A virtual power plant capacity allocation optimization method considering multiple trading markets

Abstract

The present invention belongs to the field of power system market transactions and power supply optimization configuration, and discloses a virtual power plant capacity configuration optimization method considering multiple trading markets. It divides virtual power plants into SEVPP and DRVPP according to their functionality, and models the resources they contain in turn; through the "electricity-carbon-green" joint trading market mechanism and the demand response mechanism of DRVPP participating in market transactions, a master-slave game model is established, and the optimal capacity configuration plan of SEVPP and the optimal response power plan of DRVPP are obtained by solving; the capacity proportion of different investors within the virtual power plant is allocated to SEVPP and DRVPP to achieve the optimal capacity configuration plan for each investor. The present invention can simultaneously improve the benefits of SEVPP and DRVPP, effectively increase the enthusiasm of various investors to participate in the aggregation of virtual power plants, and further tap the potential of virtual power plants for energy conservation and emission reduction.

Description

Technical Field

[0001] The present invention belongs to the field of power system market transactions and power supply optimization configuration, and specifically relates to a virtual power plant capacity configuration optimization method considering multiple transaction markets. Background Art

[0002] With the large-scale deployment of distributed power sources, the power system is transitioning toward a new power system with a high proportion of renewable energy generation. However, when these scattered distributed renewable energy sources attempt to integrate into the larger power grid, their capacity is often too small to meet the grid connection requirements. Therefore, the concept of virtual power plants has emerged. With the development of advanced measurement and communication technologies, virtual power plants can centrally dispatch and manage large-scale distributed power sources. Furthermore, by aggregating a sufficiently large number of distributed power sources, virtual power plants can serve as market participants and compete with traditional power plants in day-ahead and intraday power market bidding. At the same time, to achieve the dual carbon goals, the power industry is undergoing a campaign to reduce energy consumption and emissions. The aggregation characteristics of virtual power plants demonstrate significant potential for energy conservation and emission reduction. The renewable energy generation equipment aggregated within them can not only participate in the carbon trading market, but also participate in the green certificate trading market, adding a strong impetus to the achievement of the dual carbon goals.

[0003] Currently, virtual power plants can be divided into two categories according to the type of aggregated resources: sell energy virtual power plants (SEVPPs) and demand response virtual power plants (DRVPPs). These two types of virtual power plants have obvious complementary characteristics when participating in the "electricity-carbon-green" joint trading market. However, too little or too much capacity configuration in the virtual power plant capacity configuration will lead to reduced efficiency of the virtual power plant. In this regard, the existing technology lacks an effective method for optimal capacity configuration for the two types of virtual power plants, making it difficult to maximize the interests of both parties and achieve a win-win cooperation effect. Summary of the Invention

[0004] To solve the above problems, the present invention provides a virtual power plant capacity configuration optimization method considering multiple trading markets. By modeling the power output of the internal resources of SEVPP and DRVPP, as well as a reasonable "electricity-carbon-green" joint trading market mechanism and a demand response mechanism for DRVPP to participate in market transactions, the specific optimal capacity configuration of the aggregated resources within SEVPP and DRVPP is finally obtained, so as to maximize the benefits when the virtual power plant is in operation.

[0005] The present invention provides a method for optimizing virtual power plant capacity configuration considering multiple trading markets, comprising the following steps:

[0006] Step 1: Based on their functionality, virtual power plants are divided into two categories: sell energy virtual power plants (SEVPPs) and demand response virtual power plants (DRVPPs). Output power models are then performed for wind power investors, photovoltaic investors, thermal power investors, energy storage investors, and load aggregators in SEVPPs, as well as load aggregators in DRVPPs. Load aggregators in DRVPPs are divided into Class A load aggregators (shiftable and non-interruptible load aggregators), Class B load aggregators (interruptible and non-shiftable load aggregators), and Class C load aggregators (interruptible and shiftable load aggregators) based on their response time and response depth.

[0007] Step 2: Based on the established output model, a joint trading mechanism that considers electricity market trading, carbon trading, and green certificate trading, as well as a stepped incentive demand response mechanism, is proposed, in which both parties use the response price and the response power as decision variables in the game.

[0008] Step 3: SEVPP and DRVPP, as two market entities, participate in the bidding for day-ahead load demand. A game model is established between the two parties. Using the master-slave game method in game theory, a master-slave game model is established with SEVPP as the leader and DRVPP as the follower, using the response price and response power as the game quantities. A distributed algorithm is used to solve the model until a Nash equilibrium solution is obtained, i.e., the optimal bid amounts of SEVPP and DRVPP.

[0009] Step 4: The three types of load aggregators within the DRVPP all aim to maximize their own benefits, with the total demand-side load response power as a constraint. A non-cooperative game model is constructed using a non-cooperative game approach, and the model is solved using a heuristic intelligent algorithm until a Nash equilibrium solution, i.e., the optimal response power of each type of load aggregator, is obtained.

[0010] Step 5: The internal aggregated resources of SEVPP include photovoltaic battery units, wind turbines, thermal power units, and energy storage battery units. The improved SHAPLEY value method in cooperative game is used to establish a cooperative alliance for the internal investors of the virtual power plant to distribute benefits and establish a cooperative game model. The particle swarm algorithm is used to solve the model until the Nash equilibrium solution is obtained to determine the optimal capacity configuration plan required for each type of unit.

[0011] Furthermore, in step 1, the day-ahead load demand model is determined as follows:

[0012] The day-ahead load forecast published by the power market needs to maintain a supply-demand balance with the bid volume of virtual power plants and the response volume of demand response load aggregators:

[0013] P Forecasting =P VPP +P DR (1),

[0014]

[0015]

[0016] Where: P Forecasting is the day-ahead load forecast electricity, P VPP The amount of electricity bid for SEVPP, P DR is the DRVPP participation response power, The wind turbines participate in the response of electricity. The amount of electricity that the photovoltaic unit participates in responding to, For the energy storage unit to participate in the response power, The amount of electricity that thermal power units participate in responding to the The response power of Class A load aggregator, The response power of Class B load aggregator, The response capacity of Class C load aggregator.

[0017] Furthermore, since various distributed resources belong to different investors, the internal resource modeling in SEVPP is as follows:

[0018] Photovoltaic cells utilize the photoelectric effect to achieve photoelectric conversion, so solar radiation intensity is the primary factor affecting the output of a photovoltaic unit. Multiple photovoltaic cells are connected in series and parallel to form a photovoltaic unit. Light intensity and ambient temperature both affect the output of photovoltaic cells. The output power equation is as follows:

[0019]

[0020] Where, is the output power of the photovoltaic unit at time t, N PV is the number of integrated photovoltaic cells, S(t) is the actual solar radiation intensity at time t, S ref As a reference to solar radiation intensity, 1000W / m is generally taken. 2 , P st is the output power of the photovoltaic array at rated state, T q (t) is the actual operating point temperature at time t, T st is the temperature under rated conditions.

[0021] The output constraints of the photovoltaic system are as follows:

[0022]

[0023] Where, is the maximum installed capacity of the photovoltaic unit;

[0024] Wind turbines are set to different wind speed levels according to wind speed, including cut-in wind speed, cut-out wind speed, and rated wind speed. When the input wind speed is lower than the cut-in wind speed or higher than the cut-out wind speed, the controller controls the wind turbine output to 0 based on the received wind speed information. When the wind speed is between the cut-in wind speed and the rated wind speed, the controller controls the wind turbine output to be positively correlated with the wind speed. When the wind speed is between the rated wind speed and the cut-out wind speed, the controller sets the wind turbine system to a constant rated power output. The mathematical model of wind turbine output is as follows:

[0025]

[0026] Where, is the output power of the wind turbine at time t, v ci 、v co 、v rated , are cut-in and cut-out wind speeds, rated wind speed, v(t) is the wind speed at time t, P rated is the rated output power, k1, k2, k3 are the corresponding system parameters;

[0027] The output constraints of wind turbines are as follows:

[0028]

[0029] Where, is the maximum installed capacity of the wind turbine;

[0030] Energy storage devices can effectively suppress the randomness and volatility of renewable energy generation and increase the utilization rate of renewable energy. The State of Charge (SOC) represents the remaining energy capacity of the battery and is an important battery indicator. During charging, the SOC value increases continuously; during discharge, the SOC value decreases continuously. The SOC value is closely related to the charge and discharge power. The charging process relationship is shown in the following equation:

[0031] SOC SC (t) = (1-ε)SOC(t-1)-P BS.c Δtη c / E BS (8)

[0032] The discharge process is as follows:

[0033] SOC SD (t) = (1-ε)SOC(t-1)-P BS.d Δt / (E BS η d ) (9)

[0034]

[0035] Where, SOC SC (t), SOC SD SOC(t) and SOC(t-1) are the charging and discharging processes at time t and the remaining battery energy at time t-1 respectively; ε is the self-discharge rate, which indicates the amount of electricity lost by the energy storage battery itself; E BS is the capacity of the battery; P BS.c and P BS.d Represent the charging and discharging power of the battery, taking negative and positive values ​​respectively; η c ,η d They represent the charging and discharging efficiency of energy storage respectively. Generally, the charging efficiency is 0.6-1, and the discharging efficiency is 0.8-1.

[0036] Thermal power units rely mainly on micro gas turbines for power generation. Although a certain amount of pollutants are still released during the power generation process, the pollution level is greatly reduced compared to traditional coal-fired power generation, and the output power is adjustable. Its output model is shown below:

[0037]

[0038] Where, is the output power of the thermal power unit at time t, η CN is the conversion efficiency of the controller, η G is the power generation efficiency of the generator, P T is the output power of the working element, P C is the power consumed by the compressor, P FC is the power consumed by the fuel compressor, P CF The power consumed by the cooling fan.

[0039] The output constraints of thermal power units are as follows:

[0040]

[0041] Where, is the minimum output power of the thermal power unit, It is the maximum output power of thermal power unit.

[0042] Further modeling of the load aggregator in DRVPP:

[0043] Class A load aggregators (shiftable and non-interruptible load aggregators) aggregate resources such as household washing machines;

[0044] Class B load aggregators (interruptible but non-shiftable load aggregators) aggregate resources including building air conditioners, etc.

[0045] Class C load aggregators (interruptible and shiftable load aggregators), whose aggregated resources include electric vehicles, etc.

[0046] Therefore, the demand response load aggregator's response power is as follows:

[0047]

[0048] Where, is the amount of electricity actually participated in demand response by the load aggregator of type j at time t, and Δt is the number of hours participating in demand response in a day;

[0049] The response constraints of the load aggregator are as follows:

[0050]

[0051] Where, is the maximum response power of class j load aggregator.

[0052] Furthermore, in step 2, a “electricity-carbon-green” joint trading market mechanism is proposed, and the SEVPP participation in the carbon trading market and the green certificate trading market is modeled.

[0053] Currently, most carbon trading markets issue initial carbon allowances based on historical power generation. If a power generator's carbon emissions exceed its carbon allowance, it must purchase the excess allowances in the carbon trading market. Otherwise, it can sell its excess allowances. Therefore, the SEVPP carbon trading model in the carbon trading market is as follows:

[0054]

[0055]

[0056] λ C,sell ≤λ C,buy (17)

[0057] Where U VPP,C C is the benefit obtained by SEVPP from selling carbon quotas in the carbon trading market. VPP,C The cost of purchasing carbon quotas for SEVPP to participate in the carbon trading market, λ C,sell ,λ C,buy are the selling price and purchasing price of carbon quotas in the carbon trading market, The initial amount of carbon allowances issued by SEVPP based on historical power generation.

[0058] The green certificate trading market is an important means of encouraging the expansion of renewable energy capacity. It also offers a new way for renewable energy generators to further increase their profits after the wave of renewable energy subsidies has faded. Its core trading mechanism is that the government issues green certificates based on the amount of electricity generated by renewable energy units. Each green certificate can be sold to buyers in the green certificate trading market, who are typically traditional thermal power generators. Therefore, the model for SEVPP participation in the green certificate trading market is as follows:

[0059]

[0060] Where U VPP,G is the benefit obtained by SEVPP through green certificate trading, G It is the clearing price for green certificate transactions, and the price range is usually {0-800} / MW·h.

[0061] To increase the enthusiasm of load aggregators, a tiered incentive mechanism for calculating demand response subsidies is proposed. Different subsidy price discounts are determined based on the ratio of the load aggregator's response power to the day-ahead bid power. The greater the deviation between the user's response power and the day-ahead bid power, the lower the tiered subsidy unit price.

[0062]

[0063] Where: DR j is the demand response subsidy price for load aggregator j, λ j is the incentive coefficient of load aggregator j, ρ clr is the clearing price; P DR,j is the actual response power of load aggregator j, P exp,j is the expected response power of load aggregator j; y i is the demand response incentive coefficient for the i-th step interval, δ i-1 is the left boundary of the i-th step interval, δ i is the right boundary of the i-th step interval; δ0 is the expected compliance ratio, δ k is the expected capping ratio, where k is the number of capping steps. In China, δ0 is generally taken as 0.8, δ k =1.2.

[0064] The closer the ratio of actual demand-side user power consumption to expected power consumption is to 100%, the greater the step incentive coefficient will be. However, to prevent users from participating in demand-side response in an disorderly manner, a cap on the number of steps is set to avoid irrational user behavior. To improve the efficiency of the mechanism settlement, the step demand response incentive coefficient is designed to be symmetrical with 100% as the center. The discount coefficient for positive and negative deviations between the actual user response and expected power generation is the same. Therefore, the step incentive coefficient meets the following constraints:

[0065] (y i -y i+1 )(δ i -100%)≥0 (21)

[0066] Where y i+1 is the demand response incentive coefficient for the i+1th step interval.

[0067] Finally, we establish a master-slave game model. Compared to the classic game model, the master-slave game is a dynamic process. That is, while every participant in the classic game has the same status, the status of participants in the master-slave game is inconsistent, and the strategy choices of followers depend on the strategy choices of the leader.

[0068] The master-slave game method includes three elements: participants, strategy set, and utility function:

[0069] Participants include leaders (SEVPP) and followers (DRVPP). The leader selects a price strategy from the virtual power plant strategy set and publishes it to followers based on the day-ahead load forecast released by the power market and the output information of each unit collected in the virtual power plant management platform.

[0070] The follower adjusts the corresponding response power and electricity price according to the price strategy released by the leader, and selects a response strategy from the demand-side response strategy set to feed back to the leader.

[0071] The leader adjusts its own pricing strategy again based on the response strategy fed back by the demand-side load aggregator and publishes it to the followers, repeating the above steps until the Nash equilibrium is reached.

[0072] The SEVPP policy set is as follows:

[0073]

[0074] Where: Ω st,VPP is the strategy set of the virtual power plant in the master-slave game, is the bidding price of the virtual power plant, is the minimum bidding price of the virtual power plant, is the maximum bid price of the virtual power plant.

[0075] The DRVPP policy set is represented as follows:

[0076]

[0077] Where: Ω st,DR The set of demand-side response strategies in the master-slave game, P DR,t is the response power of DRVPP at time t, is the minimum response power of DRVPP, It is the maximum response power of DRVPP.

[0078] The benefits of SEVPP are mainly obtained through participation in the electricity market, carbon trading market and green certificate market. To maximize its objective function, it is expressed as follows:

[0079]

[0080] Where: U VPP This is the benefit of SEVPP in one day. is the response electricity price at time t, P VPP,t is the bid electricity of the virtual power plant at time t, and the cost includes the cost of the virtual power plant to purchase carbon quotas C VPP,C , average daily initial investment cost C VPP,inv , average daily operation and maintenance cost C VPP,mat .

[0081] The utility function for maximizing DR benefits is as follows:

[0082]

[0083] Where: U DR The benefit of DRVPP per day, C j,loss It is the subsidy cost for the implicit losses caused by the participation of different types of load aggregators in demand response.

[0084] According to the above formula, the Nash equilibrium solution is:

[0085]

[0086]

[0087] Where, is the optimal response electricity price at time t, P DR,t * is the optimal response power at time t.

[0088] When neither party can improve their own benefits by changing the response price and response power, a Nash equilibrium is reached, and the optimal bid power of SEVPP and the optimal response power of DRVPP can be obtained.

[0089] The master-slave game model uses a distributed algorithm. The specific steps are as follows:

[0090] (1) Input raw data, including but not limited to annual wind speed, light intensity, temperature, load and equipment parameters.

[0091] (2) Leaders release initial pricing strategy To the followers, the followers will feedback the corresponding power response strategy P according to the price strategy released by the leader DR,t To the leader.

[0092] (3) Perform iterative solutions and dynamic game to achieve independent optimization of each participant's strategy.

[0093] (4) Determine whether the current solution meets the conditions of Nash equilibrium. If not, return to step 3 to continue iterating. If it does, the solution is considered is the Nash equilibrium solution.

[0094] After obtaining the optimal bidding power of SEVPP and the optimal response power of DRVPP, we enter the next stage of capacity configuration and propose a load aggregator capacity allocation method based on non-cooperative game.

[0095] Since the demand response subsidy price of load aggregators is jointly determined by the three types of load aggregators, the benefits of a single load aggregator depend not only on the amount of its own bid for demand response, but also on the strategies of other load aggregators. Based on this, the non-cooperative game model of load aggregators is as follows:

[0096] (1) Participants: A, B, and C load aggregators

[0097] (2) Strategy set:

[0098]

[0099] Where: Ω nc,DR,j is the strategy set of non-cooperative demand response load aggregator j, is the bid amount of demand response load aggregator j at time t, is the minimum bid amount of demand response load aggregator j, is the maximum bid amount of demand response load aggregator j.

[0100] (3) The utility function is as follows:

[0101]

[0102] Various load aggregators maximize their own benefits by changing their bidding amounts until any load aggregator

[0103] If the bidder changes its bidding strategy, its own benefits will not increase, and the Nash equilibrium solution is:

[0104]

[0105] Where, is the optimal bid amount of demand response load aggregator j at time t.

[0106] The heuristic intelligent algorithm is used to solve the non-cooperative game model. The steps are as follows:

[0107] (1): Input system-related parameters, including but not limited to annual wind speed, light intensity, temperature, load and equipment parameters, and also take the optimal response quantity of demand response in the master-slave game as one of the input conditions.

[0108] (2): Initialize the strategy set of each load aggregator and randomly initialize the capacity configuration within the constraints.

[0109] (3): Each load aggregator's response power is optimized separately. Based on the strategy set of other load aggregators in the previous round, each load aggregator optimizes its own optimal capacity configuration plan for this round with the goal of maximizing its own benefits.

[0110] (4): Determine whether a Nash equilibrium solution is obtained. If so, output the optimal response capacity of each load aggregator. If not, return to (3) and continue iterative optimization until the optimal capacity configuration plan for each load aggregator is obtained.

[0111] Furthermore, after the optimal bidding power of SEVPP is obtained, the next stage of capacity configuration is entered, and a virtual power plant internal capacity allocation method based on the improved SHAPLEY value method based on cooperative game is proposed.

[0112] The allocation scheme based on the Shapley value method has been proven to satisfy both individual rationality and overall rationality, and is a more reasonable method for solving the problem of benefit allocation in multi-person cooperation. The Shapley value method allocates benefits based on the contribution of participants in the alliance. The benefit allocated to participant i from the alliance Ψ is recorded as x i (Ψ):

[0113]

[0114]

[0115] Where: x i (Ψ) is the benefit obtained after the alliance member i is allocated, I(S\i) is the benefit after the alliance S removes member i, is the probability of alliance S occurring, also known as the weighting factor.

[0116] The classic SHAPLEY value method has the following defects:

[0117] (1) Contribution rate is the only basis for members to participate in benefit distribution.

[0118] (2) Absolute benefit is the only criterion for measuring contribution.

[0119] (3) The Shapley value method does not take into account the risk factors and actual contributions of the participants.

[0120] Based on this, this method introduces risk coefficient and contribution coefficient to improve the classic SHAPLEY value method, and further refines it by combining the analytic hierarchy process and entropy weight method.

[0121] (1) Risk coefficient indicator

[0122] Since the resources aggregated by virtual power plants are very different, different investors face different levels of risk in actual operation. The SHAPLEY value method assumes that all participants share the risk. Therefore, it is considered to introduce a risk factor to improve the SHAPLEY value method.

[0123] In order to accurately assess the risks borne by each investor in the actual operation of a virtual power plant, this paper mainly constructs a comprehensive indicator system for comprehensively evaluating the actual risk rates borne by different investors from the two aspects of explicit risks and implicit risks.

[0124] Table 1 Investor risk assessment indicators

[0125]

[0126] Based on the different actual risk coefficients borne by different investors, the profit distribution model improved based on the risk coefficient is as follows:

[0127]

[0128] Where: R k is the actual risk borne by the kth investor, n is the total number of n different investors, In order to comprehensively consider the risk coefficient after explicit risk and implicit risk, is the benefit of the kth investor before risk assessment improvement, is the benefit of the kth investor obtained based on the modified model after risk assessment;

[0129] (2) Contribution coefficient

[0130] Since the resources aggregated by virtual power plants are very different, the actual contributions made by different investors in actual operation are different. Therefore, it is considered to introduce a contribution coefficient to improve the SHAPLEY value method.

[0131] In order to accurately measure the contribution of each investor in the actual operation of the virtual power plant, the present invention mainly constructs a comprehensive indicator system for comprehensively evaluating the actual contribution rate of different investors from two aspects: explicit contribution and implicit contribution, as shown in Table 2:

[0132] Table 2 Investor contribution evaluation indicators

[0133]

[0134] Based on the different actual contribution coefficients of different investors, the profit distribution model improved based on the contribution coefficient is as follows:

[0135]

[0136] Where: D k The actual contribution made by the k-th investor, In order to comprehensively consider the contribution coefficient after explicit contribution and implicit contribution, The benefit of the kth investor before the improvement based on contribution evaluation; is the benefit of the kth investor obtained based on the modified model after contribution evaluation.

[0137] The improved model of comprehensive risk coefficient and contribution coefficient is as follows:

[0138]

[0139] Where: The weight coefficients of η and μ should be selected in combination with the actual situation, and the weights should be assigned using the hierarchical analysis method and entropy weight method.

[0140] The cooperative game model in the virtual power plant is as follows:

[0141] Participants: Photovoltaic investors, wind power investors, energy storage investors, thermal power investors

[0142] Policy Collection:

[0143]

[0144] Investor benefit function:

[0145]

[0146] Where, For the benefit of the j-th type of investors, is the bid electricity of type j investor at time t.

[0147] Nash equilibrium solution:

[0148]

[0149] Where, is the optimal amount of electricity required for the photovoltaic unit to respond, is the optimal amount of electricity required for wind turbines to participate in the response, is the optimal amount of energy storage unit to participate in the response, It is the optimal amount of electricity required for thermal power units to participate in the response.

[0150] The particle swarm algorithm is used to solve the cooperative game model. The specific steps are as follows:

[0151] (1): Input raw data, including but not limited to annual wind speed, light intensity, temperature, load and equipment parameters, and also take the optimal bid amount of the virtual power plant in the master-slave game as one of the input conditions.

[0152] (2): Set the initial value of the configuration capacity and randomly assign it to various investors in the virtual power plant.

[0153] (3): Perform iterative solution, adopt the strategy of “eliminating inferior” solution, and continuously iterate to eliminate strictly evil strategies.

[0154] (4): Determine whether the current solution satisfies the Nash equilibrium condition. If not, return to (3) and continue iterative solution. If it does, the solution is considered to be a Nash equilibrium solution, and the optimal capacity configuration plan bid by various investors in the virtual power plant is output.

[0155] The beneficial effects of the present invention are: the method described in the present invention uses the master-slave game method to effectively increase the benefits of virtual power plant investors and load aggregators, while taking into account the "electricity-carbon-green" joint market trading mechanism, further promoting the absorption of new energy power generation, reducing the wind and solar power abandonment rate, and reducing carbon dioxide emissions while ensuring power generation tasks, laying a solid foundation for achieving dual carbon goals; at the same time, the method described in the present invention uses the SHAPLEY method in the improved cooperative game to make the benefit distribution method of different investors within the virtual power plant more fair and reasonable, and uses non-cooperative game to make different types of load aggregators respond to price signals more accurately. BRIEF DESCRIPTION OF THE DRAWINGS

[0156] Figure 1 Framework diagram of the “Electricity-Carbon-Green” joint trading mechanism;

[0157] Figure 2 This is a schematic diagram of a typical daily load curve;

[0158] Figure 3 Iterate the settlement process for virtual power plants and load aggregators;

[0159] Figure 4 Schematic diagram of the impact of carbon trading prices on virtual power plant capacity configuration and wind and solar power curtailment rates;

[0160] Figure 5 Schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0161] In order to make the contents of the present invention more clearly understood, the present invention is further described in detail below based on specific embodiments in conjunction with the accompanying drawings.

[0162] like Figure 5 As shown, the virtual power plant capacity configuration optimization method considering multiple trading markets described in the present invention includes the following steps:

[0163] S1, such as Figure 1 As shown in the figure, virtual power plants are divided into electricity sales virtual power plants (SEVPPs) and demand response virtual power plants (DRVPPs) according to their functionality, and output power modeling is performed on wind power investors, photovoltaic investors, thermal power investors, energy storage investors and load aggregators included in SEVPPs and DRVPPs in turn; load aggregators in DRVPPs are divided into Class A load aggregators (i.e., load aggregators that can be shifted but not interrupted), Class B load aggregators (i.e., load aggregators that can be interrupted but not shifted), and Class C load aggregators (i.e., load aggregators that can be interrupted but shifted) according to their response time and response depth;

[0164] S2. Based on the established output model, a joint trading mechanism that considers electricity market trading, carbon trading, and green certificate trading, as well as a stepped incentive demand response mechanism, is proposed, so that both parties in the game can use the response price and the response quantity as decision variables in the game;

[0165] S3, SEVPP, and DRVPP are two market entities participating in the bidding for day-ahead load demand. A game model is established between the two parties using the master-slave game method in game theory, with SEVPP as the leader and DRVPP as the follower. The master-slave game model is established with response price and response power as the game quantities. A distributed algorithm is used to solve the model until a Nash equilibrium solution is obtained, which is the optimal bid amount for SEVPP and DRVPP.

[0166] The three types of load aggregators in S4 and DRVPP all aim to maximize their own benefits, with the total demand-side load response power as a constraint. They use a non-cooperative game method to build a non-cooperative game model and use a heuristic intelligent algorithm to solve the model until a Nash equilibrium solution is obtained, that is, the optimal response power of each type of load aggregator.

[0167] The internal aggregated resources of S5 and SEVPP include photovoltaic battery units, wind turbines, thermal power units, and energy storage battery units. The improved SHAPLEY value method in cooperative game is used to establish a cooperative alliance for the internal investors of the virtual power plant to distribute benefits and establish a cooperative game model. The particle swarm algorithm is used to solve the model until the Nash equilibrium solution is obtained, which can determine the optimal capacity configuration plan required for each type of unit.

[0168] Taking the capacity configuration of a virtual power plant in a certain region as an example for analysis, the specific system information parameters are shown in Table 3-5 below:

[0169] Table 3 New energy unit system parameters

[0170]

[0171] Table 4 Energy storage unit system parameters

[0172]

[0173] Table 5 Thermal power unit system parameters

[0174]

[0175] The typical daily load curve of the area is as follows: Figure 2 shown.

[0176] To verify the feasibility of this technical solution, the following four scenario cases are used to illustrate:

[0177] Scenario 1: Virtual power plant power supply benefit allocation and sizing without considering carbon market trading, green certificate trading, and demand response;

[0178] Scenario 2: Considering carbon market transactions and green certificate transactions, but not considering demand response, the distribution and sizing of virtual power plant power benefits.

[0179] Scenario 3: Virtual power plant power supply benefit allocation and sizing taking into account demand response, but not carbon market trading or green certificate trading;

[0180] Scenario 4: Virtual power plant power supply benefit distribution and sizing taking into account carbon market trading, green certificate trading and demand response.

[0181] Table 6: Comparison of planning results for each scenario

[0182]

[0183] It can be seen from Table 6 that the capacity planning of virtual power plants after considering carbon trading, green certificate trading and demand response mechanisms has changed mainly in the following aspects.

[0184] (1) The capacity of thermal power units and energy storage units is reduced. This is because before carbon trading, green certificate trading and demand response mechanisms are considered, thermal power units are the only controllable power generation resources within the virtual power plant. The controllable and stable characteristics make them occupy an important position within the virtual power plant. In addition, the uncertainty of renewable energy power generation will cause frequency fluctuations, and thermal power units can alleviate this situation to a certain extent. However, after considering carbon trading, green certificate trading and demand response mechanisms, a large number of thermal power units will cause the virtual power plant to pay additional costs to purchase sufficient carbon quotas. In addition, the demand response mechanism further optimizes the load curve and realizes "peak shaving and valley filling" on the demand side. Therefore, there is no need for a large number of thermal power units and energy storage units to be on standby.

[0185] (2) The proportion of new energy units in the capacity configuration of virtual power plants has increased. On the one hand, more new energy units can bring more carbon quotas to virtual power plants and sell more green certificates. This additional income greatly improves the efficiency of virtual power plants. On the other hand, the complementary characteristics of wind and solar power generation can be further enhanced with the help of demand response mechanisms, further reducing the rate of wind and solar power curtailment. In addition, the cost of new energy power generation is much lower than that of traditional thermal power generation. Therefore, after considering the above factors, virtual power plants prefer to configure more new energy units.

[0186] By comparing Scenario 1 and Scenario 4, it can be seen that after taking into account carbon trading, green certificate trading and demand response mechanisms, the total installed capacity of the virtual power plant is reduced by 40MW while being able to meet load demand, the benefits are increased by 15.524 million yuan, and the actual carbon emissions are reduced by 68.8 tons, further tapping the potential of virtual power plants for low-carbon emission reduction and new energy absorption.

[0187] Scenario 4 is selected for simulation analysis. Figure 3 As can be seen, the VPP's benefit curve reaches its optimal value after 13 iterations, while the load aggregator's optimal value is achieved after 73 iterations. This concludes the master-slave game, as neither the VPP nor the load aggregator can alter their own benefits by changing their respective electricity prices or consumption, resulting in a Nash equilibrium solution.

[0188] Thermal power units, as the largest dispatchable units on the power generation side, are affected by carbon trading prices. Setting the carbon trading price to 0 to 300 yuan / t, the power planning model proposed in this paper is solved, and the planning results are as follows: Figure 4 shown.

[0189] As can be seen, when the carbon trading price was 0 yuan / t, the planned capacity of thermal power units accounted for a high proportion. At this time, the low carbon trading price caused virtual power plants to still rely primarily on relatively stable and controllable thermal power output. However, as the carbon trading price increased, the planned capacity of thermal power units gradually decreased, while the planned capacity of wind power and energy storage units, as well as the demand for load aggregators to participate in DR, gradually increased. When the carbon trading price rose to 300 yuan / t, virtual power plants significantly increased the capacity of renewable energy generators to improve overall profits, thereby increasing the curtailment rate of wind and solar power. At the same time, more energy storage was required to cope with the uncertainty of renewable energy generation.

[0190] The above description is only a preferred embodiment of the present invention and is not intended to further limit the present invention. All equivalent changes made using the contents of the present invention description and drawings are within the scope of protection of the present invention.

Claims

1. A virtual power plant capacity configuration optimization method considering multiple trading markets, characterized in that: The following steps are involved: Step 1: Based on the functionality of virtual power plants, virtual power plants are divided into electricity sales virtual power plants (SEVPPs) and demand response virtual power plants (DRVPPs). Output power models are then performed for wind power investors, photovoltaic investors, thermal power investors, energy storage investors, and load aggregators included in SEVPPs, as well as load aggregators included in DRVPPs. Load aggregators in DRVPPs are divided into Class A load aggregators (i.e., load aggregators that can be shifted and not interrupted), Class B load aggregators (i.e., load aggregators that can be interrupted and not shifted), and Class C load aggregators (i.e., load aggregators that can be interrupted and shifted), based on their response time and response depth. Step 2: Based on the established output model, a joint trading mechanism that considers electricity market trading, carbon trading, and green certificate trading, as well as a stepped incentive demand response mechanism, is proposed, in which both parties use the response price and the response power as decision variables in the game. Step 3: SEVPP and DRVPP, as two market entities, participate in the bidding for day-ahead load demand. A game model is established between the two parties. Using the master-slave game method in game theory, a master-slave game model is established with SEVPP as the leader and DRVPP as the follower, with response price and response power as the game quantities. A distributed algorithm is used to solve the model until a Nash equilibrium solution is obtained, i.e., the optimal bid amounts of SEVPP and DRVPP. Step 4: The three types of load aggregators within the DRVPP all aim to maximize their own benefits, with the total demand-side load response power as a constraint. A non-cooperative game model is constructed using a non-cooperative game approach, and the model is solved using a heuristic intelligent algorithm until a Nash equilibrium solution, i.e., the optimal response power of each type of load aggregator, is obtained. Step 5: Aggregate resources within the SEVPP, including photovoltaic battery units, wind turbines, thermal power units, and energy storage battery units. Utilize the improved SHAPLEY value method in cooperative game theory to establish a cooperative alliance among investors within the virtual power plant to distribute benefits. Establish a cooperative game model and use the particle swarm algorithm to solve the model until a Nash equilibrium solution is obtained, thereby determining the optimal capacity allocation plan for each unit. In step 3, a master-slave game model is established; the master-slave game method includes three elements: participants, strategy set, and utility function: Participants include the leader SEVPP and the follower DRVPP. The leader selects a price strategy from the virtual power plant strategy set and publishes it to the followers based on the day-ahead load forecast released by the power market and the output information of each unit collected in the virtual power plant management platform. The followers adjust the corresponding response power and electricity price according to the price strategy released by the leader, and select a response strategy from the demand-side response strategy set to feed back to the leader. The leader adjusts its own price strategy again based on the response strategy fed back by the demand-side response load aggregator and publishes it to the followers. The above steps are repeated until a Nash equilibrium is reached. The SEVPP policy set is as follows: Where: Ω st,VPP is the strategy set of the virtual power plant in the master-slave game, is the bidding price of the virtual power plant, is the minimum bidding price of the virtual power plant, is the maximum bid price of the virtual power plant; The DRVPP policy set is represented as follows: Where: Ω st,DR The set of demand-side response strategies in the master-slave game, P DR,t is the response power of DRVPP at time t, is the minimum response power of DRVPP, is the maximum response power of DRVPP; The utility function of SEVPP is expressed as follows: Where: U VPP This is the benefit of SEVPP in one day. is the response electricity price at time t, P VPP,t is the bid electricity of the virtual power plant at time t, and the cost includes the cost of the virtual power plant to purchase carbon quotas C VPP,C , average daily initial investment cost C VPP,inv , average daily operation and maintenance cost C VPP,mat ; The utility function of DRVPP is expressed as follows: Where: U DR The benefit of DRVPP per day, C j,loss It is the subsidy cost for the implicit losses caused by the participation of different types of load aggregators in demand response; According to the above formula, the Nash equilibrium solution is: Where, is the optimal response electricity price at time t, P DR,t * is the optimal response power at time t; When both parties cannot improve their own utility functions by changing the response price and response power, Nash equilibrium is reached, that is, the optimal bidding power of SEVPP and the optimal response power of DRVPP are obtained; In step 4, the non-cooperative game model of the load aggregator is as follows: (1) Participants: A, B, and C load aggregators (2) Strategy set: Where: Ω nc,DR,j is the strategy set of non-cooperative demand response load aggregator j, is the bid amount of demand response load aggregator j at time t, is the minimum bid amount of demand response load aggregator j, is the maximum bid amount of demand response load aggregator j; (3) The utility function is as follows: All load aggregators maximize their own benefits by changing their bidding amounts, until any load aggregator changes its bidding strategy without increasing its own utility function, and the Nash equilibrium solution is obtained as follows: Where, is the optimal bid amount of demand response load aggregator j at time t; In step 5, the distribution is based on the contribution of the participants in the alliance. The benefit allocated to participant i from the alliance Ψ is recorded as x i (Ψ): Where: x i (Ψ) is the benefit obtained after the alliance member i is allocated, I(S\i) is the benefit of alliance S after removing member i, is the probability of alliance S occurring, also known as the weighting factor; (1) Based on the different actual risk coefficients borne by different investors, the profit distribution model based on the improved risk coefficient is as follows: Where: R k is the actual risk borne by the kth investor, n is the total number of n different investors, In order to comprehensively consider the risk coefficient after explicit risk and implicit risk, is the benefit of the kth investor before risk assessment improvement, is the benefit of the kth investor obtained based on the modified model after risk assessment; (2) Based on the different actual contribution coefficients of different investors, the improved profit distribution model based on the contribution coefficient is obtained as follows: Where: D k The actual contribution made by the k-th investor, In order to comprehensively consider the contribution coefficient after explicit contribution and implicit contribution, The benefit of the kth investor before the improvement based on contribution evaluation; is the benefit of the kth investor obtained by the modified model based on contribution evaluation; (3) The improved model of comprehensive risk coefficient and contribution coefficient is as follows: Where: The weight coefficients of η and μ should be selected in accordance with the actual situation, and the weights should be assigned using the hierarchical analysis method and entropy weight method; The cooperative game model in the virtual power plant is as follows: Participants: Photovoltaic investors, wind power investors, energy storage investors, thermal power investors Policy Collection: Investor utility function: Where, For the benefit of the j-th type of investors, is the bid electricity of type j investor at time t; Nash equilibrium solution: Where, is the optimal amount of electricity required for the photovoltaic unit to respond, is the optimal amount of electricity required for wind turbines to participate in the response, is the optimal amount of energy storage unit to participate in the response, It is the optimal amount of electricity required for thermal power units to participate in the response.

2. The virtual power plant capacity configuration optimization method considering multiple trading markets according to claim 1 is characterized in that: In step 1, the day-ahead load demand model is determined as follows: The day-ahead load forecast published by the power market needs to maintain a supply-demand balance with the bid volume of virtual power plants and the response volume of demand response load aggregators: P Forecasting =P VPP +P DR (1), Where: P Forecasting is the day-ahead load forecast electricity, P VPP The amount of electricity bid for SEVPP, P DR The amount of DRVPP response power, The wind turbines participate in the response of electricity. The amount of electricity that the photovoltaic unit participates in responding to, For the energy storage unit to participate in the response power, The amount of electricity that thermal power units participate in responding to the The response power of Class A load aggregator, The response power of Class B load aggregator, The response power of Class C load aggregator.

3. The virtual power plant capacity configuration optimization method considering multiple trading markets according to claim 2 is characterized in that: Since various distributed resources belong to different investors, the internal resource modeling in SEVPP is as follows: Multiple photovoltaic cells are connected in series and parallel to form a photovoltaic unit, and its output power equation is as follows: Where, is the output power of the photovoltaic unit at time t, N PV is the number of integrated photovoltaic cells, S(t) is the actual solar radiation intensity at time t, S ref is the reference solar radiation intensity, P st is the output power of the photovoltaic array at rated state, T q (t) is the actual operating point temperature at time t, T st is the temperature under rated conditions; The output constraints of the photovoltaic system are as follows: Where, is the maximum installed capacity of the photovoltaic unit; The mathematical model of wind turbine output power is as follows: Where, is the output power of the wind turbine at time t, v ci 、v co 、v rated , respectively cut-in and cut-out wind speeds, rated wind speed, v(t) is the wind speed at time t, P rated is the rated output power, k1, k2, k3 are the corresponding system parameters; The output constraints of wind turbines are as follows: Where, is the maximum installed capacity of the wind turbine; The state of charge (SOC) represents the remaining capacity of the battery energy. During the charging process, the SOC value increases continuously; the discharging SOC value decreases continuously. The relationship of the charging process is shown in the following formula: SOC SC (t)=(1-ε)SOC(t-1)-P BS.c Δtη c / IN BS (8), The discharge process is as follows: SOCIETY SD (t)=(1-ε)SOC(t-1)-P BS.d Δt / (E BS the d ) (9), Where, SOC SC (t), SOC SD SOC(t) and SOC(t-1) are the battery remaining energy during the charging process, discharging process and t-1 respectively; ε is the self-discharge rate, which indicates the amount of energy lost by the energy storage battery itself; E BS is the capacity of the battery; P BS.c and P BS.d Represent the charging and discharging power of the battery, taking negative and positive values ​​respectively; η c ,η d Respectively represent the charging and discharging efficiency of energy storage; The output model of thermal power units is as follows: Where, is the output power of the thermal power unit at time t, η CN is the conversion efficiency of the controller, η G is the power generation efficiency of the generator, P T is the output power of the working element, P C is the power consumed by the compressor, P FC is the power consumed by the fuel compressor, P CF The power consumed by the cooling fan; The output constraints of thermal power units are as follows: Where, is the minimum output power of the thermal power unit, It is the maximum output power of thermal power unit.

4. The virtual power plant capacity configuration optimization method considering multiple trading markets according to claim 3 is characterized in that: Modeling the load aggregator in DRVPP: The demand response load aggregator's response power is calculated as follows: Where, is the amount of electricity actually participated in demand response by the load aggregator of type j at time t, and Δt is the number of hours participating in demand response in a day; The response constraints of the load aggregator are as follows: Where, is the maximum response power of class j load aggregator.

5. The virtual power plant capacity configuration optimization method considering multiple trading markets according to claim 4 is characterized in that: A joint trading mechanism that considers electricity market trading, carbon trading, and green certificate trading is proposed, and SEVPP's participation in the carbon trading market and the green certificate trading market is modeled. The model for SEVPP to participate in the carbon trading market is as follows: l C,sell ≤λ C,buy (17), Where U VPP,C The benefits of SEVPP participating in the carbon trading market to sell carbon quotas, C VPP,C The cost of purchasing carbon quotas for SEVPP to participate in the carbon trading market, λ C,sell ,λ C,buy are the selling price and purchasing price of carbon quotas in the carbon trading market, The initial amount of carbon allowances issued by SEVPP based on historical power generation; The model of SEVPP participating in the green certificate trading market is as follows: Where U VPP,G For the benefit of SEVPP participating in the green certificate trading market, G is the clearing price of green certificate transactions; A tiered incentive mechanism for calculating demand response subsidies is proposed, which determines different subsidy price discounts based on the ratio of the load aggregator's response power to the day-ahead bid power. DR j =λ j r clr P DR,j (19), Where: DR j is the demand response subsidy price for load aggregator j, λ j is the incentive coefficient of load aggregator j, ρ clr is the clearing price; P DR,j is the actual response power of load aggregator j, P exp,j is the expected response power of load aggregator j; y i is the demand response incentive coefficient for the i-th step interval, δ i-1 is the left boundary of the i-th step interval, δ i is the right boundary of the i-th step interval; δ0 is the expected compliance ratio, δ k is the expected capping ratio, where k is the number of capping steps; The step excitation coefficient satisfies the following constraints: (and i -and i+1 )(δ i -100%)≥0 (21), Where y i+1 is the demand response incentive coefficient for the i+1th step interval.

Images