power grid company. In the formula, and are the normalized load values ​​of the demand response market entity and the power grid company in the tth period, respectively. The accurate quasi-linear incentive is obtained through the similarity index and the total response of the demand response entity, which is convenient for constructing a quasi-linear demand response model based on the accurate linear incentive, and improves the accuracy and flexibility of the constructed quasi-linear demand response model.

[0024] S2: In the day-ahead stage, based on the dynamic response incentive and the first constraint, an adaptive control algorithm is used to construct the day-ahead optimization objective function. In the intraday stage, based on the dynamic response incentive and the second constraint, an adaptive control algorithm is used to construct the intraday optimization objective function. Based on the day-ahead optimization objective function and the intraday optimization objective function, an improved quasi-linear demand response model is constructed.

[0025] In S2, the day-ahead optimization objective function is: min C=C buy +C ES +C TL -M before -M sale +Q t , where C is the cost of the virtual power plant in the day-ahead stage, C buy is the electricity purchase cost of the virtual power plant in the day-ahead stage, CES is the operation and maintenance cost of energy storage in the virtual power plant in the day-ahead phase, C TL The compensation cost for industrial users to transfer load in the day-ahead phase, M before is the dynamic response incentive of the demand response subject in the day-ahead stage, M sale is the revenue from electricity sales of the virtual power plant in the day-ahead phase, Q t It represents the economic losses that may be caused by the uncertainty of wind and solar power output of the virtual power plant at time t in the day-ahead phase.

[0026] In S2, the intraday optimization objective function is min B=ΔP AC K AC +ΔP EV K EV +ΔPb grid K grid +ΔPb W K W +ΔPb V K V +ΔP tr K tr +ΔP ES K ES , where B is the adjustment cost of the virtual power plant during the intraday stage, ΔP AC is the total amount of adjustment of the air conditioner in the virtual power plant compared with the day-ahead dispatch plan, K AC is the penalty adjustment coefficient of the air conditioner, ΔP EV is the total amount of electric vehicles in the virtual power plant adjusted compared to the day-ahead dispatch plan, K EV is the penalty adjustment coefficient of the electric vehicle, ΔPb grid K is the total amount of grid interaction power in the virtual power plant compared to the day-ahead dispatch plan, grid is the penalty adjustment coefficient of the grid interaction power, ΔPb W K is the total amount of wind power adjusted in the virtual power plant compared to the day-ahead dispatch plan, W is the penalty adjustment coefficient of wind power, ΔPb V is the total amount of adjustment of PV in the virtual power plant compared to the day-ahead dispatch plan, K V is the penalty adjustment coefficient of the photovoltaic power generation, ΔP tr K is the total amount of industrial transfer load in the virtual power plant compared to the day-ahead dispatch plan, tr is the penalty adjustment coefficient for the industrial load transfer, ΔP ES K is the total amount of energy storage in the virtual power plant compared to the day-ahead dispatch plan, ES is the penalty adjustment coefficient for the energy storage.

[0027] In this embodiment, The virtual power plant may suffer economic losses due to the uncertainty of wind power w and photovoltaic v output at time t in the day-ahead stage. In the day-ahead stage, the virtual power plant optimizes demand response based on the load guidelines released by the power grid company, new energy forecast data, and the response enthusiasm of internal flexible loads, with the goal of maximizing overall benefits. The day-ahead optimization objective function is defined as the overall benefit obtained by the virtual power plant on the day-ahead, including the following parts: ① dynamic response incentives; ② virtual power plant electricity sales income and electricity purchase costs; ③ ES operation and maintenance costs; ④ industrial users' compensation costs for load transfer. The calculation formulas for each parameter in the day-ahead optimization objective function are as follows: C is the day-ahead cost; M sale , C buy C are the revenue from electricity sales and the cost of electricity purchase of the virtual power plant respectively; ES is the charging and discharging cost of ES; C TL is the compensation cost for load transfer; Q t is the economic loss that may be caused by the uncertainty of wind power w and photovoltaic v output of the virtual power plant at time t in the day-ahead stage, n EV 、n AC is the number of EVs and ACs; P W , P V are the electricity purchase prices from the grid company, wind power w, and photovoltaic power v at time t; Pb t , is the amount of electricity purchased by the virtual power plant from the power grid company, wind power w, and photovoltaic power v at time t; re To provide incentives for electricity prices of new energy sources; ES is the ES operation and maintenance cost coefficient; is the charge and discharge power of ES at time t; TL is the cost coefficient of industrial load; is the response power of industrial load at time t, β is the given confidence level, f(P Xjt ,P Rjt ) is the output deviation of energy X at time t, P Xjt Contribute to the plan of energy X at time t, P Rjt is the actual output of energy X at time t.

[0028] The first constraint condition includes electric vehicle constraint, air conditioning model constraint, energy storage constraint, industrial load constraint, wind power and photovoltaic power generation power constraint and power balance constraint. The electric vehicle constraint includes: assuming that the daily mileage distribution of EV and conventional fuel vehicles is similar. Without considering the EV queuing model, the initial charging time satisfies the following probability density function: In the formula, x represents time; μ is 17.47; σ is 3.41.

[0029] The probability density function of the end charging time is: In the formula, μ is taken as 8.92 and σ is taken as 3.24.

[0030] The probability density of daily mileage of electric vehicles is: Where S represents the daily mileage; μ is 2.98; σ is 1.14.

[0031] The expected power can be obtained by using the daily mileage L of the EV Where η EV,ch is the transformation efficiency; Q EV,km The power consumption per kilometer.

[0032] In addition, considering the user's participation in demand response, the EV charging capacity should be within a reasonable range to ensure the user's daily travel needs and meet the following constraints: In the formula, is the response positivity of EV user i; is the power of EV user i when he leaves; They represent the charging power and maximum charging power of EV user i at time t respectively; For charging efficiency.

[0033] The constraints of the air conditioning model include: To ensure a comfortable indoor environment and user comfort, the temperature of public buildings needs to be controlled within an appropriate range. The first-order thermal equivalent model of public buildings is shown as follows: In the formula, T in,max With T in,min are the maximum and minimum values ​​of the ambient temperature and human comfort temperature of building i at time t, The building i is responsive to the positive response, They represent the cooling power of air-conditioning unit j of building i at time t and the maximum cooling power, R i and C i is the equivalent thermal resistance and heat capacity of building i.

[0034] Energy storage constraints include: Because the energy storage system has multi-time scale coupling characteristics, it is necessary to make a good charging and discharging strategy in the day-ahead stage. The day-ahead constraints are as follows: In the formula, Q t , SOC t They represent the charge and state of charge of ES at time t respectively; Q is the rated capacity of ES, η ch , η dis They represent the charge and discharge efficiency, are the charge and discharge power at time t, They represent the upper and lower limits of the state of charge at time t, P ch,max , P dis,max Respectively represent the maximum value of charging and discharging power, is a 0-1 variable at time t, 1 means charging, 0 means not charging, It is a 0-1 variable at time t, 1 means discharge, and 0 means no discharge.

[0035] Industrial load constraints include: industrial load is regarded as transferable load and participates in virtual power plant regulation. Its constraints are shown in the following formula: In the formula, are the initial power of industrial load at time t, the transfer response power and the power after transfer response, λ be is the industrial load response positivity, ΔP TL,in,max , ΔP TL,out,max They are the upper and lower limits of the power that can be transferred in and out of the industrial load respectively.

[0036] Wind power and photovoltaic power generation power constraints include: In the formula, is the predicted power generation of wind power / photovoltaic power at time t, indicating that the amount of electricity purchased by the virtual power plant from new energy should be less than or equal to the predicted amount of new energy.

[0037] The power balance constraint is:

[0038] In the intraday optimization stage, in order to give full play to the guiding significance of the day-ahead scheduling plan, the intraday rolling optimization aims to minimize the deviation from the day-ahead plan, and incorporate the deviation as a penalty term into the objective function to minimize the total penalty within the cycle. ΔP AC , ΔP EV , ΔPb grid , ΔPb W , ΔPb V , ΔP tr , ΔP ESThey are the total amount of adjustments made to the air conditioners, electric vehicles, grid-interactive electricity, wind power, photovoltaic power, transfer loads, and energy storage in the virtual power plant compared to the day-ahead dispatch plan; ΔPb W , ΔPb V , They are the daily electric power of air conditioning, electric vehicles, grid interaction, wind power, photovoltaics, transfer loads, and energy storage.

[0039] The second constraint condition includes energy storage constraint, industrial load reduction constraint, electric vehicle constraint and power balance constraint. The constraint conditions involved are exactly the same as those formulated by the day-ahead dispatch, so they will not be repeated here. By constructing an improved quasi-linear demand response model through dynamic response incentives and constraints, it is possible to dynamically adjust the two-stage demand response strategy according to real-time monitoring demand, and thus flexibly respond to the uncertainty and volatility of large-scale renewable energy integration. Through continuous optimization on the day-ahead and within the day, it ensures the balance between power supply and demand, and improves the adaptability and response speed of the power market; by constructing an improved quasi-linear demand response model through two-stage constraints, it realizes the integration of user-side distributed resources, breaks through the limitation of small-scale single-unit capacity, and significantly improves the regulation potential of the power system; by improving the quasi-linear demand response model, it also provides a full-time response benchmark for user-side resources, simplifies the operation process of traditional demand response, reduces the difficulty of implementation, and improves the economy and efficiency of demand response.

[0040] S3: Based on the improved quasi-linear demand response model, the Shapley value allocation method considering risk preference is adopted to obtain the benefits of the demand response entity.

[0041] The S3 includes: S31: Based on the improved quasi-linear demand response model, the optimal target value is obtained using the historical data of the virtual power plant; S32: Obtain the risk preference corresponding to the demand response subject based on the state variable of the demand response subject, and obtain the benefit of the demand response subject based on the optimal target value and the risk preference corresponding to the demand response subject.

[0042] In S32, the expression for obtaining the risk preference corresponding to the demand response subject based on the state variable of the demand response subject is: In the formula, ψ z is the fair revenue distribution that participant z deserves in the demand response subject, S represents the number of participants in the subset S of the demand response subject, n is the total number of flexible loads aggregated by the virtual power plant, l is the risk adjustment coefficient based on the improved load criterion, and M Z represents the risk-weighted marginal contribution of the participant z.

[0043] In S32, the expression for obtaining the revenue of the demand response subject based on the optimal target value and the risk preference corresponding to the demand response subject is R x =(1-φ x )R, where R x is the benefit of participant x in the demand response entity, R represents the optimal target value, φ x is the profit proportion of the participant x in the virtual power plant.

[0044] In this embodiment, M z =(1-θ z )(R(S∪{z})-R(S)),(1-θ z ) is used to adjust the marginal contribution, θ z reflects the risk preference of participant z, R(S∪{z})-R(S) represents the marginal contribution of participant z to subset S, and R(S) is the overall benefit of aggregated subset S, i.e., the overall flexible load; the optimal target value is the overall benefit of the virtual power plant, and the response enthusiasm is related to the benefit R allocated by the virtual power plant to each alliance member. u and form alliances to gain benefits R u It is related to the exponential function, which is normalized to the range of 0-1, as follows: Where: A u is the response positivity, α act is the response positivity coefficient, β A is the response coefficient, A u The larger it is, the more motivated the user is and the more willing they are to participate in the response. This approach ensures that members who take greater risks or play a more critical role in demand response can obtain a share of the revenue that matches their contribution, thereby motivating all members to actively participate in demand response and jointly improve the operational efficiency and market competitiveness of virtual power plants.

[0045] S4: Predict the power generation of renewable energy based on conditional value at risk, obtain the accuracy factor based on the power generation, and optimize the demand response based on the accuracy factor and the benefit.

[0046] In S4, the expression for obtaining the accuracy factor based on the generated power is: In the formula, F represents new energy, is the accuracy factor of the new energy at time t, are the predicted renewable energy power generation power in the day-ahead stage and the intraday stage respectively, and μ and η are the parameters of the accuracy factor.

[0047] In this embodiment, if the accuracy of the forecast of renewable energy power generation is insufficient, the virtual power plant may face greater operational uncertainty. To deal with this situation, the virtual power plant adopts a strategy to share part of the risk caused by forecast errors by optimizing demand response, that is, adjusting the response mechanism of flexible loads. At the same time, in order to reflect the uncertainty of renewable energy power generation forecasts, the virtual power plant will also adjust the electricity price strategy of renewable energy to ensure that the electricity price can change dynamically according to the forecast accuracy of the day-ahead market.

[0048] The decision-making mechanism based on sharing and sharing ensures that all participants within the virtual power plant can share the benefits fairly according to their contributions, and reasonably share the risks brought about by the uncertainty of new energy, thereby achieving fair distribution of risks. At the same time, through the risk-sharing mechanism, the virtual power plant can better cope with the uncertainty of renewable energy and enhance the resilience and robustness of the power system in the face of various operational risks.

[0049] In order to evaluate the feasibility and superiority of this method, the following two scenarios are set up for comparative analysis. Scenario 1: No consideration of improved quasi-linear demand response; Scenario 2: Considering improved quasi-linear demand response and Shapley value benefit distribution considering risk attitude. Taking Shaoxing City, Zhejiang Province as the research object, it is assumed that a certain park in the area contains industrial loads, air-conditioning loads and 400 electric vehicles. The time-of-use prices of the upper power grid are shown in Table 1, and the first-order equivalent thermal parameters and the number of air conditioners of public buildings are shown in Table 2. The maximum regulation capacity of the virtual power plant is 35MW, the adjustable power of air-conditioning load and industrial load are both 5MW, the adjustable power of energy storage is 10.5MW, the adjustable power of electric vehicles is 4.5MW, the adjustable power of wind and solar is 10MW, and the incentive coefficient is set to 1.5. The day-ahead forecast and ultra-short-term forecast curves for industrial load, wind power and photovoltaic power are shown as follows: Figure 3 , 4 , as shown in Figure 5. Table 1. Time-of-use electricity price of the upper power grid (yuan / kWh) Time Electricity purchase price Electricity sales price 10:00-22:00 1 0.65 7:00-9:00、23:00-24:00 0.75 0.42 1:00-6:00 0.4 0.22 The load of industrial load and the predicted power generation of wind and solar power are as follows: Figure 2 As shown, the first-order equivalent thermal parameters and specific parameters of air conditioners for commercial buildings are shown in Table 2: Table 2. First-order equivalent thermal parameters and number of air conditioners for commercial buildings Mall No. C / (℃ / kW) R / (℃ / kW) Number of air conditioners 1 11 21.18 75 2 10.19 50.82 70 3 7.5 29.08 60 4 11.64 30.38 75 5 9.96 28.54 72 6 10.46 42.1 65 Based on the above simulation examples and parameter settings, simulation verification and analysis are carried out from the perspective of improving the quasi-linear two-stage demand response decision and sharing-sharing decision, and improving the quasi-linear day-ahead optimal demand response decision: In the day-ahead market optimization process, the virtual power plant uses the upcoming wind and solar power generation forecast data and the expected load guidance issued by the grid operator, with every 15 minutes as a decision cycle, striving to maximize the benefits, and accordingly formulates the optimal power generation plan. The relevant results are shown in Figure 6 At the same time, the virtual power plant also actively participated in the load-based demand response and compared the per-unit load curve when it did not participate in these demand response measures. The comparison results are shown in Figure 7 In the present invention, the incentive provided by the power grid company to users participating in the two types of demand response is consistent. Figure 6 It can be seen that air-conditioning load usually reaches its peak in the afternoon or evening on hot days, because users have an increased demand for cooling during these periods. The peak hours of industrial loads may be related to production plans, and usually reach their maximum during daytime working hours. The charging demand for electric vehicles increases at night, and the energy storage system charges when the electricity price is low and discharges when the electricity price is high according to the electricity price signal to maximize cost-effectiveness. VPP adjusts the load according to the load guidelines and electricity prices released by the power grid. During periods of high electricity prices, VPP may reduce load to reduce costs; when electricity prices are low, it will increase load to take advantage of low-cost electricity, and these are all scheduled under the premise of ensuring that the per-unit load curve fits the load guidelines as closely as possible. In addition, VPP manages the uncertainties and forecast errors associated with renewable energy, and reduces these risks by flexibly scheduling flexible loads. From Figure 7 Clear differences can be seen. The curves that participate in demand response show conscious load adjustment, reducing electricity consumption during peak hours through demand-side management, thereby achieving a smoother load distribution. This adjustment contributes to the stable operation of the power grid and reduces overall energy costs. In contrast, the curves that do not participate in demand response reflect the natural electricity consumption pattern of users, with obvious peaks and valleys, and lack of response to electricity price signals. This causes the power grid to be under greater pressure during peak hours, increasing energy consumption and costs.

[0050] Improved quasi-linear intraday optimal demand response decision: In the intraday scheduling stage, VPP takes 15 minutes as the time interval and 4 hours as the rolling cycle. The objective function is to minimize the sum of the adjustment amount penalties within the rolling scheduling cycle. The optimal scheduling of intraday flexible load, the adjustment amount of the day-ahead scheduling, and the day-ahead-intraday standard load curve are solved as follows: Figure 8 , 9 , as shown in 10. Figure 8 , Fig. 9 and Fig.10It can be seen that intraday rolling optimization corrects errors through adjustments based on day-ahead scheduling, responds more flexibly to real-time changes in demand and supply, and can reduce cost increases caused by inaccurate forecasts, thereby maximizing cost-effectiveness. Intraday load adjustments in the power market are usually intended to better adapt to natural load fluctuations and market signals. Air conditioning use increases with rising temperatures, industrial production may change according to planned adjustments, electric vehicle charging may surge at night, and energy storage systems charge and discharge flexibly according to demand. Improved quasi-linear demand response encourages users to reduce electricity consumption during peak hours. This results in the intraday per-unit load curve being more in line with the load quasi-line, so that the more demand response incentives there are, the higher the benefits.

[0051] Revenue sharing model: The degree of user participation in the virtual power plant is closely related to the proportion of revenue they receive. The Shapley value method considering risk preference is used to fairly distribute the benefits of flexible loads participating in demand response to stimulate users' enthusiasm for response. Fig.11 The profit distribution ratio shown reflects the optimization result of the distribution scheme. At the same time, it further quantifies the impact of different profit sharing strategies on user response behavior. This profit distribution mechanism based on fairness aims to increase the enthusiasm of users to participate in the operation of virtual power plants, thereby enhancing the flexibility and benefits of the entire system. Fig.11 It can be seen that considering the Shapley value profit distribution of risk attitude, both electric vehicle load and energy storage system increase, while air conditioning load and industrial load decrease. This is due to the adaptability of electric vehicles and energy storage systems to market fluctuations and their higher risk tolerance. They can flexibly respond to market supply and demand, and their active market participation strategies receive more rewards; while the high demand for stability and comfort and risk aversion of air conditioning load and industrial load lead to their less response to market demand, and the profit distribution follows a decline.

[0052] Fig.11 It is revealed that after adopting the Shapley value method that takes risk preference into consideration, the profit distribution of flexible loads in intraday demand response has been significantly improved, especially the growth of EV and industrial load profits, which provides a stronger incentive for users to bear intraday volatility risks, as shown in Table 3: Table 3. Flexible load responsiveness under different revenue ratios Proportion of income Air conditioning load EV Load Industrial load 0.01 0.835 0.856 0.812 0.1 0.814 0.836 0.765 0.3 0.775 0.816 0.526 0.5 0.557 0.789 0.381 0.7 0.365 0.584 0.135 0.9 0.256 0.356 0.091 The response enthusiasm of flexible loads decreases as the proportion of revenue increases. In the initial stage, virtual power plants attract users to participate in improved demand response by increasing incentives. As the number of users increases, virtual power plants optimize their own cost-effectiveness while maintaining user participation by reducing incentives.

[0053] In order to reveal the benefit comparison of different entities under the two modes, the benefits of virtual power plants and new energy in different scenarios are shown in Table 4: Table 4. Comparison of benefits of virtual power plants and new energy under different modes model Virtual power plant revenue / yuan New energy income / yuan 1 17216 65654 2 32158 66763 Under Mode 2, the revenue of the virtual power plant increased by 86.56%, and the revenue of new energy increased by 1.69%, indicating that the use of improved quasi-linear demand response and Shapley value profit distribution taking into account risk attitude is conducive to achieving a win-win situation for both parties.

[0054] Embodiment 2: This embodiment also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the steps of the virtual power plant two-stage demand response optimization method based on the adaptive control algorithm are implemented.

[0055] The specific implementation described above is a preferred implementation of the two-stage demand response optimization method of a virtual power plant based on an adaptive control algorithm of the present invention, but it is not intended to limit the specific implementation scope of the present invention. The scope of the present invention includes but is not limited to this specific implementation. All equivalent changes made in accordance with the shape and structure of the present invention are within the protection scope of the present invention.

Claims

1. A two-stage demand response optimization method for virtual power plants based on an adaptive control algorithm, characterized in that: The following steps are involved: S1: Obtain the dynamic response incentive of the demand response subject based on the similarity index between the demand response subject and the load criterion and the total response amount of the demand response subject; S2: In the day-ahead stage, based on the dynamic response incentive and the first constraint, an adaptive control algorithm is used to construct the day-ahead optimization objective function. In the intraday stage, based on the dynamic response incentive and the second constraint, an adaptive control algorithm is used to construct the intraday optimization objective function. An improved quasi-linear demand response model is constructed based on the day-ahead optimization objective function and the intraday optimization objective function. S3: Based on the improved quasi-linear demand response model, the Shapley value allocation method considering risk preference is adopted to obtain the benefits of the demand response subject; S4: Predict the power generation of renewable energy based on conditional value at risk, obtain the accuracy factor based on the power generation, and optimize the demand response based on the accuracy factor and the benefit.

2. The two-stage demand response optimization method for virtual power plants based on adaptive control algorithm according to claim 1 is characterized in that: In S1, the similarity index χ = 1-νe -λd , ν is the adjustment factor, e is the base of the natural logarithm, λ is the attenuation coefficient, and d is the Euclidean distance between the demand response subject and the load directrix.

3. The two-stage demand response optimization method for virtual power plants based on adaptive control algorithm according to claim 2 is characterized in that: In S1, the dynamic response excitation M award =δ b χP D , where δ b is the excitation coefficient, P D is the total response amount of the demand response entity.

4. The two-stage demand response optimization method for virtual power plants based on adaptive control algorithm according to claim 1 is characterized in that: In S2, the day-ahead optimization objective function is min C = C buy +C ES +C TL -M before -M sale +Q t , where C is the cost of the virtual power plant in the day-ahead stage, C buy is the electricity purchase cost of the virtual power plant in the day-ahead stage, C ES is the operation and maintenance cost of energy storage in the virtual power plant in the day-ahead phase, C TL The compensation cost for industrial users to transfer load in the day-ahead phase, M before is the dynamic response incentive of the demand response subject in the day-ahead stage, M sale is the revenue from electricity sales of the virtual power plant in the day-ahead phase, Q t It represents the economic losses that may be caused by the uncertainty of wind and solar power output of the virtual power plant at time t in the day-ahead phase.

5. The two-stage demand response optimization method for virtual power plants based on adaptive control algorithm according to claim 1 is characterized in that: In S2, the intraday optimization objective function is min B = ΔP AC K AC +ΔP EV K EV +ΔPb grid K grid +ΔPb W K W +ΔPb V K V +ΔP tr K tr +ΔP ES K ES , where B is the adjustment cost of the virtual power plant during the intraday stage, ΔP AC K is the total amount of adjustment of the air conditioner in the virtual power plant compared with the day-ahead dispatch plan, AC is the penalty adjustment coefficient of the air conditioner, ΔP EV is the total amount of electric vehicles in the virtual power plant adjusted compared to the day-ahead dispatch plan, K EV is the penalty adjustment coefficient of the electric vehicle, ΔPb grid K is the total amount of grid interaction power in the virtual power plant compared to the day-ahead dispatch plan, grid is the penalty adjustment coefficient of the grid interaction power, ΔPb W K is the total amount of wind power in the virtual power plant adjusted compared to the day-ahead dispatch plan, W is the penalty adjustment coefficient of wind power, ΔPb V is the total amount of adjustment of PV in the virtual power plant compared to the day-ahead dispatch plan, K V is the penalty adjustment coefficient of the photovoltaic power generation, ΔP tr K is the total amount of industrial transfer load in the virtual power plant compared to the day-ahead dispatch plan, tr is the penalty adjustment coefficient for the industrial load transfer, ΔP ES K is the total amount of energy storage in the virtual power plant compared to the day-ahead dispatch plan, ES is the penalty adjustment coefficient for the energy storage.

6. According to the virtual power plant two-stage demand response optimization method based on adaptive control algorithm according to claim 1, S3 comprises: S31: Based on the improved quasi-linear demand response model, the optimal target value is obtained using the historical data of the virtual power plant; S32: Obtain the risk preference corresponding to the demand response subject based on the state variable of the demand response subject, and obtain the benefit of the demand response subject based on the optimal target value and the risk preference corresponding to the demand response subject.

7. The virtual power plant two-stage demand response optimization method based on adaptive control algorithm according to claim 6 is characterized in that: In S32, the expression for obtaining the risk preference corresponding to the demand response subject based on the state variable of the demand response subject is: In the formula, ψ z is the fair revenue distribution that participant z deserves in the demand response subject, S represents the number of participants in the subset S of the demand response subject, n is the total number of flexible loads aggregated by the virtual power plant, l is the risk adjustment coefficient based on the improved load criterion, M Z represents the risk-weighted marginal contribution of the participant z.

8. The virtual power plant two-stage demand response optimization method based on adaptive control algorithm according to claim 6 is characterized in that: In S32, the expression for obtaining the revenue of the demand response subject based on the optimal target value and the risk preference corresponding to the demand response subject is R x =(1-φ x )R, where R x is the benefit of participant x in the demand response entity, R represents the optimal target value, φ x is the profit proportion of the participant x in the virtual power plant.

9. The virtual power plant two-stage demand response optimization method based on adaptive control algorithm according to claim 1 is characterized in that: In S4, the expression for obtaining the accuracy factor based on the generated power is: In the formula, F represents new energy, is the accuracy factor of the new energy at time t, are the predicted renewable energy power generation power in the day-ahead stage and the intraday stage respectively, and μ and η are the parameters of the accuracy factor.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the two-stage demand response optimization method of a virtual power plant based on an adaptive control algorithm as described in any one of claims 1 to 9 are implemented.

Images