Comprehensive energy scheduling method and system considering wind and light load uncertainty

Abstract

The invention discloses a comprehensive energy scheduling method and system considering wind and light load uncertainty, and the method comprises the steps: constructing a comprehensive energy system model covering energy supply equipment, energy conversion equipment and energy storage equipment, and carrying out the coordinated scheduling of various types of energy. A load demand response mechanism is introduced to flexibly adjust an energy supply-demand relationship; in the aspect of describing uncertain factors, the method generates a wind and light output and load demand scene covering an uncertain space based on Monte Carlo sampling of multiple probability distribution functions, and carries out scene reduction in combination with a K-means clustering algorithm and a synchronous back substitution elimination technology so as to obtain a typical scene; in the aspect of processing uncertain factors, a two-stage robust optimization method is introduced, and a distribution fuzzy set is constrained through a comprehensive norm, so that the fluctuation range of probability distribution is limited, the influence of deviation on an optimization result is reduced, and the uncertainty of wind and light output and load requirements is processed.

Description

Technical Field

[0001] The present invention belongs to the technical field of integrated energy system scheduling, and particularly relates to an integrated energy scheduling method and system considering the uncertainty of wind and light loads. Background Art

[0002] As an innovative energy supply system integrating various energy forms, the integrated energy system (IES) can achieve the collaborative complementarity between different energy carriers, and has significant advantages in improving the comprehensive energy utilization efficiency, enhancing economic feasibility, and reducing environmental pollution. The advanced optimization scheduling strategy of the integrated energy system integrating renewable energy can effectively improve the reliability and flexibility of the supply side. Therefore, this integration method is considered to be one of the effective methods to solve the current resource shortage and environmental challenges in the energy field. Renewable energy has been applied in various occasions, such as rural areas, communities, and industrial parks. However, the uncertainty of renewable energy in the IES poses a hidden danger to the demand for stable energy supply on the user side. Similarly, the volatility of load demand also brings great challenges to the stable operation and optimal scheduling of the system. Traditional deterministic optimization methods are difficult to ensure robustness. Stochastic optimization methods rely on accurate probability distributions, but the actual data is scarce and the distribution is difficult to accurately characterize, resulting in conservative or unstable optimization results. Robust optimization methods only consider the worst-case scenario, which may sacrifice economy and are difficult to balance the complex constraints of multi-energy coupling. Therefore, how to effectively manage the impact brought by uncertainty factors to the system, realize the minimization of the system operation cost and the stable and optimal operation of the IES, and efficiently utilize energy has become an urgent problem to be solved. Summary of the Invention

[0003] The purpose of the present invention is to provide an integrated energy scheduling method and system considering the uncertainty of wind and light loads.

[0004] In a first aspect, the present invention provides an integrated energy scheduling method considering the uncertainty of wind and light loads, which includes the following steps: Taking the start-stop cost of the micro gas turbine and the total system operation cost as the optimization objectives, the objective function and constraint conditions of the integrated energy system model covering energy supply equipment, energy conversion equipment, and energy storage equipment are constructed; the probability distribution models of photovoltaic output, wind turbine output, and electric and heat loads are respectively constructed, and random sampling is carried out according to different probability distribution models and constraint conditions to obtain scenarios of the individual outputs of photovoltaic units, wind turbine units, electric loads, and heat loads with different probabilities, so as to obtain the typical scenario set of the combined output of wind and light loads and their corresponding probabilities; a distributionally robust optimization model is constructed, and the probability distribution of the typical scenarios is constrained to obtain the fuzzy set of the uncertain probability distribution; the distributionally robust optimization model is decomposed into a master problem and a subproblem, and the master problem and the subproblem are continuously iterated through the column-and-constraint generation decomposition algorithm until the termination condition is reached, and then the optimal integrated energy scheduling scheme is obtained.

[0005] Preferably, the probability distribution model of photovoltaic output is constructed based on the beta distribution; the probability distribution model of wind turbine output is constructed based on the Weibull distribution; the probability distribution model of electric and heat loads is constructed based on the normal distribution.

[0006] Preferably, the objective function of the constructed integrated energy system model is as follows: In the formula, is the start-stop cost of the micro gas turbine; is the start-stop cost coefficient of the micro gas turbine; is the state variable of the micro gas turbine at time taking values of 0 or 1; is the total system operation cost; is the cost of purchasing electricity from the power grid; is the cost of purchasing gas from the gas network; is the energy storage cost; is the cost of curtailed wind; is the cost of curtailed light; is the revenue from selling electricity;

[0007] Preferably, the demand response cost is introduced into the objective function corresponding to the total system operation cost; the demand response cost includes the compensation cost for shiftable loads participating in demand response, the compensation cost for transferable loads participating in demand response, and the compensation cost for reducible loads participating in demand response.

[0008] Preferably, the specific process of obtaining the typical scenario set of the combined output of wind and light loads is as follows: The K-means clustering algorithm is used to compress the sampling scenarios, extract the typical scenario sets of the individual outputs of photovoltaic, wind turbines, electrical loads, and thermal loads, and perform a Cartesian join on these different types of scenarios. Then, the synchronous back substitution elimination technique is used to further reduce the scenarios, obtaining a typical scenario set composed of multiple groups of typical scenarios of the combined output of wind-solar loads.

[0009] Preferably, the number of scenario clusters in the K-means clustering algorithm is obtained by the silhouette coefficient method.

[0010] Preferably, the probability distribution of the said typical scenarios is constrained by the comprehensive norm.

[0011] Preferably, the objective function of the main problem includes the operating cost, compensation cost, and variables related to uncertainty.

[0012] Preferably, the specific process of iteratively solving the two-stage distributionally robust optimization model is as follows: Initialize the iteration times of the uncertain variables; set the convergence accuracy, upper bound value, and lower bound value respectively; set the occurrence probability of each typical scenario to its reference probability distribution value; solve the main problem to obtain the first-stage variables and the optimal solution, and update the lower bound value according to the optimal solution corresponding to the main problem; solve the sub-problem based on the first-stage variables to obtain the second-stage variables and the optimal solution, and update the upper bound value according to the optimal solution corresponding to the sub-problem; if the difference between the upper bound value and the lower bound value is greater than the convergence accuracy, then add the second-stage variables and the related constraints to the main problem, and re-solve the main problem until the difference between the upper bound value and the lower bound value is less than or equal to the convergence accuracy, stop the iteration, and obtain the optimal solution.

[0013] In the second aspect, the present invention provides a comprehensive energy scheduling system considering the uncertainty of wind-solar loads, which includes energy supply equipment, energy conversion equipment, energy storage equipment, and a control module; the energy supply equipment includes photovoltaic units and wind turbine units; the energy conversion equipment includes electric boilers and combined heat and power units; the energy storage equipment includes storage batteries and heat storage pools; the said control module is used to execute the above-mentioned comprehensive energy scheduling method and control the energy supply equipment, energy conversion equipment, and energy storage equipment based on the generated optimal comprehensive energy scheduling plan.

[0014] The beneficial effects of the present invention are as follows: 1. The present invention combines the Monte Carlo sampling method based on multiple probability distribution functions with the K-means clustering algorithm to construct typical scenarios for describing the uncertainty of wind-solar loads; Monte Carlo can capture extreme scenarios through completely random sampling to test the performance of the system in the worst case, ensure the reliability of the scheduling plan, solve the problem of insufficient scenario coverage in the prior art that cannot fully capture extreme situations, enhance the robustness of the model, and reduce the high-cost risk in actual operation.

[0015] 2. The present invention adopts a two-stage distributionally robust optimization method to handle the uncertainties of wind and light loads in the system, achieving a balance between economy and robustness during the optimization dispatch process. At the same time, by introducing the operation mechanism of demand response, the distribution of electric and thermal loads is optimized, further optimizing the system operation and reducing the system operation cost, enabling the present invention to timely adjust the unit output to find the optimal solution to optimize the cost under different scenarios.

[0016] 3. The present invention adopts a comprehensive norm to constrain the probability distribution of typical scenarios, and has better performance in terms of economy compared to only considering single-norm constraints. At the same time, compared with the two methods of stochastic optimization and robust optimization for traditional uncertainty optimization problems, the present invention has better robustness than stochastic optimization and is more prominent in terms of economy compared to robust optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] To more clearly illustrate the technical solutions of the present invention, the drawings required for use in accordance with the embodiments will be briefly introduced below. Obviously, the descriptions in the drawings are only some embodiments of the present invention, and those of ordinary skill in the art can obtain other drawings based on these drawings without investing creative labor.

[0018] Figure 1 It is the overall flowchart of the present invention.

[0019] Figure 2 It is the flowchart of joint scenario generation in the present invention.

[0020] Figure 3 It is the flowchart of solving the distributionally robust optimization model in the present invention.

[0021] Figure 4 It is the diagram of the combined output of wind and light and load scenarios in the present invention; among them, (a) is the schematic diagram of photovoltaic output; (b) is the schematic diagram of wind turbine output; (c) is the schematic diagram of electrical load; (d) is the schematic diagram of thermal load.

[0022] Figure 5 It is the diagram of the combined output of wind and light and load scenarios based on Latin hypercube sampling and K-means clustering algorithm; among them, (a) is the scenario diagram of photovoltaic output; (b) is the scenario diagram of wind turbine output; (c) is the scenario diagram of electrical load; (d) is the scenario diagram of thermal load.

[0023] Figure 6 It is the schematic diagram of the optimized electric and thermal load results of combined demand response under different energy storage configurations; among them, (a) is the schematic diagram of the scenario without energy storage; (b) is the schematic diagram of the combined scenario of electrical energy storage and thermal energy storage; (c) is the schematic diagram of the electrical energy storage scenario; (d) is the schematic diagram of the thermal energy storage scenario.

[0024] Figure 7This is the graph of the optimized results for electric heating. Among them, (a) is the schematic diagram of the optimized results for electric energy; (b) is the schematic diagram of the optimized results for thermal energy. Detailed implementation manners

[0025] The present invention will be further described below with reference to the accompanying drawings.

[0026] As Figure 1 shown, a comprehensive energy scheduling method considering the uncertainty of wind and light loads includes the following steps: Step 1: Construct a comprehensive energy system model 1-1. Objective function The comprehensive energy system supplies energy and stores energy through energy supply equipment, energy conversion equipment, and energy storage equipment; the energy supply equipment includes photovoltaic units (PV) and wind turbine units (WT); the energy conversion equipment includes electric boilers (EB) and combined heat and power units (CHP); the energy storage equipment includes storage batteries and heat storage pools, and each equipment operates coordinately to meet the electric and thermal loads of the system under the demand response mechanism; in addition, the comprehensive energy system also supplies energy by purchasing electricity from the power grid and purchasing gas from the gas network. Taking the start-stop cost of the micro gas turbine in the combined heat and power unit and the lowest total system operation cost as the goal, the objective function of the comprehensive energy system model covering the energy supply equipment, energy conversion equipment, and energy storage equipment is: In the formula, is the start-stop cost of the micro gas turbine; is the start-stop cost coefficient of the micro gas turbine; is the state variable of the micro gas turbine at time with a value of 0 or 1; is the total system operation cost; is the cost of purchasing electricity from the power grid; is the cost of purchasing gas from the gas network; is the cost of wind curtailment; is the cost of light curtailment; is the revenue from selling electricity; is the cost of demand response.

[0027] The cost of purchasing electricity from the power grid , the cost of natural gas required for combined heat and power , the cost of energy storage , the cost of wind curtailment , the cost of light curtailment , the revenue from selling electricity and the cost of demand response are expressed as follows: In the formula, is the electricity purchase price of the IES from the superior power grid; is the electricity purchase quantity of the IES from the superior power grid at time is the gas purchase price of the IES from the superior gas grid; is the gas purchase quantity of the IES from the superior gas grid at time is the degradation coefficient of the electrical energy storage; is the degradation coefficient of the thermal energy storage; is the charging power of the electrical energy storage device at time is the discharging power of the electrical energy storage device at time is the charging power of the thermal energy storage device at time is the discharging power of the thermal energy storage device at time is the operation cost coefficient of the wind turbine; is the predicted wind power at time is the actual wind power participating at time , kW; is the operation cost coefficient of the photovoltaic; is the predicted photovoltaic power at time is the actual photovoltaic power participating at time is the electricity selling price of the IES to the superior power grid; is the electricity selling quantity of the IES to the superior power grid at time is the compensation cost for the shiftable load participating in the demand response; is the compensation cost for the transferable load participating in the demand response; is the compensation cost for the curtailable load participating in the demand response.

[0028] The compensation cost for the shiftable load participating in the demand response model is: In the formula, is the compensation price for the unit power load shift; is the set of start time periods of the load shift; is the power distribution vector before participating in the dispatch; is the power distribution vector translated from the starting time period to the power distribution vector with the starting time period being When , the load is not translated; is the power distribution vector sum of the consumed electric powers; is the power distribution vector at a certain time period translation state, taking values of 0 or 1; and are respectively the starting and ending times of the translatable time period; is the duration; is .

[0029] Compensation cost for the transferable load participating in demand response model is: In the formula, is the compensation price for the transfer of unit power load; is transferable load power at the time period; and are respectively the starting and ending times of the transfer time period; is transfer state at the time period, taking values of 0 or 1.

[0030] Compensation cost for the curtailable load participating in demand response model is: In the formula, is the compensation price for the curtailment of unit power load; is a dummy variable representing the curtailment state of the curtailable load at a certain time period ; is the power after participating in the dispatch at the time period ; is the power before participating in the dispatch at the time period ; is load curtailment coefficient at the time period.

[0031] 1 - 2. Constraint conditions Power of the transferable load has the following power constraints: In the formula, and are respectively the minimum and maximum values of the power of the transferable load; Meanwhile, to prevent the equipment from starting and stopping frequently due to overly dispersed load, it is necessary to impose constraints on the minimum continuous operation time of the transferable load. Let's start observing the operation of the transferred load from time. If , then the operating state of the equipment where the transferable load is located remains unchanged at time. If , then the equipment where the transferable load is located starts to transfer the load at time. During the period from , the sum of the equipment operating states should satisfy: In the formula, is the minimum continuous operation time.

[0032] Impose constraint control on the maximum continuous reduction time, minimum continuous reduction time, and reduction times: In the formula, is the maximum reduction times; is the maximum continuous reduction time; is the minimum continuous reduction time.

[0033] Wind turbine power constraint: In the formula, is the maximum wind power; is the maximum photovoltaic power.

[0034] CHP unit operation constraint: In the formula, is the electrical power output at time; is the gas-electric conversion efficiency; is the input gas power; and are respectively the upper and lower limit powers of the input gas energy; is the thermoelectric ratio produced by CHP; and are respectively the upper and lower limits of the thermoelectric ratio.

[0035] Ramp constraint: In the formula, is the upper limit of the ramping rate of the combined heat and power unit; is the lower limit of the ramping rate of the combined heat and power unit.

[0036] Operating constraints of the energy storage device: In the formula, and are the charging and discharging power of the energy storage device at time and are the charging and discharging flags at time is the upper limit of the energy storage charging and discharging, kW; is the energy storage capacity at time and are the upper and lower limits of the energy storage capacity; and are the energy storage charging and discharging efficiency.

[0037] Electric power balance constraint: In the formula, is the electric load during the is the shiftable electric load during the is the transferable electric load during the

[0038] Thermal power balance constraint: In the formula, is the electro-thermal conversion efficiency of the electric boiler; is the power of the electric boiler at time is the thermal load at time is the reducible thermal load at time

[0039] Step 2. For the uncertainty problems existing in the wind and solar power output and the electric and thermal loads in the integrated energy system, generate multiple representative wind and solar load scenarios and their probability distributions, and incorporate the uncertainty factors into the optimization model for consideration, so that the optimized scheduling strategy can, to a certain extent, cope with the random changes in the renewable energy output and the load, and improve the system's ability to absorb renewable energy and the robustness of its operation.

[0040] According to the historical data characteristics of photovoltaic output, wind turbine output and electro-thermal load, Monte Carlo (MC) random sampling is used to generate multiple scenarios, and combined with the following probability distribution models to cover the uncertainty space: Photovoltaic output is significantly affected by weather conditions (such as irradiance, cloud cover changes), and its normalized power value is usually distributed in the interval [0, 1]. Therefore, the Beta distribution is selected to describe the probability characteristics of photovoltaic output, and its probability density function is: In the formula, , are two shape parameters, and the distribution form (such as symmetry, skewness) can be adjusted by fitting historical data; is the Beta function, and its calculation formula is: , and the gamma function satisfies in the integer domain, supporting efficient calculation of parameters.

[0041] The output of the wind turbine is closely related to the wind speed, and the probability distribution of the wind speed usually follows the Weibull distribution. Therefore, the Weibull distribution is selected to describe the probability characteristics of the wind turbine output, and its probability density function is: In the formula, is the scale parameter, which determines the typical value range of the wind speed and reflects the "average wind speed" characteristic; is the shape parameter. When , the Weibull distribution degenerates into an exponential distribution; when , the Weibull distribution is similar to the Rayleigh distribution, and can flexibly adapt to the wind speed characteristics of different regions.

[0042] The uncertainty of electro-thermal load demand is usually affected by multiple factors such as user behavior and temperature changes, and its prediction error conforms to a normal distribution. Therefore, the normal distribution is selected to describe the probability characteristics of the load demand. Its probability density function is: In the formula, represents the expected value of the load, reflecting the typical energy consumption pattern of users; represents the load fluctuation range, The larger it is, the higher the uncertainty.

[0043] Step 3: Since the number of scenarios generated by Monte Carlo sampling is huge, directly using them for optimization calculation will lead to extremely high computational complexity. Therefore, the K-means clustering algorithm (K-means clustering algorithm) is used to compress the sampled scenarios and extract a set of typical scenarios. When determining the number of scenario clusters, it is not that the larger the number, the better. If the number is too high, it may cause the calculation rate of the optimization model to decrease. Therefore, it is necessary to find an appropriate number of clusters according to the data itself. The present invention selects the silhouette coefficient method to determine the number of scenario clusters.

[0044] As Figure 2 shown, based on a large amount of scenario data generated by MC sampling, the K-means clustering algorithm is used to reduce and obtain a finite number of scenarios with different probabilities of the individual outputs of photovoltaic units, wind turbine units, electrical loads, and thermal loads. These four types of scenarios are subjected to a Cartesian join, and the synchronous back substitution elimination technique is used to further reduce the scenarios. Finally, Y typical scenarios of the combined output of wind-solar loads and their corresponding probabilities are obtained; where Y is the number of typical scenarios of the combined output of wind-solar loads. Then, the objective function values of each group of combined scenarios can be calculated and added up to facilitate the analysis of the system.

[0045] Step 4: Build a distributionally robust optimization model Given the possible limitations of the historical data of wind-solar loads, the typical scenarios obtained through scenario clustering may not fully reflect the actual fluctuations. Therefore, the distributionally robust optimization method is introduced. Taking the typical scenarios as inputs, a distributionally fuzzy set is defined through comprehensive norm constraints, thereby restricting the fluctuation range of the probability distribution, reducing the impact of deviations on the optimization results, and balancing the robustness and economy in the optimization process.

[0046] Construct a two-stage distributionally robust optimization model; the first stage lays the foundation for the normal operation of the system and conducts preliminary cost optimization to minimize the fixed cost. Its objective function comprehensively considers various costs, such as operating costs, compensation costs, and variables related to uncertainties ; the second stage mainly adjusts the decision-making for uncertain scenarios, that is, according to the decision-making results of the first stage and the values of the uncertain variables, calculates the costs or benefits of the system under different uncertain scenarios, and feeds them back to the first stage to adjust the decision variables of the first stage to minimize the expected adjustment cost in the worst case. The matrix representation of the two-stage distributionally robust optimization model is as follows: In the formula, xis the variable of the first stage; y b is the variable of the second stage; m, n is the cost coefficient corresponding to the variables of the first and second stages; is the scenario occurrence probability; is the set that satisfies; is the scenario probability distribution value that maximizes the objective function; is the unit deviation penalty coefficient; is the scenario wind power and photovoltaic actual output deviation and load fluctuation vector; is the number of scenario clusters; is the matrix or vector that matches the variables in the constraints of the optimization model; Equation is all the constraints related to the variables of the first stage; Equation is the related constraint of the variables of the second stage; Equation and Equation is the coupling constraint of the two-stage variables.

[0047] Step 5: Based on the probability distribution of the typical scenarios constrained by the comprehensive norm, obtain the fuzzy set of the uncertain probability distribution. The probability value of each discrete scenario can belong to any feasible region . In order to balance robustness and economy during the optimization process and ensure that the actual probability distribution fluctuates within a reasonable range and is closer to the characteristics of the actual data, a comprehensive norm is introduced as a constraint condition to limit the fluctuation range of the wind power, photovoltaic, and load probability distributions, enabling the optimization model to remain robust under the worst probability distribution. Specifically, the comprehensive norm includes the 1-norm and the ∞-norm; by setting the constraint conditions based on the comprehensive norm, a distribution fuzzy set containing various possible probability distributions is defined. This distribution fuzzy set reflects the distribution uncertainty in the understanding of uncertain factors, that is, although the occurrence probability of each scenario cannot be accurately determined, the reasonable fluctuation range of the probability distribution can be limited through the comprehensive norm constraint, avoiding scenario deviations caused by the limitations of historical data. During the optimization process, the model minimizes the expected cost under the worst distribution to balance economy and robustness and reduce the impact of extreme scenarios on the system operation cost. This method enables the optimization model to consider a wider range of uncertainty situations and enhances the adaptability of the model to uncertainty. If only the norm or is considered alone, it may lead to extreme or one-sided situations in probability fluctuations. Therefore, a comprehensive norm constraint is constructed, and its distribution fuzzy set satisfies the following constraints: In the formula, is the typical scenario Reference probability distribution value; is the constraint radius, that is, the allowable value of the probability deviation under the corresponding norm constraint; is the number of typical scenarios.

[0048] The deviation of the scenario probability satisfies the following probability guarantee: In the formula: represents the probability of the event occurring; is the number of historical scenarios.

[0049] Let the right side of the above formula be equal to and , then it is transformed into the formula: Among them, α1 and α ∞ represent the confidence level that the probability distribution needs to satisfy.

[0050] Step 6: Iteratively solve the two-stage distributionally robust optimization model to obtain the optimal integrated energy scheduling plan.

[0051] As Figure 3 shown, for the constructed two-stage distributionally robust optimization model matrix, it is disassembled into two parts, namely the master problem and the sub-problem, and then a decomposition algorithm based on column and constraint generation (C&CG) is used to solve it. By continuously iterating the master problem and the sub-problem through the C&CG algorithm, the upper and lower bounds are gradually tightened, and finally the iteration stops after obtaining the optimal solution that meets certain accuracy requirements.

[0052] The objective function L of the master problem can be expressed in matrix form as: In the formula, is the given threshold; is the total number of model iterations.

[0053] The sub-problem is a two-layer structure of max-min, which is solved after the first-stage variables are given, and its objective function is expressed as: In the sub-problem, the discrete scenario probability values are irrelevant to other variables. Therefore, the sub-problem can be decomposed into two steps for solution, that is, first find the inner-layer minimum value , and then solve the outer-layer max problem U. The specific formula is as follows: As Figure 3 shown, the specific process of iterative solution for the two-stage distributionally robust optimization model is as follows: (1) Initialize the iteration number i of the uncertain variable to 1; set the convergence accuracy ε, the upper bound UB = ∞, the lower bound LB = 0; set the occurrence probability of each typical scenario to its reference probability distribution value.

[0054] (2) By solving the master problem, obtain the first-stage variable and the optimal solution . According to the optimal solution , update the lower bound value ; (3) Based on the first-stage variable , solve the sub-problem to obtain the scenario occurrence probability and the optimal solution . According to the optimal solution , update the upper bound value ; (4) Perform convergence judgment. If UB - LB ≤ ε, stop the iteration and obtain the optimal solution; otherwise, update = in the master problem, and add the new variable and the related constraint conditions to the master problem, and then solve the master problem again.

[0055] Example 1 Apply this model to a community-level integrated energy system. The time-of-use electricity price, gas price, and model parameters of each device are shown in Tables 1-3. Then, compare different scenario generation methods, different data volume scenarios, different norms, other uncertainty methods, and analyze the comprehensive cost of joint demand response.

[0056] Table 1 Energy and natural gas prices of the superior power grid Table 2 Parameter settings of energy conversion equipment Equipment Name Energy Conversion Efficiency Capacity / kW CHP 0.92 300 EB 0.9 250 Table 3 Parameter settings of energy storage devices (1) Comparison of different scenario generation methods 1000 sets of initial scene data of wind and solar loads were generated by MC sampling and Latin hypercube sampling (LHS) based on different distribution functions, and K-means clustering algorithm was used to reduce them to 20 scenarios with different probabilities of photovoltaic units, wind turbine units, electric loads and thermal loads output separately. These four types of scenarios were Cartesian connected, and the synchronous back-substitution elimination technology was used to further reduce the scenarios. Finally, four sets of typical wind and solar load joint output typical scenarios were obtained, such as Figure 4 and Figure 5 shown.

[0057] from Figure 4 From the wind, solar and electric heat load scene diagrams shown, the distribution of each scene data at different times and scene numbers is relatively more dispersed. Taking photovoltaic output as an example, the different scenes represented by different color curves have obvious numerical differences at different times, covering a wide range. This dispersion shows that the scenes generated by Monte Carlo sampling can capture a wider range of possible situations, reflecting more uncertainty and variability. Since the changes of these parameters in actual operation are diverse and random, it has advantages in simulating the actual complex wind and solar joint output and load conditions; there are also differences in the output value range between scenes. The output value fluctuation range of some scenes is relatively wide, which may cover a large span from lower power to higher power, while the output value fluctuation range of some scenes is relatively narrow, reflecting the characteristics of relatively more stable output or smaller fluctuation range, which also indirectly reflects the representativeness of the scenes after reduction in reflecting the diversity of the overall scenes, including both drastic output changes and relatively stable situations; the volatility of the data curves of each scene is large. Taking the electric load scene as an example, the curve fluctuates significantly, which can reflect the rapid changes and uncertainty of the load at different times. This larger volatility is more in line with the actual characteristics of load changes at any time, and can more realistically simulate the operating status of the system under different load fluctuations.

[0058] exist Figure 5 Among them, the distribution of typical scenario data is relatively concentrated, and the values ​​of different scenario curves are relatively close at certain moments. For example, in the wind turbine output scenario diagram, some curves almost overlap at some moments, indicating that the scenarios generated by LHS lack diversity and may not be able to fully cover all possible working conditions, and some extreme or special situations are not reflected enough; in addition, the data curve is relatively smooth and has low volatility. For example, in the electric load scenario diagram, the curve changes relatively slowly, which may not accurately reflect the rapid changes in electric load in actual operation. For research that needs to consider the impact of rapid load fluctuations on the system, Figure 5 The data may not provide sufficiently precise information.

[0059] Five scenarios as shown in Table 4 were designed respectively while keeping other factors unchanged. In Scenario 1, the photovoltaic output scenarios were generated using the LHS method, and the wind turbine output, electrical load, and heat load scenarios were all generated using the MC method; in Scenario 2, the photovoltaic output scenarios were generated using the MC method, the wind turbine output scenarios were generated using the LHS method, and the electrical load and heat load scenarios were generated using the MC method; in Scenario 3, the photovoltaic output, wind turbine output, and heat load scenarios were generated using the MC method, and the electrical load scenario was generated using the LHS method; in Scenario 4, the photovoltaic output, wind turbine output, and electrical load scenarios were generated using the MC method, and the heat load scenario was generated using the LHS method; in Scenario 5, the photovoltaic output, wind turbine output, electrical load, and heat load scenarios were all generated using the MC method. By calculating the sum of the cost optimization results of all typical combined scenarios in the above scenarios, a scenario generation method with better optimization effect can be explored, and the calculation results are shown in Table 5.

[0060] Table 4 Scenario Scheme Design Table 5 Cost Optimization Results of Each Scenario Scheme Scenario Plan Cost Optimization Result (Unit: Yuan) Plan 1 11812.5782 Plan 2 12290.1790 Plan 3 12980.7465 Plan 4 12113.1033 Plan 5 11812.5713 As can be seen from Table 5, when other parameters remain constant, the cost optimization result of Scenario 5 is 11,812.5713 yuan, which is the lowest among the 5 scenarios. The cost of Scenario 1 is 11,812.5782 yuan. Although it is close to Scenario 5, the cost of Scenario 5 is lower. The cost of Scenario 2 is 12,290.1790 yuan, the cost of Scenario 3 is as high as 12,980.7465 yuan, and the cost of Scenario 4 is 12,113.1033 yuan, all of which are significantly higher than that of Scenario 5. This shows that in terms of cost control, Scenario 5 has significant advantages, can further reduce the conservativeness of the obtained results, achieve the optimization goal at a lower cost, helps to formulate a more accurate scheduling plan, saves some unnecessary additional expenses, and is more attractive economically. And LHS may introduce additional sampling bias in the optimization solution process, making the convergence speed unstable, thus affecting the optimization calculation efficiency and optimization results.

[0061] (2) Comparison of the Amount of Data Generated in Different Scenarios To highlight the impact of the scale of generating N initial scenario data using MC on the comprehensive cost, on the premise that other factors remain unchanged, according to Figure 2 the method described in the process, scenario reduction was carried out based on generating 200, 500, 1000, and 1500 groups of wind-solar output and load scenario data respectively, and the optimization results are shown in Table 6 as REF _Ref7195 \h.

[0062] Table 6 Result Comparison of Different Numbers of Wind-Solar Output and Load Scenario Data Number of Scenarios of Wind and Photovoltaic Output and Load Comprehensive Cost / Yuan 200 11892.3943 500 11873.4966 1000 11812.5713 1500 11779.3077 As can be seen from Table 6 REF _Ref7195 \h, as the number of initial scenarios generated by MC gradually increases, the comprehensive cost calculated by the system shows an obvious downward trend. The reason is that when the number of scenarios continues to increase, the deviation between the probability distribution constructed based on these data and the actual true probability distribution gradually decreases. In the optimization process of the integrated energy system, the accuracy of the probability distribution plays a crucial role in decision-making. When the constructed probability distribution can approximate the real situation more accurately, the information on which the optimization process is based is more reliable, thus effectively reducing the conservatism of the optimization results. This means that the system can allocate resources more reasonably during operation, avoid unnecessary cost increases caused by overly conservative decisions, and thus achieve a reduction in the comprehensive cost.

[0063] (3)Comparison of different norms To show the difference between the comprehensive norm constraint and only considering the 1-norm or ∞-norm constraint, when and take values of 0.6 and 0.99 respectively, the calculation results of the single norm are shown in Tables 7 and 8. The comparison results show that using the comprehensive norm method can reduce the conservatism brought by the optimization and is more economical than only considering the 1-norm or -norm constraint, which indicates that the comprehensive norm constraint has a lower conservatism degree compared to only considering a single norm constraint.

[0064] Table 7 Comparison of comprehensive norm and 1-norm results Table 8 Comparison of comprehensive norm and -norm results (4)Comparison with other uncertainty methods The DRO (distributionally robust optimization) algorithm proposed in the present invention is compared with SO (stochastic optimization) and RO (robust optimization) respectively, and the comparison results are shown in Table 9 REF _Ref8491 \h.

[0065] Table 9 System operation costs under different uncertainty optimization methods Method Comprehensive Cost / Yuan Abandoned Wind and Photovoltaic Energy / % SO 10213.8842 0.00 RO 14742.3211 0.00 DRO 11812.5713 0.00 As can be seen from Table 9, the comprehensive cost obtained by using the RO optimization model is the largest, the comprehensive cost obtained by the SO optimization model is the smallest, and the comprehensive cost obtained by the DRO is between the two. The random optimization model has a low conservatism in its results because it considers the uncertainty of the distribution parameters of the uncertain variables. The robust optimization model does not consider the distribution parameters, resulting in overly conservative results. However, the distributionally robust optimization model takes into account the distribution in the worst-case scenario, while taking into account both conservatism and economy, achieving relatively ideal results.

[0066] (5)Analysis of the Comprehensive Cost of Joint Demand Response For Figure 4 the typical scenario 1 in, the comprehensive cost analysis of joint demand response is carried out. As Figure 7 shown, the load transfer, load shifting, and load curtailment in the typical scenario 1 mainly occur from 10:00 to 13:00, 14:00 to 15:00, and 17:00 to 21:00, mostly during the peak electricity price periods, relieving the peak electricity consumption pressure and playing a role in peak shaving and valley filling. Also, combined with Figure 6 it can be seen that the battery is always charged during the low electricity price period and discharged during the peak electricity price period throughout the scheduling period, using the difference in time-of-use electricity prices to reduce the system operation cost; the heat storage tank also always charges heat during the low heat load period and releases heat during the peak heat load period, assisting the electric boiler and the combined heat and power unit to meet the heat demand. To further explore the impact of energy storage devices on the two-stage robust optimization model of joint demand response, while keeping other factors unchanged, the experimental scenario scheme shown in Table 10 is set: Table 10 Scenario Design Plan 1 Plan 2 Plan 3 Plan 4 Energy Storage Configuration No Energy Storage Electric Energy Storage + Thermal Energy Storage Electric Energy Storage Thermal Energy Storage The optimal electro-thermal load results of joint demand response for the four scenario schemes with different energy storage configurations are as Figure 6 shown.

[0067] From Figure 6 the curve changes of the 4 schemes, it can be seen that the impact of different energy storage configurations on the final cost optimization is mainly larger during the peak load period. And the peak load period is affected by the peak electricity price, and it is necessary to increase the electricity purchase volume or increase the power supply of the micro gas turbine to ensure the electro-thermal load demand of users, resulting in an increase in cost. The specific final optimized costs of each scenario scheme are: 12,850.1467 yuan, 11,812.5713 yuan, 12,624.5927 yuan, and 12,012.1309 yuan. It can be seen that configuring appropriate energy storage devices can play a positive role in reducing the system cost.

[0068] From Figure 7As can be seen from (a) in [reference], in typical scenario 1, photovoltaic wind turbines and micro gas turbines are mainly used for power generation, and a large amount of electricity is purchased during the low - valley period of electricity price from 05:00 to 08:00. On the premise of meeting the optimization of its own electricity load dispatching, according to the real - time electricity demand, the surplus electricity in the process of photovoltaic wind turbine power generation can be sold externally; the energy storage device mainly charges during the low - valley period of electricity price and discharges during the peak period of electricity load and peak - time electricity price. From Figure 7 As can be seen from (b) in [reference], in typical scenario 1, CHP burns natural gas and an electric boiler is mainly used for heating. Among them, the electric boiler converts electrical energy into heat energy to supply the demand of heat load. The heat energy storage device stores heat from 06:00 to 08:00 and releases heat when the output of the heating device is weak or the heat load is relatively large. The addition of the heat storage device can effectively reduce the processing of the heat recovery system, thus further reducing the system operation cost during the period when the heat storage device outputs; during the periods of 17:00 - 18:00, 20:00 - 21:00, and 22:00 - 23:00, the output of the heat storage tank device and the electric boiler device reduces the output of the micro gas turbine, and the system can consume more wind power output during this period.

[0069] The present invention uses a typical scenario constructed by combining the Monte Carlo sampling method based on multiple probability distribution functions with the K - means clustering algorithm to describe the uncertainty of wind - solar load. Monte Carlo can capture extreme scenarios, such as extremely high or low wind - solar output and load demand, through completely random sampling, to test the performance of the system in the worst - case scenario, ensure the reliability of the dispatching scheme, enhance the robustness of the model, and reduce the high - cost risk in actual operation. A two - stage distributionally robust optimization model is used to handle the uncertainty of wind - solar load in the system, achieving a balance between economy and robustness during the optimization dispatching process, and flexibly adjusting the energy supply - demand relationship by introducing a load demand response mechanism. In the first stage of this model, a relatively conservative but feasible decision - making scheme is determined for the normal operation of the system to cope with possible uncertainty situations; the second stage is mainly responsible for calculating the cost or benefit of the system under different uncertainty scenarios according to the decision - making results of the first stage and the values of uncertainty variables, and feeding it back to the first stage to adjust the decision - making variables of the first stage. At the same time, in order to ensure that the actual probability distribution fluctuates within a reasonable range and is closer to the characteristics of actual data, the 1 - norm and ∞ - norm are introduced as constraint conditions to limit the probability distribution values of each scenario, enabling the optimization model to consider a wider range of uncertainty situations and enhancing the adaptability of the model to uncertainty.

[0070] The above are only the preferred specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, making equivalent substitutions or changes, shall be covered by the protection scope of the present invention.

Claims

1. A comprehensive energy scheduling method considering the uncertainty of wind and light loads, characterized in that: The method includes the following steps: Taking the start-stop cost of the micro gas turbine and the total system operation cost as the optimization objectives, constructing the objective function and constraint conditions of the integrated energy system model covering energy supply equipment, energy conversion equipment and energy storage equipment; respectively constructing the probability distribution models of photovoltaic output, wind turbine output and electric and heat loads, and performing random sampling according to different probability distribution models and constraint conditions to obtain scenarios with different probabilities of individual output of photovoltaic, wind turbine, electric load and heat load, so as to obtain the typical scenario set of the combined output of wind and light loads and their corresponding probabilities; constructing a distributionally robust optimization model, and constraining the probability distribution of the typical scenarios to obtain an uncertain probability distribution fuzzy set; decomposing the distributionally robust optimization model into a master problem and a sub-problem, and continuously iterating the master problem and the sub-problem through a column-and-constraint generation-based decomposition algorithm until the termination condition is reached, and then obtaining the optimal integrated energy scheduling scheme.

2. The integrated energy dispatch method considering the uncertainty of wind and light loads according to claim 1, characterized in that: The probability distribution model of photovoltaic output is constructed based on the beta distribution; the probability distribution model of wind turbine output is constructed based on the Weibull distribution; the probability distribution model of electric and heat loads is constructed based on the normal distribution.

3. A comprehensive energy dispatch method considering the uncertainty of wind and light loads according to claim 1, characterized in that: The objective function of the constructed integrated energy system model is as follows: Wherein, is the start-stop cost of the micro gas turbine; is the start-stop cost coefficient of the micro gas turbine; is the state variable of the micro gas turbine at time , taking a value of 0 or 1; is the total operating cost of the system; is the cost of purchasing electricity from the power grid; is the cost of purchasing gas from the gas network; is the energy storage cost; is the cost of curtailed wind; is the cost of curtailed light; is the revenue from selling electricity; is the cost of demand response.

4. A comprehensive energy dispatch method considering the uncertainty of wind and light loads according to claim 1, characterized in that: The demand response cost is introduced into the objective function corresponding to the total system operation cost; the demand response cost includes the compensation cost for shiftable loads participating in demand response, the compensation cost for transferable loads participating in demand response, and the compensation cost for curtailable loads participating in demand response.

5. The integrated energy scheduling method considering the uncertainty of wind and light loads according to claim 1, characterized in that: The specific process of obtaining the typical scenario set of the combined output of wind and light loads is as follows: Using the K-means clustering algorithm to compress the sampling scenarios, extracting the typical scenario set of individual output of photovoltaic, wind turbine, electric load and heat load, and performing Cartesian connection on these different types of scenarios, and then further reducing the scenarios using the synchronous back substitution elimination technique to obtain a typical scenario set composed of multiple groups of typical scenarios of the combined output of wind and light loads.

6. The integrated energy dispatch method considering the uncertainty of wind and light loads according to claim 5, characterized in that: The number of scenario clusters in the K-means clustering algorithm is obtained by the silhouette coefficient method.

7. The integrated energy dispatch method considering the uncertainty of wind and light loads according to claim 1, characterized in that: The probability distribution of the described typical scenarios is constrained by the integrated norm.

8. A comprehensive energy dispatch method considering the uncertainty of wind and light loads according to claim 1, characterized in that: The objective function of the master problem includes the operation cost, the compensation cost and the variables related to uncertainty.

9. The integrated energy scheduling method considering the uncertainty of wind and light loads according to claim 1, characterized in that: The specific process of iteratively solving the two-stage distributionally robust optimization model is as follows: Initializing the iteration times of the uncertain variables; respectively setting the convergence accuracy, the upper bound value and the lower bound value; setting the occurrence probability of each typical scenario to its reference probability distribution value; Solving the master problem to obtain the first-stage variables and the optimal solution, and updating the lower bound value according to the optimal solution corresponding to the master problem; Solving the sub-problem based on the first-stage variables to obtain the second-stage variables and the optimal solution, and updating the upper bound value according to the optimal solution corresponding to the sub-problem; if the difference between the upper bound value and the lower bound value is greater than the convergence accuracy, then adding the second-stage variables and the related constraint conditions to the master problem, and re-solving the master problem until the difference between the upper bound value and the lower bound value is less than or equal to the convergence accuracy, stopping the iteration, and obtaining the optimal solution.

10. An integrated energy scheduling system considering the uncertainty of wind and light loads, comprising an energy supply device, an energy conversion device, an energy storage device and a control module; the energy supply device includes a photovoltaic unit and a wind turbine unit; the energy conversion device includes an electric boiler and a combined heat and power unit; the energy storage device includes a storage battery and a heat storage tank; characterized in that: The described control module is used to execute the integrated energy scheduling method described in claim 1, and control the energy supply equipment, energy conversion equipment and energy storage equipment based on the generated optimal integrated energy scheduling scheme.

Images