A robust expansion optimization scheduling method for a composite microgrid

Abstract

This invention proposes a robust augmentation optimization scheduling method for hybrid microgrids, addressing the limitations of existing methods in effectively handling the uncertainty of wind power prediction errors and the low efficiency of scheduling model solutions in hybrid wind-thermal-storage microgrids. This method constructs a fuzzy set of wind power prediction errors based on historical data statistical characteristics to characterize wind power uncertainty; establishes a two-stage partially robust optimization scheduling model encompassing the coordination of multiple stakeholders—wind turbines, thermal power units, and energy storage systems—with the goal of minimizing system operating costs; designs a constrained augmentation optimization solution algorithm to improve solution efficiency and accuracy; and finally obtains the optimal scheduling scheme for the hybrid wind-thermal-storage microgrid, achieving economical and stable operation of the microgrid.

Description

Technical Field

[0001] This invention presents a robust augmentation optimization scheduling method for composite microgrids, which solves the problems of existing methods being unable to effectively address the uncertainty of wind power prediction errors and the low efficiency of scheduling model solving in wind-thermal-storage composite microgrids. By constructing a fuzzy set of wind power prediction errors, establishing a two-stage partially robust optimization scheduling model, and designing a constrained augmentation optimization algorithm, the method reduces operating costs while ensuring robust and stable system operation. This is of great significance for promoting efficient optimization scheduling of wind-thermal-storage composite microgrids and belongs to the field of power system optimization scheduling. Background Technology

[0002] With the large-scale integration of distributed renewable energy and the profound transformation of power system architecture, the problem of coordinated dispatch of multiple entities (source, load, and storage) in composite microgrids is becoming increasingly prominent. Composite microgrids can flexibly absorb renewable energy sources such as wind power, improve regional energy structure, and promote the achievement of carbon emission reduction targets, which is of great significance to the construction and development of new power systems. However, the randomness and volatility of wind power output in composite microgrids pose a severe challenge to dispatch decisions. To improve the robustness and economy of dispatch schemes, two-stage partial Bruker optimization technology has been widely applied in microgrid optimization dispatch research.

[0003] Most microgrid dispatching research employs robust optimization or stochastic optimization methods to establish dispatching models. However, robust optimization methods suffer from poor economic efficiency due to overly conservative uncertainty set construction, while stochastic optimization methods rely on precise probability distribution assumptions, making it difficult to accurately characterize the true stochastic characteristics of wind power output prediction errors. Therefore, how to reasonably model the uncertainty of wind power prediction errors in wind-thermal-storage integrated microgrids is an important research topic. During the optimization dispatching process, the probability distribution information of wind power prediction errors is difficult to obtain accurately, and uncertainty sets constructed solely based on limited historical data are often distorted. This results in dispatching models that fail to guarantee the robustness of the scheme and struggle to achieve coordinated and optimized operation among wind, thermal, and storage entities. Therefore, characterizing the uncertainty of wind power prediction errors based on fuzzy sets and establishing a two-stage partially Bruker robust optimization dispatching model is of great significance for improving the robustness and economy of dispatching schemes in wind-thermal-storage integrated microgrids. Furthermore, the minimax-minimax nested optimization problem formed by the two-stage partially Bruker model has a complex structure and large scale. Traditional solution algorithms suffer from low computational efficiency and poor convergence, making it difficult to meet the engineering requirements of practical dispatching. Therefore, designing an efficient constrained augmentation optimization algorithm can not only significantly improve the solution efficiency and accuracy of the two-stage sub-Bruker scheduling model, but is also the key to achieving stable and efficient operation of wind-thermal-storage hybrid microgrids.

[0004] This invention analyzes the uncertainty characteristics of wind power output in a wind-thermal-storage hybrid microgrid, constructs a fuzzy set of wind power prediction errors based on historical data, establishes a two-stage sub-Bruker optimal scheduling model for the wind-thermal-storage hybrid microgrid, designs a constrained augmentation optimization solution algorithm, decomposes the complex minimax nested problem into a main subproblem and solves iteratively and efficiently, and finally obtains the optimal scheduling scheme under various uncertain scenarios, thus realizing the economical and stable operation of the wind-thermal-storage hybrid microgrid. Summary of the Invention

[0005] This invention proposes a robust augmentation optimization scheduling method for hybrid microgrids, addressing the limitations of existing methods in effectively handling the uncertainty of wind power prediction errors and the low efficiency of scheduling model solutions in hybrid wind-thermal-storage microgrids. This method constructs a fuzzy set of wind power prediction errors based on historical data statistical characteristics to characterize wind power uncertainty; establishes a two-stage partially robust optimization scheduling model encompassing the coordination of multiple stakeholders—wind turbines, thermal power units, and energy storage systems—with the goal of minimizing system operating costs; designs a constrained augmentation optimization solution algorithm to improve solution efficiency and accuracy; and finally obtains the optimal scheduling scheme for the hybrid wind-thermal-storage microgrid, achieving economical and stable operation of the microgrid.

[0006] The present invention adopts the following technical solution and implementation steps:

[0007] 1. A robust augmentation and optimization scheduling method for composite microgrids, specifically including the following steps:

[0008] (1) Constructing a fuzzy set of wind power prediction errors for a wind-thermal-storage hybrid microgrid: Taking a wind-thermal-storage hybrid microgrid as the research object, this microgrid consists of three core units: wind turbine generators, thermal power generators, and energy storage systems. Wind turbine generators generate electricity using wind energy, which is random and volatile. Thermal power generators, as the main controllable power source, can provide stable basic output. The energy storage system smooths out wind power fluctuations and improves the system's operational flexibility through charging and discharging operations. The three units operate in coordination, jointly connecting to the distribution network and supplying power to local loads while interacting with the external power grid. Due to the significant uncertainty in wind power output, the optimal scheduling of the microgrid needs to fully consider the impact of wind power prediction errors. Therefore, a fuzzy set of wind power prediction errors is constructed to characterize the uncertainty range of the errors.

[0009] ① Taking a wind-thermal-storage hybrid microgrid consisting of wind turbines, thermal power units, and energy storage systems as the research object, wind power is predicted using historical wind power data. Calculate the error between the actual and predicted wind power values. :

[0010]

[0011] in, The actual wind power output at time t is expressed in megawatts. The predicted wind power at time t is expressed in megawatts.

[0012] ② Construct a fuzzy set of wind power prediction errors:

[0013]

[0014]

[0015]

[0016] in, Let be the fuzzy set of wind power prediction errors at time t. This represents the probability distribution of wind power prediction errors. This is the set of probability distributions for all wind power prediction errors. The empirical distribution of historical errors in wind power at time t. The optimal transmission distance between the true distribution and the empirical distribution. Let t be the optimal transmission radius of a sliding window with a fixed width. The optimal transmission radius at time t is... This represents the 95th percentile of the wind power prediction error at time t. The 5th percentile of the wind power prediction error at time t;

[0017] (2) Establish a robust optimization scheduling model for a wind-thermal-storage hybrid microgrid:

[0018] ① Construct the day-ahead pre-dispatch cost objective function for a wind-thermal-storage hybrid microgrid:

[0019]

[0020]

[0021]

[0022]

[0023]

[0024]

[0025]

[0026] in, y represents the total day-ahead pre-scheduling cost, expressed in ten thousand yuan, and y represents the day-ahead pre-scheduling decision variable. The operating cost of thermal power units is expressed in ten thousand yuan. The operating cost of the wind turbine is expressed in ten thousand yuan. The operating cost of the energy storage system is expressed in ten thousand yuan. The cost of purchasing electricity from the power grid is expressed in ten thousand yuan. Cost reduction measures are implemented to reduce load, with the unit being 10,000 yuan. The cost of wind curtailment penalty is expressed in ten thousand yuan. The calculation methods for each cost are shown in formulas (6)-(11). Let t be the output power of the thermal power unit, in megawatts. Let t be the output power of the wind turbine, in megawatts. The charging power of the energy storage system at time t is expressed in megawatts. Let t be the discharge power of the energy storage system at time t, in megawatts. The power purchased by the grid at time t is expressed in megawatts. The load reduction power at time t is expressed in megawatts. Let t be the wind curtailment power, in megawatts. The standby capacity for adjusting the output of thermal power units at time t is expressed in megawatts (MW). The standby capacity for adjusting the wind turbine output at time t is expressed in megawatts (MW). This is the power generation cost coefficient for thermal power units, expressed in RMB 10,000 per megawatt-hour. This is the cost coefficient for standby capacity of thermal power units, expressed in RMB 10,000 per megawatt-hour. This is the power generation cost coefficient for wind turbine generators, expressed in RMB 10,000 per megawatt-hour. This is the cost coefficient for standby capacity of wind turbine units, expressed in RMB 10,000 per megawatt-hour. This refers to the charge / discharge cost coefficient for energy storage systems, expressed in tens of thousands of yuan per megawatt-hour. This is the load reduction cost factor, expressed in RMB 10,000 per megawatt-hour. This is the cost factor for wind curtailment, expressed in tens of thousands of yuan per megawatt-hour. The charging efficiency of the energy storage system is defined, with a value ranging from [0.5, 1]. Let be the discharge efficiency of the energy storage system, with a value ranging from [0.5, 1]. Let t be the electricity price at time t, expressed in ten thousand yuan per megawatt-hour.

[0027] ② Construct the objective function for the real-time rescheduling cost of a wind-thermal-storage hybrid microgrid:

[0028]

[0029] in, Let x be the real-time rescheduling cost under a given day-ahead decision y and wind power error scenario S, expressed in ten thousand yuan. Let t be the regulating power of the thermal power unit, in megawatts. The wind turbine's regulating power at time t is expressed in megawatts. The charging regulation power of the energy storage system at time t, in megawatts. Let t be the discharge regulation power of the energy storage system at time t, in megawatts. The power purchased and regulated by the power grid at time t is expressed in megawatts. This is the power regulation cost coefficient for thermal power units, expressed in RMB 10,000 per megawatt-hour. This is the power regulation cost coefficient for wind turbine units, expressed in tens of thousands of yuan per megawatt-hour. This is the power regulation cost coefficient for energy storage systems, expressed in RMB 10,000 per megawatt-hour.

[0030] (3) Design a constrained augmentation optimization algorithm for a wind-thermal-storage hybrid microgrid. The specific steps are as follows:

[0031] ① Set the total number of iterations of the algorithm to be Convergence tolerance Number of scenes Initialize the iteration counter Initial cutting constraint set ;

[0032] ② Initialize the fuzzy set: Independently sample the error distribution at each time step , Combine to generate a set of daily error scenarios ;

[0033] ③ Construction and solution of the extended principal problem: Construct a mixed integer programming model that includes the decision variables of the current stage and the upper bound variable of the worst expected cost of the real stage, as shown in formulas (13)-(15):

[0034]

[0035]

[0036]

[0037] Wherein, formula (13) is the objective function. The day-ahead pre-scheduling cost is shown in formula (5). Let be the upper bound variable of the worst expected cost in the real-time stage, and let formula (14) be the set of cutting constraints generated iteratively. Let l be the intercept term of the l-th cutting plane. The decision variable corresponding to the l-th cutting plane The coefficients, L is the currently generated set of cutting constraints, formula (15) is the set of operating constraints for each unit, and the model is solved using the commercial mathematical programming solver Gurobi to obtain the current day-ahead stage decision variables. and the lower realm ;

[0038] ④ Subproblem Construction and Solving: Constructing subproblems based on the current day-ahead decision and each scene The linear programming subproblems are shown in equations (16)-(17):

[0039]

[0040]

[0041] in, For the current decision and scene The real-time rescheduling cost is given by b, where b is the real-time rescheduling cost coefficient vector, and formula (17) is the model constraint. The commercial mathematical programming solver Gurobi is used for each scenario. Solve the optimal value of the subproblem and the dual variables of the constraints ;

[0042] ⑤ Determine the extension boundary: Construct a linear programming model that includes the optimal values ​​of all scenario subproblems, as shown in formulas (18)-(20):

[0043]

[0044]

[0045]

[0046] in, For the scene probability weights, For a set of probability distributions, Let the distance function be the distance between scenes. Solving this linear programming model yields the worst-case scenario probability distribution. And update the upper bound. ;

[0047] ⑥ Multidimensional cutting constraint extension: Extracting the optimal value of each scenario's sub-problem and dual variables Combining the worst-case scenario probability distribution The worst-case expected cost linear cutting constraint is constructed as shown in formula (22). This constraint is added to the main problem constraint set (15), and the number of generated cutting constraints is updated. Continue until the convergence condition is met;

[0048]

[0049]

[0050]

[0051] in, For the scene The intercept term of the subproblem, For the scene Decision variables for the next day The sensitivity coefficient, For the current day decision variable The corresponding coefficient column vector in the real-time rescheduling constraint matrix;

[0052] ⑦ Convergence judgment: Calculate the relative gap g between the upper and lower bounds as shown in formula (24). If or number of iterations If the algorithm converges, it outputs the minimum daily scheduling cost target value of the microgrid and the corresponding optimal decision variable value; otherwise, it lets... And return to step ③;

[0053]

[0054] (4) Optimize the scheduling scheme of the wind-thermal-storage composite microgrid: Construct a robust optimization model of the wind-thermal-storage composite microgrid using the fuzzy set of wind power prediction error, and use the constrained augmentation optimization algorithm to iteratively solve the model described by formulas (5)-(12) to obtain the optimal daily pre-scheduling stage decision variable y and the real-time rescheduling stage decision variable x, and calculate the minimum daily scheduling cost target value of the wind-thermal-storage composite microgrid, thereby achieving the stable operation of the wind-thermal-storage composite microgrid. Attached Figure Description

[0055] Figure 1 This is the fuzzy set of wind power prediction errors in this invention;

[0056] Figure 2This is the thermal power unit output plan of the present invention;

[0057] Figure 3 This is the wind turbine output plan of the present invention;

[0058] Figure 4 This is the energy storage system output plan of the present invention. Detailed Implementation

[0059] 1. A robust augmentation and optimization scheduling method for composite microgrids, specifically including the following steps:

[0060] (1) Constructing a fuzzy set of wind power prediction errors for a wind-thermal-storage hybrid microgrid: Taking a wind-thermal-storage hybrid microgrid as the research object, this microgrid consists of three core units: wind turbine generators, thermal power generators, and energy storage systems. Wind turbine generators generate electricity using wind energy, which is random and volatile. Thermal power generators, as the main controllable power source, can provide stable basic output. The energy storage system smooths out wind power fluctuations and improves the system's operational flexibility through charging and discharging operations. The three units operate in coordination, jointly connecting to the distribution network and supplying power to local loads while interacting with the external power grid. Due to the significant uncertainty in wind power output, the optimal scheduling of the microgrid needs to fully consider the impact of wind power prediction errors. Therefore, a fuzzy set of wind power prediction errors is constructed to characterize the uncertainty range of the errors.

[0061] ① Taking a wind-thermal-storage hybrid microgrid consisting of wind turbines, thermal power units, and energy storage systems as the research object, wind power is predicted using historical wind power data. Calculate the error between the actual and predicted wind power values. :

[0062]

[0063] in, The actual wind power output at time t is expressed in megawatts. The predicted wind power at time t is expressed in megawatts.

[0064] ② Construct a fuzzy set of wind power prediction errors:

[0065]

[0066]

[0067]

[0068] in, Let be the fuzzy set of wind power prediction errors at time t. This represents the probability distribution of wind power prediction errors. This is the set of probability distributions for all wind power prediction errors. The empirical distribution of historical errors in wind power at time t. The optimal transmission distance between the true distribution and the empirical distribution. Let t be the optimal transmission radius of a sliding window with a fixed width. The optimal transmission radius at time t is... This represents the 95th percentile of the wind power prediction error at time t. The 5th percentile of the wind power prediction error at time t;

[0069] (2) Establish a robust optimization scheduling model for a wind-thermal-storage hybrid microgrid:

[0070] ① Construct the day-ahead pre-dispatch cost objective function for a wind-thermal-storage hybrid microgrid:

[0071]

[0072]

[0073]

[0074]

[0075]

[0076]

[0077]

[0078] in, y represents the total day-ahead pre-scheduling cost, expressed in ten thousand yuan, and y represents the day-ahead pre-scheduling decision variable. The operating cost of thermal power units is expressed in ten thousand yuan. The operating cost of the wind turbine is expressed in ten thousand yuan. The operating cost of the energy storage system is expressed in ten thousand yuan. The cost of purchasing electricity from the power grid is expressed in ten thousand yuan. Cost reduction measures are implemented to reduce load, with the unit being 10,000 yuan. The cost of wind curtailment penalty is expressed in ten thousand yuan. The calculation methods for each cost are shown in formulas (6)-(11). Let t be the output power of the thermal power unit, in megawatts. Let t be the output power of the wind turbine, in megawatts. The charging power of the energy storage system at time t is expressed in megawatts. Let t be the discharge power of the energy storage system at time t, in megawatts. The power purchased by the grid at time t is expressed in megawatts. The load reduction power at time t is expressed in megawatts. Let t be the wind curtailment power, in megawatts. The standby capacity for adjusting the output of thermal power units at time t is expressed in megawatts (MW). The standby capacity for adjusting the wind turbine output at time t is expressed in megawatts (MW). This is the power generation cost coefficient for thermal power units, expressed in RMB 10,000 per megawatt-hour. This is the cost coefficient for standby capacity of thermal power units, expressed in RMB 10,000 per megawatt-hour. This is the power generation cost coefficient for wind turbine generators, expressed in RMB 10,000 per megawatt-hour. This is the cost coefficient for standby capacity of wind turbine units, expressed in RMB 10,000 per megawatt-hour. This refers to the charge / discharge cost coefficient for energy storage systems, expressed in tens of thousands of yuan per megawatt-hour. This is the load reduction cost factor, expressed in RMB 10,000 per megawatt-hour. This is the cost factor for wind curtailment, expressed in tens of thousands of yuan per megawatt-hour. The charging efficiency of the energy storage system is defined, with a value ranging from [0.5, 1]. Let be the discharge efficiency of the energy storage system, with a value ranging from [0.5, 1]. Let t be the electricity price at time t, expressed in ten thousand yuan per megawatt-hour.

[0079] ② Construct the objective function for the real-time rescheduling cost of a wind-thermal-storage hybrid microgrid:

[0080]

[0081] in, Let x be the real-time rescheduling cost under a given day-ahead decision y and wind power error scenario S, expressed in ten thousand yuan. Let t be the regulating power of the thermal power unit, in megawatts. The wind turbine's regulating power at time t is expressed in megawatts. The charging regulation power of the energy storage system at time t, in megawatts. Let t be the discharge regulation power of the energy storage system at time t, in megawatts. The power purchased and regulated by the power grid at time t is expressed in megawatts. This is the power regulation cost coefficient for thermal power units, expressed in RMB 10,000 per megawatt-hour. This is the power regulation cost coefficient for wind turbine units, expressed in tens of thousands of yuan per megawatt-hour. This is the power regulation cost coefficient for energy storage systems, expressed in RMB 10,000 per megawatt-hour.

[0082] (3) Design a constrained augmentation optimization algorithm for a wind-thermal-storage hybrid microgrid. The specific steps are as follows:

[0083] ① Set the total number of iterations of the algorithm to be Convergence tolerance Number of scenes Initialize the iteration counter Initial cutting constraint set ;

[0084] ② Initialize the fuzzy set: Independently sample the error distribution at each time step , Combine to generate a set of daily error scenarios ;

[0085] ③ Construction and solution of the extended principal problem: Construct a mixed integer programming model that includes the decision variables of the current stage and the upper bound variable of the worst expected cost of the real stage, as shown in formulas (13)-(15):

[0086]

[0087]

[0088]

[0089] Wherein, formula (13) is the objective function. The day-ahead pre-scheduling cost is shown in formula (5). Let be the upper bound variable of the worst expected cost in the real-time stage, and let formula (14) be the set of cutting constraints generated iteratively. Let l be the intercept term of the l-th cutting plane. The decision variable corresponding to the l-th cutting plane The coefficients, L is the currently generated set of cutting constraints, formula (15) is the set of operating constraints for each unit, and the model is solved using the commercial mathematical programming solver Gurobi to obtain the current day-ahead stage decision variables. and the lower realm ;

[0090] ④ Subproblem Construction and Solving: Constructing subproblems based on the current day-ahead decision and each scene The linear programming subproblems are shown in equations (16)-(17):

[0091]

[0092]

[0093] in, For the current decision and scene The real-time rescheduling cost is given by b, where b is the real-time rescheduling cost coefficient vector, and formula (17) is the model constraint. The commercial mathematical programming solver Gurobi is used for each scenario. Solve the optimal value of the subproblem and the dual variables of the constraints ;

[0094] ⑤ Determine the extension boundary: Construct a linear programming model that includes the optimal values ​​of all scenario subproblems, as shown in formulas (18)-(20):

[0095]

[0096]

[0097]

[0098] in, For the scene probability weights, For a set of probability distributions, Let the distance function be the distance between scenes. Solving this linear programming model yields the worst-case scenario probability distribution. And update the upper bound. ;

[0099] ⑥ Multidimensional cutting constraint extension: Extracting the optimal value of each scenario's sub-problem and dual variables Combining the worst-case scenario probability distribution The worst-case expected cost linear cutting constraint is constructed as shown in formula (22). This constraint is added to the main problem constraint set (15), and the number of generated cutting constraints is updated. Continue until the convergence condition is met;

[0100]

[0101]

[0102]

[0103] in, For the scene The intercept term of the subproblem, For the scene Decision variables for the next day The sensitivity coefficient, For the current day decision variable The corresponding coefficient column vector in the real-time rescheduling constraint matrix;

[0104] ⑦ Convergence judgment: Calculate the relative gap g between the upper and lower bounds as shown in formula (24). If or number of iterations If the algorithm converges, it outputs the minimum daily scheduling cost target value of the microgrid and the corresponding optimal decision variable value; otherwise, it lets... And return to step ③;

[0105]

[0106] (4) Optimize the scheduling scheme of the wind-thermal-storage composite microgrid: Construct a robust optimization model of the wind-thermal-storage composite microgrid using the fuzzy set of wind power prediction error, and use the constrained augmentation optimization algorithm to iteratively solve the model described by formulas (5)-(12) to obtain the optimal daily pre-scheduling stage decision variable y and the real-time rescheduling stage decision variable x, and calculate the minimum daily scheduling cost target value of the wind-thermal-storage composite microgrid, thereby achieving the stable operation of the wind-thermal-storage composite microgrid.

Claims

1. A robust augmentation and optimization scheduling method for composite microgrids, characterized in that: The process involves constructing a fuzzy set of wind power prediction errors for a wind-thermal-storage hybrid microgrid, establishing a robust optimization scheduling model for the hybrid microgrid, designing a constrained augmentation optimization algorithm for the hybrid microgrid, and optimizing the scheduling scheme for the hybrid microgrid. The specific steps include: (1) Constructing the fuzzy set of wind power prediction errors for a wind-thermal-storage hybrid microgrid: This study focuses on a wind-thermal-storage hybrid microgrid consisting of wind turbines, thermal power units, and energy storage systems. Historical wind power data is used to predict wind power output. Calculate the error between the actual and predicted wind power values. :

2. Among them, The actual wind power output at time t is expressed in megawatts. The predicted wind power at time t is expressed in megawatts. Constructing a fuzzy set of wind power prediction errors: ； ； ； in, Let be the fuzzy set of wind power prediction errors at time t. Let represent the probability distribution of wind power prediction errors. This is the set of probability distributions for all wind power prediction errors. The empirical distribution of historical errors in wind power at time t. The optimal transmission distance between the true distribution and the empirical distribution. Let t be the optimal transmission radius of a sliding window with a fixed width. The optimal transmission radius at time t is... This represents the 95th percentile of the wind power prediction error at time t. The 5th percentile of the wind power prediction error at time t; (2) Establish a robust optimization scheduling model for a wind-thermal-storage hybrid microgrid: Construct the day-ahead pre-scheduling cost objective function for a wind-thermal-storage hybrid microgrid: ； ； ； ； ； （10）； ； in, y represents the total day-ahead pre-scheduling cost, expressed in ten thousand yuan, and y represents the day-ahead pre-scheduling decision variable. The operating cost of thermal power units is expressed in ten thousand yuan. The operating cost of the wind turbine is expressed in ten thousand yuan. The operating cost of the energy storage system is expressed in ten thousand yuan. The cost of purchasing electricity from the power grid is expressed in ten thousand yuan. Cost reduction measures to reduce load, unit: 10,000 yuan. The cost of wind curtailment penalty is expressed in ten thousand yuan. The calculation methods for each cost are shown in formulas (6)-(11). Let t be the output power of the thermal power unit, in megawatts. Let t be the output power of the wind turbine, in megawatts. The charging power of the energy storage system at time t is expressed in megawatts. Let t be the discharge power of the energy storage system at time t, in megawatts. The power purchased by the grid at time t is expressed in megawatts. The load reduction power at time t is expressed in megawatts. Let t be the wind curtailment power, in megawatts. The standby capacity for adjusting the output of thermal power units at time t is expressed in megawatts (MW). The standby capacity for adjusting the wind turbine output at time t is expressed in megawatts. This is the power generation cost coefficient for thermal power units, expressed in RMB 10,000 per megawatt-hour. This is the cost coefficient for standby capacity of thermal power units, expressed in RMB 10,000 per megawatt-hour. This is the power generation cost coefficient for wind turbine generators, expressed in RMB 10,000 per megawatt-hour. This is the cost coefficient for standby capacity of wind turbine units, expressed in RMB 10,000 per megawatt-hour. The charging and discharging cost coefficient for energy storage systems is expressed in tens of thousands of yuan per megawatt-hour. This is the load reduction cost factor, expressed in RMB 10,000 per megawatt-hour. This is the cost factor for wind curtailment, expressed in tens of thousands of yuan per megawatt-hour. The charging efficiency of the energy storage system is defined, with a value ranging from [0.5, 1]. Let be the discharge efficiency of the energy storage system, with a value ranging from [0.5, 1]. Let t be the electricity price at time t, expressed in ten thousand yuan per megawatt-hour. Construct the objective function for the real-time rescheduling cost of a wind-thermal-storage hybrid microgrid: ； in, Let x be the real-time rescheduling cost under a given day-ahead decision y and wind power error scenario S, expressed in ten thousand yuan. Let t be the regulating power of the thermal power unit, in megawatts. The wind turbine's regulating power at time t is expressed in megawatts. The charging regulation power of the energy storage system at time t, in megawatts. Let t be the discharge regulation power of the energy storage system at time t, in megawatts. The power purchased and regulated by the power grid at time t is expressed in megawatts. This is the power regulation cost coefficient for thermal power units, expressed in RMB 10,000 per megawatt-hour. This is the power regulation cost coefficient for wind turbine units, expressed in tens of thousands of yuan per megawatt-hour. This is the power regulation cost coefficient for energy storage systems, expressed in RMB 10,000 per megawatt-hour. (3) Design a constrained augmentation optimization algorithm for a wind-thermal-storage hybrid microgrid. The specific steps are as follows: ① Set the total number of iterations of the algorithm to be Convergence tolerance Number of scenes Initialize the iteration counter Initial cutting constraint set ; ② Initialize the fuzzy set: Independently sample the error distribution at each time step , Combine to generate a set of daily error scenarios ; ③ Construction and solution of the extended principal problem: Construct a mixed integer programming model that includes the decision variables of the current stage and the upper bound variable of the worst expected cost of the real stage, as shown in formulas (13)-(15): ； ； ； Wherein, formula (13) is the objective function. The day-ahead pre-scheduling cost is shown in formula (5). Let be the upper bound variable of the worst expected cost in the real-time stage, and let formula (14) be the set of cutting constraints generated iteratively. Let l be the intercept term of the l-th cutting plane. The decision variable corresponding to the l-th cutting plane The coefficients, L is the currently generated set of cutting constraints, formula (15) is the set of operating constraints for each unit, and the model is solved using the commercial mathematical programming solver Gurobi to obtain the current day-ahead stage decision variables. and the lower realm ; ④ Subproblem Construction and Solving: Constructing subproblems based on the current day-ahead decision and each scene The linear programming subproblems are shown in equations (16)-(17): ； ； in, For the current decision and scene The real-time rescheduling cost is given by b, where b is the real-time rescheduling cost coefficient vector, and formula (17) is the model constraint. The commercial mathematical programming solver Gurobi is used for each scenario. Solve the optimal value of the subproblem and the dual variables of the constraints ; ⑤ Determine the extension boundary: Construct a linear programming model that includes the optimal values ​​of all scenario subproblems, as shown in formulas (18)-(20): ； ； ； in, For the scene probability weights, For a set of probability distributions, Let the distance function be the distance between scenes. Solving this linear programming model yields the worst-case scenario probability distribution. And update the upper bound. ; ⑥ Multidimensional cutting constraint extension: Extracting the optimal value of each scenario's sub-problem and dual variables Combining the worst-case scenario probability distribution The worst-case expected cost linear cutting constraint is constructed as shown in formula (22). This constraint is added to the main problem constraint set (15), and the number of generated cutting constraints is updated. Continue until the convergence condition is met; ； ； ； in, For the scene The intercept term of the subproblem, For the scene Decision variables for the next day The sensitivity coefficient, For the current day decision variable The corresponding coefficient column vector in the real-time rescheduling constraint matrix; ⑦ Convergence judgment: Calculate the relative gap g between the upper and lower bounds as shown in formula (24). If or number of iterations If the algorithm converges, it outputs the minimum daily scheduling cost target value of the microgrid and the corresponding optimal decision variable value; otherwise, it lets... And return to step ③; ； (4) Optimize the scheduling scheme of the wind-thermal-storage composite microgrid: Construct a robust optimization model of the wind-thermal-storage composite microgrid using the fuzzy set of wind power prediction error, and use the constrained augmentation optimization algorithm to iteratively solve the model described by formulas (5)-(12) to obtain the optimal daily pre-scheduling stage decision variable y and the real-time rescheduling stage decision variable x, and calculate the minimum daily scheduling cost target value of the wind-thermal-storage composite microgrid.

Images