Income distribution method for internal alliance cooperative game of virtual power plant

Abstract

The invention provides a revenue distribution method for an internal alliance cooperative game of a virtual power plant, and the method comprises the steps: forming an alliance by distributed energy resources, and constructing a multi-agent optimization scheduling model in the virtual power plant with the maximized electricity selling revenue; constructing constraint conditions of a multi-body optimization scheduling model in the virtual power plant; determining a multi-agent cooperative game model of the game participants, the strategy set and the feature function; determining the overall electricity selling income of the virtual power plant, and designing a two-stage refined income distribution scheme of a multi-factor improved shapley value method and a kernel method based on odd-even circulation; and calling a CLPEX solver of the YALMIP on a Matlab platform to solve the cooperative game scheduling model and the income distribution strategy. According to the method, cooperative complementary operation among the alliances in the virtual power plant is realized, and fairness and rationality of income distribution among multiple subjects are ensured.

Description

Technical Field

[0001] The present invention belongs to the technical field of distributed energy optimization operation, and specifically relates to a profit distribution method for internal alliance cooperation game of a virtual power plant. Background Art

[0002] Traditional fossil energy is gradually withdrawing from the operation of the power system. In the future, a large number of distributed energy resources (DERs) will participate in power market transactions. However, decentralized distributed energy has problems such as small capacity, large quantity, and unstable output. It is impossible to trade electricity independently, resulting in a waste of resources. As a unified agent for distributed energy to participate in the power market, the virtual power plant (VPP) aggregates distributed energy flexibly and efficiently through cooperative and complementary means to ensure the maximization of overall benefits. With the increase in the number of aggregated DERs, the balance of benefits among members in the VPP affects their enthusiasm for cooperation. Therefore, establishing a fair and reasonable benefit distribution mechanism is of great significance to the long-term development of virtual power plants. Summary of the invention

[0003] The technical problem to be solved by the present invention is to provide a profit distribution method for the cooperative game of alliances within a virtual power plant, which realizes the cooperative and complementary operation among alliances within the virtual power plant and ensures the fairness and rationality of profit distribution among multiple entities.

[0004] In order to solve the above technical problems, the technical solution adopted by the present invention is: a profit distribution method of the internal alliance cooperation game of a virtual power plant, comprising the following steps: S1. Distributed energy forms an alliance to build a multi-agent optimization scheduling model within a virtual power plant to maximize electricity sales revenue; S2. Construct the constraints of the multi-agent optimization scheduling model in the virtual power plant; S3. Determine the multi-agent cooperative game model of game participants, strategy sets and characteristic functions; S4. Determine the overall electricity sales revenue of the virtual power plant and design a two-stage detailed revenue distribution plan based on the multi-factor improved Shapley value method and the odd-even cycle-based kernel method; S5. The cooperative game scheduling model and revenue distribution strategy are solved by calling the CLPEX solver of YALMIP on the Matlab platform.

[0005] The invention provides a method for distributing profits of a virtual power plant internal alliance cooperation game, which has the following beneficial effects: 1. Virtual power plants coordinate and aggregate wind farms, photovoltaic power stations, pumped storage power stations, and demand users through advanced control, communication, and metering technologies, and establish a multi-agent optimization scheduling model with the goal of maximizing benefits. This scheduling model can make full use of decentralized and small-capacity energy individuals, promote the rational consumption of wind and solar resources, and achieve a balance between supply and demand in the electricity market.

[0006] 2. By constructing different constraints of the dispatch model, these constraints mainly include: operation constraints of wind and solar power output and demand response, pumped hydropower generation balance constraints, and power supply and demand balance constraints. The virtual power plant can be in the best operating state. These constraints can not only ensure mathematical feasibility, but also make the dispatch model meet the requirements of actual operation.

[0007] 3. In the multi-alliance cooperative game model of wind power, photovoltaic power, pumped storage, and demand response, when the wind-solar alliance generates excess power, a small amount of electricity will be transferred to the pumped storage alliance for storage. When the wind-solar alliance is under-powered, the pumped storage and demand response alliances will adjust the power generation. This will achieve the coordination and complementarity of multiple resources, greatly improve the overall power generation benefits of each alliance, and provide the power market with high-quality electricity with high cost performance.

[0008] 4. The improved Shapley value method considering risk factors, profit factors and cost factors comprehensively distributes the benefits among wind power, photovoltaic, pumped storage and demand response alliances, ensuring the fairness and complementarity of the operation of each alliance. The nucleolus method based on odd-even cycles refines the power generation benefits among alliance members, making the benefits of each alliance member more balanced, which helps to enhance the enthusiasm of members to participate in the operation of the alliance.

[0009] 5. CPLEX solver is a leading mathematical programming optimization software developed by IBM. It can complete complex problem preprocessing, efficient restart, sensitivity analysis and infeasibility finder, etc. The CPLEX solver can solve the cooperative game scheduling model and profit distribution strategy in a shorter time and with stronger robustness, thus achieving a better solution and improving scheduling efficiency and the overall performance of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] The present invention will be further described below in conjunction with the accompanying drawings and embodiments: Figure 1 This is a diagram of the operation framework of multiple DERs in a virtual power plant; Figure 2 It is a schematic diagram of the solution process of the present invention; Figure 3 The power load demand diagram of the present invention; Figure 4 It is a typical day wind and solar output curve diagram of the present invention; Figure 5The output diagrams of each alliance under two different operation modes of the present invention are: (a) the output result of each alliance operating independently, and (b) the output result of each alliance cooperating and complementing each other; Figure 6 A comparison chart of different profit distribution methods among alliances of the present invention; Figure 7 This is a comparison chart of the variance of the overall satisfaction of each alliance of the present invention. DETAILED DESCRIPTION

[0011] In order to make the purpose, technical solution 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 embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0012] Combination Figure 1 to Figure 7 The specific embodiments of the present invention are described in further detail.

[0013] The present invention discloses a method for distributing the revenue of a cooperative game of a distributed energy alliance within a virtual power plant. First, a mechanism is proposed for multiple types of DERs to form an alliance and aggregate in a VPP to participate in power market transactions. Then, a cooperative game scheduling model for VPP multi-alliance is constructed with the goal of maximizing the revenue from electricity sales. At the same time, in order to ensure the stability of the aggregated alliance within the VPP and the fairness of the revenue, a multi-factor improved Shapley value method and a two-stage refined revenue distribution scheme based on the odd-even cycle nucleolus method are also designed. Finally, the cooperative game scheduling model and the revenue distribution strategy are solved by calling the CLPEX solver of YALMIP on the Matlab2018b platform.

[0014] Specifically, Figure 2 As shown, the following steps are included: S1. Distributed energy resources of the same type form an alliance and build a multi-alliance optimization scheduling model within a virtual power plant to maximize electricity sales revenue.

[0015] The total revenue of the virtual power plant in one cycle is , which includes the revenue from wind, solar, and pumped storage electricity sales, and the net revenue from demand response users when shaving peaks and filling valleys. The specific calculation is as follows: ; (1) In the formula, is the net income from the sale of electricity from the wind farm, is the net income from electricity sales of the photovoltaic power plant, is the net income from the pumped storage power station selling electricity to the VPP, It is the net benefit of load changes when demand response users are shaving peak loads and filling valley loads.

[0016] (1) Wind farm entity benefit model ; (2) In the formula for t The active power delivered by the wind farm to the load in the virtual power plant during the period, The electricity price provided to the virtual power plant, is the wind abandonment penalty coefficient, They are t The predicted output power and actual power generation of wind turbines in each period, is the wind turbine operation and maintenance cost coefficient, is the total capacity of the wind turbine.

[0017] (2) Photovoltaic power plant main body profit model ; (3) In the formula for t The active power delivered by the photovoltaic power station to the load in the virtual power plant during the period, is the light abandonment penalty coefficient, for t The predicted output power and actual power generation of the photovoltaic power field during the period, is the photovoltaic unit operation and maintenance cost coefficient, is the total capacity of the photovoltaic field.

[0018] (3) Pumped-storage power station main benefit model ; (4) In the formula They are t The power generation and pumping power of the pumped storage power station during the period, are the comprehensive cost coefficients of auxiliary services paid by wind farms and photovoltaic power stations to pumped storage power stations, for t The pumped electricity price of the period pumped storage power station, is the operation and maintenance cost coefficient of the pumped storage power station, is the total capacity of the pumped storage power station; (4) Demand response user subject benefit model ; (5) In the formula, are the price incentive coefficients for load reduction and electricity increase in demand response, Demand response users u The load increases and decreases, Demand response users u The increase and decrease of the load cost coefficient, is the cost coefficient of the standby load, Demand response users utotal capacity.

[0019] S2. Constraints for constructing the multi-agent optimization scheduling model within the virtual power plant, including: 1) Power balance constraints ; (6) In the formula is the load demand of the virtual power plant within a scheduling cycle.

[0020] 2) Wind power operation constraints ; (7) ; (8) ; (9) In the formula They are t The actual power generation of the wind farm at the moment and the power generation transmitted to the pumped storage power station, is the maximum value of wind power prediction.

[0021] 3) Photovoltaic output constraints ; (10) ; (11) ; (12) In the formula They are t The actual power generation of the photovoltaic power station at the moment and the power generation transmitted to the pumped storage power station, is the maximum value predicted for PV power.

[0022] 4) Pumped storage constraints ; (13) ; (14) ; (15) ; (16) ; (17) ; (18) In the formula for t The pumping power of the pumped-storage power station when it operates independently at any time. are the upper and lower limits of pumped storage power generation, are the upper and lower limits of pumping power respectively. are 0-1 variables for the start and stop states of the units under the working conditions (when the pumped storage power station is pumping water for power generation, Take 1, Take 0, when the pumped storage power station is performing pumped storage (energy storage), Take 1, Take 0), , They are t Moment and t- The reservoir capacity of the pumped storage power station at time 1, are the operating efficiencies for power generation and water pumping, are the minimum storage capacity and the maximum storage capacity respectively.

[0023] 5) Demand response constraints ; (19) ; (20) ;(twenty one) ;(twenty two) ;(twenty three) In the formula, Demand response users u The maximum input and output power, They are t The load transfer state variables at time, Demand response users u The start and end time of the transferable load power input, Demand response users u The start and end time of the transferable load power transfer.

[0024] S3. Determine the multi-agent cooperative game model of game participants, strategy sets and characteristic functions.

[0025] The specific content of the multi-agent cooperative game model is: (1) Game Participants The game model is composed of VPP internal sub-alliance members N 1 wind farm, N 2 photovoltaic power stations, N 3 pumped storage power stations, N Four demand response users act as participants in a four-party game, using express.

[0026] (2) Participant Strategies When VPP conducts economic transactions in the electricity market, the output power of each alliance in each period include The transferable load and interruptible load of the demand response user are used as game strategies. The strategy space is expressed as ;(twenty four) In the formula Demand response users i The maximum transfer-in and transfer-out power of VPP is , where the strategy space of each sub-alliance of VPP should satisfy the constraint conditions of the corresponding variables (8)-(23).

[0027] (3) Characteristic function The characteristic function of the game is the objective function of the profit model of each subject in this paper. .

[0028] Whether the participants have reached a cooperative alliance: whether the cooperative alliance meets the superadditivity condition, that is, the benefits generated by the cooperative alliance should be greater than the sum of the benefits of each participant in the alliance under the non-cooperative model. The feasibility of cooperation is ; (25) In the formula, is the profit of the virtual power plant when the sub-alliance does not cooperate, In order to generate benefits for members within the sub-alliance, Virtual Power Plant Alliance N The total revenue is N Includes all possible alliance states that can be formed by all DER entities in the VPP.

[0029] S4. Determine the overall electricity sales revenue of the virtual power plant and design a two-stage detailed revenue distribution plan based on the multi-factor improved Shapley value method and the nucleolus method based on odd-even cycles.

[0030] The profit distribution plan of each alliance based on the multi-factor improved Shapley value method is based on the theory of Shapley value method, taking into account the risk factors of different participants and their different contributions to the comprehensive profit and cost of VPP. Based on the above situation, a correction factor is introduced into the original profit distribution model to optimize the large alliance N The profit distribution model below.

[0031] (1) Comprehensive risk preference For the unregulated wind and solar alliance, the prediction error directly affects the overall income change of the alliance. Therefore, the overall power prediction accuracy is selected as the risk function and defined as the risk-averse utility. The details are as follows: ; (26) ; (27) In the formula, and For wind power t The predicted and actual values ​​at the time, is the total number of uncontrollable alliances, is the risk-averse exponential utility function, is the base of the natural logarithm, approximately 2.71828.

[0032] For adjustable alliance pumped storage and demand response, the uncontrollable wind and solar output errors can be adjusted and more benefits can be obtained during peak load regulation. Therefore, a risk preference utility function is selected to quantify the risk. The formula is as follows: ; (28) ; (29) In the formula The adjustable maximum capacity of the controllable alliance, is the total number of controllable alliances, is the risk preference exponential utility function The risk preference factor is: ; (30) (2) Contribution of comprehensive profit When distributing alliance profits, we need to consider not only the marginal profit in the Shapley value method, but also the profit contribution of the sub-alliance to the grand alliance. N The comprehensive profit contribution of various cooperation modes with sub-alliances S The ratio of power generation revenue before and after joining to the incremental revenue of the Grand Alliance as a whole is calculated as follows: ; (31) In the formula, Indicates sub-alliance S Profits before joining the alliance, For the Alliance N Remove sub-alliance S The subsequent revenue gains from cooperative operation; (3) Comprehensive cost contribution The wind and solar operation and maintenance costs and the comprehensive cost of pumped storage ancillary services will also change before and after the alliances participate in the cooperation, so the comprehensive cost contribution is included in the profit distribution model. The comprehensive cost contribution is defined as the ratio of the cost change under the cooperative operation of each sub-alliance to the total cost change of the alliance, that is, ; (32) ; (33) In the formula For sub-alliance S The cost required when not joining the cooperative alliance, is the total cost of each sub-alliance when operating in cooperation, It is the sum of the total cooperative operation cost of each alliance of wind farms, photovoltaic power stations, pumped storage power stations, and demand response users, including operation and maintenance costs, loss costs, and ancillary service fees; 4) After comprehensively considering risk preference, profit contribution and comprehensive cost, the entropy weight method is used to assign a weighted value to each correction factor and define the profit distribution correction coefficient as follows: ; (34) In the formula, They are the risk preference factor, comprehensive profit contribution factor and comprehensive cost contribution factor of the cooperative sub-alliance. is the weighting factor, and , after improvement, the difference between the old and new weights of the sub-alliance is obtained for: ; (35) in, n is the total number of sub-alliances participating in the allocation in the cooperative alliance, and the difference in weights satisfies , and further obtain the initial allocation result based on the Shapley value method under the grand coalition N And the improved distribution results , the specific formula is as follows: ; (36) ; (37) In the formula, For the major leagues N Neutron Alliance S The preliminary allocation results based on the Shapley value method are: N.S. The Grand Alliance for Collaborative Operations N and sub-alliance S The number of subjects, It is the profit distribution result after the Shapley value method based on multiple factors improvement; In the theoretical basis of the nucleolus method, the vector It is expressed as the overall satisfaction of the sub-alliance, and the core solution is the solution with the highest overall satisfaction. The overall satisfaction is measured in lexicographic order. By improving the overall satisfaction, the benefits among alliance members can be made more fair and reasonable. Therefore, the improved core method based on odd-even cycle is used to refine the benefits of each alliance member.

[0033] The specific steps are: in the odd linear programming, the goal is to minimize the maximum value in the overall satisfaction vector, and in the even linear programming, the goal is to maximize the minimum value in the overall satisfaction vector, and alternately solve the problem until a unique solution is found. The alliance equilibrium solution with a higher overall satisfaction than the core solution is calculated, that is, the profit distribution plan at this time is the optimal solution. The specific model is as follows: No. n An odd linear programming model LP with maximization as the goal n -Odd ; (38) In the formula The optimal solution with the minimum overall satisfaction for solving an odd linear cycle, For the n -1 odd linear cycle i The second lowest overall satisfaction, G The set is 2 n Sub-alliance, collection P To be included in the alliance G Sub-alliances within For model LP n-1 -Even solves the problem of maximizing overall satisfaction n -1 minimum value; When LP n -Odd When a unique solution is found, the solution obtained is the alliance equilibrium solution; otherwise, continue to solve the linear programming model LP n -Even minimizes the maximum value in the overall satisfaction vector .

[0034] No. n An even-numbered linear programming model LP with minimization as the objective n -Even ; (39) In the formula, The optimal solution for maximizing overall satisfaction for solving even linear cycles, The solution of the odd linear cycle minimizes the overall satisfaction n Second maximum value.

[0035] For the above models, all of them satisfy , indicating that the individuals participating in the cooperation are rational. Used to constrain the satisfaction values ​​of all sub-alliances to be less than the model LP n-1 -Odd solved This ensures that when finding the min function, there will be no n- 1 small satisfaction Smaller satisfaction value. Used to constrain the satisfaction values ​​of all sub-alliances to be greater than the model LP n-1 -Even solved This ensures that when calculating the max function, there will be no n The second highest satisfaction Greater satisfaction value. Indicates i Linear Programming Model LP i -Odd's solution, the satisfaction of all sub-alliances is obtained The alliance collection at that time. Indicates i Linear Programming Model LP i -Even’s solution, the satisfaction of all sub-alliances is obtained The alliance collection at that time.

[0036] S5, the cooperative game scheduling model and the revenue distribution strategy are solved by calling the ClPEX solver of YALMIP on the Matlab2018b platform, which specifically includes the following steps: 1) Input the basic data of wind farms, photovoltaic power stations, pumped storage power stations, and demand response users, and import the data into the multi-alliance cooperative game scheduling model within the virtual power plant.

[0037] 2) Given the initial value of the cooperative game equilibrium solution, randomly select a strategy set ,in They represent the initial power values ​​of the wind power alliance, photovoltaic alliance and pumped storage alliance cooperation games respectively, They are respectively represented as the initial values ​​of the total user load reduction and increase in the demand response alliance. Then, the CPLEX solver is used to solve the optimal strategy equilibrium point of the cooperative game.

[0038] 3) Determine whether the strategy sets of each alliance have reached cooperative equilibrium. If "yes", the electricity sales revenue of each alliance will be calculated; if "no", the cooperative game strategy will be iteratively updated again.

[0039] 4) The Shapley value method is improved by introducing risk, profit and cost factors to calculate the benefits of each alliance. The benefits of alliance members are calculated using the improved nucleolus method with odd-even cycles.

[0040] Example The case object is Figure 1 The virtual power plant multi-DERs operation framework shown in the figure. Specific operation mechanism: In the day-ahead power market transaction cycle, the VPP dispatch center announces the power demand in advance, and the VPP makes plans based on the predicted output and reports the purchase and sale prices to the information center. Multiple wind farms, photovoltaic power stations, pumped storage power stations and demand response users form a multi-type DERs alliance aggregated in the virtual power plant. The multi-type DERs alliance conducts cooperative games and multi-source power generation based on the output plan, and conducts transactions in the power market through the VPP. The capacity of wind farm 1 is set to 75MW, wind farm 2 is set to 80MW, the capacities of photovoltaic power station 1 and photovoltaic power station 2 are both 65MW, and the capacity of photovoltaic power station 3 is 80MW. The typical daily output curve of wind and solar units is as follows: Figure 4As shown. The capacity of pumped storage power station 1 is 80MW, the capacity of pumped storage power station 2 is 70MW, the capacity of demand response 1 is 45MW, and the capacity of demand response 2 is 55MW. The incentive compensation for demand response users to increase load and reduce load is 160 yuan / MWh and 320 yuan / MWh respectively. Assume that on a certain trading day, the virtual power plant participates in the 24-hour time-of-use electricity price of the day-ahead power market as shown in Table 1 and the power load demand of the system is as follows Figure 3 shown.

[0041]

[0042] In the example, wind power, photovoltaic power, pumped storage, and demand response in VPP are set to operate independently as mode 1, and entities of the same type form an alliance to cooperate and complement each other in power generation as mode 2. The comparison of the total power generation revenue results in one dispatch cycle under the two operating modes is shown in Table 2.

[0043]

[0044] As can be seen from Table 2, the same type of entities within the VPP form an alliance for cooperative and complementary power generation, and the total income of the alliance and the income of each small alliance have been significantly improved. Under the cooperative and complementary mode, the wind power alliance and the photovoltaic alliance store a large amount of electricity in the pumped storage alliance to fill the power generation during peak and valley periods. As a result, their respective income increased by 19,340 yuan and 9,540 yuan. The pumped storage alliance and the demand response alliance have the effect of energy storage when the power demand is shaving peaks and filling valleys. They produce power during peak loads and transfer energy storage and load during valleys, thereby increasing their respective income by 87,100 yuan and 95,400 yuan.

[0045] The output results of wind-solar pumped storage independent operation and cooperative game are as follows Figure 5 As shown in the figure, in the independent operation mode, each alliance fully generates electricity to maximize its own interests. Among them, the pumped storage alliance did not cooperate and complement with the wind power and photovoltaic alliances, and the overall output showed a significant increase, which led to a large amount of wind and solar power abandonment in the system. This operation mode does not have economic and environmental benefits. In the cooperative and complementary mode, the wind power alliance's output decreased during the peak period of the system, and the "anti-peak" phenomenon occurred. However, at this time, the photovoltaic alliance fully generated electricity, and the excess electricity of the wind and solar alliance was stored in the pumped storage alliance for pumped storage, which further improved the utilization rate of renewable energy. The demand response alliance takes into account the user's electricity usage habits to maximize load transfer and reduction. It can not only guarantee the power demand of load users, but also improve its own benefits. In normal and valley periods, the pumped power generation of the pumped storage alliance and the load transfer of demand response effectively smoothed the intermittent output of the wind and solar alliance, and realized the optimized operation of electric energy complementarity.

[0046] According to the operation results of the two modes and the benefits of each alliance under various combinations, three benefit distribution methods are considered for comparison: direct distribution according to output, traditional Shapley value distribution, and improved Shapley distribution. Figure 6 shown.

[0047] The fairness and rationality of the three allocation results are analyzed in detail, and the results are as follows: (1) For the wind power alliance and the photovoltaic alliance, the weight of the comprehensive risk factor of the volatility and intermittency of their output gradually increased, resulting in a decrease in the improved allocation results compared with the direct allocation and Shapley value allocation results. However, as the core power generation members participating in the day-ahead power trading, their profit contribution and cost expenditure for the entire alliance are greater than those of other alliances, thus improving the overall benefits. In the end, the allocation results for the wind power alliance and the photovoltaic alliance were RMB 125,330 and RMB 99,570 respectively.

[0048] (2) The allocation result of the pumped storage alliance in the improved Shapley value is 114,390 yuan, which is 10,370 yuan more than the Shapley value allocation. This is because under the cooperative and complementary model, the pumped storage alliance is a risk-preference adjustable member and gains benefits in the process of dealing with risks. At the same time, a large amount of electricity from the wind power and photovoltaic alliance is transferred to the pumped storage alliance for pumped storage, and the profit growth rate during peak and valley periods increases accordingly. Although the risk fluctuation of the demand response alliance is relatively small, the demand side needs to consider the randomness of user electricity satisfaction and electricity consumption behavior, resulting in uncertainty in both comprehensive profits and cost expenditures. Therefore, the improved allocation result of the demand response alliance is 14,870 yuan less than the Shapley value allocation result, and the final allocation result of the demand response alliance is 94,560 yuan.

[0049] By comparing the income distribution results between alliance members before and after the improvement of the core method, the fairness and rationality of income distribution among members are analyzed, as shown in Table 3. The overall satisfaction variance analysis of each alliance is used to analyze the degree of balance of income distribution results, as shown in Table 3. Figure 7 shown.

[0050]

[0051] From the distribution results in Table 3, we can see that the distribution results obtained by the improved kernel method are more balanced than those of the kernel method. In order to further prove the advantages of the profit distribution scheme of the improved kernel method, the variance of the overall satisfaction of each alliance is solved. The smaller the variance, the more stable the distribution of satisfaction of each sub-alliance. Figure 7As shown in the figure, the variance of the overall satisfaction of each alliance before and after the improvement was reduced by 0.039, 0.021, 0.044 and 0.031, respectively, proving that the equilibrium degree of the improved nucleolus solution is higher than that of the nucleolus solution, that is, the fairness of the improved nucleolus solution is higher than that of the nucleolus solution. The improved nucleolus method based on odd-even cycles provides a better income distribution scheme for the cooperative game of alliance members, which not only ensures fairness and rationality among members, but also enhances the enthusiasm and stability of members to participate in the alliance.

Claims

1. A method for distributing profits in a virtual power plant internal alliance cooperation game, characterized in that: The following steps are involved: S1. Distributed energy forms an alliance to build a multi-agent optimization scheduling model within a virtual power plant to maximize electricity sales revenue; S2. Construct the constraints of the multi-agent optimization scheduling model in the virtual power plant; S3. Determine the multi-agent cooperative game model of game participants, strategy sets and characteristic functions; S4. Determine the overall electricity sales revenue of the virtual power plant and design a two-stage detailed revenue distribution plan based on the multi-factor improved Shapley value method and the odd-even cycle-based kernel method; S5. The cooperative game scheduling model and revenue distribution strategy are solved by calling the CLPEX solver of YALMIP on the Matlab platform.

2. The method for distributing profits of a virtual power plant internal alliance cooperation game according to claim 1 is characterized in that: In step S1, multiple wind farms, photovoltaic power stations, pumped storage power stations and demand response users form a multi-type distributed energy alliance and aggregate them in a virtual power plant.

3. The profit distribution method of the internal alliance cooperation game of a virtual power plant according to claim 1 is characterized in that: In step S1, a multi-agent optimization scheduling model within a virtual power plant is constructed to maximize the electricity sales revenue, including: The goal of constructing a virtual power plant is to maximize the economic benefits within a 24-hour dispatch cycle, and to maximize the electricity sales revenue of each entity in the operation of the virtual power field, including the sales revenue of wind, solar, and pumped storage, and the net revenue of demand response users in peak shaving and valley filling. The calculation formula is: ；（1） In the formula, is the net income from the sale of electricity from the wind farm, is the net income from electricity sales of the photovoltaic power plant, is the net income from the pumped storage power station selling electricity to the VPP, The net benefit of load changes when demand response users are shaving peaks and filling valleys; The main profit model of wind farm is: ；（2） In the formula, for t The active power delivered by the wind farm to the load in the virtual power plant during the period, The electricity price provided to the virtual power plant, is the wind abandonment penalty coefficient, They are t The predicted output power and actual power generation of wind turbines in each period, is the wind turbine operation and maintenance cost coefficient, is the total capacity of the wind turbines; The main benefit model of photovoltaic power field is: ；（3） In the formula, for t The active power delivered by the photovoltaic power station to the load in the virtual power plant during the period, is the light abandonment penalty coefficient, for t The predicted output power and actual power generation of the photovoltaic power field during the period, is the photovoltaic unit operation and maintenance cost coefficient, is the total capacity of the photovoltaic field; The main revenue model of the pumped storage power station is: ；（4） In the formula, They are t The power generation and pumping power of the pumped storage power station during the period, are the comprehensive cost coefficients of auxiliary services paid by wind farms and photovoltaic power stations to pumped storage power stations, for t The pumped electricity price of the period pumped storage power station, is the operation and maintenance cost coefficient of the pumped storage power station, is the total capacity of the pumped storage power station; The main revenue model of demand response users is: ；（5） In the formula, are the price incentive coefficients for load reduction and electricity increase in demand response, Demand response users u The load increases and decreases, Demand response users u The increase and decrease of the load cost coefficient, is the cost coefficient of the standby load, Demand response users u total capacity.

4. The method for distributing profits of a virtual power plant internal alliance cooperation game according to claim 1 is characterized in that: In step S2, the constraints of the multi-agent optimization scheduling model within the virtual power plant include: 1) Power balance constraints: ；（6） In the formula, is the load demand of the virtual power plant within a dispatch cycle; 2) Wind power operation constraints: ；（7） ；（8） ；（9） In the formula, They are t The actual power generation of the wind farm at the moment and the power generation transmitted to the pumped storage power station, is the maximum value of wind power prediction; 3) Photovoltaic output constraints: ；（10） ；（11） ；（12） In the formula, They are t The actual power generation of the photovoltaic power station at the moment and the power generation transmitted to the pumped storage power station, is the maximum value predicted for photovoltaic power; 4) Pumped storage constraints: ；（13） ；（14） ；（15） ；（16） ；（17） ；（18） In the formula, for t The pumping power of the pumped storage power station when it is operating independently at any time. are the upper and lower limits of pumped storage power generation, are the upper and lower limits of the pumping power, are 0-1 variables of the start and stop states of the units under working conditions. When the pumped storage power station is pumping water for power generation, Take 1, Take 0, when the pumped storage power station is performing pumped storage, Take 1, Take 0, , They are t Moment and t- The reservoir capacity of the pumped storage power station at time 1, are the operating efficiencies for power generation and water pumping, are the minimum storage capacity and the maximum storage capacity respectively; 5) Demand response constraints: ；（19） ；（20） ；（21） ；（22） ；（23） In the formula, Demand response users u The maximum input and output power, They are t The load transfer state variables at time, Demand response users u The start and end time of the transferable load power input, Demand response users u The start and end time of the transferable load power transfer.

5. The method for distributing profits of a virtual power plant internal alliance cooperation game according to claim 1 is characterized in that: In step S3, the game participants, strategy sets and characteristic functions of the multi-agent cooperative game model are: 1) Game participants The game model is composed of members of the virtual power plant internal sub-alliance N 1 wind farm, N 2 photovoltaic power stations, N 3 pumped storage power stations, N Four demand response users act as participants in a four-way game, using express; 2) Participant Strategies When virtual power plants conduct economic transactions in the electricity market, the output power of each alliance in each period include The transferable load and interruptible load of the demand response user are used as game strategies, and the strategy space is expressed as: ；（24） 3) Characteristic function The characteristic function of the game is the objective function of the profit model of each subject ; Conditions for participants to form a cooperative alliance: Whether the cooperative alliance meets the superadditivity condition, that is, the benefits generated by the cooperative alliance should be greater than the sum of the benefits of each participant in the alliance under the non-cooperative mode, expressed as: ；（25） In the formula, is the profit of the virtual power plant when the sub-alliance does not cooperate, In order to generate benefits for members within the sub-alliance, Virtual Power Plant Alliance N total revenue.

6. The method for distributing profits of a virtual power plant internal alliance cooperation game according to claim 1 is characterized in that: In step S4, the profit distribution plan of each alliance of the multi-factor improved Shapley value method is to consider the comprehensive risk preference of different participants and their different contributions to the comprehensive profit and comprehensive cost of the virtual power plant on the theoretical basis of the Shapley value method, introduce correction factors into the original profit distribution model based on the above situation, and optimize the virtual power plant alliance. N The profit distribution model below is operated as follows: 1) Consider the overall risk preference The overall power prediction accuracy is selected as the risk function and defined as the risk-averse utility, expressed as: ；（26） ；（27） In the formula, and For wind power t The predicted and actual values ​​at the time, is the total number of uncontrollable alliances, is the risk-averse exponential utility function, is the base of natural logarithms; The risk preference utility function is selected to quantify the risk, and the expression is: ；（28） ；（29） In the formula, The adjustable maximum capacity of the controllable alliance, is the total number of controllable alliances, is the risk preference exponential utility function; The risk preference factor is: ；（30） 2) Consider the contribution of comprehensive profit Defining the Major League N The comprehensive profit contribution of various cooperation modes with sub-alliances S The ratio of power generation revenue before and after joining to the incremental revenue of the Grand Alliance as a whole is given by the formula: ；（31） In the formula, Indicates sub-alliance S The income before joining the alliance, For the Alliance N Remove sub-alliance S The subsequent revenue gains from cooperative operation; 3) Consider the overall cost contribution The comprehensive cost contribution is defined as the ratio of the cost change of each sub-alliance under cooperative operation to the total cost change of the alliance, and the expression is: ；（32） ； （33） In the formula, For sub-alliance S The cost required when not joining the cooperative alliance, is the total cost of each sub-alliance when operating in cooperation, It is the total cost of cooperative operation of wind farms, photovoltaic power stations, pumped storage power stations, and demand response users, including operation and maintenance costs, loss costs, and ancillary service fees; 4) After comprehensively considering risk preference, profit contribution and comprehensive cost, the entropy weight method is used to assign a weighted value to each correction factor and define the profit distribution correction coefficient as follows: ； （34） In the formula, They are the risk preference factor, comprehensive profit contribution factor and comprehensive cost contribution factor of the cooperative sub-alliance. is the weighting factor, and , after improvement, the difference between the old and new weights of the sub-alliance is obtained for: ； （35） In the formula, n is the total number of sub-alliances participating in the allocation in the cooperative alliance, and the difference in weights satisfies , and further obtain the initial allocation result based on the Shapley value method under the grand coalition N And the improved distribution results , the specific formula is as follows: ；（36） ；（37） In the formula, For the major leagues N Neutron Alliance S The preliminary allocation results based on the Shapley value method are: N.S. The Grand Alliance for Collaborative Operations N and sub-alliance S The number of subjects, This is the profit distribution result based on the multi-factor improved Shapley value method.

7. The method for distributing profits of a virtual power plant internal alliance cooperation game according to claim 1 is characterized in that: In step S4, the profit distribution plan among the alliance members is refined based on the odd-even cycle kernel method, including: When calculating the optimal nucleolus solution, the vector It is expressed as the overall satisfaction of the sub-alliance, the core solution is the solution with the highest overall satisfaction, and the standard for measuring the overall satisfaction is the lexicographic order; In the odd-numbered linear programming, the goal is to minimize the maximum value in the overall satisfaction vector, and in the even-numbered linear programming, the goal is to maximize the minimum value in the overall satisfaction vector. Alternate and cyclic solutions are performed until a unique solution is obtained. The alliance equilibrium solution with a higher overall satisfaction than the nucleus solution is calculated, that is, the profit distribution plan at this time is the optimal solution. The specific model is as follows: No. n An odd linear programming model LP with maximization as the goal n -Odd: ；（38） In the formula, The optimal solution with the minimum overall satisfaction for solving an odd linear cycle, For the n -1 odd linear cycle i The second lowest overall satisfaction, G The set is 2 n Sub-alliance, collection P To be included in the alliance G Sub-alliances within For the linear programming model LP n-1 -Even solves the problem of maximizing overall satisfaction n -1 minimum value; When LP n -Odd When a unique solution is found, the solution obtained is the alliance equilibrium solution; otherwise, continue to solve the linear programming model LP n -Even minimizes the maximum value in the overall satisfaction vector ; No. n An even-numbered linear programming model LP with minimization as the objective n -Even: ；（39） In the formula, The optimal solution for maximizing overall satisfaction for solving even linear cycles, The solution of the odd linear cycle minimizes the overall satisfaction n submaximum; For the above models, all of them satisfy , indicating that the individuals participating in the cooperation are rational; Used to constrain the satisfaction values ​​of all sub-alliances to be less than the model LP n-1 -Odd solved This ensures that when finding the min function, there will be no n- 1 small satisfaction Smaller satisfaction value; Used to constrain the satisfaction values ​​of all sub-alliances to be greater than the model LP n-1 -Even solved This ensures that when calculating the max function, there will be no n The second highest satisfaction Greater satisfaction value; Indicates i Linear Programming Model LP i -Odd's solution, the satisfaction of all sub-alliances is obtained The alliance set at that time; Indicates i Linear Programming Model LP i -Even’s solution, the satisfaction of all sub-alliances is obtained The alliance collection at that time.

8. The method for distributing profits of a virtual power plant internal alliance cooperation game according to claim 1 is characterized in that: The step S5 includes the following steps: S501, input the basic data of wind farms, photovoltaic power plants, pumped storage power plants, and demand response users, and import the data into the multi-alliance cooperative game scheduling model within the virtual power plant; S502: Given the initial value of the cooperative game equilibrium solution, randomly select a strategy set ,in They represent the initial power values ​​of the wind power alliance, photovoltaic alliance and pumped storage alliance cooperation games respectively, It is expressed as the initial value of the total user load reduction and increase in the demand response alliance. Then, the CPLEX solver is used to solve the optimal strategy equilibrium point of the cooperative game. S503: Determine whether the strategy sets of each alliance have reached a cooperative equilibrium. If yes, calculate the electricity sales revenue of each alliance; if no, iterate and update the cooperative game strategy again; S504. Risk, profit and cost factors are introduced to improve the Shapley value method and then calculate the income of each alliance. The income of alliance members is calculated using the improved nucleolus method of odd-even cycle.

Images