A pricing method for shared energy storage in a park based on Stackelberg game

Abstract

The present invention provides a pricing method for shared energy storage in a park based on the Stackelberg game, which relates to the technical field of new power systems. The method includes: constructing a park solar-storage-charging system, constructing an upper-layer model with a shared energy storage power station as a leader, constructing a lower-layer model with a user group as a follower, using the Stackelberg game mathematical model to combine the upper and lower models into a double-layer optimization model, using the KKT optimality condition to convert the double-layer optimization model into a single-layer optimization model, using a solver to solve the single-layer optimization model to obtain shared energy storage pricing. The present invention can simulate the behavior of leaders and followers in the market, implement a more reasonable and efficient pricing strategy, find the global optimal solution under the KKT optimality condition, and reduce the energy consumption cost of the user group. At the same time, a differential privacy perturbation mechanism is introduced to prevent individual data leakage and achieve a balance between privacy protection and optimization utility.

Description

Technical Field

[0001] The present invention relates to the technical field of novel power systems, and in particular to a pricing method for park photovoltaic, storage and charging shared energy storage based on Stackelberg game. Background Art

[0002] As the primary energy source for future power systems, renewable energy, with its volatility and indirect nature, will pose multiple challenges to the grid's ability to absorb and absorb power. Furthermore, the rapid adoption of electric vehicles (EVs) and the emergence of EV charging stations as new power load points mean that their concentrated charging demand can lead to a sharp increase in grid load during peak hours, impacting grid stability and power supply capacity. The construction of fast-charging stations, in particular, places even higher demands on the grid's instantaneous power supply due to their high-power charging requirements. Furthermore, current EV charging stations are typically powered solely by the grid. Because peak grid demand overlaps with peak charging demand at EV charging stations, charging prices at EV charging stations are relatively high, a significant pain point for EV users.

[0003] To address one of the aforementioned issues, the park photovoltaic, storage, and charging system has emerged. This system primarily includes industrial users within the park, photovoltaic power generation, energy storage systems, and charging facilities (such as electric vehicle charging stations), aiming to achieve efficient energy management and utilization. The park photovoltaic, storage, and charging system can increase the access ratio of new energy sources and is of great significance for promoting energy transformation and low-carbon development. However, the energy storage systems within the current park photovoltaic, storage, and charging systems face challenges, such as high investment costs, difficulty in customizing the energy storage capacity required by industrial users within the park, and low utilization rates of energy storage equipment, which significantly hinder the development of energy storage. Furthermore, the high charging prices associated with these charging stations have not been fully considered within the park's charging facilities (such as electric vehicle charging stations). In other words, since charging stations within the park are powered solely by the grid, electric vehicles face high charging prices.

[0004] At present, there have been some studies on the issue of energy utilization in industrial parks. For example, Wang Yongli et al., in the article "Two-layer Game Optimization Model of Multi-park Integrated Energy System Based on Electricity-Carbon Coupling", Renewable Energy, 2021, 23(02), took the multi-park integrated energy system as the research object and constructed a two-layer game optimization scheduling model. First, it comprehensively considered the carbon emissions generated by the industrial park in its production and operation activities, and constructed a ladder carbon-green certificate trading model that considered the equivalent offset mechanism; secondly, based on the actual cooperation situation of the industrial park, it constructed a multi-park game optimization model to study the dynamic pricing of system operators and the optimization operation scheduling of the industrial park; finally, through case analysis, it was verified that the constructed model can reduce the carbon emissions of the system while ensuring economy, and achieve the unity of economic and carbon emission reduction benefits. However, this paper mainly focuses on the game between multi-park integrated energy systems, and does not involve the energy scheduling problem within a single industrial park. Moreover, in the two-layer game model it adopts, the upper layer uses the improved Harris Hawk algorithm to iteratively solve dynamic pricing. The Harris Hawk algorithm is an intelligent algorithm. The mathematical convergence proof and stability analysis are still imperfect, lacking rigorous theoretical support, and requires multiple iterative calculations. The calculation not only requires a lot of computing power and time, but the calculation results may converge to the local optimal solution too early, and the global search capability is poor, which ultimately leads to unreasonable pricing results.

[0005] The shared energy storage model of a park's solar-storage-charging system allows all users to share the benefits of energy storage by sharing the storage costs and leveraging the complementarity of electricity loads. This model is expected to overcome the bottlenecks faced by energy storage and the high charging costs of charging facilities. In this context, considering the current grid and photovoltaic electricity prices, how to design a reasonable shared energy storage pricing mechanism to increase the benefits of shared energy storage in the park and reduce user energy costs, especially charging costs for industrial and commercial users and electric vehicle users within the park, has become a pressing issue. Summary of the Invention

[0006] The purpose of this invention is to provide a pricing method for shared energy storage in a park based on Stackelberg game, provide a reasonable shared energy storage pricing mechanism, improve the benefits of shared energy storage in the park, reduce the energy costs of the user group, optimize resource allocation, and achieve a balance between privacy protection and optimization utility.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A pricing method for shared energy storage in a campus solar-storage-charging system based on Stackelberg game theory includes:

[0009] Build a campus solar-storage-charging system, including shared energy storage power stations and user groups;

[0010] Construct an upper-level model with the shared energy storage power station as the leader, and determine the objective function and constraints of the upper-level model;

[0011] Construct a lower-level model with user groups as followers and determine the objective function and constraints of the lower-level model;

[0012] The upper and lower models are combined into a two-layer optimization model using the mathematical model of the Stackelberg game. A differential privacy perturbation mechanism is introduced to protect the privacy of the response data transmitted by the user group in the lower model to the upper model.

[0013] The KKT optimality condition is used to transform the two-level optimization model into a single-level optimization model;

[0014] The single-layer optimization model is solved using the solver to obtain the shared energy storage pricing.

[0015] The present invention provides a preferred solution, wherein the park shared energy storage pricing method further includes: constructing a dynamic pricing adjustment mechanism, and upon receiving feedback from a user group indicating dissatisfaction with the shared energy storage pricing or upon receiving a total power sold by the shared energy storage power station to the user group within a preset period being lower than the expected value, starting the dynamic pricing adjustment mechanism to regain the shared energy storage pricing.

[0016] The present invention provides a preferred solution, wherein the user group includes multiple industrial and commercial users equipped with photovoltaic power stations and electric vehicle charging stations; the upper-level model is constructed with the shared energy storage power station as the leader, and the objective function and constraints of the upper-level model are determined, specifically including:

[0017] Total revenue from shared energy storage power stations Maximize the objective function of the upper model to establish the target:

[0018]

[0019] Where, The income from selling electricity to the grid for shared energy storage power stations, The income from selling electricity to user groups for shared energy storage power stations, To generate revenue from selling electricity to electric vehicles for shared energy storage power stations, The cost of purchasing electricity from the grid for the shared energy storage power station, The cost of purchasing excess PV from the user base for shared energy storage plants;

[0020] Set the constraints of the upper-level model: set the state of charge continuity constraint, the non-simultaneous charging and discharging constraint, and the maximum charging and discharging power constraint of the shared energy storage power station.

[0021] The present invention provides a preferred solution, wherein the construction of a lower-level model with a user group as followers and the determination of the objective function and constraints of the lower-level model specifically include:

[0022]

[0023] Where, The cost of purchasing electricity from the grid for the user group, The cost of purchasing electricity from shared energy storage power stations for user groups; Revenue from selling excess PV to the grid for the user group; Revenue from selling excess photovoltaic power to shared energy storage power stations for user groups; the cost of charging electric vehicles from the grid; the cost of charging electric vehicles from shared energy storage stations;

[0024] Set the constraints of the lower-level model: set the load balance constraints of each user, the photovoltaic power balance constraints, the price range constraints for the electricity sold by the shared energy storage power station to the user group, the electric vehicle charging balance constraints, the electric vehicle charging power constraints and the power value range constraints.

[0025] The present invention provides a preferred solution, wherein the load balance constraints of each user and the photovoltaic power balance constraints are equality constraints, the electricity price range constraints sold by the shared energy storage power station to the user group and the power value range constraints are inequality constraints, and corresponding Lagrange multipliers are set in each constraint.

[0026] The present invention provides a preferred solution, which introduces a differential privacy perturbation mechanism to protect the privacy of response data transmitted by the user group in the lower-level model to the upper-level model, specifically by adding Gaussian noise to the response data.

[0027] The present invention provides a preferred solution, in which the KKT optimality condition is used to convert a double-layer optimization model into a single-layer optimization model, and the KKT optimality condition is used to convert the lower-layer model in the double-layer optimization model into an additional constraint of the upper-layer model, thereby realizing the conversion of the double-layer optimization model into a single-layer optimization model; the KKT optimality condition refers to that the derivative of the lower-layer model at the extreme point is zero, which is one of the KKT conditions.

[0028] The present invention provides a preferred solution, which uses the KKT optimality condition to convert a double-layer optimization model into a single-layer optimization model, specifically including the following steps: constructing a Lagrangian function of the lower-layer model; using the Lagrangian function to calculate the partial derivatives of each variable of the lower-layer model so that the lower-layer model meets the KKT optimality condition; according to the inequality constraints of the lower-layer model and the Lagrangian multipliers corresponding to the inequality constraints, obtaining the complementary relaxation conditions of the lower-layer model as additional constraints of the upper-layer model, thereby obtaining the single-layer optimization model.

[0029] The present invention provides a preferred solution, wherein a dynamic pricing adjustment mechanism is constructed, and when feedback is received from a user group indicating dissatisfaction with the shared energy storage pricing or the total power sold by the shared energy storage power station to the user group within a preset period is lower than the expected value, the dynamic pricing adjustment mechanism is activated to regain the shared energy storage pricing. The dynamic pricing adjustment mechanism is constructed, specifically including the following steps: data collection, including real-time collection of the output of the photovoltaic power station, the load of industrial and commercial users, the load of the electric vehicle charging station, the state of charge of the shared energy storage power station, and the power grid electricity price; prediction model construction, training, and output, including construction of a photovoltaic output prediction value model and a user demand load prediction model. The photovoltaic output prediction model and the user demand load prediction model are trained using historical data, and then the real-time collected data are input into the trained photovoltaic output prediction model and the user demand load prediction model to output the photovoltaic output prediction value and the user load demand prediction value for a period of time in the future; a dynamic pricing model is constructed and output, and a dynamic pricing model is constructed with rolling time domain optimization as the framework. The parameters are initialized according to the current state of charge of the shared energy storage power station, the grid electricity price, the photovoltaic output prediction value and the user load demand prediction value, and then the solution is optimized to output the shared energy storage electricity price sequence for a period of time in the future.

[0030] Compared with the existing technology, the above technical solution has the following advantages:

[0031] This paper proposes a pricing method for shared energy storage in a campus solar-storage-charging system based on the Stackelberg game. Leveraging Stackelberg game theory, this paper provides a framework for analyzing the behavior of market leaders and followers. Within this framework, a shared energy storage pricing mechanism is proposed, in which shared energy storage power stations serve as leaders and user groups (electric vehicle charging stations, industrial and commercial users) serve as followers. By incorporating Stackelberg game theory, this mechanism simulates the behavior of leaders and followers in the market, enabling a more rational and efficient pricing strategy. A two-layer optimization model is constructed, with an upper-level model centered on shared energy storage as the leader and a lower-level model centered on the campus user groups as followers. This model more comprehensively considers the objectives and constraints of all participants in the system. The KKT optimality condition is used to transform the two-layer optimization model into a single-layer optimization model. The interaction between the upper and lower models leads to a global optimal solution, thereby optimizing resource allocation. This allows all parties in the campus to reach an optimal profit distribution plan through competition and cooperation, maximizing the benefits of the energy storage operator while reducing energy costs for the user groups. Not only that, it can also promote the commercial operation of the park's photovoltaic storage and charging system, achieving efficient energy utilization and sustainable environmental development.

[0032] Furthermore, the present invention introduces a differential privacy perturbation mechanism to protect the privacy of the response data transmitted by the user group in the lower-level model to the upper-level model, prevent individual data leakage, protect commercially sensitive information, and achieve a balance between privacy protection and optimization utility. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0034] Figure 1 A flowchart of a method for pricing solar-storage-charging shared energy storage in a park based on Stackelberg game provided in a specific embodiment of the present invention;

[0035] Figure 2 A specific embodiment of the present invention provides a pricing method for a campus photovoltaic storage and charging shared energy storage based on Stackelberg game, which is applied to the load and photovoltaic output diagram of user 1 in a scenario;

[0036] Figure 3 A specific embodiment of the present invention provides a campus photovoltaic storage and charging shared energy storage pricing method based on Stackelberg game, which is applied to the load and photovoltaic output diagram of user 2 in a scenario;

[0037] Figure 4 A specific embodiment of the present invention provides a pricing method for a campus photovoltaic storage and charging shared energy storage based on Stackelberg game theory, which is applied to the load and photovoltaic output diagram of user 3 in a scenario;

[0038] Figure 5 A specific embodiment of the present invention provides a pricing method for a campus photovoltaic storage and charging shared energy storage based on Stackelberg game theory, which is applied to the load and photovoltaic output diagram of user 4 in a scenario;

[0039] Figure 6 A specific embodiment of the present invention provides a method for pricing a campus photovoltaic storage and charging shared energy storage based on Stackelberg game theory, applied to a photovoltaic power consumption diagram of user 1 in a scenario;

[0040] Figure 7 A specific embodiment of the present invention provides a method for pricing a campus photovoltaic storage and charging shared energy storage based on Stackelberg game theory, applied to the photovoltaic power consumption diagram of user 2 in a scenario;

[0041] Figure 8 A photovoltaic power consumption diagram of user 3 C in a scenario where a campus photovoltaic storage and charging shared energy storage pricing method based on Stackelberg game is applied, provided in a specific embodiment of the present invention;

[0042] Figure 9 A specific embodiment of the present invention provides a method for pricing a campus photovoltaic storage and charging shared energy storage based on Stackelberg game theory, applied to a photovoltaic power consumption diagram of user 4 in a scenario;

[0043] Figure 10 A specific embodiment of the present invention provides a method for pricing a campus solar-storage-charging shared energy storage based on Stackelberg game theory, applied to a load balance diagram of user 1 in a scenario;

[0044] Figure 11 A specific embodiment of the present invention provides a method for pricing a campus solar-storage-charging shared energy storage based on Stackelberg game theory, applied to a load balance diagram of user 2 in a scenario;

[0045] Figure 12 A specific embodiment of the present invention provides a method for pricing a campus solar-storage-charging shared energy storage based on Stackelberg game theory, applied to a load balance diagram of user 3 in a scenario;

[0046] Figure 13 A specific embodiment of the present invention provides a method for pricing a campus solar-storage-charging shared energy storage based on Stackelberg game theory, applied to a user 4 load balance diagram in a scenario;

[0047] Figure 14 A specific embodiment of the present invention provides a pricing method for a campus photovoltaic storage and charging shared energy storage based on Stackelberg game, which is applied to the charging and discharging power of a shared energy storage power station in a scenario;

[0048] Figure 15 A specific embodiment of the present invention provides a method for pricing a campus solar-storage-charging shared energy storage system based on Stackelberg game theory and is applied to a load diagram of an electric vehicle charging station in a scenario;

[0049] Figure 16 A specific embodiment of the present invention provides a method for pricing a campus photovoltaic storage and charging shared energy storage based on Stackelberg game, which is applied to a load balance curve of an electric vehicle charging station in a scenario. DETAILED DESCRIPTION

[0050] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0051] Example 1

[0052] Please refer to Figure 1 In a preferred embodiment, a pricing method for shared energy storage based on the Stackelberg game is provided, which is mainly implemented through the following steps:

[0053] S1. Build a campus solar-storage-charging system, including a shared energy storage power station and a user group; the user group includes multiple commercial and industrial users equipped with photovoltaic power stations and electric vehicle charging stations. The shared energy storage power station is referred to below and in the accompanying drawings as "energy storage" or "energy storage power station," and the photovoltaic power station is referred to below and in the accompanying drawings as "photovoltaic."

[0054] S2. Build an upper-level model with the shared energy storage power station as the leader and determine the objective function and constraints of the upper-level model;

[0055] S3. Build a lower-level model with the user group as followers and determine the objective function and constraints of the lower-level model;

[0056] S4. Using the mathematical model of the Stackelberg game, we combine the upper and lower models into a two-layer optimization model. We also introduce a differential privacy perturbation mechanism to protect the privacy of the response data transmitted by the user group in the lower model to the upper model.

[0057] S5. Use KKT optimality conditions to transform the two-level optimization model into a single-level optimization model;

[0058] S6. Use the solver to solve the single-layer optimization model and obtain the shared energy storage pricing.

[0059] Based on steps S1 to S5 of the above embodiment, a more preferred and detailed implementation method is provided:

[0060] S1. Construct a park-based solar-storage-charging system, including a shared energy storage power station and a user group. In one embodiment, taking a user group consisting of multiple commercial and industrial users equipped with a photovoltaic power station and an electric vehicle charging station as an example, the shared energy storage power station is designated as the leader, and the user group of multiple commercial and industrial users equipped with a photovoltaic power station and an electric vehicle charging station is designated as the follower. The power load of all electric vehicle charging stations within the park can also be considered the power demand of all electric vehicles within the park. Electric vehicles obtain power from the shared energy storage power station through the electric vehicle charging stations for charging.

[0061] S2. Construct an upper-level model with the shared energy storage power station as the leader, and determine the objective function and constraints of the upper-level model.

[0062] S21. Total revenue from shared energy storage power stations Maximize the objective function of the upper model to establish the target:

[0063] (1)

[0064] Where, To share the income from the sale of electricity from the energy storage power station to the grid, To generate revenue from electricity sales to user groups for shared energy storage power stations, To generate revenue from selling electricity to electric vehicles for shared energy storage power stations, The cost of purchasing electricity from the grid for shared energy storage power stations, Purchase excess photovoltaic costs from the user base for shared energy storage power stations.

[0065] In a preferred embodiment, the revenue from electricity sales from the shared energy storage power station to the grid is , Shared energy storage power station sales revenue to user groups , Revenue from electricity sales from shared energy storage power stations to electric vehicles 2. The cost of purchasing electricity from the grid for a shared energy storage power station 2. Shared energy storage power stations purchase excess photovoltaic costs from user groups The calculation formulas are as follows:

[0066] (2)

[0067] (3)

[0068] (4)

[0069] (5)

[0070] (6)

[0071] Where, The price of electricity sold by the shared energy storage power station to the grid at time t; The power sold by the shared energy storage power station to the grid at time t; The electricity price sold by the shared energy storage power station to the user group (industrial and commercial users or electric vehicles) at time t; The power sold by the shared energy storage power station to user i at time t; is the “peak-flat-valley” electricity price sold by the power grid to the user group at time t; is the charging power of the i-th electric vehicle through the shared energy storage power station at time t; The price of electricity sold by the power grid to the shared energy storage power station at time t; The power sold by the grid to the shared energy storage power station at time t; The price at which the user group sells excess photovoltaic power to the shared energy storage power station at time t; The power generated when user i sells excess photovoltaic power to the shared energy storage power station at time t, where T is the time period and I is the user group.

[0072] In a more preferred implementation, the calculation of the total revenue of a shared energy storage power station can include additional variables, such as user service fee revenue, energy storage investment, and operation and maintenance costs. This allows for a multi-faceted revenue structure and cost-sharing mechanism. This can further enhance the economic viability and risk resilience of the shared energy storage pricing model, making it particularly suitable for power systems with a high proportion of renewable energy access. Future applications of technologies such as virtual power plants and blockchain ledger sharing will further unlock this revenue potential.

[0073] S22. Set the constraints of the upper-level model (upper-level constraints): Set the state of charge continuity constraint, the non-simultaneous charging and discharging constraint, and the maximum charging and discharging power constraint of the shared energy storage power station.

[0074] 1) State of charge continuity constraint:

[0075] (7)

[0076] (8)

[0077] (9)

[0078] (10)

[0079] Where, , Respectively represent the charging and discharging efficiency of the shared energy storage power station; , They represent the charging and discharging power of the shared energy storage power station at time t respectively; Indicates the maximum capacity of the shared energy storage power station; set Indicates the initial power of the shared energy storage power station before use; Share the power of energy storage power stations at the last moment; The power of the shared energy storage power station at time t; is the amount of electricity consumed by the shared energy storage power station at the last moment t. A time period (dispatching period) T has 24 moments in total, each moment lasting one hour, so we set , that is, the initial power of the shared energy storage power station before use is equal to the power of the shared energy storage power station at the last moment.

[0080] 2) Simultaneous charging and discharging constraints and maximum charging and discharging power constraints:

[0081] (11)

[0082] (12)

[0083] (13)

[0084] (14)

[0085] Where, 、 These are the charging and discharging Boolean variables of the shared energy storage power station, respectively. Boolean variables are 0 and 1, which can represent status bits; Indicates the maximum charge and discharge power of the shared energy storage station. When the shared energy storage station is charging, the charging status is The value is 1, then is 0, The lower limit is 0, which means that the shared energy storage power station cannot discharge. Similarly, when the shared energy storage is discharged, the charging state is The value is 1, then is 0, If the lower limit is 0, the shared energy storage station cannot be charged. This makes it impossible for the shared energy storage station to provide charging and discharging services at the same time.

[0086] S3. Construct a lower-level model with user groups as followers, and determine the objective function and constraints of the lower-level model.

[0087] S31. Based on the electricity cost of user groups The objective function of the lower model is established by minimizing the electricity cost of industrial and commercial users + the charging cost of electric vehicles:

[0088] (15)

[0089] Where, The cost of purchasing electricity from the grid for the user group, The cost of purchasing electricity from shared energy storage power stations for user groups; Revenue from selling excess PV to the grid for the user group; Revenue from selling excess photovoltaic power to shared energy storage power stations for user groups; the cost of charging electric vehicles from the grid; The cost of charging electric vehicles from shared energy storage stations.

[0090] In a preferred embodiment, the cost of electricity purchased by the user group from the power grid is , the cost of electricity purchased by user groups from shared energy storage power stations , the user group will sell excess photovoltaic power to the grid income , user groups sell excess photovoltaic power to shared energy storage power stations to generate income , the cost of charging electric vehicles from the grid , the cost of charging electric vehicles from shared energy storage power stations The calculation formulas are as follows:

[0091] 1) Cost of electricity purchased from the power grid by the user group :

[0092] (16)

[0093] Where, is the peak-flat-valley electricity price sold by the power grid to the user group at time t; is the power sold by the grid to the user group at time t.

[0094] 2) Cost of electricity purchased by user groups from shared energy storage power stations :

[0095] (17)

[0096] Where, The electricity price sold by the shared energy storage power station to the user group at time t; The shared energy storage power station sells electricity to user i at time t.

[0097] 3) User groups sell excess photovoltaic power to the grid for income :

[0098] (18)

[0099] Where, is the excess photovoltaic grid-connected electricity price of the user group at time t; is the excess photovoltaic grid-connected power of user i at time t.

[0100] 4) User groups sell excess photovoltaic power to shared energy storage power stations to generate revenue :

[0101] (19)

[0102] Where, The price at which the user group sells excess photovoltaic power to the shared energy storage power station at time t; The power generated by user i at time t when he sells excess photovoltaic power to the shared energy storage power station.

[0103] 5) Electric vehicle charging cost from the grid :

[0104] (20)

[0105] Where, is the peak-flat-valley electricity price sold by the power grid to the user group at time t; is the charging power of the electric vehicle through the grid at time t.

[0106] 6) Charging costs of electric vehicles from shared energy storage power stations :

[0107] (twenty one)

[0108] Where, The electricity price sold by the shared energy storage power station to the user group at time t; is the charging power of the i-th electric vehicle through the shared energy storage power station at time t.

[0109] S32. Set the constraints of the lower-level model (lower-level constraints): set the load balance constraints of each user, the photovoltaic power balance constraints, the price range constraints of the shared energy storage power station sold to the user group, the electric vehicle charging balance constraints, the electric vehicle charging power constraints and the power value range constraints.

[0110] 1) Load balancing constraints for each user:

[0111] (twenty two)

[0112] Where, is the load power of user i at time t, Provides photovoltaic power to users. The power purchased by user i from the shared energy storage power station at time t, The power purchased by user i from the grid at time t; is the Lagrange multiplier for the load balance constraint.

[0113] 2) Photovoltaic power balance constraints:

[0114] (twenty three)

[0115] Where, is the photovoltaic power generation power of user i at time t, The photovoltaic power supply of user i at time t is: is the power sold by user i to the grid at time t, The power sold by user i to the shared energy storage power station at time t, is the Lagrange multiplier for the photovoltaic power balance constraint.

[0116] 3) Electric vehicle charging balance constraints:

[0117] (twenty four)

[0118] Where, is the charging power of the i-th electric vehicle at time t, is the charging power of the i-th electric vehicle through the shared energy storage station at time t, is the charging power of the i-th electric vehicle through the grid at time t; Lagrange multipliers for the charging balance constraints of electric vehicles.

[0119] 4) Constraints on the price range of electricity sold by shared energy storage power stations to user groups:

[0120] (25)

[0121] is the electricity price sold by the shared energy storage power station to the user group at time t, : The Lagrange multiplier corresponding to the price range constraint of electricity sold by the shared energy storage power station to the user group.

[0122] 5) Electric vehicle charging power constraints:

[0123] (26)

[0124] Where, 、 are the power of the i-th electric vehicle at time t and when it is connected to charging; Charging efficiency for electric vehicles; The charging power of the i-th electric vehicle at time t; Charging time for the i-th electric vehicle; 、 The time when charging starts and ends for the i-th electric vehicle; 、 The minimum and maximum values ​​of the power of the i-th electric vehicle at the end of charging.

[0125] 6) Constraints on the range of power values:

[0126] (27)

[0127] Where, is the load power of user i at time t, The maximum power of photovoltaic power supply to the grid for user i, The maximum power supplied by user i's photovoltaic system to the shared energy storage station, The maximum power that the shared energy storage power station supplies to the grid, The maximum power that the grid supplies to the shared energy storage power station, is the Lagrange multiplier for the power range constraint.

[0128] S4. Using the mathematical model of Stackelberg game, the upper and lower models are combined into a two-layer optimization model. The KKT optimality condition is used to transform the two-layer optimization model into a single-layer optimization model. A differential privacy perturbation mechanism is introduced to protect the privacy of the response data transmitted by the user group in the lower model to the upper model. The two-layer optimization model using the mathematical model of Stackelberg game is expressed as:

[0129] (28)

[0130] is the objective function of the leader's upper model, which is the above formula (1). is the objective function of the follower's lower model, which is the above formula (15). is the leader’s inequality constraint, is the leader's equality constraint, represents the set of inequality constraint indices of the leader, represents the set of equality constraint indices of the leader, is the inequality constraint for the follower, is the equality constraint for the follower, represents the set of inequality constraint indices of the followers, represents the set of equality constraint indices of followers, and are the indices of inequalities and equalities, respectively, and denote the decision variables of the objective functions of the two models, and and All satisfied , is the strategy set of the Stackelberg game, that is, the set of all variables involved in the game.

[0131] In a preferred embodiment, a differential privacy perturbation mechanism is introduced to protect the privacy of the response data transmitted by the user group in the lower-level model to the upper-level model. This is achieved by adding Gaussian noise to the response data. The specific implementation process is as follows:

[0132] S4a. Determine the response data (sensitive data) and the disturbance timing. In this embodiment, first, the response data transmitted by the user group to the upper model can be: the power purchased by the user group from the shared energy storage power station (such as the power sold by the shared energy storage power station to user i at time t). ), the photovoltaic power sold by the user group to the shared energy storage power station (for example, at time t, user i sells excess photovoltaic power to the shared energy storage power station ), the power purchased by the user group from the grid (e.g., the power sold by the grid to the user group at time t ), the photovoltaic power sold by the user group to the grid (such as the excess photovoltaic power of user i at time t ). Secondly, for the perturbation position, the user group uses the time before passing the response data to the upper-level model as the perturbation timing, and a differential privacy mechanism is added to the corresponding data.

[0133] S4b. Design a differential privacy perturbation mechanism. Use a Gaussian noise mechanism that satisfies differential privacy, that is, add Gaussian noise to the corresponding data above, and the noise intensity is determined by the privacy budget. and Take the purchase of electricity from the shared energy storage power station by the user group as an example, that is, at time t in the above embodiment, the shared energy storage power station sells electricity to user i. For example, the corresponding data after disturbance is , obtained by the following formula:

[0134] (29)

[0135] Where, represents Gaussian distribution (normal distribution), where the mean (0) indicates that the expected value of the noise is 0, that is, the noise is symmetrically distributed around the true value; the variance ( ) describes the discreteness of the noise, The larger it is, the stronger the noise added, the better the privacy protection effect, but the data accuracy decreases; The smaller it is, the weaker the noise is and the data is closer to the true value, but the privacy protection ability is reduced. According to the privacy budget and and sensitivity Determine, and obtain it through the following formula:

[0136] (30)

[0137] Where, Indicates the strength of privacy protection. The smaller the value, the stricter the privacy protection and the lower the risk of privacy leakage. The larger the value, the looser the privacy protection and the higher the data utility, but the greater the privacy risk. represents the probability of violating strict privacy protection, which is usually set to a very small value (such as ); Indicates that differential privacy is strictly satisfied, without exceptions; Indicates that the probability Privacy leakage occurs, which is used to relax the constraints in the Gaussian mechanism. Example: Assume that the sensitivity of the user's purchased electricity power is , , set a privacy budget , then the noise standard deviation is: At this time, Gaussian noise with a mean of 0 and a standard deviation of 1114 kW is added to the user response data to ensure that attackers cannot reversely infer individual data, while the optimization results still maintain statistical validity.

[0138] S4c. Adaptive adjustment of the objective function and constraints of the upper model. In a more preferred embodiment, since noise will affect the profit calculation of the shared energy storage power station, it is necessary to adaptively adjust the objective function and constraints of the upper model, including objective function correction, such as the revenue from selling electricity to the user group in the shared energy storage power station. , Revenue from electricity sales from shared energy storage power stations to electric vehicles In the term, use the perturbed data Instead of original data At the same time, in order to address the power limit problem that may be caused by noise, the original hard constraints are relaxed into probabilistic constraints or robust constraints. For example, the constraints are satisfied with a certain confidence level, that is, the constraints are allowed to be satisfied within a certain probability range, rather than strictly and absolutely. For hard constraints, for example, P≤100 kW must be strictly satisfied. The probabilistic constraints adopted in this preferred embodiment allow the constraints to be satisfied with a certain probability, for example, , that is, the constraint is established in 95% of cases. The above probability is the confidence level, which can be 95%, indicating the degree of confidence in the establishment of the constraint. When Gaussian noise is added to the user response data, the original data becomes a random variable , , at this time: hard constraint Frequent violations may occur due to noise fluctuations. By using probabilistic constraints, a small number of violations, such as 5%, can be tolerated while ensuring that the constraints hold in most cases, thus balancing data perturbations with system security.

[0139] Take the power purchased by the user group from the shared energy storage power station as an example, that is, at time t in the above embodiment, the shared energy storage power station sells power to user i For example, such as:

[0140] Original hard constraints: . Probability constraint after noise addition: ;in is the confidence level, such as 95% corresponds to The more specific transformation steps are as follows:

[0141] Step 1. Assume the noise distribution: If the noise obeys Gaussian distribution ,but obey .

[0142] Step 2: Standardization: Definition ,but .

[0143] Step 3: Rewrite the probability constraint: .

[0144] Step 4: Use the quantile function to transform and obtain the constraints: For the standard normal distribution, we have is a critical value, such as hour, The final constraint is transformed into: Example: Assume that the load power of user i at time t is , which is also the upper limit of the user's power purchase, and the noise standard deviation , requiring a confidence level of 95%, that is , calculate the adjusted constraints: .

[0145] In the preferred implementation described above, by introducing probabilistic constraints to ensure that constraints are met at a certain confidence level, the problem of constraining random variables after noise perturbation is transformed into a deterministic optimization problem. This approach ensures the security and economic efficiency of system operation while protecting privacy, and is a key technical means to balance privacy protection and optimization utility.

[0146] S5. Use the KKT optimality condition to transform the two-level optimization model into a single-level optimization model.

[0147] The KKT optimality condition is used to transform the lower-level model into an additional constraint on the upper-level model, thereby converting the two-level optimization model into a single-level optimization model. The KKT optimality condition, which states that the derivative of the lower-level model at the extreme point is zero, is one of the KKT conditions. The KKT conditions consist of a system of equations consisting of the original feasibility condition, the optimality condition, and the complementary relaxation condition.

[0148] In a preferred embodiment, the KKT optimality condition is used to transform the two-level optimization model into a single-level optimization model, which specifically includes the following steps:

[0149] S51. Construct the Lagrangian function of the lower model. Combining formulas (16), (14), (18), (19), (20), (21), (22), (23), (24), (25), (26) and (27), the Lagrangian function of the lower model is obtained. The specific expression is as follows:

[0150] (31)

[0151] S52. Use the Lagrangian function to find the partial derivatives of the variables of the lower model. The partial derivatives of the Lagrangian function are 0. , so that the lower model meets the KKT optimality conditions, as follows:

[0152] (32)

[0153] S53. Based on the constraints of the lower model and its corresponding Lagrange multiplier, the complementary relaxation condition of the lower model is obtained as an additional constraint of the upper model to obtain the single-layer optimization model. The complementary relaxation condition expression is:

[0154] (33)

[0155] Where: and At most one of them can be strictly greater than 0; and At most one of them can be strictly greater than 0. At this point, the lower-level model is converted into an additional constraint of the upper-level model through the KKT condition, completing the transformation of the two-level optimization model into a single-level optimization model using the KKT optimality condition.

[0156] S54. Linearize the transformed single-layer optimization model.

[0157] S541. Use the Big-M method to linearize the complementary relaxation conditions (nonlinear terms) in the transformed single-layer optimization model.

[0158] The details are as follows: By introducing Boolean variables and a sufficiently large positive number M transforms the constraint (26) of the complementary slack condition into the following linear inequality:

[0159] (34)

[0160] S542. Use the Boolean expansion method to linearize the objective function of the single-layer optimization model to eliminate the bilinear term obtained by multiplying the electricity price and power. The details are as follows: Shared energy storage power station sells power to user i at time t After discretization, it can be expressed as:

[0161] (35)

[0162] On this basis, the new constraints are introduced as follows:

[0163] (36)

[0164] In formulas (35) and (36), Indicates the length of the segment; Indicates the maximum length of the charging power segment. Indicates the minimum length of the charging power segment, K is the number of segments for the charging power; and are the maximum and minimum prices of electricity sold by the shared energy storage power station to users; is the selected Boolean variable; represents a continuous variable.

[0165] The Boolean expansion method discretizes continuous variables into combinations of Boolean variables, "decomposing" the original bilinear terms into multiple simple Boolean product terms; then new variables and linear constraints are introduced to ensure that the calculation results of these "decomposed forms" are equivalent to the original bilinear expressions.

[0166] Problems involving bilinear terms are typically non-convex optimization problems. Due to their non-convex nature, they often have multiple local optimal solutions, making it difficult to guarantee a global optimal solution through direct solution. By optimizing and processing these complex problems involving bilinear terms, we can transform them into mixed-integer quadratic programming problems and directly solve them using an established mixed-integer optimization solver (Gurobi).

[0167] S6. Use the solver to solve the single-layer optimization model to obtain the shared energy storage pricing. In a preferred embodiment, the single-layer optimization model is solved by calling the YALMTP toolbox and the GUROBI commercial solver using Matlab software to obtain the shared energy storage pricing. The selection of the above solvers and toolboxes can solve complex optimization problems efficiently and accurately. The YALMTP toolbox and the GUROBI solver perform well in processing mixed-integer quadratic programming problems and can ensure the accuracy and reliability of the solution results.

[0168] In combination with the above embodiments, the present invention can achieve the following beneficial technical effects:

[0169] 1. The present invention proposes a pricing method for shared energy storage in a park based on the Stackelberg game based on the above-mentioned specific implementation methods. In this mechanism, shared energy storage acts as a leader and the user group acts as a follower. By introducing the Stackelberg game theory, it is possible to simulate the behavior of leaders and followers in the market and achieve a more reasonable and efficient pricing strategy. An upper-level model with shared energy storage as the leader and a lower-level model with the park user group as the follower are constructed to form a two-layer optimization model. The two-layer optimization model can more comprehensively consider the goals and constraints of each participant in the system, and use the KKT optimality condition to transform the two-layer optimization model into a single-layer optimization model. The global optimal solution is found through the interaction between the upper and lower models, thereby optimizing resource allocation, ensuring the maximization of the interests of energy storage operators, and reducing the energy costs of the user group.

[0170] 2. Model simplification and solution efficiency: The present invention uses the KKT optimality condition to transform the double-layer optimization model into a single-layer optimization model. By applying the KKT condition, the complex double-layer optimization problem is simplified to a single-layer optimization problem, which can be directly solved by a mature optimization solver. After the double-layer model is transformed into a single-layer model, the present invention can directly solve it by a mature optimization solver, avoiding the complexity of nested iteration required to solve the double-layer model through intelligent algorithms, reducing the difficulty of solution, and significantly improving computational efficiency.

[0171] 3. Guarantee of global optimal solution: The KKT condition ensures the consistency of the optimal solution between the converted single-layer model and the original two-layer model, avoiding the local optimal problem caused by information asymmetry in the two-layer model, thereby making the pricing of shared energy storage in the solar-storage-charging system within the park more reasonable.

[0172] 4. Privacy-preserving design: This paper introduces a differential privacy perturbation mechanism to protect the privacy of the response data transmitted by the user group in the lower-level model to the upper-level model. By embedding differential privacy into the response transmission link of the Stackelberg game, this solution prevents individual data leakage and protects commercially sensitive information without reconstructing the original method framework, thus achieving a balance between privacy protection and optimization utility.

[0173] 5. This invention linearizes nonlinear problems under complementary relaxation conditions, converting nonlinear terms in a single-level programming model into linear terms using the Big-M method. This linearization makes the model easier to solve while ensuring the stability and convergence of the solution. Furthermore, by introducing Boolean variables and new constraints, the problem containing bilinear terms is successfully transformed into a mixed-integer quadratic programming problem, further simplifying the solution process.

[0174] In summary, the present invention constructs a two-layer optimization model through the Stackelberg game framework, which can more accurately simulate the interactive relationship between the various participants in the campus photovoltaic storage and charging system. The leader (shared energy storage) influences the behavior of followers (user groups) by formulating pricing strategies, and the followers adjust their own electricity consumption behavior according to the leader's pricing strategy. This interactive relationship makes the pricing strategy more reasonable and efficient. The application of KKT conditions enables the two-layer optimization model to be converted into a single-layer optimization model for solution. The linearization process further simplifies the solution process of the single-layer model. The synergistic effect of the two makes the solution of complex optimization problems more efficient and accurate.

[0175] The following examples are given in combination with specific application scenarios:

[0176] The peak-flat-valley electricity prices sold by the power grid to users , the price of electricity sold by the power grid to shared energy storage power stations , the price of electricity sold to the grid by shared energy storage power stations , the price of photovoltaic electricity sold to shared energy storage power stations , Photovoltaic grid-connected electricity price are all given, and the price of electricity sold by the shared energy storage power station to the user group is the quantity to be optimized, that is, the pricing of the shared energy storage.

[0177] Among them, the specific values ​​of the given electricity prices are shown in Table 1:

[0178] Table 1: Electricity price parameters

[0179]

[0180] The relevant parameters of the shared energy storage power station are shown in Table 2:

[0181] Table 2: Shared energy storage power station parameters

[0182]

[0183] Each of the four industrial users is equipped with photovoltaic power generation, which can supply part of their own electricity. The remaining electricity can be connected to the grid or sold to a shared energy storage station.

[0184] The load and photovoltaic output curves of users 1 to 4 are as follows: Figure 2 、 Figure 3 、 Figure 4 、 Figure 5 As shown, users 1 through 4 all have photovoltaic power generation, but users 1 through 3 have larger photovoltaic systems. During periods of peak photovoltaic output (11:00 AM to 2:00 PM), photovoltaic power generation exceeds the user's own load. At this time, the user needs to consider selling excess photovoltaic power to energy storage or the grid. Energy storage systems must meet two operational constraints: the principle of mutual exclusion between charging and discharging, which prohibits simultaneous charging and discharging at any given time; and maximum charge and discharge power limits, which require that power input and output not exceed the rated upper limit. During other periods, when photovoltaic output is insufficient to cover the user's load, the user must purchase electricity from the energy storage system or the grid to meet their electricity needs. This is also subject to the energy storage system's discharge power limit. Therefore, the rational allocation and scheduling of electricity requires an optimization model that comprehensively considers load, photovoltaic output, electricity price, and energy storage constraints. For user 4, the photovoltaic system is relatively small, and even at peak photovoltaic output, it cannot meet the user's load.

[0185] The distribution curves of photovoltaic power flow from user 1 to user 4 (photovoltaic A, photovoltaic B, photovoltaic C, photovoltaic D) are as follows: Figure 6 、 Figure 7 、 Figure 8 、 Figure 9As shown, the electricity generated by a user's own photovoltaic power generation can be supplied to their own loads, sold to the grid, or sold to shared energy storage power stations, with priority given to self-use. To encourage local consumption of photovoltaic power, the price of photovoltaic power sold to shared energy storage power stations is higher than the grid-connected price. Therefore, photovoltaic power generation is preferentially sold to shared energy storage power stations. Once a shared energy storage power station reaches its maximum charge or capacity limit, photovoltaic power generation will also be connected to the grid. In the figure, the amount of electricity supplied by photovoltaic power station A refers to the amount of electricity supplied by photovoltaic power station A to user 1, the amount of electricity supplied by photovoltaic power station B to user 2, the amount of electricity supplied by photovoltaic power station C to user 3, and the amount of electricity supplied by photovoltaic power station D to user 4. The users here refer to the industrial and commercial users in the user group.

[0186] For users 1 through 3, when PV output is low, all of this power is used by the user. From 11:00 to 14:00, when PV output is high, any surplus power left after using it will be sold to the shared energy storage power station or the grid. This is consistent with the inherent characteristics of PV and shared energy storage power stations. For user 4, however, due to its low PV output, all of this power is used by the user.

[0187] The load balance curves of users 1 to 4 are as follows: Figure 10 、 Figure 11 、 Figure 12 、 Figure 13 As shown, Figures 10 to 13 Here, "electric power supplied by photovoltaics" refers to "electric power supplied by photovoltaic A to user 1, electric power supplied by photovoltaic B to user 2, electric power supplied by photovoltaic C to user 3, and electric power supplied by photovoltaic D to user 4." For users 1 to 3, when the photovoltaic output can meet their own load, photovoltaics will supply electricity for their own use. When the photovoltaic output cannot fully meet their own load, users will choose to purchase electricity from the grid or energy storage power stations. When the grid's peak electricity price is high, users tend to purchase electricity from shared energy storage; when the grid's valley electricity price is low, users tend to purchase electricity from the grid. At this time, the shared energy storage power station also needs to purchase electricity from the grid to charge itself. Due to the constraint that the shared energy storage power station cannot charge and discharge simultaneously, the energy storage power station will choose the charging and discharging behavior that will benefit its own profitability. For user 4, due to its own small photovoltaic scale, it cannot meet its own load demand at any time, and it needs to consider purchasing electricity from the grid or the shared energy storage power station.

[0188] The charging and discharging power and charge status of the shared energy storage power station are as follows: Figure 14As shown in the figure, "PV power sold to energy storage" refers to the total PV power sold to the shared energy storage station by PV A, PV B, PV C, and PV D; "energy storage power sold to users" refers to the total power sold by the shared energy storage station to all commercial and industrial users, including users 1, 2, 3, and 4. The charging and discharging behavior of the shared energy storage station shows that during off-peak hours (00:00 to 8:00 and 21:00 to 24:00), when grid prices are low, the shared energy storage station purchases electricity from the grid and sells it to the grid or users when grid prices are higher. This overall strategy demonstrates a "low storage, high discharge" strategy, which aligns with the actual operating characteristics of energy storage stations. Due to the maximum charge and discharge power limits of the energy storage station, both the maximum charge and discharge power are 1600kW; the maximum state of charge of the shared energy storage station is 1600kWh.

[0189] Combined with the given parameters in Table 1 and Table 2, and Figures 2 to 5 The specific data of load and photovoltaic output of each user at each time are used. The optimal electricity price sold by the shared energy storage power station to the user group is obtained by applying the campus photovoltaic storage and charging shared energy storage pricing method based on Stackelberg game in the above embodiment of the present invention, as shown in Table 3.

[0190] Table 3 Electricity prices sold to users by shared energy storage power stations

[0191]

[0192] During peak and off-peak electricity prices, shared energy storage power stations sell electricity to users at a price lower than the grid price, attracting users to purchase electricity from them. During off-peak electricity prices, shared energy storage power stations will appropriately raise their own prices to avoid users purchasing electricity from them during off-peak periods. However, due to the shared energy storage station's inability to charge and discharge simultaneously, shared energy storage power stations must charge at night when grid prices are lower. This strategy, known as "low charge, high discharge," maximizes profits.

[0193] In addition, considering the rapid popularization of electric vehicles (EVs) and the fact that EV charging stations are new types of power load points, their concentrated charging demand during peak hours can lead to a sharp increase in grid load, impacting the stability and power supply capacity of the grid. In particular, the construction of fast charging stations, due to their high-power charging requirements, places higher demands on the instantaneous power supply of the grid. Therefore, in a more preferred embodiment, as one of the member types of the park's solar storage and charging system, EV charging stations can also be included in the user group and participate in the game. Figure 15 The load of the electric vehicle charging station is characterized by low load in the morning and evening, and high load at other times. This is because electric vehicle users charge their vehicles after get off work at night, so the load is also high at night. The load balance curve of the electric vehicle charging station is as follows: Figure 16As shown in the figure, during peak hours of grid electricity prices (8:00-12:00 and 17:00-21:00), consumers will choose to purchase electricity from shared energy storage power plants. The results of this example show that during these hours, the price of electricity sold by shared energy storage power plants to users is lower than the grid price, thus attracting users to purchase electricity from shared energy storage power plants. During off-peak and normal times of grid electricity prices, consumers will purchase more electricity from the grid.

[0194] Example 2: During the actual application of the above embodiment, it will be found that the pricing given by the shared energy storage power station may not satisfy the users, resulting in users being unwilling to use the shared energy storage power station for power supply. For example, a user interaction module can be set up in the shared energy storage power station, and users can provide feedback on the pricing at any time, or set an expected value for the total power sold by the shared energy storage power station to the user group within a preset period. If it is lower than the expected value within a certain period, it can also indirectly indicate that the pricing given by the shared energy storage power station may not satisfy the users, resulting in a decrease in the user's utilization rate of the shared energy storage power station. This embodiment only provides two conditions for triggering the dynamic pricing adjustment mechanism, which is an example of the actual application of the mechanism of the present invention, but is not limited to this. To this end, considering that it is difficult to respond to photovoltaic output fluctuations and changes in user demand in real time based on fixed-time or periodic optimization, a more preferred solution is given on the basis of Example 1, which provides a pricing dynamic adjustment mechanism. That is, after using the solver to solve the single-layer optimization model to obtain the shared energy storage pricing, an additional step is added: S7. Construct a pricing dynamic adjustment mechanism, and when feedback is received from the user group that they are dissatisfied with the shared energy storage pricing or the total power sold by the shared energy storage power station to the user group within the preset period is lower than the expected value, the pricing dynamic adjustment mechanism is activated to regain the shared energy storage pricing. The specific implementation process is as follows:

[0195] S71. Data collection: Real-time data collection includes information on photovoltaic power plant output, industrial and commercial user loads, electric vehicle charging station loads, the state of charge of shared energy storage power plants, and grid electricity prices (in this embodiment, specifically the price of electricity sold by the grid to shared energy storage power plants). Furthermore, the collected data can be cleaned and preprocessed, including removing outliers such as negative power and exceeding limits, and aligning the data to a unified timestamp.

[0196] S72. Forecast model construction, training, and output: Build a photovoltaic output forecast model and a user demand load forecast model. These models are trained using historical data. Real-time data is then fed into the trained photovoltaic output forecast model and user demand load forecast model to output photovoltaic output forecasts and user load demand forecasts for the future. More specifically, the photovoltaic output forecast model can be implemented using a long short-term memory (LSTM) network. The inputs are historical data such as historical photovoltaic output, weather data, and temporal characteristics. The output is photovoltaic output forecasts for each time point in the future. The user demand load forecast model can be implemented using an XGBoost or Transformer algorithm architecture. The inputs are historical load data, electricity price sensitivity, user type labels (commercial and industrial users or electric vehicle charging stations), and real-time weather data. The output is load demand forecasts for each time point in the future. In actual application, the prediction time window (T) and time step (t) are set according to actual needs. For example, for the next 4 hours, every 15 minutes is a time step, that is, a period (t), for a total of 16 periods.

[0197] S73. Dynamic pricing model construction and output. A dynamic pricing model is constructed using the rolling horizon optimization framework. Parameters are initialized based on the current state of charge of the shared energy storage power station, grid electricity prices, photovoltaic output forecast values, and user load demand forecast values. Optimization is then performed to output a shared energy storage electricity price sequence for a period of time in the future. More specifically, for the dynamic pricing model, a rolling horizon optimization framework is used. First, a time window is set. For the prediction window, it can be the next 4 hours, with each 15 minutes as a period. For the execution window, the electricity price of the next period is used as the execution result, and subsequent periods are used as optimization references. Secondly, the optimization frequency is set, which can be a period that is re-solved once. The optimization objective function is designed as follows:

[0198] (37)

[0199] in, represents the shared energy storage electricity price during period t (yuan / kWh), represents the electricity sales of the shared energy storage power station in period t (kWh), represents the amount of electricity purchased from the grid by the shared energy storage power station during period t (kWh), represents the user's photovoltaic power generation during period t (kWh), represents the total electricity consumption of users during period t (kWh), It represents the supply-demand balance penalty coefficient (yuan / kWh), which is used to adjust the cost of curtailed solar power or insufficient power supply. It can be adjusted based on historical data. represents the electricity sales power of the shared energy storage power station during period t (kW), represents the electricity price sold by the power grid to the shared energy storage power station during period t (yuan / kWh), represents the power purchased from the grid by the shared energy storage during period t (kW), represents the photovoltaic output forecast value (kW) during period t, which is obtained from the photovoltaic output forecast value model output in step S72. Represents the user demand load forecast value (kW) for period t—output by the user demand load forecast model in step S72.

[0200] The constraints set for the above model mainly include the following constraints:

[0201] (1) Energy storage operation constraints: mainly include SOC continuity, charge and discharge power limits, and SOC safety range.

[0202] (2) Electricity price range constraints: , which can avoid malicious pricing.

[0203] The rolling time domain optimization process of the dynamic pricing model in this embodiment is as follows:

[0204] Step 1: Initialization: Load initial data, including the current state of charge (SOC) of the shared energy storage power station, photovoltaic output forecast, user load forecast, and grid electricity price.

[0205] Step 2: Regular data update: Update data every other period, including actual PV output, user load, and grid electricity price, to correct the prediction model error.

[0206] Step 3: Solve the optimization problem: Use a mixed integer programming solver to solve the dynamic pricing model, preferably Gurobi or CPLEX, to obtain the optimal shared energy storage electricity price sequence for the future period. .

[0207] Step 4: Pricing release and feedback: Release the shared energy storage electricity price for the next period to users. Then monitor the actual electricity purchasing behavior of users in the next period - reflected in the actual power sales of the shared energy storage power station in the next period. , obtain feedback, and thus update the above objective function.

[0208] Here are two specific example scenarios:

[0209] Scenario 1: Sudden increase in photovoltaic output: When the photovoltaic output forecast value is significantly higher than the user's demand load, the above rolling time domain optimization is used to reduce the shared energy storage electricity price, incentivize users to increase electricity purchases and reduce solar curtailment. In the above objective function, the supply and demand balance term Increase, forcing the optimizer to lower the shared energy storage electricity price to increase the power sales of the shared energy storage power station.

[0210] Scenario 2: Grid electricity prices soar: Grid electricity prices rise sharply due to peak loads. Through the above rolling time domain optimization, the shared energy storage price is increased to match the high-price period of the grid and maximize the benefits. In the above objective function, the cost term As energy storage prices rise, optimizers tend to reduce their purchases of electricity from the grid and instead increase the price of electricity sold by energy storage to increase revenue.

[0211] This embodiment uses data-driven real-time optimization, namely real-time data collection. IoT devices can be deployed to monitor PV output, user load, shared energy storage power station SOC (state of charge), grid electricity prices, and other data in real time. Through predictive model integration and combined with machine learning, PV output and user load demand over a period of time are predicted. A dynamic pricing model is then introduced to optimize and update pricing strategies over a rolling time domain. When PV output suddenly increases, the price of shared energy storage electricity sold to users is dynamically reduced to promote consumption. When grid electricity prices soar, the energy storage discharge price is increased to match market changes.

[0212] In summary, this embodiment, through a dynamic pricing mechanism that combines real-time data collection, machine learning prediction, rolling time domain optimization, and user response feedback, can effectively respond to fluctuations in photovoltaic output and user load demand, improve the economy and flexibility of shared energy storage, and provide reliable support for new power systems.

[0213] Those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing the relevant hardware using a computer program. The program can be stored in a computer-readable storage medium, and when executed, the program can include the processes in the above-described method embodiments. The storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM). The technical features of the above-described embodiments can be combined in any manner. To simplify the description, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as such combinations do not conflict, they should be considered within the scope of this specification. The above-described embodiments represent only a few implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. Those skilled in the art will appreciate that various variations and improvements can be made without departing from the spirit of the present invention, and these variations and improvements fall within the scope of protection of the present invention.

Claims

1. A pricing method for shared energy storage in a park based on Stackelberg game, characterized by: include: Build a campus solar-storage-charging system, including a shared energy storage power station and a user group; the user group includes multiple industrial and commercial users equipped with photovoltaic power stations and electric vehicle charging stations; Construct an upper-level model with the shared energy storage power station as the leader, and determine the objective function and constraints of the upper-level model; Construct a lower-level model with user groups as followers and determine the objective function and constraints of the lower-level model; The upper and lower models are combined into a two-layer optimization model using the mathematical model of the Stackelberg game. A differential privacy perturbation mechanism is introduced to protect the privacy of the response data transmitted by the user group in the lower model to the upper model. The KKT optimality condition is used to transform the two-level optimization model into a single-level optimization model; Use the solver to solve the single-layer optimization model and obtain the shared energy storage pricing; The upper-level model with the shared energy storage power station as the leader is constructed, and the objective function and constraints of the upper-level model are determined, specifically including: Total revenue from shared energy storage power stations Maximize the objective function of the upper model to establish the target: ； Where, The income from selling electricity to the grid for shared energy storage power stations, The income from selling electricity to user groups for shared energy storage power stations, To generate revenue from selling electricity to electric vehicles for shared energy storage power stations, The cost of purchasing electricity from the grid for the shared energy storage power station, The cost of purchasing excess PV from the user base for shared energy storage plants; Set the constraints of the upper-level model: set the state of charge continuity constraint, the non-simultaneous charging and discharging constraint, and the maximum charging and discharging power constraint for the shared energy storage power station; The construction of the lower-level model with the user group as followers and the determination of the objective function and constraints of the lower-level model specifically include: Electricity cost by user group Minimize the objective function that models the underlying target: ； Where, The cost of purchasing electricity from the grid for the user group, The cost of purchasing electricity from shared energy storage power stations for user groups; Revenue from selling excess PV to the grid for the user group; Revenue from selling excess photovoltaic power to shared energy storage power stations for user groups; the cost of charging electric vehicles from the grid; the cost of charging electric vehicles from shared energy storage stations; Set the constraints of the lower-level model: set the load balance constraints of each user, the photovoltaic power balance constraints, the price range constraints for the electricity sold by the shared energy storage power station to the user group, the electric vehicle charging balance constraints, the electric vehicle charging power constraints and the power value range constraints.

2. The method for pricing a solar-storage-charging shared energy storage system in a park based on Stackelberg game according to claim 1, characterized in that: Also includes: Build a dynamic pricing adjustment mechanism, and when receiving feedback from the user group that they are dissatisfied with the shared energy storage pricing or when the total power sold by the shared energy storage power station to the user group within a preset period is lower than expected, activate the dynamic pricing adjustment mechanism to regain the shared energy storage pricing.

3. The method for pricing a shared energy storage system for photovoltaic storage and charging in a park based on Stackelberg game according to claim 1, characterized in that: The load balance constraint of each user and the photovoltaic power balance constraint are equality constraints, the electricity price range constraint and the power value range constraint of the shared energy storage power station to the user group are inequality constraints, and a corresponding Lagrange multiplier is set in each constraint.

4. The method for pricing a solar-storage-charging shared energy storage system in a park based on Stackelberg game according to claim 3, characterized in that: The mathematical model of Stackelberg game is used to form a two-layer optimization model with the upper model and the lower model, and a differential privacy perturbation mechanism is introduced to protect the privacy of the response data transmitted by the user group in the lower model to the upper model. The two-layer optimization model using the mathematical model of Stackelberg game is expressed as: ； is the objective function of the leader’s upper model, is the objective function of the follower’s lower model, is the leader’s inequality constraint, is the leader's equality constraint, represents the set of inequality constraint indices of the leader, represents the set of equality constraint indices of the leader, is the inequality constraint for the follower, is the equality constraint for the follower, represents the set of inequality constraint indices of the followers, represents the set of equality constraint indices of followers, and are the indices of inequalities and equalities, respectively, and denote the decision variables of the objective functions of the two models, and and All satisfied , is the strategy set of the Stackelberg game, that is, the set of all variables involved in the game.

5. The method for pricing a solar-storage-charging shared energy storage system in a park based on Stackelberg game according to claim 1, characterized in that: The differential privacy perturbation mechanism is introduced to protect the privacy of the response data transmitted by the user group in the lower-level model to the upper-level model, which is specifically achieved by adding Gaussian noise to the response data.

6. The method for pricing solar-storage-charging shared energy storage in a park based on Stackelberg game according to claim 3, characterized in that: The KKT optimality condition is used to convert the double-layer optimization model into a single-layer optimization model, and the KKT optimality condition is used to convert the lower model in the double-layer optimization model into an additional constraint of the upper model, thereby realizing the conversion of the double-layer optimization model into a single-layer optimization model; the KKT optimality condition refers to the derivative of the lower model at the extreme point being zero, which is one of the KKT conditions.

7. The method for pricing solar-storage-charging shared energy storage in a park based on Stackelberg game according to claim 5, characterized in that: The method of converting the double-layer optimization model into a single-layer optimization model by using the KKT optimality condition specifically includes the following steps: Construct the Lagrangian function of the underlying model; Use the Lagrangian function to find the partial derivatives of the variables of the lower model so that the lower model meets the KKT optimality condition; According to the constraints of the lower model and the corresponding Lagrange multipliers, the complementary relaxation conditions of the lower model are obtained as additional constraints of the upper model to obtain the single-layer optimization model.

8. The method for pricing solar-storage-charging shared energy storage in a park based on Stackelberg game according to claim 2, characterized in that: The dynamic pricing adjustment mechanism is constructed, and when feedback is received from the user group that the user group is dissatisfied with the shared energy storage pricing or the total power sold by the shared energy storage power station to the user group within a preset period is lower than the expected value, the dynamic pricing adjustment mechanism is activated to regain the shared energy storage pricing. The dynamic pricing adjustment mechanism is constructed specifically including the following steps: Data collection: real-time collection of photovoltaic power station output, industrial and commercial user loads, electric vehicle charging station loads, shared energy storage power station charge status, and grid electricity prices; Prediction model construction, training and output: Build a photovoltaic output prediction model and a user demand load prediction model, and train the photovoltaic output prediction model and the user demand load prediction model with historical data. Then, input the real-time collected data into the trained photovoltaic output prediction model and the user demand load prediction model to output the photovoltaic output prediction value and the user load demand prediction value for a period of time in the future. Dynamic pricing model construction and output: A dynamic pricing model is constructed based on the rolling time domain optimization framework. Parameters are initialized based on the current state of charge of the shared energy storage power station, grid electricity price, photovoltaic output forecast value, and user load demand forecast value. Then, the solution is optimized and the shared energy storage electricity price sequence for a period of time in the future is output.

Images