A kl divergence-based robust optimization scheduling method for microgrid distribution in industrial park

Abstract

The application discloses a kind of industrial park microgrid distribution robust optimization scheduling method based on KL divergence, steps include: obtaining the operating parameter of each flexibility resource of industrial park microgrid, establishes industrial park microgrid deterministic model;Obtain the historical sample of wind and light output, construct the fuzzy set of wind and light output based on KL divergence;According to fuzzy set, the deterministic model of industrial park microgrid is converted into two-stage distribution robust optimization model;Two-stage distribution robust optimization model is decoupled into main problem for obtaining the lower bound of original problem and sub-problem for obtaining the upper bound of original problem, the main problem and sub-problem are iteratively solved until the difference between upper and lower bound meets the convergence condition, and the optimal scheduling strategy is obtained.The application realizes the robust optimization scheduling of industrial park microgrid in uncertain environment, improves the economy of system operation and the accommodation capacity of renewable energy, and considers the robustness and economy of scheduling scheme.

Description

Technical Field

[0001] This invention relates to microgrid optimization scheduling technology, specifically to a KL divergence-based sub-Bluerg bar optimization scheduling method for industrial park microgrids. Background Technology

[0002] Industrial parks, as energy-intensive areas, have complex energy consumption structures and large load fluctuations, placing high demands on power supply reliability and economy. Microgrids, as small-scale power generation and distribution systems integrating distributed power sources, energy storage systems, flexible loads, and control devices, can effectively promote the local consumption of renewable energy and enhance the flexibility and resilience of the park's energy supply. With the increasing penetration of intermittent renewable energy sources such as photovoltaic and wind power in industrial park microgrids, the uncertainty of wind and solar power output poses a severe challenge to the safe and economical operation of microgrids. How to formulate robust and economical dispatch strategies under uncertain environments has become a research hotspot in the field of optimized dispatching of industrial park microgrids. Summary of the Invention

[0003] The technical problem to be solved by this invention is to provide a KL divergence-based sub-Blu-ray robust optimization scheduling method for industrial park microgrids, which, while taking into account both economy and robustness, achieves coordinated and optimized scheduling of multiple types of flexible resources in industrial park microgrids, improves the system's adaptability to wind and solar uncertainties, and reduces overall operating costs.

[0004] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A KL divergence-based distributed bar optimization scheduling method for industrial park microgrids includes the following steps: Obtain the operating parameters of each flexible resource in the industrial park microgrid, and establish a deterministic model of the industrial park microgrid with the operating constraints of each flexible resource as the constraint condition and the goal of minimizing operating costs; Obtain historical samples of wind and solar power output, and construct a fuzzy set of wind and solar power output based on KL divergence; Based on the fuzzy set, the deterministic model of the industrial park microgrid is transformed into a two-stage sub-Brussels bar optimization model. The two-stage sub-Brussels bar optimization model uses the fuzzy set as the uncertainty set of scenario probabilities, the operating constraints as the constraint conditions, and aims to minimize the operating cost under the worst scenario distribution within the fuzzy set. The two-stage sub-Bruker optimization model is decoupled into a main problem for obtaining the lower bound of the original problem and a sub-problem for obtaining the upper bound of the original problem. The main problem and the sub-problem are solved iteratively until the difference between the upper and lower bounds satisfies the convergence condition, thus obtaining an optimized scheduling strategy. The optimized scheduling strategy is then used to schedule and control the microgrid in the industrial park.

[0005] Furthermore, the flexible resources include, but are not limited to, distributed photovoltaic, wind power, electric energy storage, thermal energy storage, demand response loads, heat pumps, gas boilers, and the interaction power between microgrids and distribution networks. The operational constraints include, but are not limited to, photovoltaic output constraints, wind power output constraints, energy storage operation constraints, demand response load constraints, power purchase and sale constraints, thermal energy storage constraints, gas boiler constraints, heat pump constraints, thermal balance constraints, and electrical energy balance constraints.

[0006] Furthermore, the operating costs include electricity purchase costs, electricity sales revenue, wind and solar curtailment penalty costs, and natural gas costs, expressed mathematically as follows:

[0007] In the formula: This represents the total number of scenes; This represents the total number of scheduling periods; For microgrid electricity purchase price; Microgrid electricity sales price; The unit penalty cost for curtailing wind and solar power; For natural gas prices; Power sold by the microgrid; Power purchased for the microgrid; Representing a scene s Down t The output of wind power at all times; Representing a scene s Down t The output of photovoltaic power at all times; This refers to the flow rate of natural gas.

[0008] Furthermore, the fuzzy set is constrained by the KL divergence between the reference distribution and the actual distribution not exceeding a set threshold, as expressed mathematically in the following formula:

[0009] In the formula: G The actual probability distribution of contributions to wind and solar power. G 0 represents the reference distribution of wind and solar power output. Reference distribution representing wind and solar power output G 0 and the actual distribution G KL distance, This is the distance threshold.

[0010] Furthermore, the mathematical expression of the two-stage sub-Bruker optimization model is as follows:

[0011] In the formula: The binary variables for the first-stage optimization model include the energy storage charging / discharging state variables and the electricity purchase / sale state variables from the deterministic model of the industrial park microgrid. The binary variables for the second-stage optimization model include the binary variables for air conditioner switching in the deterministic model of the industrial park microgrid. For all continuous variables in the decision variables of the deterministic model of the industrial park microgrid, For the scene s The probability, D For fuzzy sets, , , , , Let be a constant coefficient matrix, where It consists of the cost terms in the objective function of the deterministic model of the industrial park microgrid. It consists of constraint coefficients corresponding to the binary variables in the binary variables of the first-stage optimization model, which are part of the constraint conditions of the deterministic model of the industrial park microgrid. It consists of constraint coefficients corresponding to continuous variables in the constraint conditions of the deterministic model of the industrial park microgrid. It consists of constraint coefficients corresponding to the binary variables in the second-stage optimization model, which are part of the constraints in the deterministic model of the industrial park microgrid. It consists of terms in the constraints of the deterministic model of the industrial park microgrid that do not contain decision variables.

[0012] Furthermore, when the two-stage sub-Bruker optimization model is decoupled into a main problem for obtaining the lower bound of the original problem and a sub-problem for obtaining the upper bound of the original problem, the mathematical expression of the main problem is as follows:

[0013]

[0014] In the formula: As an auxiliary variable; For the first k The continuous variables in the next loop; For the first k The worst-case scenario distribution probability obtained in the next iteration; For the first k The binary variables of the second-stage optimization model in the next iteration; Index for the number of iterations; This represents the maximum number of loops. S This represents the total number of scenes.

[0015] Furthermore, when the two-stage sub-Bruker optimization model is decoupled into a main problem for obtaining the lower bound of the original problem and a sub-problem for obtaining the upper bound of the original problem, the mathematical expression of the sub-problem is as follows:

[0016] In the formula: These are the binary variables already obtained in the first stage; S This represents the total number of scenes.

[0017] Furthermore, before iteratively solving the main problem and subproblems, the method further includes: transforming the subproblems into inner subproblems and outer subproblems, wherein the mathematical expression of the inner subproblems is as follows:

[0018] The mathematical expression for the outer subproblem is as follows:

[0019] In the formula: inner sub-problem scenario s The objective function is to find the minimum value.

[0020] Furthermore, the iterative solution of the main problem and subproblems includes: Set the initial values ​​for the lower bound and the upper bound of the original problem; Solve the main problem to obtain the optimal solution of the binary variables of the first-stage optimization model and the optimal value of the objective function of the main problem, and update the lower bound of the original problem to the optimal value of the objective function. Substitute the optimal solution of binary variables of the first-stage optimization model into the inner sub-problem, solve the inner sub-problem to obtain the optimal solution of continuous variables under each scenario, the optimal solution of binary variables of the second-stage optimization model and the minimum value of the objective function of the corresponding inner sub-problem, substitute the minimum value of the objective function of the inner sub-problem into the outer sub-problem, solve the outer sub-problem to obtain the worst scenario distribution probability and the corresponding objective function value of the outer sub-problem, and update the upper bound of the original problem to the objective function value of the outer sub-problem corresponding to the worst scenario distribution probability; If the difference between the upper and lower bounds of the original problem satisfies the convergence condition, then the optimal solution of the binary variables of the first-stage optimization model, the optimal solution of the binary variables of the second-stage optimization model, and the optimal solution of the continuous variables are output as the optimization scheduling strategy. Otherwise, the worst-case scenario distribution probability is returned to the main problem, the main problem is solved again, and the next iteration is carried out.

[0021] Furthermore, it also includes the step of adjusting the distance threshold, which specifically involves obtaining the actual and predicted values ​​of wind and solar power output in the previous scheduling cycle, calculating the prediction deviation, and then adjusting the distance threshold to the corresponding threshold level based on the error range in which the prediction deviation is located.

[0022] Compared with the prior art, the advantages of the present invention are as follows: 1. To address the challenges posed by the uncertainty of wind and solar power output in existing technologies to the safe and economical operation of microgrids, this invention acquires historical samples of wind and solar power output and constructs fuzzy sets of wind and solar power output based on KL divergence. By establishing a measurable probability distance between the reference distribution and the actual distribution through KL divergence, it avoids strong assumptions about the precise probability distribution, thereby making full use of the distribution information in the limited historical sample data, improving the accuracy of characterizing the uncertainty of wind and solar power output and the flexibility of modeling, and providing a data foundation for robust optimization scheduling of industrial park microgrids under uncertain environments.

[0023] 2. This invention establishes a deterministic model for industrial park microgrids, with operational constraints of various flexible resources as constraints and minimizing operating costs as the objective. It then transforms this model into a two-stage sub-Bruker optimization model based on fuzzy sets constructed using KL divergence. The fuzzy sets serve as the uncertainty set of scenario probabilities, with operational constraints as the constraints, and the objective is to minimize the operating costs under the worst-case scenario distribution within the fuzzy sets. This collaboratively optimizes the discrete decision variables of the first stage and the continuous decision variables of the second stage, achieving unified scheduling and efficient coordination of multiple types of flexible resources within the industrial park microgrid. While balancing economy and robustness, it effectively reduces overall operating costs and enhances the system's adaptability to wind and solar uncertainties.

[0024] 3. This invention decouples the two-stage bibliometric optimization model into a main problem for obtaining the lower bound of the original problem and a subproblem for obtaining the upper bound of the original problem. The main problem and the subproblem are solved iteratively until the difference between the upper and lower bounds satisfies the convergence condition. Thus, the optimal solution is efficiently converged through the upper and lower bound approximation mechanism, reducing the complexity of the model solution, quickly obtaining the optimized scheduling strategy and using it for scheduling and control of industrial park microgrids. This enables efficient and robust optimized scheduling of microgrids under uncertain environments, improving the economic efficiency of system operation and the ability to absorb renewable energy. Attached Figure Description

[0025] Figure 1 This is a flowchart of a method according to an embodiment of the present invention.

[0026] Figure 2 This is a diagram of the microgrid structure in the industrial park.

[0027] Figure 3 To predict electrical load.

[0028] Figure 4 To predict outdoor temperature.

[0029] Figure 5 The results of microgrid power purchase and sale and energy storage operation.

[0030] Figure 6This refers to the output of photovoltaic and wind power.

[0031] Figure 7 For different distance thresholds ρ The impact. Detailed Implementation

[0032] The present invention will be further described below with reference to the accompanying drawings and specific preferred embodiments, but this does not limit the scope of protection of the present invention.

[0033] This embodiment proposes a KL divergence-based sub-Blule rod optimization scheduling method for industrial park microgrids, such as... Figure 1 As shown, the method includes the following steps: S1: Obtain the operating parameters of each flexible resource in the industrial park microgrid, and establish a deterministic model of the industrial park microgrid with the operating constraints of each flexible resource as the constraint condition and the goal of minimizing operating costs; S2: Obtain historical samples of wind and solar power output, and construct a fuzzy set of wind and solar power output based on KL divergence; S3: Based on the fuzzy set, the deterministic model of the industrial park microgrid is transformed into a two-stage sub-Bruker optimization model. The two-stage sub-Bruker optimization model uses the fuzzy set as the uncertainty set of scenario probabilities, the operating constraints as the constraint conditions, and aims to minimize the operating cost under the worst scenario distribution within the fuzzy set. S4: Decouple the two-stage sub-Bruker optimization model into a main problem for obtaining the lower bound of the original problem and a sub-problem for obtaining the upper bound of the original problem; S5: The C&CG algorithm is used to iteratively solve the main problem and sub-problems until the difference between the upper and lower bounds meets the convergence condition, thereby obtaining an optimized scheduling strategy. The optimized scheduling strategy is then used to schedule and control the microgrid in the industrial park.

[0034] Through the above steps, the method in this embodiment constructs a fuzzy set of wind and solar power output using KL divergence, reasonably measures the difference between the actual distribution and the reference distribution under limited historical sample data, and achieves fine modeling of uncertainty. At the same time, it constructs a two-stage sub-Bruker optimization model and uses the C&CG algorithm for efficient solution, thereby achieving coordinated and optimized scheduling of multiple types of flexible resources in industrial park microgrids while taking into account both economy and robustness, improving the system's adaptability to wind and solar uncertainties, and reducing overall operating costs.

[0035] The following provides a detailed explanation of each step.

[0036] Step S1 of this embodiment establishes microgrid flexibility resource operation constraints and a cost minimization objective function based on the operating principles and processes of each device, thereby constructing a deterministic model of the industrial park microgrid. The specific implementation process is as follows: First, establish the constraints. The flexibility resources in this embodiment include, but are not limited to, distributed photovoltaic power, wind power generation, electrical energy storage, thermal energy storage, demand response loads, heat pumps, gas boilers, and the interaction power between the microgrid and the distribution network. Therefore, the operational constraints include: (1) Photovoltaic output constraint: During the microgrid dispatching process, the actual output of the photovoltaic array should not exceed its maximum available power, while allowing the system to perform curtailment operation when necessary.

[0037] (1) (2) In the formula: Representing a scene s Down t The output of photovoltaic power at all times; Representing a scene s Down t Predicted photovoltaic power at any given time; Representing a scene s Down t The amount of curtailed solar power at any given time.

[0038] (2) Wind power output constraint: During the microgrid dispatching process, the actual active power output of the wind turbine is limited by the maximum available power at the current moment. At the same time, the system can actively curtail wind power to maintain power balance according to the operation needs.

[0039] (3) (4) In the formula: Representing a scene s Down t The output of wind power at all times; Representing a scene s Down t Forecast wind power output at any given time; Representing a scene s Down t The amount of wind power curtailed at any given time.

[0040] (3) Energy storage operation constraints: The energy storage system has the ability to flow power in both directions and shift energy over time, and can charge during off-peak hours and discharge during peak hours, thereby reducing the system operating cost.

[0041] (5) (6) (7) (8) (9) (10) In the formula: , Representing the scene respectively s Down t time, t -1 is the state of charge of the energy storage battery; Representing a scene s Down t The charging power of the energy storage battery at all times; Representing a scene s Down t The discharge power of the energy storage battery at all times; This indicates the battery's charge and discharge efficiency; Indicates a time interval; , Indicates the minimum and maximum states of charge of the energy storage battery; , Representing a scene s Down t The charging and discharging state variables of the energy storage battery at all times; , This indicates the maximum charging and discharging power of the energy storage battery; Indicates the initial state of charge of the battery; Representing a scenario s Down t = The state of charge of the energy storage battery at time T. T The scheduling time period.

[0042] (4) Demand response load ① Demand response load is divided into transferable load and interruptible load. The transferable load model is represented as follows: (11) (12) (13) (14) (15) (16) (17) (18) In the formula: Representing a scene s Down t Load following immediate demand response; Representing a scene s Down tLoad before immediate demand response; , It is a continuous auxiliary variable; Indicates the demand response coefficient; Representing a scene s Down t The increased load due to the constant demand response; , Indicates the minimum and maximum proportion of demand response to load changes; This represents the maximum value of the demand response coefficient.

[0043] ② Interruptible loads are mainly air conditioners, which can be interrupted according to electricity prices, as shown below: (19) (20) In the formula: , Representing the scene respectively s Down t time, t The indoor temperature at time +1; R , C These represent indoor thermal resistance and heat capacity, respectively. A state variable indicating whether the air conditioner is on or off; This refers to the air conditioner's power rating. Indicates the outdoor temperature; , Indicates the minimum and maximum indoor temperatures.

[0044] (5) Power purchase and sale constraints: The interaction power between the microgrid and the distribution network is expressed as follows: (twenty one) In the formula: This refers to the interaction power between the microgrid and the distribution network. Power sold by the microgrid; Power purchased for the microgrid.

[0045] In this context, microgrid power purchase and sale cannot be performed simultaneously; that is, only one operation, such as power purchase or sale, can be performed at any given time, as shown below: (twenty two) (twenty three) In the formula: Represents the state variables of electricity sales; Represents the electricity purchase status variable; This indicates the maximum power output of the microgrid. This indicates the maximum power purchased by the microgrid.

[0046] (6) Energy balance constraint: At any time, the output power of all power sources in the system, the net discharge power of energy storage devices, the exchange power with the distribution network, and the load demand are kept in real time.

[0047] (twenty four) In the formula: This refers to the number of air-conditioned rooms. The heat pump consumes electrical power.

[0048] (7) Thermal energy storage (25) (26) (27) (28) (29) (30) In the formula: , Representing a scene s Down t time, t The thermal energy level of thermal energy storage at time -1; , Indicates the charging and releasing power of thermal energy storage; Indicates the heat dissipation efficiency; , Indicates the minimum and maximum thermal energy storage levels; , Represents the state variables of thermal energy storage charging and releasing heat; , This indicates the maximum power of thermal energy storage for charging and releasing heat; Indicates thermal energy storage t = T The thermal energy level at that time; This indicates the initial thermal energy level of the thermal storage.

[0049] (8) Gas-fired boiler: Gas-fired boilers convert the chemical energy of natural gas into thermal energy.

[0050] (31) (32) In the formula: The thermal power of the boiler; For boiler efficiency; The calorific value of natural gas; Natural gas flow rate; This represents the boiler's maximum heat output.

[0051] (9) Heat pump: A heat pump converts electrical energy into heat energy.

[0052] (33) In the formula: The heat output power of the heat pump; For heat pump efficiency; The heat pump consumes electrical power.

[0053] (10) Thermal equilibrium (34) In the formula: This is the heat load of the microgrid.

[0054] Next, the objective function is established. The microgrid operation objective is to minimize the microgrid's operating costs. In this embodiment, the operating costs include electricity purchase costs, electricity sales revenue, wind and solar curtailment penalties, and natural gas costs. The mathematical expression of the objective function is as follows: (35) In the formula: This represents the total number of scenes; This represents the total number of scheduling periods; For microgrid electricity purchase price; Microgrid electricity sales price; The unit penalty cost for curtailing wind and solar power; For natural gas prices; Power sold by the microgrid; Power purchased for the microgrid; Representing a scene s Down t The output of wind power at all times; Representing a scene s Down t The output of photovoltaic power at all times; This refers to the flow rate of natural gas.

[0055] Step S2 in this embodiment, based on the deterministic model constructed in S1, considers the uncertainty of wind and solar power output and establishes a fuzzy set based on KL divergence. The specific implementation process is as follows: ① Construct a reference distribution G 0 Assuming selection N Using historical data from the past 24 days as samples, the features are classified as follows: R If there are intervals, then the samples for each corresponding interval are... N 1, N 2, , N R Each interval represents the expected value of a sample within that interval, and its corresponding probability can be expressed as: (36) In the formula: For reference distribution number r The sample probability of the interval.

[0056] The reference distribution is (37) In the formula: This is the reference distribution.

[0057] ②Establish fuzzy sets Based on the constructed reference distribution, KL divergence is used to characterize the reference distribution. G 0 and the actual distribution G The smaller the distance, the more similar the two distributions are. Using discrete probability distributions, the KL distance formula is: (38) In the formula: Represents the reference distribution G 0 and the actual distribution G KL distance; Reference distribution range r The probability of.

[0058] The KL distance between two probability distributions should be within a certain threshold range. Therefore, fuzzy sets D With the constraint that the KL divergence between the reference distribution and the actual distribution is no greater than a set threshold, the mathematical expression is as follows: (39) In the formula: G The actual probability distribution of contributions to wind and solar power. G 0 represents the reference distribution of wind and solar power output. Reference distribution representing wind and solar power output G 0 and the actual distribution G KL distance, This is the distance threshold.

[0059] In this embodiment, step S3 establishes a two-stage sub-Bruker optimization model for the industrial park microgrid based on the model and fuzzy set constructed in steps S1-S2. Specifically, based on the model and fuzzy set constructed in steps S1-S2, the two-stage sub-Bruker optimization model is represented in the following compact form: (40) In the formula: The binary variables (i.e., 0-1 variables) of the first-stage optimization model include the energy storage charging and discharging state variables in the decision variables of the deterministic model of the industrial park microgrid. , ) and power purchase and sale state variables ( , ), The binary variables for the second-stage optimization model include the binary variables of air conditioner switches in the decision variables of the deterministic model of the industrial park microgrid. ), For all continuous variables in the decision variables of the deterministic model of the industrial park microgrid ( , , , , , , , , , , , , , , , , , wait), For the scene s The probability, D For fuzzy sets, , , , , Let be a constant coefficient matrix, where The cost items in the objective function of the deterministic model of the industrial park microgrid ( , , , )constitute, It consists of constraint coefficients corresponding to the binary variables in the binary variables of the first-stage optimization model, which are part of the constraint conditions of the deterministic model of the industrial park microgrid. It consists of constraint coefficients corresponding to continuous variables in the constraint conditions of the deterministic model of the industrial park microgrid. It consists of constraint coefficients corresponding to the binary variables in the second-stage optimization model, which are part of the constraints in the deterministic model of the industrial park microgrid. It consists of terms in the constraints of the deterministic model of the industrial park microgrid that do not contain decision variables (i.e., constants, known parameters, scenario-given parameters, etc. in the constraints).

[0060] In this embodiment, step S4, based on the established two-stage sub-Bruker optimization model of the industrial park microgrid, uses the C&CG algorithm to decouple the main problem and sub-problems. The specific implementation process is as follows: The two-stage sub-Bruker optimization model is difficult to solve directly, but it can be solved iteratively using the C&CG algorithm. Based on the C&CG algorithm, the model is decomposed into the following main problem for obtaining the lower bound of the original problem and sub-problems for obtaining the upper bound of the original problem: (1) Main problem (41) (42) In the formula: As an auxiliary variable; For the first k The continuous variables in the next loop; For the first k The worst-case scenario distribution probability obtained in the next iteration; For the first k The second stage of the iteration optimizes the 0-1 variables of the model; Index for the number of iterations; This represents the maximum number of loops. S This represents the total number of scenes.

[0061] The goal of the main problem is to obtain the lower bound of the original problem. The goal of the subproblem is to obtain the upper bound of the original problem. The model converges when the following conditions are met.

[0062] (43) In the formula: It is a very small positive number.

[0063] (2) Subproblems The subproblem is represented as follows: (44) In the formula: These are the binary variables already obtained in the first stage. Since the above subproblem is a max-min problem, solving it requires transforming it into two subproblems: an inner subproblem and an outer subproblem.

[0064] ①Inner layer subproblems: (45) In the formula: For inner layer problem scenarios s The objective function is to find the minimum value.

[0065] ②Outer Sub-problems: (46) The process of solving the subproblems is as follows: first solve the inner subproblems, and then obtain the optimal value. Substituting the solution into the outer subproblem yields the upper bound of the original problem. .

[0066] Step S5 of this embodiment solves the original problem based on the decoupled main problem and sub-problems, according to the designed C&CG algorithm flow. The specific steps for solving the problem include: Step 1: Set the initial values ​​for the lower bound and upper bound of the original problem. In this embodiment, the upper bound is set. The lower realm Number of outer iterations k =0; Step 2: Solve the main problem to obtain the optimal solution for the binary variables of the first-stage optimization model. The optimal value of the objective function of the main problem is obtained, and the lower bound of the original problem is updated to the optimal value of the objective function. Step 3: Find the optimal solution of the binary variables in the first-stage optimization model. Substituting the inner subproblems, we solve them to obtain the optimal solutions for continuous variables in each scenario, the optimal solutions for binary variables in the second-stage optimization model, and the minimum values ​​of the objective functions of the corresponding inner subproblems. ; Step 4: Minimize the objective function of the inner subproblems Substituting the outer subproblem, we obtain the worst-case scenario probability distribution by solving the outer subproblem. The upper bound of the original problem is updated to the worst-case scenario distribution probability by comparing the objective function value of the corresponding outer subproblem. The corresponding objective function value of the outer subproblem; Step 5: If the difference between the upper and lower bounds of the original problem satisfies the convergence condition. If the first-stage optimization model's binary variable optimal solution, the second-stage optimization model's binary variable optimal solution, and the continuous variable optimal solution are output as the optimization scheduling strategy, the iteration ends; otherwise, the worst-case scenario distribution probability is... Return to the main question and let k = k +1, proceed to step 2 to solve the main problem again, and implement the next iteration.

[0067] The effectiveness of the method proposed in this embodiment will be tested through specific experiments below.

[0068] Figure 2This is a schematic diagram of the industrial park microgrid structure used for testing. The microgrid is connected to the upper-level distribution network through a grid connection point, enabling bidirectional power exchange. Internally, the microgrid integrates a photovoltaic (PV) power generation system, a wind power generation system, an electric energy storage system, a heat pump, a gas-fired boiler, a thermal energy storage device, and demand response loads. PV and wind power are connected to the AC bus as intermittent renewable energy sources; the electric energy storage system achieves time-shifting of electrical energy through charging and discharging; the heat pump and gas-fired boiler, as thermoelectric coupling devices, convert electrical energy and natural gas into heat energy respectively, and work in conjunction with the thermal energy storage to meet heat load demands; demand response loads include transferable loads and interruptible loads (air conditioning), which can adjust electricity consumption behavior according to electricity price signals or dispatch instructions. This structure fully demonstrates the multi-energy flow coupling and source-load-storage coordinated interaction characteristics of industrial park microgrids, providing a typical application scenario for verifying the applicability of the proposed method.

[0069] Table 1 presents the main technical parameters of the industrial park microgrid used in the test. Among them, the demand response load change ratio parameter... All parameters are set to 0.3, indicating that the transferable load can fluctuate by 30% above or below the original load. The indoor heat capacity C is set to 0.525 kW·h·℃, and the indoor thermal resistance R is set to 18 ℃·kW, reflecting the typical thermodynamic characteristics of the buildings in this industrial park. The maximum charging and discharging power of the electric energy storage is 20 kW, and the initial state of charge is 0.2 pu. The power limit for interaction between the microgrid and the distribution network is 20 kW. The indoor comfort temperature range is set to 24 ℃ to 26 ℃. These parameters provide the equipment capacity and physical boundary conditions for subsequent optimized scheduling.

[0070] Table 1 Microgrid parameters

[0071] Figure 3 This is a predicted electrical load curve for a test scenario. (From...) Figure 3 It is evident that the industrial park's electrical load exhibits a distinct bi-peak characteristic: the load is at its lowest point from early morning to dawn (0:00-6:00 AM), at approximately 150 kW; subsequently, the load gradually climbs, reaching its daytime peak (approximately 300 kW) around 1:00 PM; it slightly decreases in the afternoon, and remains above 200 kW from evening to night. This load curve reflects the industrial park's production and operating patterns, providing fundamental data for the optimized allocation of resources in subsequent demand response.

[0072] Figure 4 This is a predicted outdoor temperature curve for the test scenario. (Source: [Insert Source Here]) Figure 3 As can be seen, the outdoor temperature reaches its lowest point (around 30°C) in the early morning (around 5 AM), then gradually rises, reaching its highest point (around 38°C) between 3 PM and 4 PM, before slowly decreasing. The daily variation in outdoor temperature directly affects the cooling demand of interruptible loads such as air conditioners, thus influencing the electrical load curve. Combined with... Figure 3and Figure 4 It can be seen that during the afternoon high-temperature period (13:00-16:00), the outdoor temperature is close to or exceeds the upper limit of the indoor comfortable temperature (26 ℃), and the air conditioning cooling load increases significantly, which highly overlaps with the peak period of electricity load, exacerbating the power supply pressure during this period. This also verifies the necessity of alleviating peak load through demand response and energy storage coordination optimization.

[0073] Figure 5 This is the result of microgrid power purchase and sale and energy storage operation. Figure 6 This refers to the output of solar and wind power. Figure 6 As can be seen, photovoltaic (PV) power output exhibits a significant daytime-to-night characteristic, peaking between 12:00 and 14:00 (approximately 90 kW) and decreasing to zero at night. Wind power output is relatively stable, maintaining a range of 160-280 kW throughout the day, but with random fluctuations. The combined effect of these two factors results in drastic intraday variations in the microgrid's net renewable energy output. A power surplus may occur during the peak PV generation period at midday, while insufficient wind and solar output at night necessitates reliance on the external grid or energy storage for power supply. This highlights the practical need for the proposed method in this embodiment to address the uncertainties of wind and solar power generation.

[0074] During microgrid operation, when the output of photovoltaic and wind turbines is insufficient to meet local load demand, the system will purchase electricity from the upstream grid to make up the shortfall. t During the 17-22 hour period, due to a surplus in power generation, the microgrid can sell electricity back to the upstream grid. The power variation curve of the energy storage system shows two distinct charging and discharging processes within this dispatch cycle. This is mainly guided by the electricity pricing mechanism; the energy storage system charges when electricity prices are low and discharges when prices are high. This allows the microgrid to respond more flexibly to changes in external grid prices, thereby improving its economic dispatch performance and operational efficiency.

[0075] Depend on Figure 5 Furthermore, it can be seen that during periods of low electricity prices, such as t=6~8 and t=11~14, the energy storage system primarily charges, and its energy charge level gradually increases. During periods of high electricity prices, such as t=17~22, the energy storage system switches to discharging, its energy charge level decreases, and the microgrid sells electricity to the upstream grid. These results verify that the method proposed in this embodiment can fully utilize the energy time-shifting characteristics of the energy storage system, storing energy during off-peak hours and releasing it during peak hours, thus achieving peak shaving and valley filling and reducing electricity purchase costs.

[0076] according to Figure 7 It can be seen that, with the distance parameter ρ As the size of the network increases, the overall operating cost of microgrids shows an upward trend. ρWhen ρ = 0, the cost is lowest, and the model is a deterministic model with the lowest conservatism in wind and solar prediction errors. As ρ gradually increases, the cost rises. A larger distance parameter results in more conservative outcomes and a stronger ability to handle uncertainties, but at the cost of lower economic efficiency. Therefore, appropriately increasing the distance parameter can enhance the model's ability to cope with wind and solar uncertainties, but the parameter setting should not be too large, otherwise it will lead to overly conservative scheduling results and decreased economic efficiency. In practical applications, a balance should be struck between robustness and economy to select an appropriate ρ value to obtain a better scheduling scheme.

[0077] Depend on Figure 7 Further quantification reveals that when ρ increases from 0 to 0.02, the operating cost rises from approximately 530 yuan to approximately 550 yuan, an increase of about 3.8%. This indicates that by reasonably selecting the ρ value, the system's robustness to wind and solar uncertainties can be effectively improved without significantly sacrificing economic efficiency. In this embodiment, to take into account the dynamic changes in prediction accuracy under different operating conditions, a rolling feedback mechanism is also used to adaptively adjust the distance threshold ρ. Specifically, the actual and predicted values ​​of wind and solar power output in the previous scheduling cycle are obtained, the prediction deviation is calculated (e.g., the root mean square error of prediction is calculated), and then the distance threshold is adjusted to the corresponding threshold level based on the error range of the prediction deviation. A larger root mean square error corresponds to a higher distance threshold level to enhance model conservatism; a smaller root mean square error corresponds to a lower distance threshold level to improve economic efficiency. For example, when the root mean square error of the prediction is in the interval [0, 0.05], set ρ=0.005; when it is in the interval (0.05, 0.1], set ρ=0.01; when it is in the interval (0.1, 0.15], set ρ=0.015.

[0078] Through the above dynamic adjustments, a continuous optimization trade-off between robustness and economy can be achieved between scheduling cycles, thereby obtaining a better scheduling scheme.

[0079] In summary, this invention proposes a robust optimization scheduling method for industrial park microgrids based on KL divergence. Based on the operating principles and processes of each device, a deterministic model of the industrial park microgrid is constructed by establishing operational constraints on microgrid flexibility resources and a cost minimization objective function. Based on this deterministic model, considering the uncertainty of wind and solar power output, a fuzzy set based on KL divergence is established. According to the constructed model and fuzzy set, a two-stage robust optimization model for the industrial park microgrid is established. Based on the established two-stage robust optimization model, the C&CG algorithm is used to decouple the model into a main problem and sub-problems. Based on the decoupled main problem and sub-problems, the original problem is solved according to the designed C&CG algorithm flow. This method can efficiently construct fuzzy sets for wind and solar power output by fully utilizing limited historical sample data, achieving robust optimization scheduling of industrial park microgrids under uncertain environments, improving the system's operational economy and renewable energy absorption capacity, and balancing the robustness and economy of the scheduling scheme.

[0080] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-readable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1The functions specified in one or more boxes. These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable apparatus for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0081] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A KL divergence-based distributed bar optimization scheduling method for industrial park microgrids, characterized in that, Includes the following steps: Obtain the operating parameters of each flexible resource in the industrial park microgrid, and establish a deterministic model of the industrial park microgrid with the operating constraints of each flexible resource as the constraint condition and the goal of minimizing operating costs; Obtain historical samples of wind and solar power output, and construct a fuzzy set of wind and solar power output based on KL divergence; Based on the fuzzy set, the deterministic model of the industrial park microgrid is transformed into a two-stage sub-Brussels bar optimization model. The two-stage sub-Brussels bar optimization model uses the fuzzy set as the uncertainty set of scenario probabilities, the operating constraints as the constraint conditions, and aims to minimize the operating cost under the worst scenario distribution within the fuzzy set. The two-stage sub-Bruker optimization model is decoupled into a main problem for obtaining the lower bound of the original problem and a sub-problem for obtaining the upper bound of the original problem. The main problem and the sub-problem are solved iteratively until the difference between the upper and lower bounds satisfies the convergence condition, thus obtaining an optimized scheduling strategy. The optimized scheduling strategy is then used to schedule and control the microgrid in the industrial park.

2. The KL divergence-based sub-Bluerbar optimal scheduling method for industrial park microgrids according to claim 1, characterized in that, The flexible resources include, but are not limited to, distributed photovoltaic, wind power, electric energy storage, thermal energy storage, demand response loads, heat pumps, gas boilers, and the interaction power between microgrids and distribution networks. The operational constraints include, but are not limited to, photovoltaic output constraints, wind power output constraints, energy storage operation constraints, demand response load constraints, power purchase and sale constraints, thermal energy storage constraints, gas boiler constraints, heat pump constraints, thermal balance constraints, and electrical energy balance constraints.

3. The KL divergence-based distributed bar optimal scheduling method for industrial park microgrids according to claim 1, characterized in that, The operating costs include electricity purchase costs, electricity sales revenue, wind and solar curtailment penalty costs, and natural gas costs, expressed mathematically as follows: In the formula: This represents the total number of scenes; This represents the total number of scheduling periods; For microgrid electricity purchase price; Microgrid electricity sales price; The unit penalty cost for curtailing wind and solar power; For natural gas prices; Power sold by the microgrid; Power purchased for the microgrid; Representing a scene s Down t The output of wind power at all times; Representing a scene s Down t The output of photovoltaic power at all times; This refers to the flow rate of natural gas.

4. The KL divergence-based sub-Bluerbar optimal scheduling method for industrial park microgrids according to claim 1, characterized in that, The fuzzy set is constrained by the KL divergence between the reference distribution and the actual distribution not being greater than a set threshold, and the mathematical expression is as follows: In the formula: G The actual probability distribution of contributions to wind and solar power. G 0 represents the reference distribution of wind and solar power output. Reference distribution representing wind and solar power output G 0 and the actual distribution G KL distance, This is the distance threshold.

5. The KL divergence-based distributed bar optimal scheduling method for industrial park microgrids according to claim 1, characterized in that, The mathematical expression of the two-stage sub-Bruker optimization model is as follows: In the formula: The binary variables for the first-stage optimization model include the energy storage charging / discharging state variables and the electricity purchase / sale state variables from the deterministic model of the industrial park microgrid. The binary variables for the second-stage optimization model include the binary variables for air conditioner switching in the deterministic model of the industrial park microgrid. For all continuous variables in the decision variables of the deterministic model of the industrial park microgrid, For the scene s The probability, D For fuzzy sets, , , , , Let be a constant coefficient matrix, where It consists of the cost terms in the objective function of the deterministic model of the industrial park microgrid. It consists of constraint coefficients corresponding to the binary variables in the binary variables of the first-stage optimization model, which are part of the constraint conditions of the deterministic model of the industrial park microgrid. It consists of constraint coefficients corresponding to continuous variables in the constraint conditions of the deterministic model of the industrial park microgrid. It consists of constraint coefficients corresponding to the binary variables in the second-stage optimization model, which are part of the constraints in the deterministic model of the industrial park microgrid. It consists of terms in the constraints of the deterministic model of the industrial park microgrid that do not contain decision variables.

6. The KL divergence-based distributed bar optimal scheduling method for industrial park microgrids according to claim 5, characterized in that, When the two-stage sub-Bruker optimization model is decoupled into a main problem for obtaining the lower bound of the original problem and a sub-problem for obtaining the upper bound of the original problem, the mathematical expression of the main problem is as follows: In the formula: As an auxiliary variable; For the first k The continuous variables in the next loop; For the first k The worst-case scenario distribution probability obtained in the next iteration; For the first k The binary variables of the second-stage optimization model in the next iteration; Index for the number of iterations; This represents the maximum number of loops. S This represents the total number of scenes.

7. The KL divergence-based sub-Bluerbar optimal scheduling method for industrial park microgrids according to claim 5, characterized in that, When the two-stage sub-Bruker optimization model is decoupled into a main problem for obtaining the lower bound of the original problem and a sub-problem for obtaining the upper bound of the original problem, the mathematical expression of the sub-problem is as follows: In the formula: These are the binary variables already obtained in the first stage; S This represents the total number of scenes.

8. The KL divergence-based sub-Brow bar optimization scheduling method for industrial park microgrids according to claim 7, characterized in that, Before iteratively solving the main problem and subproblems, the method further includes: transforming the subproblems into inner subproblems and outer subproblems, wherein the mathematical expression of the inner subproblems is as follows: The mathematical expression for the outer subproblem is as follows: In the formula: inner sub-problem scenario s The objective function is to find the minimum value.

9. The KL divergence-based sub-Bluerbar optimal scheduling method for industrial park microgrids according to claim 8, characterized in that, When iteratively solving the main problem and its subproblems, the following steps are included: Set the initial values ​​for the lower bound and the upper bound of the original problem; Solve the main problem to obtain the optimal solution of the binary variables of the first-stage optimization model and the optimal value of the objective function of the main problem, and update the lower bound of the original problem to the optimal value of the objective function. Substitute the optimal solution of binary variables of the first-stage optimization model into the inner sub-problem, solve the inner sub-problem to obtain the optimal solution of continuous variables under each scenario, the optimal solution of binary variables of the second-stage optimization model and the minimum value of the objective function of the corresponding inner sub-problem, substitute the minimum value of the objective function of the inner sub-problem into the outer sub-problem, solve the outer sub-problem to obtain the worst scenario distribution probability and the corresponding objective function value of the outer sub-problem, and update the upper bound of the original problem to the objective function value of the outer sub-problem corresponding to the worst scenario distribution probability; If the difference between the upper and lower bounds of the original problem satisfies the convergence condition, then the optimal solution of the binary variables of the first-stage optimization model, the optimal solution of the binary variables of the second-stage optimization model, and the optimal solution of the continuous variables are output as the optimization scheduling strategy. Otherwise, the worst-case scenario distribution probability is returned to the main problem, the main problem is solved again, and the next iteration is carried out.

10. The KL divergence-based sub-Bluerged bar optimal scheduling method for industrial park microgrids according to claim 4, characterized in that, It also includes the step of adjusting the distance threshold, which specifically involves obtaining the actual and predicted values ​​of wind and solar power output in the previous scheduling cycle, calculating the prediction deviation, and then adjusting the distance threshold to the corresponding threshold level based on the error range in which the prediction deviation is located.

Images