Multi-agent benefit equilibrium method for virtual power plant based on double-layer game model

Abstract

This invention discloses a method for balancing the interests of multiple stakeholders in a virtual power plant based on a two-layer game model, belonging to the field of power system optimization and operation technology. The method includes the following steps: constructing an upper-layer asymmetric evolutionary game model, introducing strategy entropy and habituation coefficients, simulating the long-term cooperative tendency strategy evolution of stakeholders such as distributed power sources, flexible loads, and energy storage systems, and outputting an evolutionarily stable strategy; inputting the evolutionarily stable strategy as a modulation parameter into a lower-layer stochastic Stackelberg game model, using a typical scenario method to handle uncertainty, and optimizing short-term scheduling plans and internal electricity prices; allocating cooperative surplus based on fuzzy Shapley values, with staker membership determined by historical cooperative tendencies; verifying execution using blockchain smart contracts, and implementing automatic punishment through a dynamic penalty function linking "deviation degree and credit score". This method achieves dynamic and reliable interest balance among multiple stakeholders on both long and short-term scales, significantly improving the stability and economy of the virtual power plant alliance.

Description

Technical Field

[0001] This invention relates to the field of power system optimization operation technology, and in particular to a method for balancing the interests of multiple stakeholders in a virtual power plant based on a two-level game model. Background Technology

[0002] Virtual power plants aggregate distributed resources such as distributed power sources, flexible loads, and energy storage systems through advanced information and communication technologies, participating in the electricity market as a special type of power plant. This is an important means to enhance the grid's ability to absorb renewable energy. However, the various participants within a virtual power plant are independent entities with their own interests. How to coordinate their interests, maximize overall benefits, and ensure the stability of the alliance is the core challenge for the commercial operation of virtual power plants.

[0003] Existing technologies mostly employ single-level game models for benefit distribution, such as the Shapley value method and nucleolus method in cooperative games, or the master-slave game (Stackelberg game) in non-cooperative games. These methods have significant shortcomings:

[0004] 1) Staticity: It assumes that the behavior patterns of the subjects are fixed and does not change, ignoring the dynamic process by which they continuously adjust their strategies through learning and imitation in long-term interactions, making it difficult to reflect the long-term stability of the balance of interests.

[0005] 2) Idealized assumptions: It is usually assumed that the subject has perfect rationality and information symmetry, which is inconsistent with the bounded rationality and information asymmetry of the subject in reality.

[0006] 3) Insufficient handling of uncertainty: There is a lack of effective modeling methods for the uncertainty brought about by the high proportion of renewable energy access, resulting in high risk of scheduling plans.

[0007] 4) Lack of enforcement safeguards: There is a lack of reliable technical means to ensure that the agreements reached in the game can be executed automatically and reliably, and there is a lack of effective and automated punishment mechanisms for breach of contract. Summary of the Invention

[0008] The purpose of this invention is to provide a method for balancing the interests of multiple stakeholders in a virtual power plant based on a two-layer game model. The core technical problem to be solved by this invention is: how to realistically simulate the interactive behavior of multiple stakeholders in a virtual power plant at different time scales, how to fairly allocate the cooperative surplus to incentivize long-term cooperation, and how to ensure that the game results are executed reliably, thereby achieving a systematic and dynamic balance of interests.

[0009] In a first aspect, the present invention provides a method for balancing the interests of multiple stakeholders in a virtual power plant based on a two-layer game model, comprising the following steps:

[0010] S1. Construct an upper-level asymmetric evolutionary game model to simulate the dynamic evolution of the cooperative tendencies of different types of participants in a virtual power plant over a long-term scale, and output the evolutionarily stable strategy.

[0011] S2. Construct a lower-level stochastic Stackelberg game model, and use the evolutionary stable strategy as the key parameter input to solve the optimal scheduling plan and internal clearing price of the virtual power plant in the short term, taking into account the uncertainty of renewable energy.

[0012] S3. Based on the cooperative surplus generated by the optimal scheduling plan, the fuzzy Shapley value method is used to distribute the benefits, wherein the membership degree of each participating entity in the fuzzy alliance is determined by its historical cooperative tendency strategy.

[0013] S4. Verify the execution status of each participating entity on the optimal scheduling plan based on the blockchain smart contract, and automatically execute dynamic penalties based on the execution deviation and the entity's historical credit.

[0014] Preferably, in step S1, the specific construction process of the upper-level asymmetric evolutionary game model includes:

[0015] S11. Classify the participants in the virtual power plant according to resource type, and define a discrete cooperation tendency strategy space for each type of participant;

[0016] S12. Construct an asymmetric payment matrix that depends on the principal type and strategy;

[0017] S13. Establish an asymmetric replication dynamic equation that incorporates strategy entropy and learning habit coefficients to describe the dynamic evolution of different strategy proportions among various types of agents.

[0018] Preferably, in step S13, the expression for the asymmetric replication dynamic equation is as follows:

[0019] ;

[0020] in, The type is Strategies adopted in the main body proportion, For type The coefficient of the subject's learning habits For type The strategy entropy of the subject For type China adopts strategy The individual's expected return, For type The average return of the group.

[0021] Preferably, in step S2, the specific construction process of the lower-level stochastic Stackelberg game model includes:

[0022] S21. Establish an upper-level optimization model with the virtual power plant operator as the leader and the goal of minimizing the total expected operating cost.

[0023] S22. Establish a lower-level optimization model with each participating entity as a follower and the goal of maximizing its own expected profit. The cost or utility function of each entity is modulated by its cooperative tendency strategy obtained from step S1.

[0024] S23. The typical scenario method is used to describe the uncertainty of renewable energy output, and the two-level optimization problem is transformed into a single-level stochastic programming problem based on KKT conditions.

[0025] Preferably, in step S22, the cooperative tendency strategy modulates the subject's cost or utility function through a modulation coefficient, wherein the modulation coefficient corresponding to the active cooperative strategy is... Modulation coefficient corresponding to opportunistic strategy Modulation coefficients corresponding to conservative cooperation strategies .

[0026] Preferably, in step S3, the calculation expression for the fuzzy Shapley value is as follows:

[0027] ;

[0028] in, It is the main body The share of the profits due, It is the collection of all subjects. yes Not included a subset of It is a subset The membership vector, It is the characteristic function of the fuzzy alliance.

[0029] Preferably, in step S4, the amount of the dynamic penalty is calculated using the following dynamic penalty function:

[0030] ;

[0031] in, as the main body The amount of the penalty, Basic penalty amount, This is the magnification factor. as the main body The deviation between the actual execution and the scheduling plan, as the main body Real-time updated historical credit score.

[0032] In a second aspect, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the above-described method.

[0033] Thirdly, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method.

[0034] Fourthly, the present invention also provides a virtual power plant control system configured to perform the above-described method.

[0035] Therefore, the present invention employs a virtual power plant multi-stakeholder interest equilibrium method based on a two-layer game model with the above-mentioned structure, which has the following beneficial effects:

[0036] (1) This invention simulates the long-term dynamic evolution of multi-agent strategies through upper-level asymmetric evolutionary game, enabling the interest equilibrium model to have the ability to learn and adjust itself, effectively adapt to changes in internal and external conditions of VPP, thereby significantly enhancing the long-term stability and vitality of the alliance.

[0037] (2) This invention introduces behavioral parameters (learning habits, policy entropy) to characterize bounded rationality and uses stochastic programming to handle the uncertainty of renewable energy, which significantly reduces the idealization of traditional models, making the scheduling plan obtained by optimization solution more robust and the decision results more accurate and reliable.

[0038] (3) The present invention uses the fuzzy Shapley value method for benefit distribution, taking the long-term cooperative loyalty (membership degree) of the subject as the key parameter, thereby directly linking cooperative contribution with benefits, which can effectively incentivize proactive cooperative behavior, avoid the "free-rider" phenomenon, and achieve inherent fairness.

[0039] (4) By introducing blockchain smart contract technology, this invention automates and transparently executes the agreement reached in the game, and implements dynamic penalties based on deviation degree and historical credit, which greatly improves the execution guarantee and default punishment efficiency of the scheduling plan, and ensures the reliability and trustworthiness of the entire system operation.

[0040] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0041] Figure 1 This is a flowchart illustrating a method for balancing the interests of multiple stakeholders in a virtual power plant based on a two-layer game model, according to the present invention. Detailed Implementation

[0042] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0043] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0044] Example

[0045] like Figure 1 As shown, this invention provides a method for balancing the interests of multiple stakeholders in a virtual power plant based on a two-layer game model, comprising the following steps:

[0046] S1. Construct an upper-level asymmetric evolutionary game model to simulate the dynamic evolution of cooperative tendencies (active cooperation P, conservative cooperation C, opportunism O) of distributed generation (DG), flexible load (FL), and energy storage system (ESS) on a long-term scale, and solve for the evolutionary stable strategy by replicating the dynamic equation.

[0047] S2. Construct a lower-level stochastic Stackelberg game model, using the evolutionary stable strategy as the modulation parameter of the principal cost / utility function, and use the typical scenario method to describe the uncertainty of renewable energy output, and solve for the short-term optimal scheduling plan and internal clearing price;

[0048] S3. Calculate the virtual power plant alliance cooperation surplus and allocate benefits based on the fuzzy Shapley value method; the membership degree of the subject in the fuzzy Shapley value is determined by the historical frequency of the subject adopting an active cooperation strategy in the long term;

[0049] S4. Based on the blockchain smart contract, collect the actual execution data of the subject, calculate the execution deviation, combine the subject's historical credit score, automatically execute the penalty through the dynamic penalty function, and feed the final benefit back to step S1 to form a closed-loop optimization.

[0050] In step S1, the construction and solution of the upper-level asymmetric evolutionary game model includes:

[0051] S11. Classify entities into DG, FL, and ESS types based on resource type, and define the strategy space for each type of entity. S ={ P , C , O};

[0052] S12. Construct the asymmetric payoff matrix and define the revenue function. :express Class body adopts strategy Benefits during interaction, satisfying ;

[0053] S13. Establish and introduce policy entropy. With learning habits Solving the replication dynamic equations yields the evolutionarily stable strategy:

[0054] ;

[0055] in, The type is Strategies adopted in the main body ( correspond P , C , O The proportion of ); and ;

[0056] For type The subject's habituality coefficient reflects the degree of sluggishness in strategy updating. , , ;

[0057] For type The policy entropy of the subject reflects the dispersion of policy choices;

[0058] in, ;

[0059] For type China adopts strategy The individual's expected return;

[0060] in, ( For interactive subject type, (The strategy number of the interaction subject).

[0061] For type The average income of the main group .

[0062] The evolutionary game solution in step S1 is obtained using the ode45 differential equation solver in Matlab, with the iteration termination condition being: (That is, reaching an evolutionary stable state).

[0063] In step S2, the construction and solution of the lower-level stochastic Stackelberg game model includes:

[0064] S21. Establish an optimization model that minimizes the total expected operating cost, with the virtual power plant operator as the leader:

[0065] ;

[0066] The constraints include:

[0067] ;

[0068] Main output constraints: (respectively, the main body) (Minimum and maximum output / power consumption);

[0069] in, Scheduling cycle duration (e.g., 24 hours). For time indexing; Number of typical scenarios contributing to renewable energy For the scene The probability of occurrence, ; , for The market electricity price and purchase volume before the specified time; , Scene , Real-time market electricity prices and purchase volume; , for Moment Subject Internal clearing electricity prices and trading volumes; , for Energy storage at all times The amount of discharge and charge;

[0070] S22. Establish an optimization model that maximizes the expected profit of each entity, with each entity acting as a follower:

[0071] ;

[0072] in, as the main body The basic cost / utility function (DG is the cost of generating electricity, FL is the cost of consuming electricity, and ESS is the cost of charging and discharging).

[0073] The policy modulation coefficients are determined by the evolutionarily stable policy: if the agent chooses P, then... ,select O but If you choose C, then ;

[0074] S23. Transform the bi-level optimization problem into a single-level stochastic programming problem using KKT conditions, solve it using the Gurobi solver, and output the optimal scheduling plan. With internal clearing price .

[0075] In step S3, the calculation of the cooperative residual and the fuzzy Shapley value includes:

[0076] S31. Calculate the cooperative surplus. :

[0077] ;

[0078] in, The total revenue of the Virtual Power Plant Alliance (market transaction revenue + policy subsidies); as the main body Revenue from independent operation (DG revenue from independent electricity sales, FL cost savings from independent electricity purchases).

[0079] S32, Calculation Entity membership degree :

[0080] ;

[0081] in, as the main body The total number of times a positive cooperative strategy is adopted over a long evolutionary period; This represents the total number of long-term evolutionary cycles.

[0082] S33. Calculate the fuzzy Shapley value. :

[0083] ;

[0084] in, It is the main body The share of the profits due, It is the collection of all subjects. The total number of main entities; yes Not included a subset of For subset The number of elements; It is a subset The membership vector, ; It is the characteristic function of the fuzzy alliance. ( (For any subset of subjects).

[0085] In step S4, the execution of the blockchain smart contract and dynamic penalties include:

[0086] S41, Calculation Entity execution deviation :

[0087] ;

[0088] in, as the main body exist Actual execution volume at any given moment (collected via IoT devices). as the main body exist The optimal scheduling plan size at any given time;

[0089] S42. Calculate the dynamic penalty amount :

[0090] ;

[0091] in, The base penalty amount (preset to be 5%~10% of the remaining amount in cooperation); This is the magnification factor. as the main body The deviation between the actual execution and the scheduling plan, as the main body Real-time updated historical credit score The initial value is 0.8, and the performance target is met. When in breach of contract .

[0092] S43, scheduling plan Compared with actual execution volume Write it to the blockchain, and the smart contract will automatically deduct the amount. And deposited into the alliance risk pool, the final returns will be... Feedback is sent to step S1 to revise the payoff matrix for the next round of evolutionary game.

[0093] The blockchain adopts the Hyperledger Fabric private chain architecture, IoT data is uploaded in encryption through the edge computing gateway, and smart contracts are written in Solidity language, with the trigger condition being "real-time execution of data upload completion".

[0094] Taking a virtual power plant in an industrial park as an example, the specific parameters are as follows:

[0095] Main body composition: It includes 3 categories and a total of 6 participating entities, specifically as follows:

[0096] Distributed power generation (DG): 2 x 100kW rooftop solar panels (DG1, DG2) and 1 x 150kW small wind turbine (DG3).

[0097] Flexible loads (FL): one 200kW adjustable industrial cooling load (FL1, adjustable between 50% and 100% of rated power), and one 100kW residential building flexible load (FL2, adjustable between 80% and 100% of rated power).

[0098] Energy Storage System (ESS): 1 unit of 500kWh lithium battery energy storage (charge and discharge power 100kW, SOC operating range 20%-80%).

[0099] Market and environmental parameters:

[0100] Scheduling cycle: 1 calendar day (24 hours, time step 1 hour, t=1 to t=24).

[0101] Electricity market prices: The day-ahead market price is RMB 0.52 / kWh; the real-time market is divided into three typical scenarios based on wind power output (Table 1).

[0102] Uncertainty regarding renewable energy: Wind power (DG3) output is based on local historical data, and K-means clustering is used to generate three typical scenarios and probabilities; Photovoltaic (DG1, DG2) output is set according to the local sunshine curve, with a fluctuation range of ±10%.

[0103] Evolution cycle: 100 virtual days (used for iterating upper-level evolutionary game strategies and verifying long-term stability).

[0104] Table 1 Real-time Market Electricity Price and Wind Power Output Scenarios

[0105]

[0106] Evolutionary game parameters: habituation coefficient , , Initial value of policy entropy (Reflects the initial strategy dispersion);

[0107] Modulation coefficients: Active cooperation (P) → α = 0.95, Conservative cooperation (C) → γ = 1.0, Opportunism (O) → β = 1.5;

[0108] Penalty parameters: Base penalty amount BasePenalty =200 yuan (8% of the average remaining value of the cooperation), deviation amplification factor k=3.5, initial credit score .

[0109] Specifically, 1. Modeling and solving of upper-level asymmetric evolutionary games

[0110] S11: Subject Classification and Strategy Definition The six subjects are divided into three categories based on resource type. The strategy space for each category is unified as {X1: Active Cooperation (P), X2: Conservative Cooperation (C), X3: Opportunistic (O)}. The specific strategy behaviors are defined as follows:

[0111] Active cooperation (P): DG proactively reduces generation cost quotation by 5%, FL proactively reduces load by 10%-15% during peak electricity price periods (t=12-14, t=18-20), and ESS charges more during off-peak periods (t=23-7) and discharges more during peak periods;

[0112] Conservative cooperation (C): All entities participate in scheduling only according to the basic requirements, without making additional concessions or adjustments;

[0113] Opportunism (O): DG inflated its generation costs by 15% to raise its bids, FL refused to cut load, and ESS "reluctantly discharged" during peak electricity prices (reduced its discharge by 50%).

[0114] S12: Asymmetric payment matrix construction is based on differences in subject type, constructing pairwise interactive asymmetric payment matrices (taking "DG and FL interaction" and "DG and ESS interaction" as examples):

[0115] Table 2 DG and FL Interactive Payment Matrix (Unit: Yuan / Hour)

[0116]

[0117] Note: In the matrix, "(a,b)" means that DG's profit is a and FL's profit is b. For example, when "DG chooses P and FL chooses P", both parties obtain high profits (12,8) due to collaborative optimization; when "DG chooses O and FL chooses P", DG profits 18 through speculation, while FL only obtains 3 due to concessions.

[0118] Table 3 DG and ESS Interactive Payment Matrix (Unit: Yuan / Hour)

[0119]

[0120] S13: Solving the replication dynamic equations uses Matlab's ode45 differential equation solver. Substituting the equations into the asymmetric replication dynamic equations:

[0121] ;

[0122] Initial conditions: All subjects have a strategy ratio of 1 / 3. );

[0123] Iteration termination condition: (The rate of change of the strategy proportion approaches 0);

[0124] Solution results: The evolution reached a stable state on the 82nd virtual day, and the strategy ratios of various agents are as follows:

[0125] Table 4. Strategy Proportions of Various Entities

[0126]

[0127] Conclusion: After long-term evolution, the proportion of opportunistic strategies has dropped to below 3%, while ESS has the highest proportion of active cooperation (91%) due to its high flexibility in energy storage dispatch.

[0128] Step S2: Modeling and solving the lower-level stochastic Stackelberg game

[0129] Based on the evolutionary stability strategy, a joint scheduling optimization of "day-ahead + real-time" is performed, and the specific process is as follows:

[0130] S21: Leader (VPP operator) optimization model objective function: Minimize the total expected operating cost of the virtual power plant (including day-ahead power purchase, real-time power purchase, and internal transaction costs):

[0131]

[0132] Constraints:

[0133] Supply and demand balance constraints (taking t=12 as an example):

[0134] ;

[0135] Main output constraints: such as DG1 output ESS charging and discharging power , .

[0136] S22: Follower (Agents) Optimization Model: Each agent aims to maximize its own expected profit. The cost / utility function is modulated according to the evolutionary strategy. Taking DG1 (selecting P, α=0.95) and FL1 (selecting P, α=0.95) as examples:

[0137] DG1 profit function: ;

[0138] in (The unit cost of photovoltaic power generation is 0.32 yuan / kWh);

[0139] FL1 Profit Function: (Taking cost savings as profit)

[0140] ;

[0141] in (The unit electricity cost for industrial refrigeration is 0.55 yuan / kWh).

[0142] S23: Model Solving and Results. The bi-level optimization problem was transformed into a single-level stochastic programming problem using KKT conditions. The Gurobi 10.0 solver was used to solve the problem, and the key scheduling results for t=12 (midday load peak) and t=2 (late-night load trough) were output:

[0143] Table 5 Scheduling Results

[0144]

[0145] Note: In the ESS cost function, 0.25 yuan / kWh is the unit loss cost of charging and discharging. At t=2, ESS charging utilizes the off-peak electricity price (0.42 yuan / kWh) for subsequent peak discharge arbitrage.

[0146] Step S3: Benefit allocation based on fuzzy Shapley values

[0147] S31: Cooperative Residual Calculation

[0148] Alliance Total Revenue: The total revenue of the virtual power plant on that day (including market transactions and policy subsidies) was 8,620 yuan;

[0149] Independent operating revenue: The sum of the revenue of each entity participating in the market independently is 6280 yuan (e.g., DG1's independent electricity sales revenue is 980 yuan, FL1's independent electricity cost is 1250 yuan, and ESS's independent arbitrage revenue is 820 yuan).

[0150] Remaining cooperation: .

[0151] S32: Membership Degree Calculation Membership degree is determined by the number of times the active cooperative strategy was adopted in the past 60 evolutionary cycles. The calculation results are as follows:

[0152] Table 6 Membership Calculation Results

[0153]

[0154] S33: Substituting fuzzy Shapley value assignment into the fuzzy Shapley value formula:

[0155] ;

[0156] Among them, the characteristic function Indicates fuzzy alliance The remaining cooperation within the scheduling period is calculated using a linear weighted method as follows:

[0157] ;

[0158] in, This represents the total number of time periods in the scheduling cycle (e.g., 24 hours). Indexed by time period; Belongs to a subset The main body; as the main body The degree of membership is determined by historical cooperative tendencies (see step S32). as the main body During the period The optimal scheduling plan power is obtained by solving step S2; as the main body During the period The internal clearing price is also obtained by solving step S2. This characteristic function reflects the actual contribution of each entity in the fuzzy alliance, providing a basis for subsequent fuzzy Shapley value allocation, and the final allocation results are shown in Table 7.

[0159] Table 7 Final Allocation Results

[0160]

[0161] Step S4: Blockchain Smart Contract Execution and Dynamic Penalties

[0162] S41: Data Acquisition and Deviation Calculation: Collect actual execution data from each entity using IoT devices (smart meters, inverter monitoring systems) and calculate the deviation.

[0163] For example, with t=12:

[0164] Actual output of DG3: Due to a sudden gust of wind, the actual output of DG3 was... Scheduling plan , deviation degree ;

[0165] FL1 actual output: As planned , deviation degree ;

[0166] All other entities performed as planned, with a deviation of 0%.

[0167] S42: Dynamic penalty calculation and execution - Substituting into the dynamic penalty function:

[0168] ;

[0169] DG3 penalty calculation: , k =3.5, , (No history of defaults), then:

[0170] ;

[0171] Penalty Execution: The smart contract (deployed on the Hyperledger Fabric private chain) automatically deducts 61.6 yuan from the DG3 distribution revenue (415 yuan) and deposits it into the consortium risk pool (to compensate for the grid regulation costs caused by the over-issuance of DG3).

[0172] Credit score updates: DG3's credit score dropped from 0.8 to 0.7 (0.1 deduction for default), while FL1's credit score rose from 0.8 to 0.85 (0.05 increase for fulfillment) due to no deviation.

[0173] S43: Closed-loop feedback feeds the final payoff (distributed payoff - penalty) of each subject back to the upper-level evolutionary game model, and corrects the payoff matrix for the next round: if DG3's final payoff drops to 353.4 yuan due to the penalty (lower than the group average payoff of 390 yuan), the payoff coefficient for "DG3 choosing O" in the next round will be reduced by 10%, further suppressing opportunistic tendencies.

[0174] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for multi-agent benefit equilibrium of a virtual power plant based on a double-layer game model, characterized in that, Includes the following steps: S1. Construct an upper-level asymmetric evolutionary game model to simulate the dynamic evolution of the cooperative tendencies of different types of participants in a virtual power plant over a long-term scale, and output the evolutionarily stable strategy. S2. Construct a lower-level stochastic Stackelberg game model, and use the evolutionary stable strategy as the key parameter input to solve the optimal scheduling plan and internal clearing price of the virtual power plant in the short term, taking into account the uncertainty of renewable energy. S3. Based on the cooperative surplus generated by the optimal scheduling plan, the fuzzy Shapley value method is used to distribute the benefits, wherein the membership degree of each participating entity in the fuzzy alliance is determined by its historical cooperative tendency strategy. S4. Verify the execution status of each participating entity on the optimal scheduling plan based on the blockchain smart contract, and automatically execute dynamic penalties based on the execution deviation and the entity's historical credit. 2.The virtual power plant multi-agent benefit equilibrium method based on a double-layer game model according to claim 1, wherein: In step S1, the specific construction process of the upper-level asymmetric evolutionary game model includes: S11. Classify the participants in the virtual power plant according to resource type, and define a discrete cooperation tendency strategy space for each type of participant; S12. Construct an asymmetric payment matrix that depends on the principal type and strategy; S13. Establish an asymmetric replication dynamic equation that incorporates strategy entropy and learning habit coefficients to describe the dynamic evolution of different strategy proportions among various types of agents. 3.The virtual power plant multi-agent benefit equilibrium method based on a double-layer game model of claim 2, characterized in that: In step S13, the expression for the asymmetric replication dynamic equation is as follows: ； in, The type is Strategies adopted in the main body proportion, For type The coefficient of the subject's learning habits For type The strategy entropy of the subject For type China adopts strategy The individual's expected return, For type The average return of the group.

4. The virtual power plant multi-agent benefit equilibrium method based on a double-layer game model according to claim 1, characterized in that: In step S2, the specific construction process of the lower-level stochastic Stackelberg game model includes: S21. Establish an upper-level optimization model with the virtual power plant operator as the leader and the goal of minimizing the total expected operating cost. S22. Establish a lower-level optimization model with each participating entity as a follower and the goal of maximizing its own expected profit. The cost or utility function of each entity is modulated by its cooperative tendency strategy obtained from step S1. S23. The typical scenario method is used to describe the uncertainty of renewable energy output, and the two-level optimization problem is transformed into a single-level stochastic programming problem based on KKT conditions.

5. The virtual power plant multi-agent benefit equilibrium method based on a double-layer game model according to claim 4, characterized in that: In step S22, the cooperation propensity strategy modulates the cost or utility function of the subject through a modulation coefficient, wherein the modulation coefficient corresponding to the positive cooperation strategy is , the modulation coefficient corresponding to the opportunistic strategy is , and the modulation coefficient corresponding to the conservative cooperation strategy is .

6. The virtual power plant multi-agent benefit equilibrium method based on a double-layer game model according to claim 1, characterized in that: In step S3, the expression for calculating the fuzzy Shapley value is as follows: ； wherein, is the principal the share of the benefit, is the set of all principals, is does not include a subset of is the membership vector of the subset is the membership vector of the subset is the characteristic function of the fuzzy coalition.

7. The method for balancing the interests of multiple stakeholders in a virtual power plant based on a two-layer game model according to claim 1, characterized in that: In step S4, the amount of the dynamic penalty is calculated using the following dynamic penalty function: ； in, as the main body The amount of the penalty, Basic penalty amount, This is the magnification factor. as the main body The deviation between the actual execution and the scheduling plan. as the main body Real-time updated historical credit score.

8. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the method as described in any one of claims 1-7. 9.A non-transitory computer-readable storage medium having stored thereon a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1-7.

10. A virtual power plant control system, characterized by The system is configured to perform the method as described in any one of claims 1-7.

Images