Multi-source coupled distribution robust optimization scheduling method and system

Abstract

The application discloses a kind of multi-source coupling distribution robust optimization scheduling method and system, comprising: obtaining wind light load typical scene set;Dual-layer optimization scheduling model is constructed with minimum net load variance in upper layer, minimum system operation cost in lower layer;Scene probability distribution fuzzy set is constructed based on comprehensive norm, which is converted into distribution robust optimization framework;Column and constraint generation algorithm is used to solve, and pumping storage unit and thermal power unit scheduling plan is obtained;The application deeply fuses time-space correlation scene generation and distribution robust optimization, the upper layer utilizes pumping storage to suppress fluctuation and reduces thermal power peak shaving pressure, the lower layer considers thermal power deep peak shaving cost and step carbon trading in section, while ensuring robustness, economic efficiency and low carbon are also considered.Simulation results show that the application considers the advantages of random optimization and robust optimization in net load variance, system total cost and other indicators, and provides an effective solution for power system scheduling under high proportion of new energy access.

Description

Technical Field

[0001] This invention relates to the field of power system optimization scheduling technology, specifically to a multi-source coupled distributed bar optimization scheduling method and system. Background Technology

[0002] With the large-scale grid connection of new energy sources, the random fluctuations in wind and solar power output, as well as the multi-timescale fluctuations in load demand, have brought severe challenges to the safe and stable operation of the power system. In order to smooth out fluctuations and improve system flexibility, the construction of a "wind-solar-thermal-storage multi-source coupled dispatch system" has become a key technical path to balance renewable energy consumption and system reliability. Among them, thermal power provides basic regulation as a traditional stable power source, while pumped storage power stations play a role in peak shaving, valley filling, and fluctuation smoothing, complementing wind and solar power generation.

[0003] In terms of optimization scheduling methods, existing technologies can be mainly divided into three categories. The first category is deterministic optimization, which cannot cope with the operational risks caused by parameter uncertainty. The second category is stochastic optimization (SO), which describes uncertainty by assuming a probability distribution, but it is highly dependent on the distribution assumption. When the actual distribution deviates from the assumption, the robustness of the scheduling scheme will decrease significantly. The third category is robust optimization (RO), which can cope with extreme scenarios, but the optimization results are too conservative and its applicability to non-extreme scenarios is weak. Distributedly Robust Optimization (DRO) constructs a fuzzy set of probability distributions and solves for the expected value under the worst probability distribution, balancing robustness and economy, and is an effective method for handling uncertain optimization problems.

[0004] However, most existing studies on partial-scale bar optimization are independent of scenario generation methods and lack deep integration with high-precision spatiotemporal correlation scenario generation methods, resulting in limited accuracy when characterizing uncertain parameters. Meanwhile, existing scheduling models often fail to fully explore the potential of pumped storage in upper-level stability optimization, and also fail to meticulously consider the staged peak-shaving costs of thermal power units (including basic peak-shaving, deep peak-shaving without oil injection, and deep peak-shaving with oil injection) and the impact of tiered carbon trading mechanisms on scheduling decisions in lower-level economic optimization.

[0005] Therefore, how to construct a distributed bar optimization scheduling method that can be deeply integrated with high-quality scene generation methods and can collaboratively seek optimization between upper-level stability optimization and lower-level economic optimization is a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide a multi-source coupled distributed bar optimization scheduling method and system.

[0007] To achieve the above objectives, the present invention provides the following technical solution: This application provides a multi-source coupled distributed bar optimization scheduling method, including the following steps: Step A1. Obtain the typical scene set of wind, light and lotus and the corresponding initial probability distribution; Step A2. Construct a two-layer optimization scheduling model that includes an upper-layer stability optimization model and a lower-layer economic optimization model; the upper-layer stability optimization model aims to minimize net load fluctuations and optimizes the output of pumped storage units and new energy sources; the lower-layer economic optimization model aims to minimize system operating costs and optimizes the output of thermal power units. Step A3. Based on the typical scenario set, construct a fuzzy set of scenario probability distribution using the comprehensive norm, and transform the two-layer optimization scheduling model into a distributed bar optimization framework; Step A4. Solve the bibliometric optimization framework to obtain the scheduling plans for pumped storage units and thermal power units.

[0008] Preferably, the objective function of the upper-level stability optimization model is to minimize the variance of the net load power, as shown in the formula: in, The net load power variance for Net load power at any given time This represents the average net load power. The total number of time periods in the scheduling cycle. for Load power at any given time , They are respectively Wind curtailment rate and solar curtailment rate at any given time , They are respectively Real-time wind and solar power output forecasts , They are respectively The power generation and pumping capacity of the pumped storage power station at all times.

[0009] Preferably, the constraints of the upper-level stability optimization model include: the wind and solar power output constraint formula is: in, , This refers to the actual grid-connected power of wind and solar power. , Contribute to the maximum forecast of wind power and photovoltaic power; The formula for constraining wind and solar power curtailment is: in, , The maximum allowable wind and solar curtailment rates by the system; The formula for the output constraint of a pumped storage power station is: in, , These are the lower and upper limits of pumping power. , These are the lower and upper limits of power generation capacity; a product of zero indicates that pumping water and generating electricity cannot be done simultaneously. The formula for constraining the number of start-ups and shutdowns of pumped storage power stations is as follows: in, , Let sgn( represent the start-up and shutdown status of the pumped storage power station at time t, respectively (1 indicates start-up / shutdown, 0 indicates no action). ) is a symbolic function. , These are the maximum allowed number of startups and the maximum allowed number of shutdowns, respectively. The formula for the reservoir capacity constraint of a pumped storage power station is: in, , The reservoir capacities at time t are the upper and lower reservoir capacities, respectively. , These are the water volume increase coefficient (pumping) and decrease coefficient (power generation) corresponding to a unit of electricity generated in the upper reservoir. , These are the water volume rise coefficient (for power generation) and fall coefficient (for pumping) corresponding to a unit of electricity generated in the lower reservoir. , These represent the upper and lower limits of the reservoir's capacity, respectively. , These represent the upper and lower limits of the reservoir's capacity, respectively.

[0010] Preferably, the objective function of the lower-level economic optimization model is to minimize the system operating cost, as shown in the formula: in, The total operating cost of the system, Costs associated with starting and stopping thermal power units. For peak shaving costs of thermal power units, For carbon trading costs, For the revenue from deep peak shaving subsidies; The formula for calculating the start-up and shutdown costs of thermal power units is as follows: in, This refers to the number of thermal power units. For the first The cost of a single start-up and shutdown of the unit. For the first Taiwanese unit The operating status at any given time (1 - running, 0 - stopped); The segmented calculation formula for the peak-shaving cost of the thermal power unit is as follows: The formula for coal consumption cost is: in, , , For the first The coal consumption coefficient of the Taiwanese unit This refers to the unit price of coal. For the first Taiwanese unit Efforts made at all times; The formula for lifetime attrition cost is: in, This is the lifespan loss coefficient. For the cost of purchasing the generating unit, ( () represents the number of rotor-induced cracking cycles, and the fitting formula is: The formula for oil injection cost is: in, Price per unit of oil For the first Taiwanese unit The amount of oil added at any given time.

[0011] Preferably, the carbon trading cost is calculated using a tiered carbon trading mechanism, and the relevant formula is as follows: The carbon emission quota formula is: in, For carbon emission quotas, For the first Carbon emission allowance coefficient for the Taiwanese unit; The formula for actual carbon emissions is: in, For carbon emissions, This is the carbon emission coefficient; The formula for the quota for participating in carbon trading is: The formula for the tiered carbon trading cost is: in, The base price for carbon trading. For price growth rate, The length of the carbon emission range; The formula for the revenue from the deep peak shaving subsidy is as follows: in, This represents the minimum load factor during the basic peak-shaving phase. For the first Rated capacity of the unit To subsidize the unit price.

[0012] Preferably, the constraints of the lower-level economic optimization model include: The system power balance constraint formula is: in, For the total output of thermal power units; The formula for constraining the output of a thermal power unit in sections is: in, , These are the state variables for the basic peak shaving and deep peak shaving stages, respectively. , For the first The maximum and minimum output of the thermal power unit; For the basic peak shaving phase Minimum output of a thermal power unit; The formula for the ramp constraint related to the operating state is: in, , These represent the maximum upward ramp rate during the basic peak shaving and deep peak shaving phases, respectively. , These represent the maximum downward ramp rate during the basic peak shaving and deep peak shaving phases, respectively. Minimum Start-up and Shutdown Time Constraints by State: Thermal power unit operating states are divided into basic peak shaving, deep peak shaving, and shutdown states. To avoid frequent transitions between different states, a minimum duration constraint is set for each state. This refers to the minimum continuous operating time during the basic peak-shaving phase. This is the minimum continuous operating time for the deep peak shaving phase. The minimum continuous downtime is expressed mathematically as follows: The system spin-off reserve constraint formula is: in, , These are the positive and negative spinning reserve coefficients for load forecasting error. , , , These are the positive and negative spinning reserve coefficients for wind power and solar power prediction errors, respectively. The formula for line transmission capacity constraint is: in, For the line Admittance, The voltage phase angle difference between the two nodes of the line. This represents the maximum transmission capacity of the line.

[0013] Preferably, the comprehensive norm in step A3 includes the 1-norm and the ∞-norm; the fuzzy set is constructed by constraining the 1-norm distance and ∞-norm distance between the probability distribution of each typical scenario and the initial probability distribution to not exceed the corresponding fluctuation threshold, as shown in the formula: in, Let be the feasible region of the scenario probability distribution. For the first The probability of each scenario. For the first The initial probability of each scenario. , These are the scene probability fluctuation thresholds under 1-norm and ∞-norm constraints, respectively. The fluctuation threshold is determined by the confidence level of the scene probability, and the formula is: in, , These represent the confidence levels for the 1-norm and the ∞-norm, respectively. This represents the total number of scenes before scene reduction. This represents the total number of typical scenarios.

[0014] Preferably, the sub-Bruker optimization framework is a two-layer, four-stage structure, and the matrix form of the sub-Bruker optimization model is as follows: in and These are the variables for the first stage (start / stop status, etc.). and For the second stage variables (power, etc.), Contribute to the initial prediction of wind-solar load; The formula for the sub-Bluer bar optimization model is: Where X and W are the feasible regions of the variables in the first stage. , Let be the feasible region of the variables in the second stage. Let be the feasible region of the scenario probability distribution. This is a typical scenario.

[0015] Preferably, step A4 uses a column and constraint generation algorithm to solve the problem, decomposing the sub-Bruker optimization framework into a main problem and sub-problems, as shown in the following formula: Upper-level main question MP1: Lower-level main question MP2:

[0016] Upper-level subproblem SP1:

[0017] Lower-level subproblem SP2:

[0018] in, For the number of iterations, , The auxiliary variables are used; the main problem solves the first-stage variables and updates the lower bound, while the subproblems solve the worst-case probability distribution and update the upper bound, iterating until (UB) is reached. LB)≤ convergence, For convergence accuracy.

[0019] A multi-source coupled distributed bar optimization scheduling system includes: The scene input module is used to input a typical scene set and initial probability distribution of wind, solar and load. The model building module is used to build a two-layer optimization scheduling model that includes upper-layer stability optimization and lower-layer economic optimization, and to configure the above-mentioned constraints. The fuzzy set construction module is used to construct fuzzy sets of scene probability distributions based on comprehensive norm and confidence constraints, and generate a fuzzy bar optimization framework. The solution module is used to solve the sub-Bruker optimization framework using a column and constraint generation algorithm and output a scheduling plan. The control module is used to control the operation of pumped storage units and thermal power units according to the scheduling plan.

[0020] Compared with the prior art, this application has the following beneficial effects: This invention proposes a multi-source coupled sub-Blu-robin optimization scheduling method and system. It constructs a two-layer nested optimization model with the upper layer aiming to minimize net load variance and the lower layer aiming to minimize system operating cost. Based on the comprehensive norms (1-norm and ∞-norm), it builds a fuzzy set of scenario probability distributions, deeply integrating high-quality spatiotemporally relevant scenario generation with sub-Blu-robin optimization. This effectively overcomes the shortcomings of traditional stochastic optimization, such as sensitivity to distribution assumptions and overly conservative robust optimization. At the upper layer, pumped storage units are used to smooth net load fluctuations, significantly reducing the peak-shaving pressure on thermal power units. At the lower layer, the coal consumption cost, lifespan loss cost, and oil injection cost of thermal power units are considered in fine segmentation for three states: basic peak-shaving, deep peak-shaving without oil injection, and deep peak-shaving with oil injection. A tiered carbon trading mechanism is introduced, enabling the scheduling scheme to achieve robustness while also being economical, stable, and low-carbon. Simulation results show that this invention combines the advantages of stochastic optimization and robust optimization in terms of net load variance and total system cost. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the double-layer nested optimization model of the present invention.

[0022] Figure 2 Flowchart for solving the CCG algorithm.

[0023] Figure 3 Diagram of the improved IEEE 39-node system architecture.

[0024] Figure 4 A bar chart comparing the optimization results with and without pumped storage power stations.

[0025] Figure 5 This is a comparison chart of net load curves under different pumped storage capacities.

[0026] Figure 6 A comparison curve of optimization results under different pumped storage capacities. Detailed Implementation

[0027] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0028] Furthermore, in this invention, an element referred to as fixed to or disposed on another element may be directly disposed on the other element, or there may be an intermediate element. When an element is considered to be connected to another element, it may be directly connected to the other element, or there may be an intermediate element present simultaneously. The terms vertical, horizontal, left, right, and similar expressions used herein are for illustrative purposes only and do not represent the only possible implementation.

[0029] See Figures 1-6 This application provides a multi-source coupled distributed bar optimization scheduling method, including the following steps: Step A1. Obtain the typical scene set of wind, light and lotus and the corresponding initial probability distribution; Specifically, the process involves obtaining a set of typical wind-solar-load (FSL) scenarios and their corresponding initial probability distributions. The typical scenario set is obtained using the aforementioned scenario generation method: based on historical FSL output data, a large number of initial scenarios are generated using Monte Carlo simulation. Subsequently, K-means clustering is used to reduce the initial scenarios, resulting in a set containing K typical scenarios and their corresponding initial probability distributions. Monte Carlo simulation generates multiple possible output scenarios by randomly sampling the probability distribution of FSL prediction errors. K-means clustering merges similar scenarios based on Euclidean distance, retaining representative typical scenarios and effectively reducing the complexity of subsequent optimization calculations. The initial probability of each scenario is typically set to be equal, or allocated proportionally to the sample size of each scenario before scenario reduction.

[0030] Step A2. Construct a two-layer optimization scheduling model that includes an upper-layer stability optimization model and a lower-layer economic optimization model; the upper-layer stability optimization model aims to minimize net load fluctuations and optimizes the output of pumped storage units and new energy sources; the lower-layer economic optimization model aims to minimize system operating costs and optimizes the output of thermal power units. Specifically: a two-layer optimization scheduling model is constructed; the upper-layer stability optimization model aims to minimize net load fluctuations, and the decision variables include wind and solar curtailment rates and pumped storage power and power generation of pumped storage units; the lower-layer economic optimization model aims to minimize system operating costs, and the decision variables are the start-up and shutdown status and output of thermal power units; the two layers are coupled through the net load curve: the optimized net load of the upper layer serves as the input for the lower-layer thermal power scheduling.

[0031] Step A3. Based on the typical scenario set, construct a fuzzy set of scenario probability distribution using the comprehensive norm, and transform the two-layer optimization scheduling model into a distributed bar optimization framework; Specifically, based on a set of typical scenarios, a fuzzy set of scenario probability distributions is constructed using a comprehensive norm. The comprehensive norm includes the 1-norm and the infinity norm, which respectively constrain the total deviation and the maximum deviation between the scenario probability and the initial probability to not exceed a given threshold. This fuzzy set transforms the deterministic scheduling model into a distributed bar optimization model, the goal of which is to find the optimal scheduling scheme under the worst probability distribution.

[0032] Step A4. Solve the bibliometric optimization framework to obtain the scheduling plans for pumped storage units and thermal power units.

[0033] Specifically: The column and constraint generation algorithm is used to solve the sub-Bruker optimization model; the algorithm decomposes the original problem into a main problem and sub-problems, and solves them iteratively: the main problem solves the first-stage decision variables under a given probability distribution, and the sub-problems solve the worst-case probability distribution under a given first-stage decision, and the iteration is repeated until convergence; finally, the scheduling plan of pumped storage units and thermal power units is obtained.

[0034] In a preferred embodiment, the objective function of the upper-level stability optimization model is to minimize the variance of the net load power, as shown in the formula:

[0035] in, The net load power variance for Net load power at any given time This represents the average net load power. The total number of time periods in the scheduling cycle. for Load power at any given time , They are respectively Wind curtailment rate and solar curtailment rate at any given time , They are respectively Real-time wind and solar power output forecasts , They are respectively The power generation and pumping capacity of the pumped storage power station at all times.

[0036] In this embodiment, the objective function of the upper-level stability optimization model is defined; the objective is to minimize the variance of net load power; net load power is defined as: the load power at time t minus the actual grid-connected power of wind power (predicted power multiplied by 1 minus the wind curtailment rate), minus the actual grid-connected power of photovoltaic power, minus the power generation of pumped storage power station, plus the pumping power of pumped storage power station; the net load variance is the sum of the squares of the differences between the net load at each time point and the average net load divided by the total number of time periods in the scheduling cycle; the smaller the variance, the flatter the net load curve, the more stable the output of thermal power units, which is conducive to reducing the frequent adjustment and deep peak shaving of thermal power.

[0037] In a preferred embodiment, the constraints of the upper-level stability optimization model include: the wind and solar power output constraint formula is:

[0038] in, , This refers to the actual grid-connected power of wind and solar power. , Contribute to the maximum forecast of wind power and photovoltaic power; The formula for constraining wind and solar power curtailment is:

[0039] in, , The maximum allowable wind and solar curtailment rates by the system; The formula for the output constraint of a pumped storage power station is:

[0040] in, , These are the lower and upper limits of pumping power. , These are the lower and upper limits of power generation capacity; a product of zero indicates that pumping water and generating electricity cannot be done simultaneously. The formula for constraining the number of start-ups and shutdowns of pumped storage power stations is as follows:

[0041] in, , Let sgn( represent the start-up and shutdown status of the pumped storage power station at time t, respectively (1 indicates start-up / shutdown, 0 indicates no action). ) is a symbolic function. , These are the maximum allowed number of startups and the maximum allowed number of shutdowns, respectively. The formula for the reservoir capacity constraint of a pumped storage power station is:

[0042] in, , The reservoir capacities at time t are the upper and lower reservoir capacities, respectively. , These are the water volume increase coefficient (pumping) and decrease coefficient (power generation) corresponding to a unit of electricity generated in the upper reservoir. , These are the water volume rise coefficient (for power generation) and fall coefficient (for pumping) corresponding to a unit of electricity generated in the lower reservoir. , These represent the upper and lower limits of the reservoir's capacity, respectively. , These represent the upper and lower limits of the reservoir's capacity, respectively.

[0043] In this embodiment, constraints are defined for the upper-level stability optimization model. Specifically, these include: Wind and solar power output constraints: The actual grid-connected power of wind and solar power cannot be negative, nor can it exceed its maximum predicted output.

[0044] Wind and solar curtailment constraints: The wind curtailment rate and solar curtailment rate cannot exceed the maximum value allowed by the system, which is stipulated by the scheduling procedures or policies.

[0045] Pumped storage power station output constraints: pumping power must be between the minimum and maximum pumping power, power generation must be between the minimum and maximum power generation, and pumping and power generation cannot be carried out simultaneously (the product of the two is zero).

[0046] Pumped storage power station start-up and shutdown constraint: The number of start-ups and shutdowns of pumped storage units is determined by a symbolic function to determine whether the pumped storage unit switches from pumping or power generation to another state or shuts down. The number of start-ups and shutdowns must be counted and must not exceed the maximum allowable number.

[0047] Pumped storage power station reservoir capacity constraints: The capacity of the upper and lower reservoirs changes over time. When pumping water, the water volume in the upper reservoir increases (multiplied by the pumping efficiency coefficient), while the water volume in the lower reservoir decreases. When generating electricity, the water volume in the upper reservoir decreases, while the water volume in the lower reservoir increases. The reservoir capacity must be between the upper and lower limits.

[0048] In a preferred embodiment, the objective function of the lower-level economic optimization model is to minimize the system operating cost, as expressed in the formula:

[0049] in, The total operating cost of the system, Costs associated with starting and stopping thermal power units. For peak shaving costs of thermal power units, For carbon trading costs, For the revenue from deep peak shaving subsidies; The formula for calculating the start-up and shutdown costs of thermal power units is as follows:

[0050] in, This refers to the number of thermal power units. For the first The cost of a single start-up and shutdown of the unit. For the first Taiwanese unit The operating status at any given time (1 - running, 0 - stopped); The formula for segmented calculation of peak-shaving costs for thermal power units is as follows: The formula for coal consumption cost is:

[0051] in, , , For the first The coal consumption coefficient of the Taiwanese unit This refers to the unit price of coal. For the first Taiwanese unit Efforts made at all times; The formula for lifetime attrition cost is:

[0052] in, This is the lifespan loss coefficient. For the cost of purchasing the generating unit, ( () represents the number of rotor-induced cracking cycles, and the fitting formula is:

[0053] The formula for oil injection cost is:

[0054] in, Price per unit of oil For the first Taiwanese unit The amount of oil added at any given time.

[0055] In this embodiment, the objective function of the lower-level economic optimization model is defined; the total operating cost of the system includes four parts: start-up and shutdown cost of thermal power units, peak shaving cost of thermal power units, carbon trading cost, minus the revenue from deep peak shaving subsidies.

[0056] Start-up and shutdown costs are equal to the cost of a single start-up or shutdown for each unit multiplied by the change in state (the absolute value of the difference between the current state and the previous state), summed over all units and time periods.

[0057] Peak-shaving costs are divided into three stages: the basic peak-shaving stage (higher output) considers only coal consumption costs; the deep peak-shaving stage without oil injection (lower output) considers coal consumption plus lifespan loss costs; and the deep peak-shaving stage with oil injection (very low output) considers coal consumption plus lifespan loss plus oil injection costs. Coal consumption costs are a quadratic function of unit output multiplied by the coal price. Lifespan loss costs consider the low-cycle fatigue of the turbine rotor during deep peak-shaving, calculated using a fitting function of the rotor's cracking cycle count, which is a cubic polynomial of unit output. Oil injection costs are the oil price multiplied by the amount of oil injected.

[0058] As a preferred implementation, the carbon trading cost is calculated using a tiered carbon trading mechanism, and the relevant formula is as follows: The carbon emission quota formula is:

[0059] in, For carbon emission quotas, For the first Carbon emission allowance coefficient for the Taiwanese unit; The formula for actual carbon emissions is:

[0060] in, For carbon emissions, This is the carbon emission coefficient; The formula for the quota for participating in carbon trading is:

[0061] The formula for the tiered carbon trading cost is:

[0062] in, The base price for carbon trading. For price growth rate, The length of the carbon emission range; The formula for the revenue from the deep peak shaving subsidy is as follows:

[0063] in, This represents the minimum load factor during the basic peak-shaving phase. For the first Rated capacity of the unit To subsidize the unit price.

[0064] This embodiment defines a method for calculating the tiered carbon trading cost. First, the carbon emission allowance is calculated by multiplying the generator output by the carbon emission allowance coefficient and then summing the results. Then, the actual carbon emissions are calculated by multiplying the generator output by the carbon emission coefficient and then summing the results. The difference between the two is the actual allowance used for carbon trading. If the actual emissions are less than the allowance, the allowance is negative, and the system can generate revenue by selling allowances. The tiered carbon trading cost is calculated in segments based on the allowance's range: base price multiplied by the allowance; when the allowance exceeds the first range, the excess is calculated by multiplying the base price by (1 + growth rate); when it exceeds the second range, the excess is calculated by multiplying the base price by (1 + 2 times the growth rate), and so on, for a total of five ranges. This tiered mechanism increases the cost of high carbon emissions, incentivizing the system to reduce carbon emissions.

[0065] The deep peak-shaving subsidy is a form of compensation for thermal power units entering a deep peak-shaving state. The subsidy income is equal to (the minimum output rate of the unit during the basic peak-shaving phase multiplied by the rated capacity minus the actual output) multiplied by the subsidy unit price, summed over all units and time periods.

[0066] In a preferred embodiment, the constraints of the lower-level economic optimization model include: The system power balance constraint formula is:

[0067] in, For the total output of thermal power units; The formula for constraining the output of a thermal power unit in sections is:

[0068]

[0069]

[0070] in, , These are the state variables for the basic peak shaving and deep peak shaving stages, respectively. , For the first The maximum and minimum output of the thermal power unit; For the basic peak shaving phase Minimum output of a thermal power unit; The formula for the ramp constraint related to the operating state is:

[0071] in, , These represent the maximum upward ramp rate during the basic peak shaving and deep peak shaving phases, respectively. , These represent the maximum downward ramp rate during the basic peak shaving and deep peak shaving phases, respectively. Minimum Start-up and Shutdown Time Constraints by State: Thermal power unit operating states are divided into basic peak shaving, deep peak shaving, and shutdown states. To avoid frequent transitions between different states, a minimum duration constraint is set for each state. This refers to the minimum continuous operating time during the basic peak-shaving phase. This is the minimum continuous operating time for the deep peak shaving phase. The minimum continuous downtime is expressed mathematically as follows:

[0072] The system spin-off reserve constraint formula is:

[0073] in, , These are the positive and negative spinning reserve coefficients for load forecasting error. , , , These are the positive and negative spinning reserve coefficients for wind power and solar power prediction errors, respectively. The formula for line transmission capacity constraint is:

[0074] in, For the line Admittance, The voltage phase angle difference between the two nodes of the line. This represents the maximum transmission capacity of the line.

[0075] In this embodiment, the constraints on the lower-level economic optimization model are defined, including: System power balance constraint: The total output of thermal power units equals the load power minus the actual grid-connected power of wind power, minus the actual grid-connected power of photovoltaic power, minus the pumped storage power, plus the pumped storage pumping power.

[0076] Thermal power unit segmented output constraints: The output range of the unit depends on whether it is in basic peak shaving or deep peak shaving state, which is distinguished by state variables; the lower limit of output in basic peak shaving state is higher than the lower limit of output in deep peak shaving state.

[0077] Ramp-up constraints related to operating status: The maximum upward and downward ramp rates of the unit are different in basic peak shaving and deep peak shaving states. Therefore, the corresponding ramp rate is selected in the constraint based on the state at the previous moment.

[0078] Minimum start-up and shutdown time constraints for different states: To avoid frequent switching between basic peak shaving, deep peak shaving and shutdown states, a minimum duration is set for each state; for example, once in deep peak shaving state, it must remain in that state for at least a certain period of time before switching to other states.

[0079] System spinning reserve constraints: To ensure sufficient reserve capacity to cope with forecast errors, the system needs to maintain adequate reserve capacity. Positive reserve requires that the sum of the maximum output and current output of all units be greater than or equal to the sum of positive reserves required for load, wind power, and photovoltaic forecast errors; negative reserve requires that the sum of the current output and minimum output of all units be greater than or equal to the negative reserve requirement.

[0080] Line transmission capacity constraint: The power on each line (determined by the phase angle difference and admittance at both ends) cannot exceed the maximum transmission capacity of the line.

[0081] In a preferred embodiment, the comprehensive norm mentioned in step A3 includes the 1-norm and the ∞-norm; the fuzzy set is constructed by constraining the 1-norm distance and ∞-norm distance between the probability distribution of each typical scenario and the initial probability distribution to not exceed the corresponding fluctuation threshold, as shown in the formula:

[0082] in, Let be the feasible region of the scenario probability distribution. For the first The probability of each scenario. For the first The initial probability of each scenario. , These are the scene probability fluctuation thresholds under 1-norm and ∞-norm constraints, respectively. The fluctuation threshold is determined by the confidence level of the scene probability, and the formula is:

[0083] in, , These represent the confidence levels for the 1-norm and the ∞-norm, respectively. This represents the total number of scenes before scene reduction. This represents the total number of typical scenarios.

[0084] In this embodiment, the method for constructing the fuzzy set is defined. The fuzzy set consists of all probability distributions that satisfy the following conditions: the sum of the absolute values ​​of the differences between the probability of each typical scenario and the initial probability (1-norm) does not exceed the threshold θ1; and the largest single-scenario probability deviation (infinity norm) does not exceed the threshold θ. ∞ The sum of the probabilities of all scenarios is 1; the probability of each scenario is non-negative; θ1 and θ ∞The confidence level is determined by the confidence level. The confidence level λ1 represents the probability that the 1-norm constraint holds, and its expression is 1 minus twice the number of scenes multiplied by an exponential function. The parameters of the exponential function are related to the number of scenes, the initial total number of scenes, and θ1; similarly, λ ∞ This represents the probability that the infinite norm constraint holds; by setting a confidence level (e.g., λ1=0.5, λ...). ∞ =0.99), which can be used to solve for θ1 and θ. ∞ The higher the confidence level, the better θ1 and θ ∞ The larger the value, the larger the fuzzy set, and the more conservative the optimization result.

[0085] In a preferred embodiment, the sub-Bruker optimization framework is a two-layer, four-stage structure, and the matrix form of the sub-Bruker optimization model is as follows:

[0086] in and These are the variables for the first stage (start / stop status, etc.). and For the second stage variables (power, etc.), Contribute to the initial prediction of wind-solar load; The formula for the sub-Bluer bar optimization model is:

[0087]

[0088] Where X and W are the feasible regions of the variables in the first stage. , Let be the feasible region of the variables in the second stage. Let be the feasible region of the scenario probability distribution. This is a typical scenario.

[0089] In this embodiment, the mathematical form of the sub-Bruker optimization model is defined. First, the two-layer deterministic scheduling model is written in compact matrix form, where the first-stage variables (such as start-stop states) are denoted as x and w, and the second-stage variables (such as power) are denoted as y and z. The objective of the sub-Bruker optimization model is: for the upper layer, minimize the first-stage net load variance plus the expected second-stage net load variance under the worst-case probability distribution; for the lower layer, minimize the first-stage cost plus the expected second-stage cost under the worst-case probability distribution; where the worst-case probability distribution refers to the distribution that maximizes the expected net load variance or cost of the second stage within the fuzzy set. The model includes parameters such as the feasible region of variables, constraint matrix, and initial predicted output.

[0090] As a preferred implementation, step A4 uses a column and constraint generation algorithm to solve the problem, decomposing the sub-Bruker optimization framework into a main problem and sub-problems, as shown in the following formula: Upper-level main question MP1:

[0091] Lower-level main question MP2:

[0092] Upper-level subproblem SP1:

[0093] Lower-level subproblem SP2:

[0094] in, For the number of iterations, , The auxiliary variables are used; the main problem solves the first-stage variables and updates the lower bound, while the subproblems solve the worst-case probability distribution and update the upper bound, iterating until (UB) is reached. LB)≤ convergence, For convergence accuracy.

[0095] In this embodiment, the solution algorithm is defined as: a column and constraint generation algorithm. This algorithm decomposes the original problem into a main problem and subproblems. The main problem solves the first-stage variables under a given probability distribution, while introducing auxiliary variables to approximate the second-stage objective and updating the lower bound. The subproblems solve the worst-case probability distribution under the given first-stage variables and update the upper bound. The specific steps are as follows: initialize the lower bound to negative infinity, the upper bound to positive infinity, and the number of iterations to 1; then iterates: solve the main problem to obtain the first-stage variables and update the lower bound; solve the subproblems based on these variables to obtain the worst-case probability distribution and update the upper bound; if the difference between the upper and lower bounds is less than the convergence precision, stop; otherwise, increase the number of iterations, add new second-stage variables and their constraints to the main problem, and return to continue iterating.

[0096] A multi-source coupled distributed bar optimization scheduling system includes: The scene input module is used to input a typical scene set and initial probability distribution of wind, solar and load. The model building module is used to build a two-layer optimization scheduling model that includes upper-layer stability optimization and lower-layer economic optimization, and to configure the aforementioned constraints. The fuzzy set construction module is used to construct fuzzy sets of scene probability distributions based on comprehensive norm and confidence constraints, and generate a fuzzy bar optimization framework. The solution module is used to solve the sub-Bruker optimization framework using a column and constraint generation algorithm and output a scheduling plan. The control module is used to control the operation of pumped storage units and thermal power units according to the scheduling plan.

[0097] In this embodiment, a scheduling system is defined. The system includes five modules: The scene input module is used to input a typical scene set of wind, solar and load and its initial probability distribution.

[0098] The model building module is used to build a two-layer optimized scheduling model, including upper-layer stability optimization (aiming to minimize net load variance) and lower-layer economic optimization (aiming to minimize system operating cost), and configure all constraints.

[0099] The fuzzy set construction module is used to construct fuzzy sets of scene probability distributions based on the comprehensive norm (1 norm and infinite norm) and confidence constraints, and generate a fuzzy bar optimization framework.

[0100] The solution module is used to solve the sub-Bruker optimization framework using the column and constraint generation algorithm, and outputs the scheduling plan for pumped storage units and thermal power units.

[0101] The control module is used to issue instructions to the actual power generation equipment according to the scheduling plan, and control the pumped storage units and thermal power units to operate according to the optimization results.

[0102] The modules work together to achieve a complete process from uncertain scenarios to robust scheduling decisions.

[0103] Example Based on the foregoing description of the solution, this embodiment provides a further detailed explanation of the solution in this application using specific examples: The solution provided in this application is applied to the day-ahead optimization dispatch of a provincial power system with a high proportion of new energy sources; the system adopts the following... Figure 3 The improved IEEE 39-node structure shown includes 7 thermal power units (total installed capacity 2400MW), one pumped storage power station (installed capacity 150MW), one wind farm (500MW), and one photovoltaic power station (300MW); the day-ahead dispatch cycle is 24 hours, and the time resolution is 1 hour. First, five typical wind-solar-load joint scenarios are obtained from the above scenario generation method, and the initial probability of each scenario is set to 0.2.

[0104] like Figure 2As shown, in the construction of the double-layer nested optimization model structure, the upper-layer stability optimization model aims to minimize the net load variance. The net load is defined as the system load minus the renewable energy grid-connected power and pumped storage power, plus the pumped storage pumping power. Taking a typical day as an example, the peak load is 2000MW, and the valley load is 1250MW. The predicted wind power output fluctuates between 200-400MW, and the predicted photovoltaic output is between 0-250MW (zero at night). The upper-layer optimization adjusts the wind and solar curtailment rate (not exceeding 10%) and the pumped storage pumping power (0-150MW) and power generation power (0-150MW) to make the net load curve as flat as possible. The reservoir capacity constraint of the pumped storage power station requires the upper reservoir capacity to be between 780,000 and 5,460,000 cubic meters, and the water volume coefficient corresponding to the unit electricity is 251.6 (pumping) and 318.2 (power generation) cubic meters per megawatt-hour. The start-stop frequency constraint stipulates that the number of start-ups and shutdowns per day shall not exceed 3.

[0105] The lower-level economic optimization model aims to minimize system operating costs, including start-up and shutdown costs of thermal power units, peak-shaving costs, and tiered carbon trading costs, minus the revenue from deep peak-shaving subsidies. The coal consumption coefficient of thermal power units is a quadratic function, for example, 0.001124, 0.2873, and 304.07 for a 200MW unit. Life-cycle attrition costs occur only during the deep peak-shaving phase and are calculated using a rotor low-cycle fatigue curve fitting function, which is a cubic polynomial of the unit's output. Fuel injection costs are incurred during the deep peak-shaving phase, with the fuel price calculated at 500 yuan / ton, and the injection volume determined based on the output level. Carbon trading adopts a tiered mechanism, with a base price of 250 yuan / ton, a price increase rate of 0.25, and a range length of 100 tons. When actual carbon emissions exceed the quota, the excess is calculated at an increasing price. Deep peak-shaving subsidies are provided at a standard of 600 yuan / MWh to thermal power units entering deep peak-shaving mode.

[0106] The lower-level constraints include power balance, segmented output, ramping, minimum start-up and shutdown time, spinning reserve, and line capacity. Taking spinning reserve as an example, positive reserve requires that the algebraic sum of the differences between the maximum output and the actual output of all thermal power units be greater than or equal to 10% of the load forecast error plus 5% of the wind power and photovoltaic forecast errors; negative reserve is similar. Ramping constraints distinguish between basic peak shaving and deep peak shaving states. The maximum upward ramp rate during the basic peak shaving stage is 250MW / h (600MW), 150MW / h (300MW), and 100MW / h (200MW), while it is 50MW / h (600MW) during the deep peak shaving stage. The minimum start-up and shutdown time requires that the duration of thermal power units in the basic peak shaving and shutdown states be no less than 6h (600MW), 5h (300MW), and 3h (200MW), respectively, and the duration of deep peak shaving be no less than 3h (600MW).

[0107] exist Figure 2In the two-layer structure, the net load curve obtained after optimization in the upper layer is passed to the lower layer, and the economic decision of the lower layer is fed back to the upper layer after optimization, forming an interactive iteration; in actual solution, the column and constraint generation (C&CG) algorithm is used, and its solution process is as follows: Figure 3 As shown; initialize the lower bound to negative infinity and the upper bound to positive infinity, with an iteration count of 1; then execute the following loop: solve the main problem to obtain the first-stage variables (pumped storage start-up and shutdown, thermal power start-up and shutdown), and update the lower bound; based on these variables, solve the subproblems to obtain the worst-case probability distribution, and update the upper bound; if the difference between the upper and lower bounds is less than the convergence accuracy (0.1%), stop; otherwise, increase the iteration count, add new second-stage variables and constraints to the main problem, and return to continue iterating.

[0108] In the sub-Bruker optimization modeling, based on the initial probabilities of five typical scenarios, a fuzzy set of probability distributions is constructed using the 1-norm and ∞-norm. The confidence level of the 1-norm is set to 0.5, and the confidence level of the ∞-norm is set to 0.99. The initial total number of scenarios N=1000. The fluctuation thresholds θ1 and θ2 are solved by inverse equations. ∞ The fuzzy set constraint ensures that the 1-norm distance and ∞-norm distance between all possible probability distributions and the initial distribution do not exceed the corresponding thresholds, and that the sum of the probabilities is 1 and each probability is non-negative. This fuzzy set ensures that the optimization results remain robust when faced with deviations between the actual distribution and the assumed distribution.

[0109] The column and constraint generation (C&CG) algorithm was used to solve the problem, which decomposed the two-level four-stage problem into a main problem and subproblems, with a convergence accuracy of 0.1%. The results of the iterative solution showed that the net load variance was 20098.5 MW², and the total system operating cost was 7.5834 million yuan. Compared with stochastic optimization (net load variance 18617.2 MW², cost 7.5095 million yuan), although the cost of this method was slightly higher (about 1.0%), the robustness was significantly enhanced. Compared with robust optimization (net load variance 25199.5 MW², cost 7.7173 million yuan), the net load variance of this method was reduced by 20.2%, and the cost was reduced by 1.7%, achieving a good balance between robustness and economy.

[0110] like Figure 4 As shown, Figure 4 The optimization results were compared between models without pumped storage (Model 1) and models with pumped storage (Model 2). It can be seen that Model 2 has a significantly reduced net load variance, and the lifespan loss cost, oil injection cost, and deep peak-shaving subsidy are all zero, resulting in a reduction in the total system cost.

[0111] Further analysis of the impact of pumped-storage installed capacity; such as Figure 5 As shown, as the pumped storage capacity increases from 50MW to 250MW, the fluctuation of the net load curve gradually weakens; Figure 6The data shows the trends in net load variance, coal consumption cost, carbon trading cost, and total system cost as a function of capacity. When the capacity exceeds 150MW, the improvement in net load variance and wind / solar curtailment slows significantly, indicating the existence of a "pumped storage capacity threshold." Meanwhile, an increase in thermal power utilization hours leads to an approximately 6% increase in total cost. This data can be used to determine the optimal installed capacity in practical engineering projects.

[0112] Compared with stochastic optimization and robust optimization, the results of the robust optimization method in this invention are: net load variance of 20098.5 MW², and total system cost of 7.5834 million yuan. Although stochastic optimization has a lower cost (7.5095 million yuan), it relies on distribution assumptions. Robust optimization is more robust, but its cost is higher (7.7173 million yuan) and it is too conservative. This method achieves a good balance between robustness and economy. Confidence analysis shows that when the confidence level increases from 0.2 to 0.99, the net load variance increases from 19912 MW² to 20742 MW², and the total cost increases from 7.572 million yuan to 7.613 million yuan. The confidence level can be flexibly set according to actual risk preferences.

[0113] In practical deployment, this method generates a 24-hour day-ahead dispatch plan daily, including pumped storage pumping / generation periods and power curves, as well as thermal power start-up / shutdown combinations and output allocation. The dispatch center executes the plan and makes rolling adjustments every 4 hours. After six months of operation, the renewable energy absorption rate increased by approximately 4.2%, the number of deep peak shaving operations for thermal power decreased by approximately 35%, peak shaving costs decreased by approximately 8.7%, carbon emission intensity decreased by 5.3%, and net load variance improved by approximately 18%. This method effectively solves the dispatching challenges under high-proportion renewable energy integration.

[0114] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0115] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A multi-source coupled distributed bar optimization scheduling method, characterized in that, Includes the following steps: Step A1. Obtain the typical scene set of wind, light and lotus and the corresponding initial probability distribution; Step A2. Construct a two-layer optimization scheduling model that includes an upper-layer stability optimization model and a lower-layer economic optimization model; the upper-layer stability optimization model aims to minimize net load fluctuations and optimizes the output of pumped storage units and new energy sources; the lower-layer economic optimization model aims to minimize system operating costs and optimizes the output of thermal power units. Step A3. Based on the typical scenario set, construct a fuzzy set of scenario probability distribution using the comprehensive norm, and transform the two-layer optimization scheduling model into a distributed bar optimization framework; Step A4. Solve the bibliometric optimization framework to obtain the scheduling plans for pumped storage units and thermal power units.

2. The method according to claim 1, characterized in that, The objective function of the upper-level stability optimization model is to minimize the variance of the net load power, as shown in the formula: in, The net load power variance for Net load power at any given time This represents the average net load power. The total number of time periods in the scheduling cycle. for Load power at any given time , They are respectively Wind curtailment rate and solar curtailment rate at any given time , They are respectively Real-time wind and solar power output forecasts , They are respectively The power generation and pumping capacity of the pumped storage power station at all times.

3. The method according to claim 1, characterized in that, The constraints of the upper-level stability optimization model include: the wind and solar power output constraint formula is: in, , This refers to the actual grid-connected power of wind and solar power. , Contribute to the maximum forecast of wind power and photovoltaic power; The formula for constraining wind and solar power curtailment is: in, , The maximum allowable wind and solar curtailment rates by the system; The formula for the output constraint of a pumped storage power station is: in, , These are the lower and upper limits of pumping power. , These are the lower and upper limits of power generation capacity; a product of zero indicates that pumping water and generating electricity cannot be done simultaneously. The formula for constraining the number of start-ups and shutdowns of pumped storage power stations is as follows: in, , Let sgn( represent the start-up and shutdown status of the pumped storage power station at time t, respectively (1 indicates start-up / shutdown, 0 indicates no action). ) is a symbolic function. , These are the maximum allowed number of startups and the maximum allowed number of shutdowns, respectively. The formula for the reservoir capacity constraint of a pumped storage power station is: in, , The reservoir capacities at time t are the upper and lower reservoir capacities, respectively. , These are the water volume increase coefficient (pumping) and decrease coefficient (power generation) corresponding to a unit of electricity generated in the upper reservoir. , These are the water volume rise coefficient (for power generation) and fall coefficient (for pumping) corresponding to a unit of electricity generated in the lower reservoir. , These represent the upper and lower limits of the reservoir's capacity, respectively. , These represent the upper and lower limits of the reservoir's capacity, respectively.

4. The method according to claim 1, characterized in that, The objective function of the lower-level economic optimization model is to minimize the system operating cost, and the formula is: in, The total operating cost of the system, Costs associated with starting and stopping thermal power units. For peak shaving costs of thermal power units, For carbon trading costs, For the revenue from deep peak shaving subsidies; The formula for calculating the start-up and shutdown costs of thermal power units is as follows: in, This refers to the number of thermal power units. For the first The cost of a single start-up and shutdown of the unit. For the first Taiwanese unit The operating status at any given time (1 - running, 0 - stopped); The segmented calculation formula for the peak-shaving cost of the thermal power unit is as follows: The formula for coal consumption cost is: in, , , For the first The coal consumption coefficient of the Taiwanese unit This refers to the unit price of coal. For the first Taiwanese unit Efforts made at all times; The formula for lifetime attrition cost is: in, This is the lifespan loss coefficient. For the cost of purchasing the generating unit, ( () represents the number of rotor-induced cracking cycles, and the fitting formula is: The formula for oil injection cost is: in, Price per unit of oil For the first Taiwanese unit The amount of oil added at any given time.

5. The method according to claim 4, characterized in that, The carbon trading cost is calculated using a tiered carbon trading mechanism, and the relevant formula is as follows: The carbon emission quota formula is: in, For carbon emission quotas, For the first Carbon emission allowance coefficient for the Taiwanese unit; The formula for actual carbon emissions is: in, For carbon emissions, This is the carbon emission coefficient; The formula for the quota for participating in carbon trading is: The formula for the tiered carbon trading cost is: in, The base price for carbon trading. For price growth rate, The length of the carbon emission range; The formula for the revenue from the deep peak shaving subsidy is as follows: in, This represents the minimum load factor during the basic peak-shaving phase. For the first Rated capacity of the unit To subsidize the unit price.

6. The method according to claim 1, characterized in that, The constraints of the lower-level economic optimization model include: The system power balance constraint formula is: in, For the total output of thermal power units; The formula for constraining the output of a thermal power unit in sections is: in, , These are the state variables for the basic peak shaving and deep peak shaving stages, respectively. , For the first The maximum and minimum output of the thermal power unit; For the basic peak shaving phase Minimum output of a thermal power unit; The formula for the ramp constraint related to the operating state is: in, , These represent the maximum upward ramp rate during the basic peak shaving and deep peak shaving phases, respectively. , These represent the maximum downward ramp rate during the basic peak shaving and deep peak shaving phases, respectively. Minimum Start-up and Shutdown Time Constraints by State: Thermal power unit operating states are divided into basic peak shaving, deep peak shaving, and shutdown states. To avoid frequent transitions between different states, a minimum duration constraint is set for each state. This refers to the minimum continuous operating time during the basic peak-shaving phase. This is the minimum continuous operating time for the deep peak shaving phase. The minimum continuous downtime is expressed mathematically as follows: The system spin-off reserve constraint formula is: in, , These are the positive and negative spinning reserve coefficients for load forecasting error. , , , These are the positive and negative spinning reserve coefficients for wind power and solar power prediction errors, respectively. The formula for line transmission capacity constraint is: in, For the line Admittance, The voltage phase angle difference between the two nodes of the line. This represents the maximum transmission capacity of the line.

7. The method according to claim 1, characterized in that, The comprehensive norm in step A3 includes the 1-norm and the ∞-norm; the fuzzy set is constructed by constraining the 1-norm distance and ∞-norm distance between the probability distribution of each typical scenario and the initial probability distribution to not exceed the corresponding fluctuation threshold, as shown in the formula: in, Let be the feasible region of the scenario probability distribution. For the first The probability of each scenario. For the first The initial probability of each scenario. , These are the scene probability fluctuation thresholds under 1-norm and ∞-norm constraints, respectively. The fluctuation threshold is determined by the confidence level of the scene probability, and the formula is: in, , These represent the confidence levels for the 1-norm and the ∞-norm, respectively. This represents the total number of scenes before scene reduction. This represents the total number of typical scenarios.

8. The method according to claim 1, characterized in that, The proposed sub-Bruker optimization framework is a two-layer, four-stage structure, and the matrix form of the sub-Bruker optimization model is as follows: in and These are the variables for the first stage (start / stop status, etc.). and For the second stage variables (power, etc.), Contribute to the initial prediction of wind-solar load; The formula for the sub-Bluer bar optimization model is: Where X and W are the feasible regions of the variables in the first stage. , Let be the feasible region of the variables in the second stage. Let be the feasible region of the scenario probability distribution. This is a typical scenario.

9. The method according to claim 1, characterized in that, Step A4 uses a column and constraint generation algorithm to solve the problem, decomposing the sub-Bruker optimization framework into a main problem and sub-problems, as shown in the following formula: Upper-level main question MP1: Lower-level main question MP2: Upper-level subproblem SP1: Lower-level subproblem SP2: in, For the number of iterations, , The auxiliary variables are used; the main problem solves the first-stage variables and updates the lower bound, while the subproblems solve the worst-case probability distribution and update the upper bound, iterating until (UB) is reached. LB)≤ convergence, For convergence accuracy.

10. A multi-source coupled distributed bar optimization scheduling system, characterized in that, include: The scene input module is used to input a typical scene set and initial probability distribution of wind, solar and load. The model building module is used to build a two-layer optimization scheduling model that includes upper-layer stability optimization and lower-layer economic optimization, and to configure the constraints as described in any one of claims 2-6; The fuzzy set construction module is used to construct fuzzy sets of scene probability distributions based on comprehensive norm and confidence constraints, and generate a fuzzy bar optimization framework. The solution module is used to solve the sub-Bruker optimization framework using a column and constraint generation algorithm and output a scheduling plan. The control module is used to control the operation of pumped storage units and thermal power units according to the scheduling plan.

Images