Non-cooperative game optimization scheduling method based on load aggregator classification

Abstract

The present invention relates to the technical field of smart grid dispatching methods, and is a non-cooperative game optimization dispatching method based on load aggregator classification, which includes establishing a load aggregator classification model and a supply-side power supply model, establishing a non-cooperative game optimization model for load aggregator classification according to game behavior, establishing a grid-side objective function and a load aggregator-side objective function, and solving the optimal configuration between Nash equilibrium and unit output state based on the NSGA-II multi-objective optimization algorithm and the algorithm combining interior point generation. The present invention proposes a load aggregator classification non-cooperative game mechanism based on the idea of ​​non-cooperative game, and establishes a multi-objective dispatching model with maximum grid revenue, maximum profit for commercial load aggregators and residential load aggregators, so that various types of load aggregators can obtain the best bidding strategy in the competition, and the two types of load aggregators then determine the dispatching participation amount of each load aggregator according to the principle of proportional distribution.

Description

Technical Field

[0001] The invention relates to the technical field of smart grid dispatching methods, and is a non-cooperative game optimization dispatching method based on load aggregator classification. Background Art

[0002] With the two-way advancement of the energy Internet, the two-way communication technology between sources and loads in the energy Internet has been continuously improved. At present, power grid companies have been able to basically monitor and control various small and medium-sized loads through monitoring equipment and communication technology, making demand-side management possible. In the context of the rapid development of communication technology and the continuous improvement of the power market, the smart grid can use the developed communication technology to better apply the demand-side management technology to the "source-load" coordinated optimization dispatching mode through demand-side management means, effectively alleviating the contradiction between supply and demand. However, the load elasticity level of a single user is low, and it is relatively difficult to participate in the dispatching of the smart grid, and it is difficult to meet the dispatching needs. If each flexible load participates in the demand response as a separate individual, it will cause a waste of resources. Therefore, load aggregators have received great attention as a new type of organization that serves small and medium-sized users to participate in grid dispatching. Load aggregators can classify and aggregate small and medium-sized users according to the prediction of various loads by the smart grid, and establish a response model based on load groups to participate in the dispatching of smart grids. Under the background of the participation of load aggregators, smart grids have gradually become more open, which leads to the inevitable existence of complex game behaviors in the process of diversified transactions on the demand side. Therefore, the application of game theory in smart grids will also be more extensive.

[0003] To sum up, in order to alleviate the contradiction between supply and demand of the power grid, ensure the safe and stable operation of the power grid, and further improve the efficiency of resource energy utilization on the demand side, the present invention, under the background of the two-way development of the supply and demand sides of smart grid technology, aims at the problem of load aggregators participating in the smart grid dispatching game, and conducts research on the non-cooperative game optimization technology for the classification of load aggregators, while optimizing the power grid revenue and the load aggregator revenue. Summary of the invention

[0004] The present invention provides a non-cooperative game optimization scheduling method based on load aggregator classification, which overcomes the shortcomings of the above-mentioned prior art. Based on the non-cooperative game mechanism of load aggregator classification based on the non-cooperative game idea, a multi-objective scheduling model is established with the maximum power grid revenue and the maximum profit of commercial load aggregators and residential load aggregators, so that various types of load aggregators can obtain the best bidding strategy in the competition. Each load aggregator class then determines the scheduling participation amount of each load aggregator according to the principle of proportional distribution, thereby optimizing the revenue on both the source and the load side.

[0005] The technical solution of the present invention is achieved through the following measures: a non-cooperative game optimization scheduling method based on load aggregator classification, the following steps:

[0006] Step (1), determining the smart grid infrastructure in which the load aggregator participates;

[0007] Step (2), establishing a load aggregator classification model, that is, classifying the flexible loads, and classifying the load aggregators according to the different types of aggregated flexible loads;

[0008] Step (3), establishing a supply-side power supply model, the supply-side power supply model includes a thermal power unit output model and a wind power output model;

[0009] Step (4), based on the classification of load aggregators, firstly, the game behavior of load aggregators in the process of participating in power grid dispatch is given, and a non-cooperative game optimization model for the classification of load aggregators is established based on the game behavior;

[0010] Step (5), establishing the grid side objective function and the load aggregator side objective function;

[0011] Step (6), based on the NSGA-II multi-objective optimization algorithm and the interior point generation combined algorithm, the optimal configuration between the Nash equilibrium and the unit output state is solved.

[0012] The following are further optimizations and / or improvements to the above technical solutions:

[0013] In the above step (1), the smart grid infrastructure includes two parts: distributed power sources and loads. The distributed power sources take into account the power generation of wind turbines and thermal power units and participate in grid dispatching at the same time. The loads take into account flexible loads. Flexible loads are classified and aggregated by load aggregators to participate in grid dispatching.

[0014] The control layer formulates a source-load coordination scheduling plan based on the unit's operating status and output, conventional load forecasts, and the amount of aggregator participation, and sends instructions to the generator units and load aggregators. The load aggregators complete the transfer and interruption of flexible loads according to the plan formulated by the power grid, forming an economic scheduling plan for load aggregators to participate in the smart grid.

[0015] In the above step (2), based on the load response capability and physical characteristics, the flexible loads are divided into three categories: Class A loads, Class B loads and Class C loads. Different types of flexible loads are modeled and analyzed. Based on the classification of flexible loads, load aggregators are divided into two categories: commercial and residential. The multidimensionality and complexity of various types of flexible loads participating in the load aggregator scheduling are studied.

[0016] Furthermore, the specific steps of step (2) are as follows:

[0017] (21) Class A load can be transferred or interrupted. Assuming H1 is the entire working period of Class A load, the power required by the load must be completed during the working period, that is, Where: Q Ais the daily power required by Class A load; Pload1 is the rated power of the load, H1={1,2,...,H};

[0018] (22) Class B loads can be transferred but not interrupted. Assuming that the participation period of Class B loads is H2, the load aggregator needs to ensure that Class B loads complete all required electricity within the H2 period and cannot interrupt this type of load during the dispatch process, that is,

[0019] Where: H2={1,2,...,H}; Q B is the daily power requirement of class B load; arg(Sch2(h)=1) means that during the h period, the load is in the on state and can participate in the scheduling. and α represent the minimum time and maximum time in the on period, respectively;

[0020] (23) Class C loads cannot be transferred but can be interrupted intermittently, that is:

[0021]

[0022] The constraint condition in this formula means that if the C load is turned off in the h-1 period and does not participate in the dispatch, it is turned on in the h period, and then the load is turned off in the h+β period, then the total dispatch period of the C load in the entire dispatch process is min(β).

[0023] In the formula, Q C is the daily power requirement of Class C load, t on,3 is the shortest opening time of Class C load, t off,3 Maximum shutdown time;

[0024] (24) Based on the classification of flexible loads, load aggregators are also classified into two categories: commercial and residential. The daily load curve of commercial buildings is mainly single-peak, and the electricity consumption is higher during the day on weekdays. The load of commercial users is mainly rigid load, and interruption will cause huge losses. Therefore, commercial load aggregators only consider air-conditioning loads as Class A loads to participate in load aggregator scheduling. To represent commercial load aggregator n b Aggregated Class A load, then commercial load aggregator n b The load volume planned for dispatching in the day-ahead dispatching phase during period t is: Residential load is a comprehensive load with a double-peak load curve. The power consumption is high in the early morning and evening, and low at night. r There are three types of loads involved in dispatching in the jurisdiction, each type of load has I r Users participated, using To represent the load aggregator n r Aggregate Class A load, with To represent the load aggregator n r Aggregate Class B load, with To represent the load aggregator n r Aggregated Class C load, then the residential load aggregator n r The load volume planned for dispatching in the day-ahead dispatching phase during period t is:

[0025] In the above step (3), the wind power output model includes the wind turbine output constraint model and the wind abandonment cost. The operation of wind power units is subject to certain constraints during the power generation process. The maximum output of a single wind turbine is the rated power P r , when exiting the power grid, the minimum output is 0, and the wind turbine output constraint model is: In order to reduce wind curtailment and increase grid revenue, the cost of wind curtailment is also taken into account when calculating revenue. The expression is: In the formula, is the wind curtailment cost, λ wf is the wind curtailment cost coefficient, is the abandoned wind power of the wind farm at time t;

[0026] The output model of thermal power units includes the upper and lower output constraints of the thermal power units, the number of start and stop times of the thermal power units, the unit climbing constraints and the total operating cost of the thermal power units. The operation of the thermal power units must meet the upper and lower output constraints: Constraints on the number of starts and stops of thermal power units: And the unit climbing constraint: There are two ways to regulate thermal power units: variable load and start-stop. Variable load regulation refers to regulating the output of thermal power units within the upper and lower limits of thermal power output. Start-stop regulation refers to regulating the output of units by starting and stopping units within the constraints of the number of starts and stops. The strategy of jointly regulating the two methods is adopted to regulate the output of thermal power units.

[0027] The operating cost of thermal power units includes coal consumption cost and start-up and shutdown cost: the coal consumption cost can describe the coal-to-electricity conversion efficiency of thermal power units. Considering the impact of grid cost on thermal power unit power dispatch, the coal consumption cost of thermal power units is expressed as a quadratic function of unit power: in, is the output of thermal power unit i at time t, a i 、b i 、c i is the coal consumption cost coefficient of unit i during power generation; in the actual power generation process of thermal power units, the start and stop of thermal power units will also bring huge losses, and frequent start and stop of units will also affect the life of the units. Therefore, the start and stop costs of thermal power units are also considered in the operating costs: where λ nfis the start-up and shutdown cost coefficient of thermal power units, n t,i is the start and stop status of thermal power unit i at time t. The total operating cost of the thermal power unit is obtained from the above formula: is the total operating cost of the thermal power unit, F1 is the coal consumption cost of the thermal power unit, and F2 is the start-up and shutdown cost of the thermal power unit.

[0028] In the above step (4), according to the load aggregator classification, a non-cooperative game optimization model G = [N, {S n},{C n}], the specific steps are as follows:

[0029] (41) The participants are the N load aggregators in the residential area, N = {1, 2, ..., n};

[0030] (42) Strategy Set The load reduction plan reported by N load aggregators to the power company is the load reduction amount L in each peak-cutting period. st,n combination of;

[0031] (43) Payment function {C n} is the profit obtained by each load aggregator, which is expressed as the decision of aggregator n and other N-1 aggregators to bid together with the goal of maximizing their own profits, that is:

[0032] (44) Information set: The game process in which load aggregators participate in power grid dispatching is a complete information game. The bidding strategy and revenue of each load aggregator are public information and are known to all load aggregators participating in the game.

[0033] (45) Game results: Each aggregator takes maximizing its own profit as the objective function and continuously changes its bidding strategy according to the objective function. During the game, each load aggregator will choose the bidding strategy that maximizes its own interests. When all aggregators do not change their bidding strategies, that is, when any aggregator cannot change its own strategy to increase its profit, the game reaches a stable state, namely the Nash equilibrium, which is the Nash equilibrium solution of the non-cooperative game model. In this stable state, the strategy set of each load user is the equilibrium solution of the game, namely {S n}; Strategy set {S n The necessary and sufficient condition for a multi-user non-cooperative game Nash equilibrium solution is: for every strategy of any load aggregator n Both X∈S i , that is, changing the strategy of any load aggregator can't reduce its final profit. For a situation, is a pure strategy Nash equilibrium point of the load aggregator non-cooperative game model G, C i (s j ) is the corresponding equilibrium result.

[0034] In the game of distributed energy users, through distributed power sources or the use of various energy storage devices, electricity is consumed at times when electricity prices are low. The market will also set up a corresponding time-of-use electricity price mechanism. The power grid will also compete with various users to maximize its own benefits. In the process of mutual influence between the interests of the power grid and users, there is a certain game process. In the game between load aggregators, each load aggregator must compete with other load aggregators to maximize its own benefits under the condition of unified compensation prices, and there is a game phenomenon. In the non-cooperative game process of load aggregators, individual load aggregators are the main players in the game, and their electricity consumption behavior characteristics have an important impact on the main game behavior, game results, and the implementation effect of demand response.

[0035] The specific steps of the above step (5) are as follows:

[0036] (51) Grid-side objective function: For the grid, the grid pays more attention to the safety and balance of the system, and also hopes to maximize its own benefits while load aggregators participate. Its objective function can be expressed as follows: in The revenue from electricity sales to the power grid, They are respectively the operating cost of thermal power units, the cost of wind power abandonment and the cost of load aggregators, and the revenue from electricity sales is the total load at time t, e t is the electricity selling price at time t; different electricity price levels are set for peak, flat and valley periods. Through the mechanism of time-of-use electricity price, users are guided to use electricity during periods with lower electricity prices, so as to achieve the effect of load shaving and valley filling, increase wind power consumption, and weaken the characteristics of reverse distribution of source and load. Peak-valley time-of-use electricity price refers to dividing 24 hours into peak, flat and valley periods according to load changes. According to domestic electricity price policies, a city's time-of-use electricity price is divided into three peak and valley periods: valley period (0:00-6:00 and 23:00-24:00); flat period (6:00-8:00 and 16:00-18:00); peak period (8:00-16:00 and 18:00-23:00). Combining price-based demand response with incentive-based demand response can better dispatch the enthusiasm of the demand side to participate in the smart grid and achieve complementary advantages;

[0037] (52) Load aggregator side objective function

[0038] As for load aggregators, they are more concerned about obtaining more suitable prices after bargaining to maximize their own profits. In the bargaining process, there will be competition between commercial load aggregators and residential load aggregators. Therefore, a model of two load aggregator bargaining is established. The objective function of commercial load aggregators is expressed as: The objective function of the residential load aggregator is expressed as:

[0039] (53) According to the three objective functions, the multi-objective problem designed in this paper can be described as follows: Here Ω is the decision space and the set of achievable goals is defined as {F(x)|x∈Ω}.

[0040] For power grid companies, the goal is to guide users to participate in demand response through time-of-use electricity prices, hoping to mobilize more resources for electricity consumption during the low-consumption period and relieve the pressure on the power grid during the peak-consumption period, while at the same time hoping to reduce power grid expenditures. Each load aggregator hopes to obtain the most appropriate compensation price through non-cooperative games to increase its profits. The load aggregator's income comes from the power grid. The increase in the load aggregator's income means that the power grid needs to increase its expenditures. These two goals are mutually restrained. By formulating a multi-objective function and optimizing these two goals at the same time, the power grid can reduce power generation costs and increase its own profits while protecting the interests of load aggregators.

[0041] In the above step (6), the non-cooperative game multi-objective optimization model of the load aggregator can essentially be classified as a two-layer optimization problem. The outer optimization takes the unit output status and the amount of wind abandonment as decision variables, and the inner optimization takes the strategy of the load aggregator participating in the dispatch as the decision variable. The optimization goal is the grid revenue, and the load aggregator revenue is maximized.

[0042] The present invention proposes a non-cooperative game mechanism for classifying load aggregators based on the idea of ​​non-cooperative game, in which the compensation electricity price of load aggregators is set by the power grid according to the load demand, and establishes a multi-objective dispatching model with the maximum profit of the power grid and the maximum profit of commercial load aggregators and residential load aggregators, so that various types of load aggregators can obtain the best bidding strategy in the competition, and the two types of load aggregators then determine the dispatching participation amount of each load aggregator according to the principle of proportional distribution. The NSGA-II multi-objective optimization algorithm is used to solve the constructed optimization model. The simulation results of the example select the solution that improves the profit of the power grid side and the load aggregator under the participation of the power grid pricing method as the optimal solution from the Pareto frontier obtained to ensure the profit of the power grid. The results show the effectiveness of the method of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Attached Figure 1 The overall structure diagram of the method of the present invention.

[0044] Attached Figure 2 Classify the game process for load aggregators.

[0045] Attached Figure 3 Flowchart for solving the NSGA-II algorithm for the non-cooperative game multi-objective scheduling model.

[0046] Attached Figure 4 Pareto front diagram for the optimization results.

[0047] Attached Figure 5 This is a comparison chart of the power grid revenue in different time periods under time-of-use pricing and non-cooperative game modes.

[0048] Attached Figure 6a This is a comparison chart of the results of compensation based on a uniform price for each time period and compensation based on the market price obtained through non-cooperative game.

[0049] Attached Figure 6b A comparison chart of the benefits obtained from uniform price compensation and non-cooperative game by time period. DETAILED DESCRIPTION

[0050] The present invention is not limited by the following embodiments, and specific implementation methods can be determined based on the technical solution of the present invention and actual conditions.

[0051] The present invention will be further described below in conjunction with embodiments:

[0052] Embodiment 1: As attached Figure 1 As shown, the non-cooperative game optimization scheduling method based on load aggregator classification includes the following steps:

[0053] Step (1), determining the smart grid infrastructure in which the load aggregator participates;

[0054] Step (2), establishing a load aggregator classification model, that is, classifying the flexible loads, and classifying the load aggregators according to the different types of aggregated flexible loads;

[0055] Step (3), establishing a supply-side power supply model, the supply-side power supply model includes a thermal power unit output model and a wind power output model;

[0056] Step (4), based on the classification of load aggregators, firstly, the game behavior of load aggregators in the process of participating in power grid dispatch is given, and a non-cooperative game optimization model for the classification of load aggregators is established based on the game behavior;

[0057] Step (5), establishing the grid side objective function and the load aggregator side objective function;

[0058] Step (6), based on the NSGA-II multi-objective optimization algorithm and the interior point generation combined algorithm, the optimal configuration between the Nash equilibrium and the unit output state is solved.

[0059] The method of the present invention takes load aggregators as the main body, and proposes a load aggregator classification non-cooperative game mechanism based on the idea of ​​non-cooperative game in the way that the load aggregator compensation electricity price is priced by the market according to the load demand, and establishes a multi-objective dispatching model with the maximum grid revenue, the maximum profit of commercial load aggregators and residential load aggregators, so that various types of load aggregators actively participate in the dispatching of smart grids, and obtain the best bidding strategy in the competition. Each load aggregator class then determines the dispatching participation amount of each load aggregator according to the principle of proportional distribution. The multi-objective optimization algorithm NSGA-II is used to solve the constructed optimization model. The simulation results of the example select the solution that improves the revenue of both the grid side and the load aggregator under the participation of the grid pricing method as the optimal solution from the obtained Pareto frontier.

[0060] Example: The non-cooperative game optimization scheduling method based on load aggregator classification has the following steps:

[0061] like Figure 1 As shown, the smart grid infrastructure includes two parts: distributed power sources and loads. Distributed power sources take into account wind turbines and thermal power units to participate in grid dispatching at the same time, and loads take into account flexible loads. Flexible loads are classified and aggregated by load aggregators to participate in grid dispatching.

[0062] The grid control layer reasonably formulates source-load coordination scheduling plans based on information such as the unit operating status and output, conventional load forecasts, and the participation of load aggregators, and issues instructions. Finally, the load aggregators issue scheduling instructions to the flexible load user group based on the instructions to complete the transfer and interruption of flexible loads, thus forming an economic scheduling plan for load aggregators to participate in the smart grid.

[0063] In the above step (2), based on the load response capability and physical characteristics, the flexible loads are divided into three categories: Class A loads, Class B loads and Class C loads. Different types of flexible loads are modeled and analyzed. Based on the classification of flexible loads, load aggregators are divided into two categories: commercial and residential. The multidimensionality and complexity of various types of flexible loads participating in the load aggregator scheduling are studied.

[0064] The specific steps of step (2) are as follows:

[0065] (21) Class A load can be transferred or interrupted. Assuming H1 is, Class A load is in all working periods. The power required to complete the load during the working period is: Where: Q A is the daily power required by Class A load; Pload1 is the rated power of the load, H1={1,2,...,H};

[0066] (22) Class B loads can be transferred but not interrupted. Assuming that the participation period of Class B loads is H2, the load aggregator needs to ensure that Class B loads complete all required electricity within the H2 period and cannot interrupt this type of load during the dispatch process, that is,

[0067] Where: H2={1,2,...,H}; Q B is the daily power requirement of class B load; arg(Sch2(h)=1) means that during the h period, the load is in the on state and can participate in the scheduling. and α represent the minimum time and maximum time in the on period, respectively;

[0068] (23) Class C loads cannot be transferred but can be interrupted intermittently, that is:

[0069]

[0070] The constraint condition in this formula means that if the C load is turned off in the h-1 period and does not participate in the dispatch, it is turned on in the h period, and then the load is turned off in the h+β period, then the total dispatch period of the C load in the entire dispatch process is min(β).

[0071] In the formula, Q C is the daily power requirement of Class C load, t on,3 is the shortest opening time of Class C load, t off,3 Maximum shutdown time;

[0072] (24) Based on the classification of flexible loads, load aggregators are also classified into two categories: commercial and residential. The daily load curve of commercial buildings is mainly single-peak, and the electricity consumption is higher during the day on weekdays. The load of commercial users is mainly rigid load, and interruption will cause huge losses. Therefore, commercial load aggregators only consider air-conditioning loads as Class A loads to participate in load aggregator scheduling. To represent commercial load aggregator n b Aggregated Class A load, then commercial load aggregator n b The load volume planned for dispatching in the day-ahead dispatching phase during period t is: Residential load is a comprehensive load with a double-peak load curve. The power consumption is high in the early morning and evening, and low at night. r There are three types of loads involved in dispatching in the jurisdiction, each type of load has I r Users participated, using To represent the load aggregator n r Aggregate Class A load, with To represent the load aggregator n rAggregate Class B load, with To represent the load aggregator n r Aggregated Class C load, then the residential load aggregator n r The load volume planned for dispatching in the day-ahead dispatching phase during period t is:

[0073] In the above step (3), the wind power output model includes the wind turbine output constraint model and the wind abandonment cost. The operation of wind power units is subject to certain constraints during the power generation process. The maximum output of a single wind turbine is the rated power P r , when exiting the power grid, the minimum output is 0, and the wind turbine output constraint model is: In order to reduce wind curtailment and increase grid revenue, the cost of wind curtailment is also taken into account when calculating revenue. The expression is: In the formula, is the wind curtailment cost, λ wf is the wind curtailment cost coefficient, is the abandoned wind power of the wind farm at time t;

[0074] The output model of thermal power units includes the upper and lower output constraints of the thermal power units, the number of start and stop times of the thermal power units, the unit climbing constraints and the total operating cost of the thermal power units. The operation of the thermal power units must meet the upper and lower output constraints: Constraints on the number of starts and stops of thermal power units: And the unit climbing constraint: There are two ways to regulate thermal power units: variable load and start-stop. Variable load regulation refers to regulating the output of thermal power units within the upper and lower limits of thermal power output. Start-stop regulation refers to regulating the output of units by starting and stopping units within the constraints of the number of starts and stops. The strategy of jointly regulating the two methods is adopted to regulate the output of thermal power units.

[0075] The operating cost of thermal power units includes coal consumption cost and start-up and shutdown cost: the coal consumption cost can describe the coal-to-electricity conversion efficiency of thermal power units. Considering the impact of grid cost on thermal power unit power dispatch, the coal consumption cost of thermal power units is expressed as a quadratic function of unit power: in, is the output of thermal power unit i at time t, a i 、b i 、c i is the coal consumption cost coefficient of unit i during power generation; in the actual power generation process of thermal power units, the start and stop of thermal power units will also bring huge losses, and frequent start and stop of units will also affect the life of the units. Therefore, the start and stop costs of thermal power units are also considered in the operating costs: where λ nf is the start-up and shutdown cost coefficient of thermal power units, n t,i is the start and stop status of thermal power unit i at time t. The total operating cost of the thermal power unit is obtained from the above formula: is the total operating cost of the thermal power unit, F1 is the coal consumption cost of the thermal power unit, and F2 is the start-up and shutdown cost of the thermal power unit.

[0076] In the above step (4), according to the load aggregator classification, a non-cooperative game optimization model G = [N, {S n},{C n}], the specific steps are as follows:

[0077] (41) The participants are the N load aggregators in the residential area, N = {1, 2, ..., n};

[0078] (42) Strategy Set The load reduction plan reported by N load aggregators to the power company is the load reduction amount L in each peak-cutting period. st,n combination of;

[0079] (43) Payment function {C n} is the profit obtained by each load aggregator, which is expressed as the decision of aggregator n and other N-1 aggregators to bid together with the goal of maximizing their own profits, that is:

[0080] (44) Information set: The game process in which load aggregators participate in power grid dispatching is a complete information game. The bidding strategy and revenue of each load aggregator are public information and are known to all load aggregators participating in the game.

[0081] (45) Game results: Each aggregator takes maximizing its own profit as the objective function and continuously changes its bidding strategy according to the objective function. During the game, each load aggregator will choose the bidding strategy that maximizes its own interests. When all aggregators do not change their bidding strategies, that is, when any aggregator cannot change its own strategy to increase its profit, the game reaches a stable state, namely the Nash equilibrium, which is the Nash equilibrium solution of the non-cooperative game model. In this stable state, the strategy set of each load user is the equilibrium solution of the game, namely {S n}; Strategy set {S n The necessary and sufficient condition for a multi-user non-cooperative game Nash equilibrium solution is: for every strategy of any load aggregator n Both X∈S i , that is, changing the strategy of any load aggregator can't reduce its final profit. For a situation, is a pure strategy Nash equilibrium point of the load aggregator non-cooperative game model G, C i (sj ) is the corresponding equilibrium result.

[0082] As attached Figure 2 As shown in the figure, in the process of load aggregators participating in smart grid dispatching, the power grid will first determine the dispatching period of the day based on the output forecast of the load aggregators, and set the compensation price for each period based on the output of the load aggregators. After announcing it to all aggregators, a contract will be signed to stipulate the amount of dispatching they need to participate in in each period, and compensation will be given to the aggregators. The compensation is calculated according to the load reduction in each period and the compensation price stipulated by the power grid.

[0083] Assume that there are T time periods that need to reduce load, and there are N load aggregators participating in demand response. Let the load reduction amount of aggregators n∈N(n={1,2,...,N}) in time period t∈T(t={1,2,...,T}) in time period t be L st,n , the total load reduction of all aggregators in time period t is L st for: Assume that the market price in time period t∈T (T={1,2,...,24}) is ei(t), and there is a linear relationship between the market price and the load level, that is: Where: is the price load factor; is a constant term; L t Forecast load for time period t. t For L st Regardless of the constant, the above formula can be transformed into ei(t) = a t L st +b t L t Where: It can be seen that as the load reduction increases, the reduction unit price gradually decreases. st,n After the price ei(t) is determined, the revenue that the load aggregator obtains from the dispatching department of the power company in time period t is ei(t)L st,n .

[0084] Based on the market price of load aggregators, the commercial aggregator n can be obtained. b income for: Similarly, the income of residential load aggregators can be obtained. The two types of load aggregators take the maximum profit as the optimization goal, and the objective function of the bidding decision (the decision-making goal of the two aggregators in the process of participating in the game) can be obtained:

[0085]

[0086] Where, L st,nmax is the maximum load reduction managed by load aggregator n during period t, Includes residential load aggregator revenue and commercial load aggregator revenue.

[0087] After the non-cooperative game, the participation of each load aggregator is allocated according to the percentage of the bid amount in the total participation of this type of load. The total participation of commercial load aggregators in time period t is: Commercial load aggregator n b The final participation amount is: in, is the load aggregator n in period t b The maximum reduction in managed load.

[0088] The specific steps of the above step (5) are as follows:

[0089] (51) Grid-side objective function: For the grid, the grid pays more attention to the safety and balance of the system, and also hopes to maximize its own benefits while load aggregators participate. Its objective function can be expressed as follows: in The revenue from electricity sales to the power grid, They are respectively the operating cost of thermal power units, the cost of wind power abandonment and the cost of load aggregators, and the revenue from electricity sales is the total load at time t, e t is the electricity selling price at time t; different electricity price levels are set for peak, flat and valley periods. Through the mechanism of time-of-use electricity price, users are guided to use electricity during periods with lower electricity prices, so as to achieve the effect of load shaving and valley filling, increase wind power consumption, and weaken the characteristics of reverse distribution of source and load. Peak-valley time-of-use electricity price refers to dividing 24 hours into peak, flat and valley periods according to load changes. According to domestic electricity price policies, a city's time-of-use electricity price is divided into three peak and valley periods: valley period (0:00-6:00 and 23:00-24:00); flat period (6:00-8:00 and 16:00-18:00); peak period (8:00-16:00 and 18:00-23:00). Combining price-based demand response with incentive-based demand response can better dispatch the enthusiasm of the demand side to participate in the smart grid and achieve complementary advantages;

[0090] (52) Load aggregator objective function: As far as load aggregators are concerned, they are more concerned with obtaining a more suitable price after bargaining to maximize their own profits. In the bargaining process, there will also be competition between commercial load aggregators and residential load aggregators. Therefore, a model of two load aggregator bargaining is established. The objective function of commercial load aggregators is expressed as: The objective function of the residential load aggregator is expressed as:

[0091]

[0092] (53) According to the three objective functions, the multi-objective problem designed in this paper can be described as follows: Here Ω is the decision space and the set of achievable goals is defined as {F(x)|x∈Ω}.

[0093] In the above step (6), the load aggregator non-cooperative game multi-objective optimization model can be essentially classified as a two-layer optimization problem. The outer optimization takes the unit output state and the amount of wind abandoned as decision variables, and the inner optimization takes the load aggregator's participation in scheduling strategy as decision variables. The optimization goal is the power grid revenue, and the load aggregator's revenue is maximized. In view of this, this paper is based on the NSGA-II multi-objective optimization algorithm and the interior point algorithm to solve the optimal configuration between Nash equilibrium and unit output state. Among them, the NSGA-II algorithm has the advantages of fast running speed and good convergence. It can quickly find the optimal solution set in multi-objective problems, and the interior point method is an algorithm for solving linear programming or nonlinear convex optimization. It can quickly find the optimal solution in polynomial problems. Therefore, this paper uses the interior point method to solve the Nash equilibrium solution of the load aggregator non-cooperative game.

[0094] As attached Figure 3 As shown, the specific solution steps of step (6) are as follows:

[0095] 1) Set the upper and lower limits of each unit's output, start-stop constraints, climbing constraints, upper and lower limits of wind power output, upper and lower limits of the load aggregator's participation, and participation status constraints, and set the parameters of the NSGA-II algorithm;

[0096] 2) Population initialization: Each individual in the population represents a set of optimization variables, among which the active output of conventional units and wind power dispatch output are continuous variables, the switching state of the load aggregator and the start and stop state of conventional units are discrete variables, and the number of iterations Gen is initialized to 0;

[0097] 3) Modify the initialized population individuals. To meet the load balance constraint, first adjust the units with large capacity according to the constraint, and then adjust the participation amount and participation state of the load aggregator until the load balance is met;

[0098] 4) Execute the individual interior point algorithm: first initialize the load aggregator bidding strategy set; then obtain the optimal strategy set of the load aggregator in the current state by solving the objective function, and repeat the solution steps until all load aggregator strategies do not change; finally return the load aggregator bidding strategy in the current state;

[0099] 5) Perform non-dominated sorting on the population to obtain a non-dominated set. During the operation of the algorithm, in order to facilitate non-dominated sorting, the objective function is minimized as the goal. Therefore, in the actual optimization process, the objective function is set to a negative number to obtain the maximum value in the optimization process;

[0100] 6) Perform crossover and mutation on each individual to form the offspring population Q.

[0101] The mutation operation is divided into polynomial mutation and basic bit mutation. The basic bit mutation is used for the binary coding of the unit start and stop and the load aggregator participation status. A random integer smaller than the length of the binary code individual is randomly generated. If the number at this position is 1, it becomes 0, otherwise it becomes 1. For the polynomial mutation of the unit output load aggregator reduction and wind abandonment, the mutation operation is to use a certain mutation strategy to modify the i-th individual in the population. Make changes to generate mutation vectors The resulting mutation vector is used to locate the local optimum: in, is the mutation vector, r i , i=(1,2,...,NP) are integers of different sizes in [1,NP], is the r2th individual in the Gth generation population, is the best individual in the G-th generation population after non-dominated sorting, and F is the polynomial mutation parameter, which is generally selected in [0,2].

[0102] There are two types of crossover operations: the binary code representing the start and stop status of the unit and the participation status of the load aggregator adopts single-point crossover, which randomly sets only one crossover point in the individual code string, and then exchanges part of the chromosomes of the two paired individuals at this point; the real number code representing the unit output and the participation amount of the load aggregator adopts simulated binary crossover;

[0103] (7) Modify the individuals in population Q until the load balance constraint is met;

[0104] (8) The parent population and the offspring population are merged, and the crowding degree is sorted. The top N individuals are selected according to the elite selection strategy to form a new population;

[0105] (9) Determine whether the termination condition is met. If so, exit the loop to obtain the optimal scheduling plan. If not, return to step (4) to reorder the new population.

[0106] Figure 4It is the Pareto frontier diagram of multiple objective functions. Objective function 1 is the revenue of the grid side, objective function 2 is the revenue of commercial load aggregators, and objective function 3 is the revenue of residential load aggregators. Because the non-dominated sorting is sorted by the minimum value of the objective function, the objective function is set to a negative value in the solution process. In the process of smart grid dispatching, the ultimate goal is to increase the revenue of the grid. Therefore, when selecting the optimal solution, the solution that increases the revenue of the grid side and the load aggregator under the participation of the grid pricing method will be selected from the obtained Pareto frontier as the optimal solution.

[0107] The power grid revenues in each period under the non-cooperative game mode of time-of-use pricing and the non-cooperative game optimization dispatching method based on load aggregator classification described in the example are compared. Figure 5 .

[0108] from Figure 5 It can be seen that in the non-cooperative game scenario, the power grid revenue increases significantly during peak hours. According to statistics, the daily revenue also increases compared to the time-of-use pricing scenario.

[0109] The results and benefits of compensation based on the unified price compensation according to time period and the market price obtained by the non-cooperative game of the non-cooperative game optimization dispatch method based on load aggregator classification described in the example are compared. The comparison results are shown in Figure 6a , Figure 6b .

[0110] from Figure 6a , Figure 6b It can be seen that under the non-cooperative game situation, the unit price and revenue of load aggregators increase during valley and peak periods and decrease during normal periods. According to statistics, the daily revenue of various types of load aggregators has increased.

[0111] The simulation results show the effectiveness of the non-cooperative game optimization scheduling method based on load aggregator classification.

[0112] The above technical features constitute the embodiments of the present invention, which have strong adaptability and implementation effect. Non-essential technical features can be added or reduced according to actual needs to meet the requirements of different situations.

Claims

1. A non-cooperative game optimization scheduling method based on load aggregator classification, characterized by Follow these steps: Step (1), determining the smart grid infrastructure in which the load aggregator participates; Step (2), establishing a load aggregator classification model, that is, classifying the flexible loads, and classifying the load aggregators according to the different types of aggregated flexible loads; Step (3), establishing a supply-side power supply model, the supply-side power supply model includes a thermal power unit output model and a wind power output model; Step (4), based on the classification of load aggregators, firstly, the game behavior of load aggregators in the process of participating in power grid dispatch is given, and a non-cooperative game optimization model for the classification of load aggregators is established based on the game behavior; Step (5), establishing the grid side objective function and the load aggregator side objective function; Step (6), solving the optimal configuration between Nash equilibrium and unit output state based on the NSGA-II multi-objective optimization algorithm and the interior point generation combined algorithm; Wherein, in step (1), the smart grid infrastructure includes two parts: distributed power sources and loads. The distributed power sources consider wind power generation and thermal power generation to participate in grid dispatching at the same time, and the load considers flexible loads. Flexible loads are classified and aggregated by load aggregators to participate in grid dispatching. In step (2), the flexible loads are divided into three categories: Class A loads, Class B loads, and Class C loads. Modeling and analysis of different types of flexible loads are performed. Based on the classification of flexible loads, load aggregators are divided into two categories: commercial and residential. The specific steps are as follows: (21) Class A load can be transferred or interrupted. Assuming H1 is the entire working period of Class A load, the power required by the load must be completed during the working period, that is, Where: Q A is the daily power required by Class A load; Pload1 is the rated power of the load, H1={1,2,...,H}; (22) Class B loads can be transferred but not interrupted. Assuming that the participation period of Class B loads is H2, the load aggregator needs to ensure that Class B loads complete all required electricity within the H2 period, that is, Where: H2={1,2,...,H}; Q B is the daily power requirement of class B load; arg(Sch2(h)=1) means that during the h period, the load is in the on state and can participate in the scheduling. and α represent the minimum time and maximum time in the on period, respectively; (23) Class C loads cannot be transferred but can be interrupted intermittently, that is: The constraint condition in this formula means that if the C load is turned off in the h-1 period and does not participate in the dispatch, it is turned on in the h period, and then the load is turned off in the h+β period, then the total dispatch period of the C load in the entire dispatch process is min(β). In the formula, Q C is the daily power requirement of Class C load, t on,3 is the shortest on time of Class C load, t off,3 Maximum shutdown time; (24) Based on the classification of flexible loads, load aggregators are also classified into commercial and residential types. Commercial load aggregators only consider air-conditioning loads as Class A loads to participate in load aggregator scheduling. To represent commercial load aggregator n b Aggregate class A load, i=1,2...,I b ; then commercial load aggregator n b The load volume planned for dispatching in the day-ahead dispatching phase during period t is: Assume that in the residential load aggregator n r There are three types of loads involved in dispatching in the jurisdiction, each type of load has I r Users participated, using To represent the load aggregator n r Aggregate class A load, i=1,2...,I r ;use To represent the load aggregator n r Aggregate Class B load, i=1,2...,I r ;use To represent the load aggregator n r Aggregate Class C load, i=1,2...,I r ; then the residential load aggregator n r The load volume planned for dispatching in the day-ahead dispatching phase during period t is: In step (3), the wind power output model includes the wind turbine output constraint model and the wind abandonment cost. The maximum output of a single wind turbine is the rated power P r , when exiting the power grid, the minimum output is 0, and the wind turbine output constraint model is: In order to reduce wind curtailment and increase grid revenue, the cost of wind curtailment is also taken into account when calculating revenue. The expression is: In the formula, is the wind curtailment cost, λ wf is the wind curtailment cost coefficient, is the abandoned wind power of the wind farm at time t; The output model of thermal power units includes the upper and lower output constraints of the thermal power units, the number of start and stop times of the thermal power units, the unit climbing constraints and the total operating cost of the thermal power units. The operation of the thermal power units must meet the upper and lower output constraints: Constraints on the number of starts and stops of thermal power units: And the unit climbing constraint: There are two ways to regulate thermal power units: variable load and start-stop. Variable load regulation refers to regulating the output of thermal power units within the upper and lower limits of thermal power output. Start-stop regulation refers to regulating the output of units by starting and stopping units within the constraints of the number of starts and stops. The strategy of jointly regulating the two methods is adopted to regulate the output of thermal power units. The operating cost of thermal power units includes coal consumption cost and start-up and shutdown cost: Taking into account the impact of power grid cost on thermal power unit power dispatch, the coal consumption cost of thermal power units is expressed as a quadratic function of unit power: in, is the output of thermal power unit i at time t, a i , b i 、c i is the coal consumption cost coefficient of unit i during power generation; in the actual power generation process of thermal power units, the start and stop of thermal power units will also bring huge losses, and frequent start and stop of units will also affect the life of the units. Therefore, the start and stop costs of thermal power units are also considered in the operating costs: where λ nf is the start-up and shutdown cost coefficient of thermal power units, n t,i is the start and stop status of thermal power unit i at time t. The total operating cost of the thermal power unit is obtained from the above formula: is the total operating cost of the thermal power unit, F1 is the coal consumption cost of the thermal power unit, and F2 is the start-up and shutdown cost of the thermal power unit; In step (4), according to the load aggregator classification, a non-cooperative game optimization model G = [N, {S n },{C n }], the specific steps are as follows: (41) The participants are N load aggregators on the residential load side, N = {1, 2, ..., n}; (42) Strategy Set The load reduction plan reported by N load aggregators to the power company is the load reduction amount L in each peak-cutting period. st,n combination of; (43) Payment function {C n } is the profit obtained by each load aggregator, which is expressed as the decision of aggregator n and other N-1 aggregators to bid together with the goal of maximizing their own profits, that is: ei(t) is the market price; (44) Information set: The game process in which load aggregators participate in power grid dispatching is a complete information game. The bidding strategy and revenue of each load aggregator are public information and are known to all load aggregators participating in the game. (45) Game results: Each aggregator takes maximizing its own profit as the objective function and continuously changes its bidding strategy according to the objective function. During the game, each load aggregator will choose the bidding strategy that maximizes its own interests. When all aggregators do not change their bidding strategies, that is, when any aggregator cannot change its own strategy to increase its profit, the game reaches a stable state, namely the Nash equilibrium, which is the Nash equilibrium solution of the non-cooperative game model. In this stable state, the strategy set of each load user is the equilibrium solution of the game, namely {S n }; Strategy set {S n The necessary and sufficient condition for a multi-user non-cooperative game Nash equilibrium solution is: for every strategy of any load aggregator n Both That is, changing the strategy of any load aggregator can't reduce its final profit. For a situation, is a pure strategy Nash equilibrium point of the load aggregator non-cooperative game model G, C i (s j ) is the corresponding equilibrium result; The specific steps of step (5) are as follows: (51) Grid side objective function: in The revenue from electricity sales to the power grid, They are respectively the operating cost of thermal power units, the cost of wind power abandonment and the cost of load aggregators, and the revenue from electricity sales is the total load at time t, e t is the electricity selling price at time t; (52) Load aggregator side objective function The objective function of a commercial load aggregator is expressed as: The objective function of the residential load aggregator is expressed as: (53) According to the three objective functions, the multi-objective problem is described as follows: Here Ω is the decision space and the set of goals to be achieved is defined as {F(x)|x∈Ω}.

Images