Method and system for optimizing bidding strategy of electric vehicle aggregator based on double-loop principal-agent game

Abstract

The application relates to an electric vehicle aggregator bidding strategy optimization method and system based on double-cycle principal-agent game, and the method comprises the following steps: collecting electric vehicle end target parameters, power grid end target parameters and electric power market economic parameters in real time; regarding regional distribution network operators as upper leaders, regarding middle-layer electric vehicle aggregators as followers of energy stations and leaders of users, regarding lower-layer electric vehicle users as followers of electric vehicle aggregators, and constructing a double-cycle principal-agent game bidding model according to the collected parameters; solving the double-cycle principal-agent game bidding model to obtain optimized electricity prices, charging plans and bidding strategies. The application can improve the operation income of EVA, the income of DSO and the regional power supply reliability, smooth the regional load curve, and the double-cycle principal-agent game solving framework of the application only needs to exchange part of boundary information among subjects to achieve game equilibrium, which can protect the information privacy of sources and loads and accelerate the solving speed.

Description

Technical Field

[0001] This invention relates to the field of power system optimization technology, and in particular to a method and system for optimizing the bidding strategy of electric vehicle aggregators based on a dual-cycle master-slave game. Background Technology

[0002] The number of electric vehicles (EVs) is currently increasing. However, EV charging behavior varies significantly, making direct dispatch by the distribution network impossible. The unregulated charging behavior of large-scale EVs inevitably increases the operational burden on the power grid, posing a challenge to power system stability. To address this, EV aggregators are introduced to aggregate regional EV resources and facilitate energy exchange with the grid. EVAs act as a "bridge" between the distribution network and users, purchasing electricity from the distribution network operator at the upper level while charging users a charging service fee and guiding orderly EV charging, thus mitigating the impact of large-scale EV integration on the grid. To encourage EV aggregators to actively participate in the electricity market and realize their value, it is necessary to develop charging pricing schemes and bidding strategies to improve their economic efficiency in the electricity market and alleviate the load management pressure on the distribution network.

[0003] Existing technologies disclose charging operators setting time-of-use charging service fees for charging stations based on distribution network time-of-use pricing, thereby guiding EV load distribution during off-peak periods in the distribution network. However, these technologies only consider fast-charging electric vehicles. Existing technologies also disclose EVA (Electric Vehicle Amplifier) ​​which considers the uncertainty of EV access and different driving modes, aggregating EV resources to participate in the ancillary services market. Case studies show that EVA participation in the electricity market avoids transmission line congestion while also bringing certain benefits to EV owners. Furthermore, existing technologies disclose a regional energy system operation mechanism based on Stackelberg game theory and considering demand response, balancing the interests of regional energy operators and electric vehicle aggregators. Case studies demonstrate that this mechanism can guarantee the interests of all participating entities while reducing peak loads, thus contributing to stable system operation.

[0004] Most of the existing technologies mentioned above consider the game relationship between one or two of the regional distribution system operator (DSO), electric vehicle aggregator (EVA), and electric vehicle (EV). Few consider the EV aggregation dispatch and EVA bidding as a whole to participate in the DSO's pricing in the electricity market. How to balance the multi-layered game relationship of the interests of the three parties still needs further research.

[0005] Furthermore, the technical parameters of distributed electric vehicles (such as SOC and charging demand) are dispersed and dynamically updated. Existing bidding strategies lack an efficient collection and synchronization mechanism that "exchanges only partial boundary information" and do not optimize the parameter update frequency by combining the peak-valley difference characteristics of DLMP (node ​​marginal price). This results in inaccurate input data for the dual-cycle master-slave game model, leading to large biases in bidding decisions. It not only fails to achieve game equilibrium but also disrupts the smoothing effect of distribution network load, affecting the stability of power grid operation. At the same time, there is a risk of leakage of source and load information privacy. Summary of the Invention

[0006] The purpose of this invention is to provide a method and system for optimizing the bidding strategy of electric vehicle aggregators based on a dual-cycle master-slave game, thereby improving the operating revenue of EVA, increasing the revenue of DSO and the power supply reliability within the region, and smoothing the load curve within the region.

[0007] To achieve the above objectives, the present invention provides the following solution:

[0008] An optimization method for bidding strategies of electric vehicle aggregators based on a dual-cycle master-slave game includes:

[0009] Real-time collection of target parameters from electric vehicles, target parameters from the power grid, and economic parameters from the electricity market;

[0010] The regional distribution network operator is regarded as the upper-level leader, the middle-level electric vehicle aggregator is regarded as the follower of the energy station and the leader of the user, and the lower-level electric vehicle user is regarded as the follower of the electric vehicle aggregator. A dual-loop master-slave game bidding model is constructed based on the collected parameters.

[0011] The method of particle swarm optimization and mixed integer linear programming is used to solve the bicyclic master-slave game bidding model to obtain the optimized electricity price, charging plan and bidding strategy.

[0012] Optionally, the dual-cycle master-slave game bidding model includes a set of participants, a set of strategies, and a utility function. The set of strategies includes the dynamic electricity price set by the regional distribution network operator, the classified service fee electricity price set by the electric vehicle aggregator for different types of electric vehicles, the electricity purchase plan and discharge plan reported to the regional distribution network operator, the hourly charging plan of electric vehicle users, and the discharge plan of V2G electric vehicles participating in demand response. The utility function includes the revenue function of the regional distribution network operator, the revenue function of the electric vehicle aggregator, and the payment function of the electric vehicle user.

[0013] Optionally, obtaining the electricity price from the regional distribution network operator includes:

[0014] Under the unified price clearing model, the dynamic electricity price set by the regional distribution network operator is obtained based on a linear function. The dynamic electricity price consists of a base price and a price sensitivity coefficient, and the base price is constrained by the set upper and lower limits of the base price.

[0015] Optionally, solving the bicyclic master-slave game bidding model using a method based on particle swarm optimization and mixed-integer linear programming includes:

[0016] By using KKT conditions, the optimization of lower-level electric vehicle users is transformed into the constraint of optimization of middle-level electric vehicle aggregators, and the three-level optimization is transformed into a two-level nonlinear programming.

[0017] The bi-level nonlinear programming is transformed into a bi-level mixed integer quadratic programming by convexification. The Stackelberg equilibrium between regional distribution network operators and electric vehicle aggregators is solved by particle swarm optimization, and the optimized electricity price, charging plan and bidding strategy are obtained.

[0018] Optionally, solving the Stackelberg equilibrium between regional distribution network operators and electric vehicle aggregators using a particle swarm optimization iterative method includes:

[0019] S1. Obtain the current daytime electric vehicle load dataset, initialize relevant parameters, and set the maximum number of iterations;

[0020] S2. Randomly initialize the particle population and the basic prices of the regional distribution network operators as inputs for the optimization of electric vehicle aggregators;

[0021] S3. Call the Gurobi Business Solver to calculate the master-slave game between electric vehicle aggregators and electric vehicle users;

[0022] S4. The optimized power purchase plan obtained by the electric vehicle aggregator is sent back to the regional distribution network operator. The regional distribution network operator re-determines the base price based on the reported power purchase plan and generates a new distribution network node marginal price DLMP scheme, which is then issued to the electric vehicle aggregator. The distribution network node marginal price is a unified retail price.

[0023] S5. If the maximum number of iterations has not been reached, return to S3. If the maximum number of iterations has been reached, determine whether the iteration precision is greater than the preset iteration precision. If the preset iteration precision has not been reached, return to S2. If the preset iteration precision has been reached, obtain the final Stackelberg equilibrium.

[0024] Optionally, different types of electric vehicles can be obtained, including:

[0025] Based on the charging characteristics of electric vehicles, charging models for different types of electric vehicles are established, including fast charging, slow charging, V2G, and flexible charging.

[0026] Optionally, the revenue function of the regional distribution network operator for:

[0027] ;

[0028] in, The compensation electricity price paid by the DSO to the V2G EVs participating in the demand response project during time period t; The actual power value that DSO purchases from the upstream power grid at time t; Electricity charges collected by electric vehicle aggregators from users; The discharge power of V2G EVs participating in the demand response project; The electricity price sold by the upstream power grid during time period t; This is a set of scenarios for electricity sales prices from the upper-level power grid. This represents the probability of the corresponding electricity price scenario;

[0029] The revenue function of the electric vehicle aggregator for:

[0030] ;

[0031] in, The total fees charged by electric vehicle aggregators to users; Let m be the charging power of the i-th EV of type m during time period t; The duration of a unit of time period; The discharge service fee charged by EVA to V2G EVs during time period t; Discharge programs for V2G electric vehicles to participate in demand response; During time period t, EVA purchases electrical energy from DSO;

[0032] Electric vehicle user payment function for:

[0033] ;

[0034] in, represent When taking the minimum value The possible values ​​of ; Let t be the charging power of the i-th EV of type m during time period t.

[0035] This invention also provides an electric vehicle aggregator bidding strategy optimization system based on a dual-loop master-slave game, comprising:

[0036] The multi-source parameter acquisition module is used to collect target parameters from electric vehicles, target parameters from the power grid, and economic parameters from the electricity market in real time.

[0037] The game model construction module is used to construct a dual-loop master-slave game bidding model based on collected parameters, with regional distribution network operators as the upper-level leaders, mid-level electric vehicle aggregators as the followers of energy stations and the leaders of users, and lower-level electric vehicle users as the followers of electric vehicle aggregators.

[0038] Bidding strategy calculation module: Based on the particle swarm optimization algorithm and mixed integer linear programming, the double-loop master-slave game bidding model is solved to obtain the optimized electricity price, charging plan and bidding strategy.

[0039] The beneficial effects of this invention are as follows: The day-ahead bidding strategy proposed in this invention can maximize the operating revenue of aggregators, and regional distribution network operators, electric vehicle aggregators, and electric vehicle users can all benefit from the transaction mechanism of the proposed dual-cycle master-slave game model; Compared with traditional peak-valley electricity pricing, DLMP can effectively transmit market electricity shortage information. Due to the introduction of market competition, DLMP is generally lower than traditional peak-valley electricity pricing, but the peak-valley price difference is greater than that of traditional peak-valley electricity pricing, which can better stimulate the enthusiasm of market participants; The classified payment scheme formulated by aggregators can guide the orderly charging and discharging of electric vehicles and smooth the fluctuation of distribution network load; The proposed dual-cycle master-slave game solution framework only requires the exchange of some boundary information between the subjects to achieve game equilibrium, which can both protect the information privacy of source and load and speed up the solution. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0041] Figure 1 This is a framework diagram of an electric vehicle aggregator trading system according to an embodiment of the present invention;

[0042] Figure 2 This is a schematic diagram of the electric vehicle aggregator transaction price model according to an embodiment of the present invention;

[0043] Figure 3 This is a framework diagram of the dual-loop master-slave game bidding model according to an embodiment of the present invention;

[0044] Figure 4 This is a diagram illustrating the framework for solving a dual-loop master-slave game according to an embodiment of the present invention.

[0045] Figure 5 This is a flowchart illustrating the solution process of the dual-loop master-slave game model in an embodiment of the present invention.

[0046] Figure 6The day-ahead basic load forecast value is provided in this embodiment of the invention.

[0047] Figure 7 This is a diagram illustrating the revenue of DSO and EVA in an embodiment of the present invention.

[0048] Figure 8 This is a diagram illustrating the revenue of DSO and EVA in an embodiment of the present invention.

[0049] Figure 9 This is the optimal DLMP for this embodiment of the invention;

[0050] Figure 10 The total charging fee for different types of EVs in this embodiment of the invention is as follows: (a) is the total charging fee for fast-charging EVs, (b) is the total charging fee for slow-charging EVs, (c) is the total charging fee for V2G EVs, and (d) is the total charging fee for flexible EVs.

[0051] Figure 11 This invention provides daytime charge / discharge plans for different types of EVs according to embodiments of the present invention.

[0052] Figure 12 The present invention provides fast-charging and slow-charging EV charging and discharging plans, wherein (a) is a fast-charging EV charging and discharging plan and (b) is a slow-charging EV charging and discharging plan;

[0053] Figure 13 The present invention provides V2G and flexible EV charging and discharging plans, wherein (a) is a V2G EV charging and discharging plan and (b) is a flexible EV charging and discharging plan. Detailed Implementation

[0054] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0055] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0056] This embodiment proposes an optimization method for bidding strategies of electric vehicle aggregators based on a dual-cycle master-slave game, including:

[0057] Real-time collection of target parameters from electric vehicles, the power grid, and the electricity market. Among them, the target parameters from electric vehicles include battery SOC, charging power limit, remaining charging time, and access status; the target parameters from the power grid include regional real-time load, voltage fluctuation range, and frequency stability index; and the electricity market economic parameters include real-time electricity price and bidding threshold.

[0058] The regional distribution network operator is regarded as the upper-level leader, the middle-level electric vehicle aggregator is regarded as the follower of the energy station and the leader of the user, and the lower-level electric vehicle user is regarded as the follower of the electric vehicle aggregator. A dual-loop master-slave game bidding model is constructed based on the collected parameters.

[0059] The method of particle swarm optimization and mixed integer linear programming is used to solve the bicyclic master-slave game bidding model to obtain the optimized electricity price, charging plan and bidding strategy.

[0060] like Figure 1 The diagram illustrates the framework of the electric vehicle aggregator trading system considered in this embodiment. It primarily consists of a regional distribution network operator (DSO), electric vehicle aggregators within the region, and electric vehicle users. The DSO connects to the upstream power grid via a tie line, purchasing electricity from the upstream grid to meet the region's electricity demand. Simultaneously, it communicates, trades, and transmits electricity with EVAs and baseloads within the region. EVAs aggregate various types of electric vehicle loads within the region, purchasing electricity from the DSO and providing charging and discharging services to EV users. When trading electricity with the DSO, they flexibly adjust their reporting plans based on the electricity price information published by the DSO and release categorized charging and discharging payment schemes to EV users to achieve profitability in the electricity market. EV users adjust their charging and discharging times according to the pricing schemes published by the EVAs and their own circumstances, thereby reducing their payments.

[0061] Figure 2 This is a schematic diagram of an electric vehicle aggregator (EVA) transaction price model. The DSO acts as a regional user agent, purchasing electricity from the upper-level power grid according to the grid's time-of-use pricing. Simultaneously, it sets dynamic pricing to trade electricity with EVAs and base loads within the region. The fees charged by EVAs to EV users include two parts: the electricity price and the service fee. The electricity price is the dynamic price set by the DSO, while the service fee is set by the EVA.

[0062] After the DSO (Distribution and Sales Organization) first publishes the time-of-use dynamic electricity price as the transaction price, the EVA (Electric Vehicle Association) formulates a categorized payment scheme and charging / discharging plan for the aggregated EVs based on the transaction price to achieve the lowest electricity cost or the highest profit for each party, and uploads the plan to the DSO. The DSO then re-formulates the electricity price based on the electricity purchase quotation plan submitted by the aggregators, aiming to maximize its own operating revenue, and publishes it to the EVA. The EVA then modifies the electric vehicle categorized payment scheme and changes its own electricity purchase quotation plan based on the market electricity price information published by the DSO. This dynamic interaction continues until all parties are satisfied, thus determining the final DSO internal transaction price, the EVA's bidding quotation plan, and the formulated EV categorized payment scheme and EV load charging / discharging plan.

[0063] Furthermore, the dual-cycle master-slave game bidding model includes a set of participants, a set of strategies, and a utility function. The set of strategies includes the dynamic electricity price set by the regional distribution network operator, the classified service fee electricity price set by the electric vehicle aggregator for different types of electric vehicles, the electricity purchase plan and discharge plan reported to the regional distribution network operator, the hourly charging plan of electric vehicle users, and the discharge plan of V2G electric vehicles participating in demand response. The utility function includes the revenue function of the regional distribution network operator, the revenue function of the electric vehicle aggregator, and the payment function of the electric vehicle user.

[0064] Furthermore, acquiring different types of electric vehicles includes:

[0065] Based on the charging characteristics of electric vehicles, charging models for different types of electric vehicles are established, including fast charging, slow charging, V2G, and flexible charging.

[0066] Specifically, based on the driving patterns of electric vehicles, EV loads are divided into three types. Fast-charging EVs are typically represented by taxis and ride-hailing vehicles, which are fully charged in a short time with high charging power and randomly connected throughout the day. Slow-charging EVs are typically represented by private cars connected at night with lower charging power, connected at night and driven out during the day. V2G EVs have relatively ample initial charge and achieve peak-hour arbitrage by charging during nighttime when electricity prices are lower and utilizing V2G technology to reverse discharge during demand response projects in the distribution network. Flexible EVs are typically represented by daytime commuter vehicles, connected to charging stations during the day with lower charging power.

[0067] After an electric vehicle connects to a charging station, EVA obtains information such as the battery's rated capacity, charging and discharging power, and state of charge through the onboard battery management system. Users upload information such as their next usage time and target state of charge via their personal terminals. This information per EV is crucial for the connection process. for:

[0068] (1);

[0069] The arrival time of the EV; EV parking duration; The moment the EV departs; For EV battery capacity; This refers to the battery's state of charge at the time of connection. The target state of charge when the EV leaves; The EV's willingness to discharge; This represents the maximum charging power of type m EVs, which is positive. Similarly, for V2G EVs, it represents the maximum discharging power. m represents the code information for the four EV types. Based on the access information, the charging demand of the i-th type m electric vehicle can be calculated. :

[0070] (2);

[0071] For EVs with m=1,2 (fast charging type, slow charging type), the charging power is 0 or the maximum value and cannot be flexibly adjusted. If it is not fully charged in the last period, it is charged according to the average power. Its charging model is shown in equation (3):

[0072] (3);

[0073] (4);

[0074] in Let t be the battery level of the i-th EV of type m during time period t; Let t be the charging power of the i-th EV of type m during time period t; Let t be the charging power of the i-th EV of type m during time period t; Let q be the average charging power during the period when the i-th EV is not fully charged at the end of the charging process for m=1,2 (fast charging type, slow charging type); q is the integer function that takes the largest integer not exceeding the value in the parentheses, which is used to represent the total number of charging periods when the i-th EV is fully charged at the end of the charging process for m=1,2 (fast charging type, slow charging type); Let represent the set of time periods in which the i-th electric vehicle is in the access state of 1.

[0075] For EVs with m=3,4 (V2G type, flexible type), their charging power can be flexibly adjusted between 0 and the maximum value, and their charging model is shown in equation (5):

[0076] (5);

[0077] Based on the above classification of EV charging models, the following constraints are derived:

[0078] (6);

[0079] (7);

[0080] (8);

[0081] (9);

[0082] (10);

[0083] Equation (6) represents the charging state constraint. and Let represent the charging and discharging states of the i-th electric vehicle of type m during time period t. For Boolean variables, A value of 1 indicates that the EV has entered the charging / discharging state; Equation (7) is the EV charging power constraint; Equation (8) is the EV discharging power constraint; Equation (9) is the power demand constraint for the i-th EV of type m. Equation (10) is used to limit the excessive discharge of V2G electric vehicles by constraining their power separately. , It represents the amount of electricity converted from its lowest and highest states of charge.

[0084] In market transactions, the pricing strategies of regional distribution network operators influence the energy purchase plans of electric vehicle aggregators, and the energy sales prices of aggregators also affect the energy demand of electric vehicle users. Conversely, changes in users' adjusted energy demand will affect the pricing strategies and energy purchase plans of electric vehicle aggregators, thereby influencing the base pricing of regional distribution network operators. These factors lead to interactions among participants, who continuously adjust their strategies to coordinate their mutual interests.

[0085] The framework of the dual-loop master-slave game bidding model is as follows: Figure 3 As shown, the energy trading process between the entities conforms to a dynamic game with a hierarchical master-slave structure. The regional distribution network operator acts as the leader at the top, the mid-level electric vehicle aggregator is both a follower of the energy station and a leader of the users, and the lowest level, the electric vehicle user, acts as a follower of the electric vehicle aggregator. This establishes a double-loop master-slave game model, represented as:

[0086] (11);

[0087] This game theory model includes three elements: participants, strategies, and utility functions, such as... Figure 3 As shown.

[0088] (1) Participants:

[0089] Regional power distribution network operators, electric vehicle aggregators, and electric vehicle users are the participants in this game, and the participants are as follows:

[0090] (12);

[0091] in This represents the number of electric vehicle users in the region.

[0092] (2) Strategy set:

[0093] 1) In this embodiment, the electricity market adopts a unified price clearing model. When regional distribution network operators (DSOs) transact with users in the electricity market, a unified retail electricity price will appear. The Distribution Locational Marginal Price (DLMP) is introduced to describe this price. Under the unified price clearing model, the market clearing price and the amount of electricity purchased by users show a linear correlation. To simplify the market clearing problem, the DSO introduces a linear function to construct a dynamic pricing mechanism, consisting of a base price and a price sensitivity coefficient, as shown in the following formula:

[0094] (13);

[0095] (14);

[0096] Dynamic electricity pricing set for regional power distribution network operators. Basic electricity price, Let t be the actual power value that DSO purchases from the upstream power grid. The power baseline value established for the DSO is the total power value of the regional base load and EVA reported during the day-ahead phase of the DSO's forecast. Let t be the total power that EVA purchases from DSO. Let be the base load in the region at time t. To simplify the example, the predicted value of the base load for the previous day is used here. denoted as the price sensitivity coefficient, it characterizes the sensitivity of dynamic electricity prices to electricity purchase plans. In this paper, it is set to a value of [value missing]. .

[0097] The dynamic electricity price published by the DSO is directly related to the underlying demand; therefore, the DSO's strategy selects the base price from the dynamic electricity price. , represented as a vector It satisfies the following constraints:

[0098] (15);

[0099] in, , These represent the upper and lower limits of the established base electricity price, respectively.

[0100] 2) The strategies of electric vehicle aggregators include categorized service fees and electricity prices for different types of electric vehicle users. and the power purchase plan reported to the regional distribution network operator and discharge plan , represented as a vector The following constraints need to be satisfied:

[0101] (16);

[0102] (17);

[0103] (18);

[0104] , These represent the upper and lower limits of the electrical energy that EVA purchases from DSO during time period t; Let m be the charging power of the i-th EV of type m during time period t; Let be the charging power of the i-th EV during time period t. Constraints (16) and (17) ensure that the power purchased and output by the electric vehicle aggregator is balanced.

[0105] Total fees charged to users by electric vehicle aggregators Including electricity costs With the established service fee :

[0106] (19);

[0107] (20);

[0108] Electricity Costs This part refers to the DLMP scheme formulated by the DSO. Pricing scheme for electric vehicle service fees in category m. The constraints are as follows:

[0109] (twenty one);

[0110] (twenty two);

[0111] The daily average pricing that the charging scheme for EV category m must meet; , These represent the upper and lower limits for the pricing of the m-th class of EVs.

[0112] 3) The strategy for electric vehicle users is an hourly charging plan. V2G electric vehicles participate in demand response discharge programs , represented as a vector After receiving the categorized charging and discharging pricing schemes from electric vehicle aggregators, electric vehicle users adjust their own charging and discharging plans.

[0113] (3) Utility function:

[0114] 1) Revenue function of regional distribution network operators To maximize its own operating profits, the DSO formulates the optimal DLMP scheme based on the electricity sales price of the upper-level power grid and considering the lower-level power purchase scheme. Its profit function is shown in equation (23):

[0115] (twenty three);

[0116] (twenty four);

[0117] in The compensation electricity price paid by the DSO to the V2G EVs participating in the demand response project during time period t; Let t be the electricity price sold at the upstream power grid during time period t. In this paper, the uncertainty of this electricity price is considered through a scenario-based stochastic optimization method. This is a set of scenarios for electricity sales prices from the upper-level power grid. This represents the probability of the corresponding electricity price scenario; Electricity charges collected by electric vehicle aggregators from users; The discharge power of V2G EVs participating in the demand response project.

[0118] 2) Revenue function of electric vehicle aggregators To maximize its own operating profits, EVA optimizes the price and power of purchased and sold electricity, purchasing electricity from DSO and selling it to electric vehicle users. Its revenue function is shown in equation (25):

[0119] (25);

[0120] in This is the discharge service fee charged by EVA to V2G EVs during time period t. The duration is for a single time period.

[0121] 3) The utility function for electric vehicle users is: EV users can optimize their charging time based on the service fee and electricity price given by EVA, thereby minimizing their charging costs.

[0122] (26);

[0123] (27);

[0124] in represent When taking the minimum value The value of ; This represents the total number of all electric vehicle types in this embodiment. It is 4. Let t be the charging power of the i-th EV of type m during time period t.

[0125] Furthermore, the solution to the bicyclic master-slave game bidding model based on particle swarm optimization and mixed-integer linear programming includes:

[0126] By using KKT conditions, the optimization of lower-level electric vehicle users is transformed into the constraint of the optimization of middle-level electric vehicle aggregators, and the three-level optimization is transformed into a two-level nonlinear programming.

[0127] The bi-level nonlinear programming is transformed into a bi-level mixed integer quadratic programming by convexification. The Stackelberg equilibrium between regional distribution network operators and electric vehicle aggregators is solved by particle swarm optimization, and the optimized electricity price, charging plan and bidding strategy are obtained.

[0128] Furthermore, solving the Stackelberg equilibrium between regional distribution network operators and electric vehicle aggregators using the particle swarm optimization iterative method includes:

[0129] S1. Obtain the current daytime electric vehicle load dataset, initialize relevant parameters, and set the maximum number of iterations;

[0130] S2. Randomly initialize the particle population and the basic prices of the regional distribution network operators as inputs for the optimization of electric vehicle aggregators;

[0131] S3. Call the Gurobi Business Solver to calculate the master-slave game between electric vehicle aggregators and electric vehicle users;

[0132] S4. The optimized power purchase plan obtained by the electric vehicle aggregator is sent back to the regional distribution network operator. The regional distribution network operator re-determines the base price based on the reported power purchase plan and generates a new distribution network node marginal price DLMP scheme, which is then issued to the electric vehicle aggregator. The distribution network node marginal price is a unified retail price.

[0133] S5. If the maximum number of iterations has not been reached, return to S3. If the maximum number of iterations has been reached, determine whether the iteration precision is greater than the preset iteration precision. If the preset iteration precision has not been reached, return to S2. If the preset iteration precision has been reached, obtain the final Stackelberg equilibrium.

[0134] Specifically, the established dual-cycle master-slave game model essentially belongs to a type of three-level game model. If For a three-level game to reach equilibrium, the following conditions must be met:

[0135] (28);

[0136] in The optimal strategies of the top-level leader, the distribution network operator (DSO), the middle-level follower, the electric vehicle aggregator (EVA), and the bottom-level follower, the electric vehicle user, are respectively. Formula (28) indicates that the optimal strategy of each player is determined by the optimal strategies of the other players, and no party can gain more by unilaterally changing its strategy. The equilibrium of the proposed three-level game model can be divided into the following two parts: the Stackelberg equilibrium between the regional distribution network operator and the electric vehicle aggregator, and the Stackelberg equilibrium between the electric vehicle aggregator and the electric vehicle user.

[0137] In the Stackelberg game model, if the following conditions (sufficient conditions) are met simultaneously, it is proven that there is a unique Stackelberg equilibrium in the game: (1) the policy sets of all leaders and followers are non-empty sets and are compact convex sets; (2) when the leader's policy is determined, each follower has a unique optimal policy; (3) when the follower's policy is determined, the leader also has a unique optimal policy.

[0138] Proof of the Stackelberg equilibrium in the proposed game model: As shown in the proposed day-ahead bidding model for aggregators, the policy sets of all participants are non-empty and compact convex sets. Furthermore, the utility functions of the DSO, EVA, and EV users are all linear functions of their policy sets. According to the definition of concavity and convexity, a linear function is both convex and concave. Given one player's policy, the other player has a unique optimal policy. Therefore, the proposed bicyclic master-slave game model has a unique Stackelberg equilibrium solution.

[0139] This embodiment proposes a method based on particle swarm optimization and mixed-integer linear programming to solve the model, the framework of which is as follows: Figure 4 As shown, the optimization of the third-layer electric vehicle users is first transformed into the optimization of the second-layer electric vehicle aggregator using KKT conditions, thus converting the three-layer optimization into a two-layer nonlinear programming problem. Then, convexity transformation is used to further transform the two-layer nonlinear programming problem into a two-layer mixed-integer quadratic programming problem. Finally, the Stackelberg equilibrium between the regional distribution network operator and the electric vehicle aggregator is solved using a particle swarm optimization method.

[0140] Specifically, in the original three-layer game, the third layer is the optimization of electric vehicle users (aiming to minimize user charging costs, with constraints such as charging power and battery capacity); the second layer is the optimization of aggregators (aiming to maximize aggregator revenue, with constraints including user response). A Lagrangian function is constructed for the user optimization problem (objective function + constraints), and then the KKT conditions are written (including: gradient of 0, constraint satisfaction, complementary relaxation, and non-negative multipliers). The KKT conditions for the third-layer user optimization are directly added to the constraints of the second-layer aggregator optimization problem. At this point, the second-layer aggregator optimization automatically includes the logic that "users will choose the optimal charging strategy," and the three-layer optimization is transformed into a two-layer optimization of "upper layer (distribution network) → middle layer (aggregator)," with the optimal response of the lower-layer users implicitly included in the middle-layer constraints.

[0141] The particle swarm optimization algorithm takes the "distribution network strategy (such as electricity price)" and the "aggregator strategy (such as charging scheduling)" as the "particle position" and updates the particle position iteratively to find an approximate optimal solution that satisfies the two-level mixed integer quadratic programming problem.

[0142] In solving this model, the optimization problem of the electric vehicle aggregator is represented as a sub-function and embedded into the optimization function of the regional distribution network operator. The algorithm flowchart is as follows: Figure 5 As shown:

[0143] The examples in this embodiment were all run on a PC equipped with an AMD R7 processor, using MATLAB R2022b, and solved using the Gurobi commercial solver. First, the day-ahead electric vehicle load dataset was imported, relevant parameters were initialized, and the maximum number of iterations was set. The process involves randomly initializing the particle swarm and the basic prices of the regional distribution network operator (DSO) as inputs for the electric vehicle aggregator's (EV) optimization. Here, the "particle position" in the particle swarm corresponds to a set of electricity price parameters for the DSO, and "particle iteration" represents the process by which the DSO adjusts its electricity price based on the aggregator's electricity purchase plan. Ultimately, the optimal electricity price strategy for the DSO is found through particle swarm search. Afterward, the Gurobi business solver is used to calculate the master-slave game between the EV aggregator and EV users. Then, the optimized electricity purchase plan from the EV aggregator is sent back to the regional distribution network operator (DSO). The DSO re-determines its basic price based on the reported electricity purchase plan, generates a new DLMP scheme, and distributes it to the EV aggregator. If the maximum number of iterations has not been reached... and iteration accuracy This process is repeated until the final Stackelberg equilibrium is reached, yielding optimized electricity prices, charging plans, and bidding strategies. In each iteration, each entity only exchanges marginal information such as electricity prices and power purchase plans, avoiding information exposure and protecting the privacy of all participants. Therefore, the dual-loop master-slave game-solving framework proposed in this embodiment only requires the exchange of partial boundary information between entities to reach game equilibrium, thus protecting the information privacy of source and load while accelerating the solution process.

[0144] Reference Figure 1 The trading system model involves a regional distribution network operator supplying power to the regional base load and electric vehicle aggregators, and engaging in power trading and information exchange with the upper-level grid via tie lines. The maximum number of iterations for the particle swarm optimization algorithm is set to 15, the particle population size to 20, and the upper and lower limits of the DSO base price to be 0.1 yuan / kW·h and 0.55 yuan / kW·h, respectively. The day-ahead base load predicted by the DSO is as follows: Figure 6 As shown, the reporting cycle T is 24. The daily access limit for electric vehicles (EVs) to the aggregator within the region is set at 200 vehicles. The EVA generates access information for different types of EVs in 15-minute intervals and guides their charging and discharging. On the day before, the hourly power consumption is reported to the DSO in one-hour intervals. Table 1 shows the parameters, basic service fee electricity price, and service fee pricing range for different types of EVs. Fast-charging EVs have shorter parking times and higher power demands, resulting in less margin for aggregators to adjust their charging time and power; therefore, their basic charging service fee is the highest, and the upper and lower limits of the service fee pricing are also the highest. Flexible EVs have flexible charging power and longer parking times, so their basic charging service fee is the lowest, and the upper and lower limits of the service fee pricing are also the smallest. Other types of EVs have their basic charging service fees and service fee pricing ranges determined according to parking time and power adjustment margin. Distribution network operators release day-ahead demand response projects during the periods of 8:00-10:00 and 17:00-21:00 daily. The discharge compensation price for V2G EVs participating in day-ahead distribution network demand response projects is RMB 0.45 / kW·h, and EVAs collect one-third of their discharge revenue as a discharge service fee.

[0145] Table 1

[0146]

[0147] Figure 7 The Chinese block line section represents the change in DSO returns for different iteration numbers. Figure 7The triangular section represents the change in EVA revenue under different iteration numbers, reflecting the equilibrium convergence process of the game between the distribution network operator and the electric vehicle aggregator. Their revenues tend to stabilize after 10 iterations, reaching game equilibrium, indicating that the proposed algorithm converges quickly. At this point, the DSO revenue is 39,962 yuan, and the EVA revenue is 1,908.9 yuan. However, as the number of iterations increases, the distribution network operator's revenue gradually increases, while the electric vehicle aggregator's revenue decreases; their different trends represent the bargaining process in the transaction. When all followers make optimal decisions based on the leader's strategy, and the leader accepts their responses, the game reaches equilibrium. After this, no participant can unilaterally increase their own interests by changing their strategy.

[0148] Figure 8 The process of setting a base electricity price for the DSO reaches a game equilibrium at the 10th iteration, at which point the optimal base electricity price set by the DSO is 0.4777 yuan / (kW·h). Figure 9 To achieve the optimal DLMP at equilibrium in the game, EVA and DSO use this electricity price for day-ahead transaction settlement. The electricity price is high between 08:00 and 10:00 in the morning and between 17:00 and 23:00 in the afternoon, when the load in the distribution network is higher than the benchmark value set by the distribution network, so a higher electricity price is set. Conversely, the electricity price is lower between 00:00 and 7:00 in the morning and between 13:00 and 16:00 in the afternoon, when the base load in the distribution network is lower, so a lower electricity price is set. The DLMP price varies in each time period, fully reflecting the load changes in the market.

[0149] The total cost of EV charging consists of two parts: electricity cost and service fee. The former is charged according to the DLMP electricity price set by DSO, while the latter is set by EVA and the service fee plan varies for different types of EVs. Figure 10 Total payment for charging different types of EVs. Figure 11 The day-ahead charge / discharge plans are provided for different types of EVs. For V2G EVs, the charging power is negative, representing their discharge. The day-ahead payment plan for fast-charging EVs is as follows: Figure 10 As shown in (a), for fast-charging EVs that are mostly connected during the daytime, their charging demand is high, their parking time is short, and their charging time is limited. Therefore, EVA can only schedule charging during periods with lower electricity prices near their connection time. Since the electricity price set by the DSO is high at 08:00, EVA schedules charging during periods with relatively lower electricity prices at 07:00 and 09:00, as well as after 12:00 noon. Because the number of fast-charging EVs connected is small and the adjustment margin is small, they are not the main research object. The day-ahead payment scheme for slow-charging EVs is as follows: Figure 10As shown in (b), for most slow-charging EVs that connect at night, the distribution network electricity price is highest during their connection period, and the charging price set by the EVA is also the highest at this time. Therefore, the EVA postpones their charging time to the early morning hours when the electricity price is lower, mainly concentrated between 00:00 and 04:00, during which the charging price is also lower. Slow-charging vehicles have the largest number of connections and the longest parking time, making them the main target of aggregators' control; the day-ahead payment scheme for V2G EVs is as follows... Figure 10 As shown in (c), for V2G EVs that earn revenue by discharging into the grid, revenue is earned by discharging during the demand response periods announced by the distribution network operator. V2G EVs discharge during peak distribution network load periods (08:00-11:00 and after 17:00), when distribution network load demand is high, and discharging through V2G EVs can compensate for some of the power shortage; the day-ahead payment scheme for flexible EVs is as follows... Figure 10 As shown in (d), for flexible EVs that are connected during the day and whose charging power is flexibly adjustable, the adjustment margin is greater. Therefore, EVA arranges for them to charge during the daytime when the DSO electricity price is lower, from 06:00 to 07:00 and from 12:00 to 16:00, and the payment plan is lower at this time.

[0150] To verify the effectiveness of the proposed dual-cycle master-slave game day-ahead bidding strategy, the following four scenarios are set up for comparison:

[0151] Option 1: EVA conducts electricity trading with DSO based on the established marginal electricity price of distribution network nodes. EVs charge in an unordered manner and do not participate in the management of EVA. EVA charges EVs for charging and discharging according to the basic service fee.

[0152] Option 2: EVA conducts electricity trading with DSO based on peak-valley electricity prices. EVA charges EVs for charging and discharging according to a basic service fee. In this case, EVs participate in the management of EVA. EVA aims to maximize its revenue, while EVs aim to minimize their total charging and discharging fees. The two form a master-slave game relationship to arrive at the final decision.

[0153] Option 3: EVA conducts electricity trading with DSO based on peak-valley electricity prices. EVA charges EVs for charging and discharging according to the proposed classification service fee scheme. At this time, EVs participate in the management and control of EVA, and the final decision result is obtained at the master-slave game equilibrium.

[0154] Option 4: EVA conducts electricity trading with DSO based on the established marginal electricity price of distribution network nodes. EVA charges EV for charging and discharging according to the proposed classification service fee scheme. At this time, EV participates in the management and control of EVA, and the final decision result is obtained at the equilibrium of the dual-cycle master-slave game.

[0155] The peak-valley electricity prices for the distribution network are shown in Table 2, and the marginal electricity prices for distribution network nodes set by the DSO are as follows: Figure 9As shown in the figure. Comparing the two, it can be found that the electricity price of DLMP is different in each time period, which can intuitively reflect the current distribution network load deficit. Moreover, due to the introduction of market competition, the marginal electricity price of distribution network nodes is generally lower than the traditional peak-valley electricity price, but the peak-valley price difference is larger, which is more conducive to market participants to participate in market arbitrage and improves the participation enthusiasm of market participants.

[0156] Table 3 compares the revenue of DSO with load volatility, EVA revenue, and EV user fees under four schemes. The total load of DSO in this paper is the sum of the base load and the EVA reporting plan. The load volatility is calculated using the standard deviation method, which is obtained by dividing the load standard deviation by the load average value. Figures 12-13 The daytime charge and discharge plans for different types of electric vehicles under four different schemes.

[0157] Table 2

[0158]

[0159] Table 3

[0160] Figure 12 (a) The charging plans for fast-charging EVs under different schemes. In Scheme 1, fast-charging EVs are randomly connected during the day and mainly concentrated in the morning peak load period, when the market electricity price is higher. The charging plans for fast-charging vehicles under Scheme 2 and Scheme 3 are similar, mainly adjusting the charging period to the midday peak-valley electricity price period. However, due to the adjustment of the electricity price between DSO and EVA in Scheme 4, the electricity price set by DSO is lower at 14:00 in the afternoon, so EVA will transfer some of its load to this period.

[0161] Figure 12 (b) Charging plans for slow-charging EVs under different schemes: In Scheme 1, slow-charging EVs are mainly connected at night and concentrated during the evening peak load period, when the market electricity price is the highest of the day. In Scheme 2, EVA schedules slow-charging EVs for charging during the off-peak hours of the peak-valley electricity price. In Scheme 3, due to EVA's time-of-use payment scheme, the charging plan for slow-charging EVs has been adjusted accordingly. In Scheme 4, DLMP has the lowest electricity price in the early morning, so EVA mainly schedules its charging during this period, primarily between 01:00 and 04:00.

[0162] Figure 13 (a) The discharge plan for V2G EVs under different schemes. In Scheme 1, V2G EVs are randomly connected and will have discharge needs when their own power is high. In Schemes 2 and 3, they will discharge during the peak period of the peak-valley electricity price. Since the discharge compensation price is the same, the distribution network operator will guide them to discharge when the grid load is most scarce. In Scheme 4, since the electricity price is the highest after 17:00, the distribution network operator will guide the EVs to discharge a large amount during this period.

[0163] Figure 13 (b) For flexible EV charging plans under different schemes, in Scheme 1, flexible EVs charge randomly during the day, with charging times mostly distributed during peak electricity price periods after 8:00 AM. In Schemes 2 and 3, flexible EVs are mostly charged during the midday period when electricity prices are normal. However, in Scheme 4, because market electricity prices are very low between 12:00 PM and 3:00 PM, flexible EVs mostly charge after 12:00 PM to avoid peak electricity prices and reduce their own charging costs. It can be seen that the introduction of the spot market's marginal electricity price and EVA-based service fee pricing scheme into the DSO maximizes the interests of all stakeholders.

[0164] Simulation results show that the proposed strategy can improve DSO revenue, reduce load volatility of DSO and the upper-level grid, smooth the load curve, stabilize the regional power supply level, improve EVA operating revenue in the region, significantly reduce total EV user fees, and actively guide EVA and EV to participate in power trading.

[0165] This embodiment also provides an electric vehicle aggregator bidding strategy optimization system based on a dual-loop master-slave game, including:

[0166] The multi-source parameter acquisition module is used to collect target parameters from electric vehicles, target parameters from the power grid, and economic parameters from the electricity market in real time.

[0167] The game model construction module is used to construct a dual-loop master-slave game bidding model based on collected parameters, with regional distribution network operators as the upper-level leaders, mid-level electric vehicle aggregators as the followers of energy stations and the leaders of users, and lower-level electric vehicle users as the followers of electric vehicle aggregators.

[0168] Bidding strategy calculation module: Based on the particle swarm optimization algorithm and mixed integer linear programming, the double-loop master-slave game bidding model is solved to obtain the optimized electricity price, charging plan and bidding strategy.

[0169] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. An optimization method for bidding strategies of electric vehicle aggregators based on a dual-cycle master-slave game, characterized in that, include: Real-time collection of target parameters from electric vehicles, target parameters from the power grid, and economic parameters from the electricity market; The regional distribution network operator is regarded as the upper-level leader, the middle-level electric vehicle aggregator is regarded as the follower of the energy station and the leader of the user, and the lower-level electric vehicle user is regarded as the follower of the electric vehicle aggregator. A dual-loop master-slave game bidding model is constructed based on the collected parameters. The method of particle swarm optimization and mixed integer linear programming is used to solve the bicyclic master-slave game bidding model to obtain the optimized electricity price, charging plan and bidding strategy.

2. The method for optimizing the bidding strategy of electric vehicle aggregators based on a dual-cycle master-slave game as described in claim 1, characterized in that, The dual-cycle master-slave game bidding model includes a set of participants, a set of strategies, and a utility function. The set of strategies includes the dynamic electricity price set by the regional distribution network operator, the classified service fee electricity price set by the electric vehicle aggregator for different types of electric vehicles, the electricity purchase plan and discharge plan reported to the regional distribution network operator, the hourly charging plan of electric vehicle users, and the discharge plan of V2G electric vehicles participating in demand response. The utility function includes the revenue function of the regional distribution network operator, the revenue function of the electric vehicle aggregator, and the payment function of the electric vehicle user.

3. The method for optimizing the bidding strategy of electric vehicle aggregators based on a dual-cycle master-slave game as described in claim 2, characterized in that, Obtaining the electricity price from the regional distribution network operator includes: Under the unified price clearing model, the dynamic electricity price set by the regional distribution network operator is obtained based on a linear function. The dynamic electricity price consists of a base price and a price sensitivity coefficient, and the base price is constrained by the set upper and lower limits of the base price.

4. The method for optimizing the bidding strategy of electric vehicle aggregators based on a dual-cycle master-slave game as described in claim 2, characterized in that, The solution to the bicyclic master-slave game bidding model based on particle swarm optimization and mixed integer linear programming includes: By using KKT conditions, the optimization of lower-level electric vehicle users is transformed into the constraint of optimization of middle-level electric vehicle aggregators, and the three-level optimization is transformed into a two-level nonlinear programming. The bi-level nonlinear programming is transformed into a bi-level mixed integer quadratic programming by convexification. The Stackelberg equilibrium between regional distribution network operators and electric vehicle aggregators is solved by particle swarm optimization, and the optimized electricity price, charging plan and bidding strategy are obtained.

5. The method for optimizing the bidding strategy of electric vehicle aggregators based on a dual-cycle master-slave game as described in claim 1, characterized in that, Solving the Stackelberg equilibrium between regional distribution network operators and electric vehicle aggregators using the particle swarm optimization iterative method includes: S1. Obtain the current daytime electric vehicle load dataset, initialize relevant parameters, and set the maximum number of iterations; S2. Randomly initialize the particle population and the basic prices of the regional distribution network operators as inputs for the optimization of electric vehicle aggregators; S3. Call the Gurobi Business Solver to calculate the master-slave game between electric vehicle aggregators and electric vehicle users; S4. The optimized power purchase plan obtained by the electric vehicle aggregator is sent back to the regional distribution network operator. The regional distribution network operator re-determines the base price based on the reported power purchase plan and generates a new distribution network node marginal price DLMP scheme, which is then issued to the electric vehicle aggregator. The distribution network node marginal price is a unified retail price. S5. If the maximum number of iterations has not been reached, return to S3. If the maximum number of iterations has been reached, determine whether the iteration precision is greater than the preset iteration precision. If the preset iteration precision has not been reached, return to S2. If the preset iteration precision has been reached, obtain the final Stackelberg equilibrium.

6. The method for optimizing the bidding strategy of electric vehicle aggregators based on a dual-cycle master-slave game as described in claim 1, characterized in that, Acquiring different types of electric vehicles includes: Based on the charging characteristics of electric vehicles, charging models for different types of electric vehicles are established, including fast charging, slow charging, V2G, and flexible charging.

7. The method for optimizing the bidding strategy of electric vehicle aggregators based on a dual-cycle master-slave game as described in claim 2, characterized in that, The revenue function of the regional distribution network operator for: ； in, The compensation electricity price paid by the DSO to the V2G EVs participating in the demand response project during time period t; The actual power value that DSO purchases from the upstream power grid at time t; Electricity charges collected by electric vehicle aggregators from users; The discharge power of V2G EVs participating in the demand response project; The electricity price sold by the upstream power grid during time period t; This is a set of scenarios for electricity sales prices from the upper-level power grid. This represents the probability of the corresponding electricity price scenario; The revenue function of the electric vehicle aggregator for: ； in, The total fees charged by electric vehicle aggregators to users; Let m be the charging power of the i-th EV of type m during time period t; The duration of a unit of time period; The discharge service fee charged by EVA to V2G EVs during time period t; Discharge programs for V2G electric vehicles to participate in demand response; During time period t, EVA purchases electrical energy from DSO; Electric vehicle user payment function for: ； in, represent When taking the minimum value The value of ; Let t be the charging power of the i-th EV of type m during time period t.

8. A bidding strategy optimization system for electric vehicle aggregators based on a dual-cycle master-slave game is used to implement the bidding strategy optimization method for electric vehicle aggregators based on a dual-cycle master-slave game as described in any one of claims 1-7, characterized in that, include: The multi-source parameter acquisition module is used to collect target parameters from electric vehicles, target parameters from the power grid, and economic parameters from the electricity market in real time. The game model construction module is used to construct a dual-loop master-slave game bidding model based on collected parameters, with regional distribution network operators as the upper-level leaders, mid-level electric vehicle aggregators as the followers of energy stations and the leaders of users, and lower-level electric vehicle users as the followers of electric vehicle aggregators. Bidding strategy calculation module: Based on the particle swarm optimization algorithm and mixed integer linear programming, the double-loop master-slave game bidding model is solved to obtain the optimized electricity price, charging plan and bidding strategy.

Images