Virtual power plant scheduling method considering source-load coupling and multi-agent game

Abstract

The invention relates to a virtual power plant scheduling method considering source-load coupling and multi-agent game. The method comprises the following steps: establishing a multi-temporal-spatial-scale model of a multi-energy system in a virtual power plant; the uncertainty of the source load is quantified; on the basis of quantification of source load uncertainty and a multi-temporal-spatial-scale model, a multi-subject scheduling model under a master-slave game framework is constructed, and in the multi-subject scheduling model, a VPP operator serves as a leader, and the market is regulated and controlled through dynamic electricity price and heat price; an energy supply aggregator, a user side aggregator and a carbon processing aggregator are used as followers, unit output, load adjustment and carbon processing strategies are optimized respectively, the two parties perform multi-stage dynamic gaming to form balance, and the user side aggregator comprises CSRLA and EVA; and solving the multi-subject scheduling model to obtain a virtual power plant scheduling optimization strategy. Compared with the prior art, the method has the advantages of realizing dynamic balance of interests of multiple parties, privacy protection and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of virtual power plant scheduling, and in particular to a virtual power plant scheduling method considering source-load coupling and multi-agent game. Background Art

[0002] With the transformation of the global energy structure and the rapid development of renewable energy, the operating model of traditional power systems has gradually exposed many limitations. In particular, the intermittent and uncertain nature of renewable energy has posed a huge challenge to the stable operation of the power grid. The widespread access to distributed energy resources has further increased the complexity of the power system, and existing virtual power plant scheduling methods still have significant shortcomings in dealing with source-load coupling, multi-agent games, uncertainty quantification, carbon emission constraints, and privacy protection. Specifically, traditional methods have difficulty effectively coordinating the supply and demand balance between multiple energy sources, resulting in low energy utilization efficiency. At the same time, virtual power plants involve multiple stakeholders, and the objective functions and decision-making behaviors of each stakeholder influence each other. Traditional centralized optimization methods have difficulty achieving a balance of interests among multiple parties. In addition, there is significant uncertainty in wind and solar power output, load demand, etc., and existing methods lack efficient technical means for scenario generation and reduction, resulting in poor robustness of scheduling results.

[0003] As the carbon trading market gradually matures, how to incorporate carbon emission rights as tradable assets into the scheduling framework and achieve low-carbon scheduling has become a technical problem that needs to be solved urgently.

[0004] After searching, Chinese invention patent application publication number CN118710148A discloses a decision-making method and system for a virtual power plant to participate in demand response. The method includes: establishing a multi-agent demand response optimization architecture based on master-slave game theory; establishing an electricity-carbon coupling model that takes into account the characteristics of peak and valley periods based on the day-ahead dynamic electricity price and day-ahead dynamic carbon emission factor of the electricity spot market; based on the electricity-carbon coupling model, constructing an upper-level master model with the goal of maximizing the economic benefits of the virtual power plant operator and a lower-level slave model with the goal of minimizing the energy purchase cost of each user, thereby obtaining a master-slave game model for the virtual power plant to participate in demand response; solving the master-slave game model in combination with the set constraints to obtain an optimization strategy for the virtual power plant to participate in demand response. This existing patent application has the problem that only virtual operators, users, and electric vehicles participate in scheduling and coordination, resulting in poor collaborative optimization strategies and benefits for virtual power plants.

[0005] How to implement a virtual power plant scheduling method that takes into account source-load coupling and multi-agent game has become a technical problem that needs to be solved. Summary of the Invention

[0006] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and provide a virtual power plant scheduling method considering source-load coupling and multi-agent game.

[0007] The purpose of the present invention can be achieved by the following technical solutions: According to one aspect of the present invention, a virtual power plant scheduling method considering source-load coupling and multi-agent game is provided, comprising: Establish a multi-time and space-scale model of the multi-energy system in the virtual power plant, where the multi-energy system includes the energy supply system, multi-energy coupling system and energy storage system; Quantify source charge uncertainty; Based on the quantification of source-load uncertainty and a multi-spatiotemporal scale model, a multi-agent scheduling model based on a master-slave game is constructed. In this multi-agent scheduling model, the VPP operator acts as the leader, regulating the market with dynamic electricity and heat prices. The energy supply aggregator, user-side aggregator, and carbon treatment aggregator act as followers, optimizing unit output, load adjustment, and carbon treatment strategies, respectively, to form a Stackelberg equilibrium. The master-slave game is a multi-stage dynamic game. VPP operators optimize electricity pricing strategies through iterative learning, while followers adjust their responses based on real-time strategies and transmit strategies through fuzzy quotes and interval responses. The NOA-QP hybrid optimization algorithm is used to solve the multi-agent scheduling model and obtain the optimization strategy for virtual power plant scheduling.

[0008] Preferably, when the multi-agent scheduling model reaches Stackelberg equilibrium, the following equation is satisfied: , Where, 、 、 、 、 They are respectively expressed as VPP operator revenue, energy supply aggregator revenue, user-side residential load aggregator consumer satisfaction, electric vehicle aggregator consumer satisfaction, and carbon treatment aggregator revenue; They are respectively represented as the VPP operator strategy, energy supply aggregator strategy, user-side residential load aggregator strategy, electric vehicle aggregator strategy, and carbon treatment aggregator strategy under the Stackelberg equilibrium of each entity; They are respectively represented as VPP operator strategy, energy supply aggregator strategy, user-side residential load aggregator strategy, electric vehicle aggregator strategy, and carbon treatment aggregator strategy.

[0009] Preferably, the objective functions of each subject are designed for the multi-subject scheduling model, specifically including: the VPP operator aims to maximize energy sales revenue and carbon trading income, while minimizing energy purchase costs and heat interruption penalties; the energy supply aggregator aims to maximize energy sales revenue, while minimizing fuel costs and carbon trading costs; the user-side aggregator aims to maximize consumer satisfaction; and the carbon treatment aggregator aims to maximize carbon treatment income.

[0010] More preferably, the objective function expression of the VPP operator is: , Where, For the benefit of VPP operators; The revenue of the VPP operator from selling energy to the energy user at time t; is the cost of purchasing energy from the energy supplier at time t; is the interaction cost between the VPP operator and the external network at time t; are the penalty costs of supply interruption at time t; 、 are the electric load on the electricity consumption side and the electric power output on the energy supply side at time t respectively; , are the heat load on the electricity consumption side and the heat power output on the energy supply side at time t respectively; 、 、 、 are the prices of electricity and heat sold to energy users at time t, and the prices of electricity and heat purchased from suppliers at time t; 、 are the electricity purchase price and on-grid electricity price at time t respectively; is the thermal interruption penalty coefficient, T is the scheduling period; The objective function expression of the energy supply aggregator is: , Where, Profits for energy supply aggregators, is the energy sales revenue of the energy supply aggregator at time t, is the fuel cost on the energy supply side at time t, are the carbon transaction costs on the energy supply side at time t; 、 、 is the cost coefficient of the micro gas turbine, 、 、 They are the cost factors of gas boilers; , are the electrical power output of the micro gas turbine and the thermal power output of the gas boiler at time t respectively; The objective function of the carbon treatment aggregator is specifically: , Where, For CTSA benefits; 、 、 They are the income from selling natural gas to the natural gas trading market at time t, the benefits of releasing carbon to the carbon market, and the CTSA incentive income; , are the carbon sequestration cost at time t and the cost of electricity purchased from the grid by CTSA; is the amount of natural gas sold to the natural gas trading market at time t: 、 They are the unit natural gas price and the carbon trading price of the day; , are the carbon volume traded with the carbon market and carbon sequestration at time t, respectively; , are the electrical power of CCS equipment and P2G equipment at time t respectively; 、 are the incentive coefficients of the carbon treatment system, are the unit carbon sequestration cost; is the electricity selling price of VPPO to energy users at time t.

[0011] More preferably, the user-side aggregator includes a user-side residential load aggregator and an electric vehicle aggregator. The user-side residential load aggregator aims to maximize consumer satisfaction and forms a two-way game with the VPP operator's electricity price strategy. The response sensitivity is quantified by the load elasticity coefficient. The expression for maximizing consumer satisfaction is: , Where, Consumer satisfaction for user-side residential load aggregators; is the residential demand response utility function at the user side at time t; 、 are the cost of energy purchased by the user-side residential load aggregator at time t and the cost of electricity purchased by the energy storage system from the grid; 、 、 They are customer electricity load, basic residential electricity load and transferable electricity load at time t respectively; 、 are the prices of electricity and heat sold to energy users at time t; 、 、 They are the heat load on the electricity consumption side, the basic heat load of the residential building on the user side and the curtailable load at time t respectively; 、 are the prices of electricity and heat sold by the VPP operator to the energy user at time t, respectively; 、 、 、 is the user's preference coefficient for electrical energy and thermal energy, T is the scheduling period; The electric vehicle aggregator aims to maximize consumer satisfaction, which is expressed as: , Where, For EVA consumer satisfaction; is the utility function of electric vehicle users at time t; is the energy purchase cost of electric vehicle users at time t; 、 、 are electric vehicle load, transferable load, and curtailable load at time t respectively; 、 is the electric vehicle user preference coefficient; is the satisfaction penalty coefficient.

[0012] Preferably, physical constraints for multi-energy coordinated scheduling are designed for the multi-agent scheduling model, including unit output constraints, ramp constraints, electric power balance constraints, thermal power balance constraints, and carbon flow constraints; Among them, the power balance constraint requires that the output of wind, solar, CHP, and energy storage meet the user-side power load, electric vehicle charging load, and carbon treatment system power demand; the thermal power balance constraint requires that the CHP waste heat be matched with the heating of the gas boiler to reduce the user-side heat load; Carbon flow constraints require that actual carbon emissions must not exceed the sum of free allowances and carbon processing volume, including carbon balance constraints for carbon processing aggregators, energy balance constraints for internal equipment, power constraints and creep constraints, constraints for power-to-gas production and utilization, and power constraints and ramp constraints for carbon capture and storage and power-to-gas as follows: , Where, is t The amount of carbon entering the carbon capture and sequestration equipment at any given moment; 、 Respectively t The amount of carbon stored in carbon capture equipment and the actual carbon emissions at each moment; is t The amount of carbon used in power-to-gas conversion at each moment; is a Boolean variable indicating the start and stop status of the carbon capture and storage equipment; is the carbon capture and storage baseload electricity consumption; is the operational consumption factor for CCS; is the density of carbon dioxide; is the amount of natural gas produced through power-to-gas conversion; is the power-to-gas operating efficiency; is the electrical power consumed by the power-to-gas equipment; is the calorific value of natural gas; and are the operating powers of the carbon capture and storage equipment and the power-to-gas equipment at time t, and are the rated power of the carbon capture and storage equipment and the power-to-gas equipment respectively; and These are the lower and upper limits of the carbon capture and storage equipment ramp-up, respectively; and They are the lower and upper limits of the power-to-gas equipment climbing.

[0013] Preferably, a NOA-QP hybrid optimization algorithm, combining the Nutcracker optimization algorithm with a quadratic programming algorithm, is used to solve the multi-agent scheduling model. The Nutcracker optimization algorithm is used to globally search for the optimal electricity pricing strategy, and the quadratic programming algorithm is used to efficiently solve the objective function of each aggregator. Specifically, the inner layer solves for the optimal price under the game of each agent within the virtual power plant, transmits the optimal price model to the upper layer, and after receiving the optimal price model, the upper layer iterates the optimal equipment output. Preferably, the user behavior and demand response modeling quantifies the flexibility potential of user-side electrical and thermal loads, formulates the total conservation rules for transferable loads, and establishes the elastic range for load reduction.

[0014] Preferably, for electric vehicle clusters, statistical methods are used to simulate differences in network access time, mileage, and charging modes, and to classify and establish fast-charging and slow-charging load curves to characterize the impact of flexibility on the system. The corresponding model is: , Where, The end time of the last trip, that is, the time when the electric vehicle joined the network; is the expected value of charging time; is the standard deviation of charging time; , Where, Indicates the power consumption of electric vehicles for every 100 kilometers traveled. 、 、 They represent the output power, charging efficiency and delay parameters of the charging pile respectively.

[0015] Preferably, the process of quantifying source-load uncertainty includes: generating multiple wind and solar output and load scenarios based on Monte Carlo simulation, reducing redundant scenarios through Manhattan distance metric, retaining the most representative load and new energy fluctuations of typical photovoltaic output, typical wind power output, user-side electric load, user-side thermal load and electricity demand of electric vehicle clusters throughout the year, and obtaining a typical scenario set.

[0016] Compared with the prior art, the present invention has the following beneficial effects: (1) Multi-agent collaboration and interest balance: This invention effectively coordinates the interest conflicts among VPP operators, energy supply aggregators, user-side aggregators, and carbon treatment aggregators by constructing a multi-agent scheduling model under the master-slave game framework, and solves the unique equilibrium solution of the multi-agent scheduling model, ensuring that each party cannot obtain additional benefits through unilateral strategy adjustments. Through the dynamic balance of the interests of multiple parties, while ensuring the stable operation of the system, it effectively improves the energy utilization efficiency and the collaborative benefits among multiple subjects, and has good scalability and practical application value.

[0017] (2) The present invention designs the conditions for the equilibrium of the multi-agent scheduling model, the physical constraints of multi-energy coordinated scheduling, and the objective functions of each agent, so as to achieve coordination among the agents to maximize their interests under complex conditions such as fluctuations in new energy output and changes in electric vehicle charging demand.

[0018] (3) The NOA-QP hybrid optimization algorithm proposed in this paper can efficiently solve the multi-agent scheduling model, significantly shortening the solution time compared with traditional methods. At the same time, each aggregator only feeds back the output strategy, thus avoiding the leakage of sensitive information through the privacy protection mechanism.

[0019] (4) This invention proposes a quantification method for the uncertainty of wind and solar power output and load demand, forming a typical scenario set, which helps to improve the robustness of the scheduling results. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 Schematic diagram of the flow of the virtual power plant scheduling method in the present invention; Figure 2 This is a schematic diagram of the unit output of the present invention; Figure 3 This is a schematic diagram of heat scheduling of a unit according to the present invention; Figure 4 This is a schematic diagram of energy purchase and sale prices according to the present invention. DETAILED DESCRIPTION

[0021] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0022] In response to the above problems, the present invention proposes a virtual power plant scheduling method that takes into account source-load coupling and multi-agent game. Through refined modeling, uncertainty quantification, carbon trading integration, game framework design and privacy protection mechanism, it achieves multi-energy coordinated scheduling and balance of interests of multiple parties, providing technical support for the efficient operation and low-carbon development of virtual power plants, while effectively protecting sensitive information such as equipment parameters and user preferences, and providing innovative solutions for the intelligent, low-carbon and sustainable development of energy systems.

[0023] This embodiment relates to a virtual power plant scheduling method considering source-load coupling and multi-agent game, including the following steps: Step S1: Detailed modeling of multi-energy devices. A multi-temporal and spatial scale model of the virtual power plant is established, including the energy supply system, multi-energy coupling system, and energy storage system. A decoupled operation mechanism for the multi-energy coupling system is also established, providing a model foundation for subsequent multi-energy coordinated scheduling.

[0024] Step S2: User Behavior and Demand Response Modeling. Based on Step S1, the flexibility potential of user-side electrical and thermal loads is quantified. Rules for the conservation of the total amount of transferable load and the elastic range of curtailable load are established to ensure the feasibility of load adjustment. EV aggregators adjust the proportion of transferable and curtailable load based on dynamic electricity prices. Gaming strategies are fed back into the satisfaction function to create a two-way incentive mechanism, providing user behavior data support for subsequent optimized scheduling.

[0025] Step S3: Integrated modeling of carbon trading and carbon treatment systems. Based on S1 and S2, carbon emission rights are embedded into the scheduling framework as tradable assets, providing a basis for calculating carbon constraints and carbon trading costs and benefits for subsequent low-carbon scheduling.

[0026] Step S4: Quantify source-load uncertainty. Based on the above step model, Monte Carlo simulation is used to generate a large number of scenarios for wind and solar power output and load. These scenarios are then reduced to obtain a typical set of scenarios to quantify uncertainty and provide a deterministic input for subsequent optimized scheduling.

[0027] Step S5: Construct a multi-agent dispatch model within a master-slave game framework. Based on the models and data from steps S1-S4, a hierarchical decision-making framework is constructed: the VPP operator (VPPO) serves as the leader, regulating the market with dynamic electricity and heat prices. The energy supply aggregator (ESA), customer-side aggregators (customer-side residential load aggregators (CSRLAs) / electric vehicle aggregators (EVAs), and carbon treatment aggregators (CTSAs) serve as followers, optimizing unit output, load adjustment, and carbon treatment strategies, respectively. The two parties engage in a multi-stage dynamic game to achieve equilibrium.

[0028] Step S6: Design of virtual power plant constraints. Within the game framework of step S5, physical constraints for multi-energy coordinated dispatch are defined, including unit output constraints, ramping constraints, power balance constraints, thermal power balance constraints, and carbon flow constraints. The power balance constraint requires that wind and solar power, CHP, and energy storage outputs meet the user-side power load, EV charging load, and the electricity demand of the carbon treatment system. The thermal power balance constraint requires that CHP waste heat and gas boiler heating match the user's heat load after reduction. The carbon flow constraint serves as a hard boundary for the game. Actual carbon emissions are dynamically allocated to aggregators through game equilibrium, realizing the market-based circulation of carbon quotas.

[0029] Step S7: Multi-energy coordinated scheduling and dynamic strategy implementation. Based on the game theory framework of step S5 and the constraints of step S6, the objective functions of each entity are constructed. The VPP operator (VPPO) aims to maximize energy sales revenue and carbon trading benefits; the energy supply aggregator (ESA) aims to maximize energy sales revenue; the user-side aggregator (CSRLA / EVA) aims to maximize customer satisfaction; and the carbon treatment aggregator (CTSA) aims to maximize carbon treatment benefits.

[0030] Step S8: Hybrid Optimization Algorithm Design. To solve the game model in step S7, a hybrid algorithm combining bionic optimization and mathematical programming is proposed. This algorithm combines the Nutcracker Optimization Algorithm (NOA) with quadratic programming (QP) to improve the efficiency and accuracy of solving the game model. A privacy protection mechanism is designed, whereby each aggregator only provides feedback on its output strategy to the VPP operator, hiding its internal cost function and user preference parameters. This approach protects commercial privacy through fuzzy responses. NOA is designed to handle the nonlinear optimization problems faced by VPP operators and possesses strong global search capabilities. The objective functions of each aggregator are convex, making them suitable for rapid and accurate solution using the QP method. The combination of these two approaches achieves a fusion innovation in the algorithmic structure. Furthermore, considering that aggregators in the real electricity market possess sensitive information such as equipment parameters and energy consumption preferences, the algorithm is designed so that aggregators only need to provide the optimal output strategy after receiving the price signal, without exposing their internal model or parameter information, effectively safeguarding the data privacy of all participants.

[0031] Furthermore, step S1 includes the following steps: Step S1.1: Based on the Weibull distribution to characterize the randomness of wind speed, a model for calculating the output power of a wind turbine is generated. The specific model for calculating the output power of a wind turbine is: , Where, is the output power of the wind turbine generator set at time t, is the rated output power of the unit, is the actual wind speed at time t, 、 and Respectively represent rated wind speed, cut-in wind speed and cut-out wind speed; Step S1.2: Use Beta distribution to fit the photovoltaic irradiance probability characteristics, combine the temperature attenuation factor to correct the output curve, and generate the photovoltaic power generation system output power model. The photovoltaic power generation system output power model is: , Where, is the output power of the photovoltaic power generation system at time t, represents the number of photovoltaic panels in the photovoltaic cluster, and are the voltage and current of the photovoltaic panel at time t respectively; and are the voltage and current at the maximum power point respectively; and They represent the open circuit voltage and short circuit current respectively.

[0032] Furthermore, step S2 also includes: for electric vehicle clusters, simulating the differences in network access time, mileage, and charging modes based on statistical methods, classifying and establishing fast charging / slow charging load curves, and describing the impact of flexibility on the system. The corresponding model is: , Where, The end time of the last trip, that is, the time when the electric vehicle joined the network; is the expected value of charging time; is the standard deviation of charging time.

[0033] , Where, Indicates the power consumption of electric vehicles for every 100 kilometers traveled. 、 、 They represent the output power, charging efficiency and delay parameters of the charging pile respectively.

[0034] Step S3 includes the following steps: allocating free carbon quotas based on the benchmark method, building a carbon treatment benefit model, and realizing the market-oriented circulation of carbon resources. The formula for the free carbon quota and carbon trading model is: , Where, is the carbon trading cost at time t, and are the free carbon emissions and actual carbon emissions at time t, is the carbon trading price of the day, 、 、 They are the unit power electricity, heat carbon emission limit coefficient, and the carbon emission coefficient of the cogeneration unit. 、 、 、 、 are the power output of the gas turbine, wind power output, photovoltaic power output, power purchased from the grid by the VPP, and the sum of the output of the gas turbine and gas boiler during period t. 、 are the thermal power output of the gas turbine and the thermal power output of the gas boiler at time t, 、 、 is the carbon emission coefficient of natural gas.

[0035] Step S4 includes the following steps: Step S4.1: Generate multiple wind and solar power output and load scenarios based on Monte Carlo simulation, reduce redundant scenarios using the Manhattan distance metric, and retain the most representative load and renewable energy fluctuation scenarios for five types of electricity demand throughout the year: typical photovoltaic output, typical wind power output, user-side electrical load, user-side thermal load, and electric vehicle clusters, to obtain a typical scenario set; The main differences between this method and the traditional K-means clustering algorithm lie in the emphasis on scenario selection and the distance measurement method. K-means divides data clusters by Euclidean distance and is suitable for continuous data. This method, on the other hand, generates multiple wind and solar power output and load scenarios based on Monte Carlo simulation, and then uses the Manhattan distance metric to remove redundant scenarios, retaining the most representative typical load and renewable energy fluctuation scenarios. Compared with K-means, Manhattan distance is more suitable for processing non-continuous and highly volatile load data, and can more accurately capture the fluctuation characteristics of multi-dimensional energy demand, thereby optimizing the selection of typical scenarios and improving the accuracy and practicality of load forecasting and wind and solar power output scheduling.

[0036] Step S4.2: Eliminate invalid data segments such as those with wind speeds below the cut-in threshold to ensure the generalization ability of the model under typical operating conditions.

[0037] Step S5 includes: the VPP operator, as the leader, regulates the market with dynamic electricity and heat prices; the energy supplier, user side, and carbon treatment aggregator, as followers, optimize unit output, load adjustment, and carbon treatment strategies, respectively, to form a Stackelberg equilibrium (a master-slave game stable state) that meets the following conditions: , Where: 、 、 、 、 They are respectively expressed as VPPO benefit, ESA benefit, CSRLA consumer satisfaction, EVA consumer satisfaction, and CTSA benefit; They are respectively represented as VPPO strategy, ESA strategy, CSLRA strategy, EVA strategy, and CTSA strategy under the Stackelberg equilibrium of each subject; They are respectively represented as VPPO strategy, ESA strategy, CSLRA strategy, EVA strategy, and CTSA strategy.

[0038] Step S6 includes the following steps: setting constraints, including VPPO purchase and sale energy price constraints, ESA electricity / heat output power constraints, internal constraints of the combined power generation unit CHP, ramp constraints of micro gas turbines and gas boiler outputs, output constraints and ramp constraints of ESA micro gas turbines and gas boilers, CSRLA transferable electric load constraints, CSRLA curtailable heat load constraints, power constraints between the CSRLA user-side demand response aggregator and other main grid participants, transferable load constraints and curtailable load constraints in EVA constraints, CTSA carbon balance constraints, energy balance constraints of internal equipment, power constraints and creep constraints, P2G preparation and utilization constraints, (carbon capture and storage) and P2G (power to gas) power constraints and ramp constraints, VPPO energy consumption side electric load power constraints, VPP system power balance constraints, and VPP system thermal power balance constraints.

[0039] VPPO energy purchase and sales price constraints: , Where, 、 、 、 are the prices of electricity and heat sold to energy users at time t, and the prices of electricity and heat purchased from suppliers at time t; 、 are the electricity purchase price and on-grid electricity price at time t respectively; and are the minimum heating price and the maximum heating price, and are the highest average prices of electricity and heat sales respectively; T is the scheduling period.

[0040] ESA electrical / thermal output power constraints, internal constraints of the CHP unit, and ramp constraints of the micro gas turbine and gas boiler outputs: , Where, and They represent the electrical output power and thermal output power of the ESA energy supply side at time t, and is the electrical and thermal power output of the cogeneration unit at time t, is the thermal power output of the gas boiler at time t, and are the output power of photovoltaic and wind power at time t respectively; is the electric power output by the low-temperature waste heat device at time t; 、 are the electrical power and thermal power output by the micro gas turbine at time t, respectively; is the heat conversion efficiency of the waste heat boiler; is the calorific value of natural gas; is the natural gas consumption at time t; is the power generation efficiency of the low-temperature waste heat power generation device; and are the proportions of waste heat generated by the micro gas turbine allocated to the waste heat power generation device and the waste heat boiler at time t; and They are the power generation efficiency and heating efficiency of the micro gas turbine respectively.

[0041] Output constraints and ramp constraints for ESA microturbines and gas boilers: , Where, 、 are the electrical power output of the micro gas turbine and the thermal power output of the gas boiler at time t, and are the minimum and maximum output electrical power of the micro gas turbine respectively; and They are the lower and upper limits of the gas boiler’s heating power output; and are the lower and upper limits of the MT climb of the micro gas turbine, respectively; and They are the lower and upper limits of GB climbing for gas boilers.

[0042] CSRLA transferable electric load constraints: , Where, is the transferable load at time t, is the upper limit of the transferable electric load at time t; It is the total amount of transferable electric load within the dispatch period T, that is, the total amount of transferable electric load before and after the client demand response is required to remain unchanged.

[0043] CSRLA can reduce heat load constraints: , Where, The heat load can be reduced at time t, It is the upper limit of heat load reduction that can be achieved at time t.

[0044] User-side energy storage constraints in CSRLA: , Where, and are the charging power and discharging power of the energy storage system (ESS) at time t; is the storage power of ESS at the initial moment; is the storage power of ESS at time t; is the energy storage self-loss coefficient, is the charge and discharge loss coefficient; is the maximum energy storage capacity of the ESS; is the state of charge of the ESS at time t; and They are the minimum state of charge and maximum state of charge under ESS safety conditions; and are the maximum charging power and maximum discharging power of ESS respectively.

[0045] The power constraints between the CSRLA user-side demand response aggregator and other participants in the main network are: , Where, for t User-side demand response aggregator load after demand response at the moment, and respectively in t The charging power and discharging power of ESS at time t, is t The power load of the customer demand response aggregator at time t.

[0046] Transferable load in EVA restraint Constraints and curtailable loads constraint: , Where, is t The upper limit of electric vehicle load that can be transferred at any time; is the scheduling period T The total amount of transferable load within; is t The upper limit of load reduction that electric vehicles can achieve at any given time.

[0047] The CTSA carbon balance constraints, internal equipment energy balance constraints, power constraints and creep constraints, P2G preparation and utilization constraints, and CCS and P2G power constraints and ramp constraints are as follows: , Where, is the amount of carbon entering the carbon capture and sequestration equipment at time t; 、 Respectively t The amount of carbon stored in carbon capture equipment and the actual carbon emissions at each moment; is t The amount of carbon used by P2G at time; It is a Boolean variable indicating the start and stop status of the CCS device; is the CCS base power load consumption; is the operating consumption coefficient of CCS; is the density of carbon dioxide; is the amount of natural gas produced by P2G; It is the P2G operation efficiency; is the electrical power consumed by the P2G device; is the calorific value of natural gas; and They are t The working power of CCS equipment and P2G equipment at all times, and They are t -1 The working power of CCS equipment and P2G equipment at the moment, and are the rated power of CCS equipment and P2G equipment respectively; and They are the lower and upper limits of the CCS equipment climb; and They are the lower and upper limits of the P2G device climb, respectively.

[0048] Electric load power constraints on the VPPO energy consumption side: , Where, 、 Respectively t The electricity load on the user side and the electricity load of the electric vehicle aggregator (EVA) at the moment, yes t The electric load of CSRLA at a given moment consists of two parts: the user's electric load and the ESS charging power; for t The sum of the power consumption of the carbon capture equipment and the P2G equipment at any given moment, for t The electrical power consumed by the carbon capture equipment at any given moment, for t The power consumed by the P2G device at a given moment.

[0049] Power balance constraints of the VPP system: , Where, It is the power exchange between the VPP and the grid. A positive value indicates that the VPP's internal power supply is less than the power load, and the VPP needs to purchase power from the grid. A negative value indicates that the power supply is greater than the power load, and the VPP can sell power to the grid for profit. 、 、 respectively in t The output power of photovoltaic power, wind power and combined heat and power generation units at the time.

[0050] Thermodynamic power balance constraints of VPP system: , Where, for t The cogeneration unit outputs thermal power at all times. for t The output thermal power of the gas boiler at any moment, for t The heat load on the electricity side at all times.

[0051] The objective function of each subject in step S7 is as follows: VPPO aims to maximize energy sales revenue and carbon trading benefits while minimizing energy purchase costs and heat interruption penalties. Its expression is: , Where, For the benefit of VPPO, is the income of VPPO from selling energy to energy users at time t, is the cost of purchasing energy from the energy supplier at time t, is the interaction cost between VPPO and external network at time t, are the penalty costs of supply interruption at time t; 、 are the electric load on the electricity consumption side and the electric power output on the energy supply side at time t respectively; , are the heat load on the electricity consumption side and the heat power output on the energy supply side at time t respectively; 、 、 、 are the prices of electricity and heat sold to energy users at time t, and the prices of electricity and heat purchased from suppliers at time t, 、 are the electricity purchase price and on-grid electricity price at time t, is the thermal interruption penalty coefficient.

[0052] The goal of ESA is to maximize energy sales revenue while minimizing fuel costs and carbon trading costs. Its expression is: , Where, For ESA benefits, is the energy sales revenue of the energy supply aggregator at time t, is the fuel cost on the energy supply side at time t, are the carbon transaction costs on the energy supply side at time t; 、 、 is the cost coefficient of the micro gas turbine, 、 、 They are the cost factors of gas boilers; , are the electrical power output of the micro gas turbine and the thermal power output of the gas boiler at time t respectively.

[0053] CSRLA aims to maximize CSRLA consumer satisfaction, and its expression is: , Where, for CSRLA consumer satisfaction; is the residential demand response utility function at the user side at time t; 、 are the cost of energy purchased by the user-side aggregator and the cost of electricity purchased by ESS from the grid at time t, respectively; 、 、 They are customer electricity load, basic residential electricity load and transferable electricity load at time t respectively; 、 are the prices of electricity and heat sold to energy users at time t; 、 、 They are the heat load on the electricity consumption side, the basic heat load of the residential building on the user side and the curtailable load at time t respectively; 、 are the prices of electricity and heat sold by VPPO to energy users at time t, respectively; 、 、 、 is the user's preference coefficient for electrical energy and thermal energy.

[0054] EVA aims to maximize EVA consumer satisfaction, and its expression is: , Where, For EVA consumer satisfaction; is the utility function of electric vehicle users at time t; is the energy purchase cost of electric vehicle users at time t; 、 、 are electric vehicle load, transferable load, and curtailable load at time t respectively; 、 is the electric vehicle user preference coefficient; is the satisfaction penalty coefficient.

[0055] CTSA aims to maximize the benefits of carbon treatment, and its expression is: , Where, For CTSA benefits; 、 、 They are the income from selling natural gas to the natural gas trading market at time t, the benefits of releasing carbon to the carbon market, and the CTSA incentive income; , are the carbon sequestration cost at time t and the cost of electricity purchased from the grid by CTSA; is the amount of natural gas sold to the natural gas trading market at time t: 、 They are the unit natural gas price and the carbon trading price of the day; , are the carbon volume traded with the carbon market and carbon sequestration at time t, respectively; , are the electrical power of CCS equipment and P2G equipment at time t respectively; 、 are the incentive coefficients of the carbon treatment system, are the unit carbon sequestration cost; is the electricity selling price of VPPO to energy users at time t.

[0056] Step S8 includes the following steps: the Stackelberg equilibrium forms a closed loop through iterative feedback between the inner and outer layers. The outer layer uses the Nutcracker Optimization Algorithm (NOA) to globally search for the optimal electricity price strategy, simulating foraging, storage, and retrieval behavior to improve search efficiency. The inner layer uses quadratic programming (QP) to quickly solve the local optimal response of the aggregator. That is, the inner layer solves the optimal price under the game of each entity within the VPP and transmits the optimal price model to the upper layer. After receiving the optimal price model, the upper layer iterates the optimal equipment output.

[0057] Example 2 This embodiment also involves a virtual power plant scheduling method that considers source-load coupling and multi-agent game theory. The example data comes from a community-level virtual power plant (VPP), which includes photovoltaic, wind power, combined heat and power (CHP) units, an energy storage system (ESS), customer-side residential power, thermal loads, electric vehicles, and a carbon treatment system. The VPP's optimized scheduling cycle for the integrated energy multi-aggregator is 24 hours, with an optimization step of 1 hour. For electric vehicle aggregators, this application assumes demand-responsive electric vehicles, including private cars, taxis, and buses.

[0058] At the same time, in order to verify the effectiveness of the proposed VPP master-slave game optimization scheduling strategy, this application sets up several scenarios for comparison: Scenario 1: Aggregation is not considered.

[0059] Scenario 2: Customer-side residential demand response and electric vehicle demand response are not considered.

[0060] Scenario 3: CTSA is not considered.

[0061] Scenario 4: No customer-side residential demand response, no electric vehicle demand response, and no CTSA.

[0062] Scenario 5: Aggregation, customer-side residential demand response, and electric vehicle demand response are not considered.

[0063] Scenario 6: No aggregation and CTSA.

[0064] Scenario 7: Scenario proposed in this application.

[0065] The optimization results of each scenario are shown in Table 1 below, where the data has been dimensionless.

[0066] Table 1

[0067] As can be seen from Table 1, the virtual power plant scheduling method of this application limits unilateral adjustments to obtain additional benefits, so that the interests of various entities are dynamically balanced. Through collaboration among multiple entities, coordination among various entities is achieved to maximize their interests.

[0068] like Figure 2 The figure shows the optimized dispatch results of power and heat from a VPP in a typical scenario, demonstrating the dynamic regulation capabilities of the constructed game optimization model in terms of multi-energy synergy and load response. The dispatch results demonstrate that, under complex conditions such as fluctuating renewable energy output and varying electric vehicle charging demand, the system achieves balanced and complementary power and heat supply and demand by coordinating multiple energy entities, including wind power (WT), photovoltaics (PV), energy storage (ESS), combined heat and power (CHP), motors (MG), and customer-side demand response (CSRL). This model fully demonstrates how it effectively improves energy efficiency and the synergy benefits among multiple entities while ensuring stable system operation, demonstrating its scalability and practical application value.

[0069] like Figure 3 The results of the VPP's optimized power and heat dispatch are shown, demonstrating the dynamic adjustment capabilities of the constructed game optimization algorithm in terms of multi-energy coordination and load response. The dispatch results show that under complex conditions such as fluctuations in renewable energy output, the system achieves a balance and complementarity between power and heat supply and demand by coordinating multiple energy entities, including combined power units (CHP), GB, and CSRLA. Specifically, CHP effectively supports the system's thermal demand by providing stable thermal output; the gas-fired boiler GB provides flexible power regulation when the system load fluctuates; and the CSRLA plays a key role in balancing load demand and power generation capacity.

[0070] Figure 4 This paper demonstrates the pricing strategy for the VPPO for electricity and heat purchase and sale within the Stackelberg master-slave game framework using the NOA-QP optimization algorithm. The fluctuation trend of the sales price closely mirrors the local grid's hourly electricity price, while the purchase price aligns with load fluctuations. To incentivize energy suppliers to generate power, reduce the VPP's dependence on the main grid, and increase revenue, the VPPO offers higher electricity purchase prices between 9:00 AM and 5:00 PM and 7:00 PM and 11:00 PM.

[0071] Table 2 lists the parameters and corresponding benefits of four optimization algorithms (GA, SMA, GTO, and NOA). The data has been dimensionlessly processed. GA converged the slowest, converging to -62,378.73 yuan after 400 iterations and -54,9076 yuan after 500 iterations. GTO converged to -41,770.71 yuan and 38,777.22 yuan after 134 and 200 iterations, respectively. The NOA solution method used in this application converged the fastest and achieved the best results.

[0072] Table 2

[0073] Example 3 If the above functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program code, such as USB flash drives, mobile hard drives, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0074] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and such modifications or substitutions are intended to be within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be subject to the scope of protection of the claims.

Claims

1. A virtual power plant scheduling method considering source-load coupling and multi-agent game, characterized in that: include: Establish a multi-temporal and multi-space scale model of the multi-energy system in the virtual power plant, where the multi-energy system includes the energy supply system, the multi-energy coupling system and the energy storage system; Quantify source charge uncertainty; Based on the quantification of source-load uncertainty and a multi-spatiotemporal scale model, a multi-agent scheduling model based on a master-slave game is constructed. In this multi-agent scheduling model, the VPP operator acts as the leader and regulates the market with dynamic electricity and heat prices. Energy supply aggregators, user-side aggregators, and carbon treatment aggregators act as followers, optimizing unit output, load adjustment, and carbon treatment strategies to form a Stackelberg equilibrium. The master-slave game is a multi-stage dynamic game. VPP operators optimize electricity pricing strategies through iterative learning, while followers adjust their responses based on real-time strategies and transmit strategies through fuzzy quotes and interval responses. The NOA-QP hybrid optimization algorithm is used to solve the multi-agent scheduling model and obtain the optimization strategy for virtual power plant scheduling.

2. A virtual power plant scheduling method considering source-load coupling and multi-agent game according to claim 1, characterized in that: When the multi-agent scheduling model reaches Stackelberg equilibrium, the following equation is satisfied: ， Where, 、 、 、 、 They are respectively expressed as VPP operator revenue, energy supply aggregator revenue, user-side residential load aggregator consumer satisfaction, electric vehicle aggregator consumer satisfaction, and carbon treatment aggregator revenue; They are respectively represented as the VPP operator strategy, energy supply aggregator strategy, user-side residential load aggregator strategy, electric vehicle aggregator strategy, and carbon treatment aggregator strategy under the Stackelberg equilibrium of each entity; They are respectively represented as VPP operator strategy, energy supply aggregator strategy, user-side residential load aggregator strategy, electric vehicle aggregator strategy, and carbon treatment aggregator strategy.

3. A virtual power plant scheduling method considering source-load coupling and multi-agent game according to claim 1, characterized in that: The objective functions of each subject in the multi-agent scheduling model are designed, specifically including: the VPP operator aims to maximize energy sales revenue and carbon trading income while minimizing energy purchase costs and heat interruption penalties; the energy supply aggregator aims to maximize energy sales revenue while minimizing fuel costs and carbon trading costs; the user-side aggregator aims to maximize consumer satisfaction; and the carbon treatment aggregator aims to maximize carbon treatment income.

4. A virtual power plant scheduling method considering source-load coupling and multi-agent game according to claim 3, characterized in that: The objective function expression of the VPP operator is: ， Where, For the benefit of VPP operators; The revenue of the VPP operator from selling energy to the energy user at time t; is the cost of purchasing energy from the energy supplier at time t; is the interaction cost between the VPP operator and the external network at time t; are the penalty costs of supply interruption at time t; 、 are the electric load on the electricity consumption side and the electric power output on the energy supply side at time t respectively; , are the heat load on the electricity consumption side and the heat power output on the energy supply side at time t respectively; 、 、 、 are the prices of electricity and heat sold to energy users at time t, and the prices of electricity and heat purchased from suppliers at time t; 、 are the electricity purchase price and on-grid electricity price at time t respectively; is the thermal interruption penalty coefficient, T is the scheduling period; The objective function expression of the energy supply aggregator is: ， Where, Profits for energy supply aggregators, is the energy sales revenue of the energy supply aggregator at time t, is the fuel cost on the energy supply side at time t, are the carbon transaction costs on the energy supply side at time t; 、 、 is the cost coefficient of the micro gas turbine, 、 、 They are the cost factors of gas boilers; , are the electrical power output of the micro gas turbine and the thermal power output of the gas boiler at time t respectively; The objective function of the carbon treatment aggregator is specifically: ， Where, For CTSA benefits; 、 、 They are the income from selling natural gas to the natural gas trading market at time t, the benefits of releasing carbon to the carbon market, and the CTSA incentive income; , are the carbon sequestration cost at time t and the cost of electricity purchased from the grid by CTSA; is the amount of natural gas sold to the natural gas trading market at time t: 、 They are the unit natural gas price and the carbon trading price of the day; , are the carbon volume traded with the carbon market and carbon sequestration at time t, respectively; , are the electrical power of CCS equipment and P2G equipment at time t respectively; 、 are the incentive coefficients of the carbon treatment system, are the unit carbon sequestration cost; is the electricity selling price of VPPO to energy users at time t.

5. A virtual power plant scheduling method considering source-load coupling and multi-agent game according to claim 3, characterized in that: The user-side aggregators include user-side residential load aggregators and electric vehicle aggregators. The user-side residential load aggregator aims to maximize consumer satisfaction and forms a two-way game with the VPP operator's electricity price strategy. The response sensitivity is quantified by the load elasticity coefficient. The expression for maximizing consumer satisfaction is: ， Where, Consumer satisfaction for user-side residential load aggregators; is the residential demand response utility function at the user side at time t; 、 are the cost of energy purchased by the user-side residential load aggregator at time t and the cost of electricity purchased by the energy storage system from the grid; 、 、 They are customer electricity load, basic residential electricity load and transferable electricity load at time t respectively; 、 are the prices of electricity and heat sold to energy users at time t; 、 、 They are the heat load on the electricity consumption side, the basic heat load of the residential building on the user side and the curtailable load at time t respectively; 、 are the prices of electricity and heat sold by the VPP operator to the energy user at time t, respectively; 、 、 、 is the user's preference coefficient for electrical energy and thermal energy, T is the scheduling period; The electric vehicle aggregator aims to maximize consumer satisfaction, which is expressed as: ， Where, For EVA consumer satisfaction; is the utility function of electric vehicle users at time t; is the energy purchase cost of electric vehicle users at time t; 、 、 are electric vehicle load, transferable load, and curtailable load at time t respectively; 、 is the electric vehicle user preference coefficient; is the satisfaction penalty coefficient.

6. A virtual power plant scheduling method considering source-load coupling and multi-agent game according to claim 1, characterized in that: Designing physical constraints for multi-energy coordinated scheduling for the multi-agent scheduling model, including unit output constraints, ramp constraints, electric power balance constraints, thermal power balance constraints, and carbon flow constraints; Among them, the power balance constraint requires that the output of wind, solar, CHP, and energy storage meet the user-side power load, electric vehicle charging load, and carbon treatment system power demand; the thermal power balance constraint requires that the CHP waste heat be matched with the heating of the gas boiler to reduce the user-side heat load; Carbon flow constraints require that actual carbon emissions must not exceed the sum of free allowances and carbon processing volume, including carbon balance constraints for carbon processing aggregators, energy balance constraints for internal equipment, power constraints and creep constraints, constraints for power-to-gas production and utilization, and power constraints and ramp constraints for carbon capture and storage and power-to-gas as follows: ， Where, is t The amount of carbon entering the carbon capture and sequestration equipment at any given moment; 、 Respectively t The amount of carbon stored in carbon capture equipment and the actual carbon emissions at each moment; is t The amount of carbon utilized in power-to-gas conversion at each moment; is a Boolean variable indicating the start and stop status of the carbon capture and storage equipment; is the carbon capture and storage baseload electricity consumption; is the operational consumption factor for CCS; is the density of carbon dioxide; is the amount of natural gas produced through power-to-gas conversion; is the power-to-gas operating efficiency; is the electrical power consumed by the power-to-gas equipment; is the calorific value of natural gas; and are the operating powers of the carbon capture and storage equipment and the power-to-gas equipment at time t, and are the rated power of the carbon capture and storage equipment and the power-to-gas equipment respectively; and These are the lower and upper limits of the carbon capture and storage equipment ramp-up, respectively; and They are the lower and upper limits of the power-to-gas equipment climbing.

7. A virtual power plant scheduling method considering source-load coupling and multi-agent game according to claim 1, characterized in that: The NOA-QP hybrid optimization algorithm, which combines the Nutcracker optimization algorithm with the quadratic programming algorithm, is used to solve the multi-agent scheduling model. The Nutcracker optimization algorithm is used to globally search for the optimal electricity price strategy, and the quadratic programming algorithm is used to efficiently solve the objective function of each aggregator. That is, the inner layer solves the optimal price under the game of each agent within the virtual power plant, and passes the optimal price model to the upper layer. After receiving the optimal price model, the upper layer iterates the optimal equipment output.

8. The virtual power plant scheduling method considering source-load coupling and multi-agent game according to claim 1 is characterized in that: The method also includes user behavior and demand response modeling, that is, quantifying the flexibility potential of user-side electricity and heat loads, formulating the total conservation rules of transferable loads and the elastic range of load reduction.

9. A virtual power plant scheduling method considering source-load coupling and multi-agent game according to claim 8, characterized in that: For electric vehicle clusters, we simulate the differences in network access time, mileage, and charging modes based on statistical methods, and establish fast-charging and slow-charging load curves to characterize the impact of flexibility on the system. The corresponding model is: ， Where, The end time of the last trip, that is, the time when the electric vehicle joined the network; is the expected value of charging time; is the standard deviation of charging time; ， Where, Indicates the power consumption of electric vehicles for every 100 kilometers traveled. 、 、 They represent the output power, charging efficiency and delay parameters of the charging pile respectively.

10. A virtual power plant scheduling method considering source-load coupling and multi-agent game according to claim 1, characterized in that: The process of quantifying source-load uncertainty includes: generating multiple wind and solar power output and load scenarios based on Monte Carlo simulation, reducing redundant scenarios through Manhattan distance metric, retaining typical scenarios of typical photovoltaic output, typical wind power output, user-side electrical load, user-side thermal load, and electricity demand of electric vehicle clusters that are most representative of load and new energy fluctuations throughout the year, and obtaining a typical scenario set.

Images