An adaptive robust balancing method to adapt to the uncertainty of power supply and demand in extreme weather

By adopting an adaptive two-stage robust optimization method, the power system supply and demand balance scheduling problem is transformed into a single-layer deterministic optimization problem. By introducing an adaptive capacity gap, the problem of computational difficulties and clean energy curtailment under extreme weather conditions in traditional methods is solved, and efficient power supply and demand balance scheduling under extreme weather conditions is achieved.

CN123121589APending Publication Date: 2026-11-17SICHUAN ENERGY INTERNET RES INST TSINGHUA UNIV +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611606525.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-10-16
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Traditional power system supply and demand balancing dispatch methods are difficult to adapt to the high uncertainty on both the source and load sides under extreme weather conditions, resulting in large computational loads, highly conservative results, and serious problems of clean energy curtailment.

Method used

An adaptive interval two-stage robust optimization method is adopted. By constructing an adaptive interval robust optimization model based on confidence intervals, the two-stage robust optimization problem is transformed into a single-layer deterministic optimization problem. Upward adaptive capacity gap and downward adaptive capacity gap are introduced, and their values ​​are dynamically adjusted to adapt to the fluctuation range of uncertain parameters, thus establishing a multi-time-scale power balance optimization model.

Benefits of technology

It significantly improves computational efficiency and economy, reduces the curtailment of clean energy, balances the conservatism and economy of computational results, and adapts to power supply and demand balance scheduling under extreme weather conditions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123121589A_ABST
    Figure CN123121589A_ABST
Patent Text Reader

Abstract

This invention relates to the field of power system optimization technology and discloses an adaptive robust balancing method for power supply and demand uncertainties under extreme weather conditions. The method includes: firstly, decomposing the uncertain parameters into load-side and power-side components within a confidence interval, employing two extreme scenarios to cover the fluctuation range, thus transforming the two-stage robust optimization into a single-layer deterministic optimization problem; secondly, introducing upward and downward adaptive capacity gaps, relaxing the supply and demand balance constraints, and dynamically adjusting the objective function to construct an adaptive interval robust optimization model. This model is then embedded into a multi-timescale scheduling framework to establish a two-layer optimization model for annual power balance and pre-week power balance, solved using a cyclic iterative bidirectional coupling method to obtain the generator output plan and the supply and demand balance scheduling scheme. This invention balances the economy and robustness of the optimization results, significantly reduces computational complexity, and is suitable for supply and demand balance scheduling of high-proportion clean energy power systems under extreme weather conditions.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of power system optimization technology, and in particular to an adaptive robust balancing method for adapting to the uncertainty of power supply and demand during extreme weather. Background Technology

[0002] Compared to traditional thermal power-dominated power grids, the new power system exhibits typical characteristics of high cleanliness, high weather dependence, and strong uncertainty on both the power source and load sides: On the power source side, the output of clean energy, mainly hydropower, wind power, and photovoltaic power, is highly dependent on weather conditions and has weak stability. New extreme droughts lead to a sharp reduction in water inflow and a significant decrease in hydropower output. Continuous rain or dust storms prevent photovoltaic power generation. The flood discharge pressure brought by continuous heavy rain may be accompanied by a large amount of water and electricity being wasted by hydropower stations. On the load side, abnormal temperature fluctuations will directly cause drastic changes in temperature-controlled loads. Extreme high temperatures in summer push up cooling electricity consumption, while low temperatures and cold waves in winter surge heating loads, causing unpredictable and drastic fluctuations in the system's electricity demand.

[0003] The new, standardized power system structure makes supply and demand balance highly dependent on weather conditions. Under extreme weather conditions, imbalances can easily occur due to the reverse coupling between power output and load, seriously affecting the safety and economy of the power system. The traditional deterministic supply and demand balance dispatch and control model that relies on thermal power reserve peak shaving is no longer suitable for the structural and meteorological coupled supply and demand uncertainty risks of the new power system. There is an urgent need to develop a new power system supply and demand balance dispatch method to cope with extreme weather processes.

[0004] Currently, a large amount of research has been conducted on the problem of power supply and demand balance scheduling. To address uncertainties on both the supply and demand sides, the mainstream optimization scheduling methods fall into two main categories: stochastic optimization and robust optimization.

[0005] Multi-scenario stochastic optimization characterizes the uncertainty of source and load using a finite probability scenario tree and adopts multi-stage decision-making to achieve the optimal expected cost. It can explicitly characterize temporal correlation and joint probability of source and load. However, this method is highly dependent on the quality of the scenario. For high-dimensional scenarios of power systems, the computational load is extremely large, and its reliability under extreme weather conditions is questionable.

[0006] The advantage of robust optimization lies in its simplicity and reliability. However, the uncertainty set of traditional robust optimization methods is mainly based on box uncertainty set, which takes into account the worst-case scenario with an extremely low probability of occurrence. This leads to overly conservative optimization results, which seriously affects the economy of normal system operation.

[0007] Based on this, the two-stage robust optimization method was developed. As a branch of robust optimization, it is naturally suited to the needs of multi-timescale supply and demand balance scheduling in power systems, and has become the preferred method for studying medium- and long-term supply and demand balance scheduling problems in recent years.

[0008] The typical mathematical model for a two-stage robust optimization problem is as follows:

[0009] ;

[0010] In the formula: x is the first-stage decision variable vector, y is the second-stage adjustment decision variable vector, ξ is the uncertain parameter vector, Ξ is the set of values ​​for the uncertain parameter; m and n are the objective function coefficient vectors, A, B, C, and D are constraint coefficient matrices, and e and h are constraint right-hand side vectors.

[0011] This model uses a three-layer optimization structure of min-max-min to obtain the optimal decision under the worst-case uncertainty scenario. The first term m of the model's objective function... T The first term, x, does not contain random variables and can be solved using conventional optimization methods. The second term, however, contains the uncertain parameter ξ, which influences the decision variables x and y through constraints. The nested optimization structure makes the solution extremely difficult, and strictly following the above formula for mathematical solution is very challenging. Currently, mainstream algorithms utilize Benders decomposition or the column sum and constraint generation algorithm (CC&G) to obtain exact solutions, but these do not fundamentally solve the problem of high solution difficulty. Furthermore, overly stringent characterization of the uncertainty set Ξ and the difficulty in accurately defining its range can also lead to excessively conservative optimization results. Summary of the Invention

[0012] The purpose of this invention is to overcome the shortcomings of the prior art and provide an adaptive robust balancing method that adapts to the uncertainty of power supply and demand in extreme weather.

[0013] The objective of this invention is achieved through the following technical solution: an adaptive robust balancing method for power supply and demand uncertainties in extreme weather, comprising the following steps:

[0014] S1. Obtain the operating parameters of the power system, including load forecast data, power supply side forecast data, and operating constraint parameters of each generator set;

[0015] S2. Construct an adaptive interval two-stage robust optimization model based on confidence intervals. The method for constructing the model includes:

[0016] S21. Determine the confidence region for the uncertain parameter at a given confidence level;

[0017] S22. Decompose the uncertain parameters into load-side uncertain parameters and power-side uncertain parameters, and use two extreme cases to cover the fluctuation range of the uncertain parameters. The two extreme cases are: the load value is at its maximum and the power value is at its minimum, and the load value is at its minimum and the power value is at its maximum.

[0018] S23. Convert the two-stage robust optimization problem into a single-level deterministic optimization problem, where the objective function of the single-level deterministic optimization problem is to minimize the sum of the cost of the first-stage decision variables and the cost of the second-stage decision variables;

[0019] S24. Extract the power supply and demand balance constraint condition from the single-level deterministic optimization problem, and calculate the maximum supply-demand imbalance based on the power supply and demand balance constraint condition;

[0020] S25. Introduce an upward adaptive capacity gap and a downward adaptive capacity gap to relax the power supply and demand balance constraint condition, multiply each of the upward adaptive capacity gap and the downward adaptive capacity gap by a corresponding weight coefficient and add them to the objective function, and dynamically adjust the values of the upward adaptive capacity gap and the downward adaptive capacity gap through the optimization process of the objective function, so as to automatically adapt to the fluctuation interval of uncertain parameters and obtain an adaptive interval robust optimization model;

[0021] S3. Based on the adaptive interval robust optimization model, establish a multi-time-scale power and electricity balance optimization model for the power system under extreme weather, and solve the multi-time-scale power and electricity balance optimization model to obtain the output plan of each generator set and the power supply and demand balance scheduling scheme.

[0022] Specifically, the step of converting the two-stage robust optimization problem into a single-level deterministic optimization problem is:

[0023] ;

[0024] In the formula, F is the objective function, x is the vector of first-stage decision variables; y is the vector of second-stage adjusted decision variables; m and n are the coefficient vectors of the objective function, and the superscript T thereon represents transpose; A, B, C and D are constraint coefficient matrices; e and h are vectors of right-hand terms of constraints, , , respectively represent the expected value and the upper and lower fluctuation intervals of the load-side uncertain parameters at a given confidence level K (0<K≤1), and similarly , , respectively represent the expected value and the upper and lower fluctuation intervals of the power supply-side uncertain parameters at the given confidence level K.

[0025] Specifically, the power supply and demand balance constraint condition is:

[0026] ;

[0027] In the formula, P G is the total output of conventional deterministic generator sets; is the positive standby power capacity reserved by the system; The negative backup power capacity reserved for the system; P lack P represents the maximum possible load gap. waste The maximum possible power surplus; D G For the power supply side uncertain parameter ξ G Factors affecting power output;

[0028] The positive and negative backup power capacity are calculated as follows:

[0029] ;

[0030] In the formula, P g (i) represents the current output of the unit; and These are the unit's rated maximum and minimum output, respectively; This is the positive backup power capacity; This is the negative backup power capacity.

[0031] Specifically, the relaxation of the power supply and demand balance constraint is as follows:

[0032] ;

[0033] In the formula, To provide an upward adaptive capacity gap; This is a downward adaptive capacity gap;

[0034] The relaxed electricity supply and demand balance constraint is decomposed as follows:

[0035] ;

[0036] By weighting the upward adaptive capacity gap and the downward adaptive capacity gap and adding them to the objective function, we obtain the adaptive interval robust optimization model:

[0037] ;

[0038] In the formula, F back To adapt to capacity gap risk; λ lack λ is the load power shortage loss coefficient; waste The cost per kilowatt-hour of clean energy curtailment; λ + and λ - These are the upward and downward adaptive capacity gap risk weight coefficients for the system, with values ​​ranging from 0 to 1.

[0039] Specifically, the multi-timescale power balance optimization model includes an annual power balance model and a weekly power balance model. The annual power balance model arranges the annual production plan with a weekly scheduling cycle, while the weekly power balance model arranges the power generation plan for one week with an hourly scheduling cycle.

[0040] Specifically, the objective functions of the annual power balance model and the pre-week power balance model are:

[0041] ;

[0042] In the formula, F1 is the load power shortage loss; F2 is the cost of clean energy curtailment; F3 is the unit power generation cost; F back To mitigate the risk of adaptive capacity gaps;

[0043] ;

[0044] In the formula, nt is the number of scheduling cycles; T is the cycle duration;

[0045] ;

[0046] In the formula, F w F s F h These represent the total costs of wind curtailment, solar curtailment, and hydropower curtailment, respectively. The maximum output capacity of the i-th wind or solar power unit during time period t; Q represents the actual output of the i-th wind or solar power unit during time period t; waste (i,t) represents the water discharge of the i-th hydropower station during time period t; Ω w Ω s Ω h Let represent the collection of wind turbines, photovoltaic units, and hydropower stations; h(i) and η(i) represent the average head and power generation efficiency of the hydropower station, respectively; ρ is the density of water; and g is the gravitational acceleration constant.

[0047] ;

[0048] In the formula, F th and These are the power generation costs of thermal power units and energy storage power stations, respectively. , They are collections of thermal power units and energy storage power stations; λ th The unit cost of electricity generation for thermal power units; λ sto The cost per unit of electricity consumed by the energy storage unit; P sto (i,t) represents the power consumption of the energy storage power station when it is in energy storage mode; η sto (i) and η g (i) represents the energy conversion efficiency of the energy storage power station in the two processes of energy storage and power generation.

[0049] Specifically, the constraints of the annual power balance model and the pre-week power balance model include:

[0050] System power balance constraints:

[0051] ;

[0052] In the formula, The predicted load for time period t;

[0053] Generator output upper and lower limit constraints:

[0054] ;

[0055] In the formula, S g (i,t) is a 0-1 variable, representing the i-th... The operating status of each unit during time period t, with 1 indicating that it is operating; This is the upper limit of the unit's output. This represents the lower limit of the unit's output. This represents the total number of weeks the unit is out of service within a year.

[0056] Water and electricity constraints:

[0057] ;

[0058] ;

[0059] In the formula, V(i,t) is the first... The water storage capacity of each reservoir at the end of time period t; Q in (i,t) and Q out (i,t) represent the average inflow and outflow of the reservoir during this period, respectively; Q(i,t) represents the water flow for power generation at the hydropower station. and These are the upper and lower limits of the reservoir's water storage capacity, respectively. and These are the upper and lower limits of the reservoir's discharge capacity, respectively.

[0060] Spatiotemporal coupling constraints of flow rates in cascade hydropower stations:

[0061] ;

[0062] In the formula, The average natural runoff of the i-th reservoir during time period t is the portion of the flow that is not controlled by the upstream reservoir, consisting of regional precipitation, domestic drainage, and tributaries within the territory. Let t be the set of upstream reservoirs of the i-th reservoir; lag (j,i) represents the water flow from the j-th hydropower station to the j-th hydropower station. The delay time required for each hydroelectric power station;

[0063] Constraints of energy storage units:

[0064] ;

[0065] In the formula, E(i,t) is the first... The energy stored in a single energy storage power station at the end of time period t; and These represent the upper and lower limits of the energy storage capacity of the power station during that time period; and These represent the upper and lower limits of the power storage capacity of the power station during that time period; S sto (i,t) is a 0-1 variable, representing the state of the energy storage power station during that time period, with 1 indicating that it is storing energy.

[0066] Specifically, the constraints of the pre-week power balance model also include:

[0067] Meet the boundary conditions specified in the annual production plan:

[0068] ;

[0069] In the formula, Ω S=0 For units that need to be shut down this week according to the annual production plan; V sta (i) and V end (i) are respectively the first The water levels in the reservoirs at the beginning and end of this week; E sta (i) and E end (i) are respectively the first The energy storage capacity of an energy storage station at the beginning and end of the week; , , , The first The water storage capacity of the reservoir at the initial and final moments, and the water storage capacity of the first reservoir at the last moment. The energy stored in a power storage station at the initial and final moments.

[0070] Thermal power unit output ramp-up rate constraint:

[0071] ;

[0072] In the formula, r(i) is the ramp rate of the i-th unit; T int (i) is a non-negative integer variable, representing the i-th... The time required for an energy storage power station to switch between energy storage and power generation states; t gap As a temporary variable, it represents any variable not greater than T in the formula. int (i) are natural numbers, and N represents the set of natural numbers.

[0073] Specifically, the solution method for the multi-timescale power balance optimization model is as follows:

[0074] S310. Collect the data required for the annual power balance model and preprocess it, and preset the core indicators and the upper limit of the number of iterations for optimization calculation;

[0075] S320. Solve the annual time-scale power balance model, formulate the annual production plan, and output the annual decision variables. The annual decision variables include the start-up and shutdown arrangements of each generator unit and the water storage plans of each large reservoir, forming the boundary conditions of the weekly power balance model.

[0076] S330. Collect this week's power supply and load forecast data, and combine them with the boundary conditions determined by the annual production plan to solve the pre-week power balance model.

[0077] S340. Multiply the deviation between the reservoir water storage and energy storage capacity in the previous week's power balance model and the annual production plan by a penalty coefficient and add it to the objective function, and feed the deviation result back to the annual production plan for adjustment.

[0078] S350. Calculate the load gap, clean energy curtailment rate and power generation cost core indicators based on the solution results, and determine whether the preset requirements are met.

[0079] S360. If the requirements are not met and the number of iterations has not reached the upper limit, then the boundary conditions of the power balance model from the previous week are corrected according to the core indicators that are not met, and S330 to S350 are repeated for iterative calculation until the requirements are met or the maximum number of iterations is reached.

[0080] S370: Output this week's production plan and arrange the actual output curves of each power supply.

[0081] The present invention has the following advantages:

[0082] 1. The adaptive interval two-stage robust optimization method proposed in this invention is a compromise algorithm between existing two-stage robust optimization methods and traditional methods (without considering uncertainty). It can balance the economy and conservatism of the calculation results, while significantly improving the problem of clean energy curtailment. Power supply and demand balance scheduling based on the method proposed in this invention achieves the lowest total cost in most scenarios. In extreme high-temperature scenarios, the total cost is slightly higher than existing two-stage robust methods, but far lower than traditional methods.

[0083] 2. Compared with the existing two-stage robust optimization method, the power supply and demand balance method proposed in this invention greatly reduces the calculation time under the same conditions and significantly improves the calculation efficiency, meeting the needs of rapid iterative calculation of multivariable parameters in power systems. Attached Figure Description

[0084] Figure 1 This is a schematic diagram of the adaptive robust balancing method of the present invention;

[0085] Figure 2 This is a typical diagram illustrating the watershed relationship of a cascade hydropower station.

[0086] Figure 3 A schematic diagram of load and clean energy resource normalization curves;

[0087] Figure 4 Schematic diagram of load curves for different scenarios;

[0088] Figure 5 Schematic diagram of normalized curves for natural runoff flow in different scenarios;

[0089] Figure 6 This is a schematic diagram of the annual simulation statistics of the adaptive robust balancing method of the present invention in an ideal scenario;

[0090] Figure 7 This is a schematic diagram illustrating the statistical data of a full-year simulation of an existing robust method in an ideal scenario.

[0091] Figure 8 This is a schematic diagram illustrating the statistical data of a year-round simulation using traditional methods in an ideal scenario.

[0092] Figure 9 This is a schematic diagram of the annual simulation statistics of the adaptive robust balancing method of the present invention in a normal scenario;

[0093] Figure 10 This is a schematic diagram illustrating the annual simulation statistics of existing robust methods in a typical scenario.

[0094] Figure 11 This is a schematic diagram illustrating the statistical data of a year-round simulation in a typical scenario using traditional methods.

[0095] Figure 12 This is a schematic diagram of the annual simulation statistics of the adaptive robust balancing method of the present invention in a high-temperature scenario;

[0096] Figure 13 This is a schematic diagram illustrating the annual simulation statistics of existing robust methods in high-temperature scenarios.

[0097] Figure 14 This is a schematic diagram illustrating the annual simulation statistics of traditional methods in high-temperature scenarios.

[0098] Figure 15 This is a schematic diagram of the annual simulated operation statistics of the adaptive robust balancing method of the present invention in a flood-prone scenario;

[0099] Figure 16 This is a schematic diagram illustrating the annual simulation statistics of existing robust methods in a high-water scenario;

[0100] Figure 17 This is a schematic diagram illustrating the annual simulation statistics of traditional methods in a high-water scenario.

[0101] Figure 18 This is a schematic diagram of the load gap curves for different methods under high-temperature scenarios. Detailed Implementation

[0102] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only for explaining the invention and are not intended to limit the invention; that is, the described embodiments are merely some embodiments of the invention, and not all embodiments. The components of the embodiments of the invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0103] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0104] It should be noted that relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0105] The present invention will be further described below with reference to the accompanying drawings, but the scope of protection of the present invention is not limited to the following description.

[0106] like Figures 1 to 18 As shown, an adaptive robust balancing method for power supply and demand uncertainties in extreme weather conditions is proposed. This method includes the following steps:

[0107] S1. Obtain the operating parameters of the power system, including load forecast data, power supply side forecast data, and operating constraint parameters of each generator set;

[0108] S2. Construct an adaptive two-stage robust optimization model based on confidence intervals;

[0109] The mathematical model of a typical two-stage robust optimization problem is as follows:

[0110] ;

[0111] where x is the vector of first-stage decision variables; y is the vector of second-stage adjusted decision variables; ξ is the vector of uncertain parameters; Ξ is the value set of the uncertain parameters; m and n are the coefficient vectors of the objective function; A, B, C and D are the constraint coefficient matrices; e and h are the right-hand side vectors of the constraints;

[0112] This model obtains the optimal decision under the worst uncertain scenario through a min-max-min three-layer optimization structure. The first term m of the objective function of the model T x is the part that does not contain random variables, which can be solved by a conventional optimization method; the second term contains the uncertain parameter ξ, which affects the decision variables x and y through constraint conditions. The nested optimization structure leads to extremely high solution difficulty, and it is very difficult to perform mathematical solving strictly according to the above formula. At present, the mainstream solving algorithms include using the Benders decomposition method or the Column-and-Constraint Generation algorithm (CC&G) to obtain an exact solution, but they do not fundamentally solve the problem of high solution difficulty. In addition, problems such as overly strict characterization of the uncertainty set Ξ and difficulty in accurately defining the range will also lead to excessively conservative optimization results. In order to solve the problems of excessive solution difficulty and overly conservative results of the existing two-stage robust optimization method, in this embodiment, an adaptive interval two-stage robust optimization method is used for uncertainty analysis of power supply and demand under extreme weather, which specifically includes:

[0113] S21. Determine the confidence region of the uncertain parameter ξ under a given confidence level K (0<K≤1):

[0114] ;

[0115] where, is the expected value of the uncertain parameter ξ; ξ K+ and ξ K- are respectively the upper and lower fluctuation intervals under this confidence level;

[0116] S22. Decompose the uncertain parameters into load-side uncertain parameters and power supply-side uncertain parameters, and use two extreme scenarios to cover the fluctuation interval of the uncertain parameters. The two extreme scenarios are: the maximum load value while the power supply value is the minimum, and the minimum load value while the power supply value is the maximum;

[0117] When all the constraint conditions are linear inequalities, as long as the inequalities hold when the uncertain parameter ξ takes values on the boundary, the inequalities will hold within the entire confidence interval. Therefore, the two-stage robust optimization problem of the mathematical model of the typical two-stage robust optimization problem is transformed into the following problem:

[0118] ;

[0119] In the formula, As a temporary variable, it represents any boundary value of ξ in the formula;

[0120] To address the curse of dimensionality caused by numerous uncertain parameters in power supply and demand analysis, this embodiment employs certain engineering simplifications and approximations. The uncertain parameters consist of two parts: load and power source. Let ξ = [ξ...]. load ; ξ G ], where ξ load and ξ G These are the uncertain parameters on the load and power supply sides, respectively. At confidence level K, the second constraint condition in the above equation can be approximately simplified to the following two inequalities:

[0121] ;

[0122] In the formula, only two cases are considered: the load value is at its maximum and the power supply value is at its minimum, and the load value is at its minimum and the power supply value is at its maximum. The load side uncertainty parameter is taken at the upper boundary of the confidence interval to maximize the load prediction value, while the power supply side uncertainty parameter is taken at the lower boundary of the confidence interval to minimize the power supply generation capacity. This case represents the maximum power supply gap risk faced by the system.

[0123] The second extreme case is that the load-side uncertain parameters are taken at the lower boundary of the confidence interval to minimize the load forecast value, while the power-side uncertain parameters are taken at the upper boundary of the confidence interval to maximize the power generation capacity of the power source. This case represents the maximum risk of clean energy consumption faced by the system. This is a conservative approximation in engineering, assuming that the random fluctuations of the source and load are independent or negatively correlated. This assumption is possible under extreme weather conditions.

[0124] S23. Transform the two-stage robust optimization problem into a single-level deterministic optimization problem. The objective function of this single-level deterministic optimization problem is to minimize the sum of the costs of the decision variables in the first stage and the decision variables in the second stage.

[0125] At this point, the constraints no longer contain the uncertain parameter ξ, so the objective function can be further simplified as follows:

[0126] ;

[0127] Therefore, the original optimization problem is transformed into a single-layer structure optimization problem containing only determined parameters:

[0128] ;

[0129] where F is the objective function; x is the vector of first-stage decision variables; y is the vector of second-stage adjusted decision variables; m and n are the coefficient vectors of the objective function, and the superscript T thereon denotes transposition; A, B, C, D are constraint coefficient matrices; e and h are the right-hand side vectors of constraints, , , are respectively the expected value and the upper and lower fluctuation intervals of the load-side uncertain parameter at a given confidence level K (0<K≤1); , , are respectively the expected value and the upper and lower fluctuation intervals of the power supply-side uncertain parameter at the given confidence level K. In this problem, y no longer has the meaning of adjusting the decision after the uncertainty is realized in the two-stage robust optimization, and needs to satisfy all possibilities (covered by two extreme cases), and this problem can be solved by a conventional optimization method.

[0130] S24. Extract the power supply and demand balance constraint condition from the single-layer deterministic optimization problem, and calculate the maximum supply-demand imbalance amount based on the power supply and demand balance constraint condition;

[0131] The power supply and demand balance constraint condition is:

[0132] ;

[0133] where P G is the total output of conventional deterministic generating units; is the positive backup power capacity reserved by the system; is the negative backup power capacity reserved by the system; P lack is the maximum possible load gap; P waste is the maximum possible power surplus; D G is the influence factor of the power supply-side uncertain parameter ξ G on the power supply output.

[0134] The positive and negative backup power capacities reserved by the system are only related to the dispatch operation mode and are deterministic parameters. On a medium-to-long time scale, restrictions such as the unit ramp rate and operation mode switching are not considered, and the positive and negative backup power capacities of each unit are determined by its rated maximum or minimum output and the current output, and the positive backup power capacity and the negative backup power capacity are calculated as:

[0135] ;

[0136] where P g (i) is the current output of the unit; and are respectively the rated maximum and minimum output of the unit; the sum of the positive and negative backup power capacities of all units that can participate in short-term peak shaving is the positive and negative backup power capacity of the system; This is the positive backup power capacity; This is the negative backup power capacity.

[0137] The maximum output power of wind power and photovoltaic power can be directly given by the prediction data of the uncertain parameter ξ, and the corresponding influence factor is 1.

[0138] The uncertainty of the average generating capacity of hydropower over a long period is mainly determined by the natural runoff, which is not controlled by upstream reservoirs. Natural runoff can also be directly obtained from forecast data; therefore:

[0139] ;

[0140] In the formula, Ω h Let represent the set of hydropower stations; h(i) and η(i) are the average head height and power generation efficiency of the hydropower station, respectively; ρ is the density of water; g is the gravitational acceleration constant. It should be noted that the following simplifications have been made to the hydropower model: changes in reservoir head height are ignored, assuming the head height of the same hydropower station remains essentially constant; the influence of other factors on the power generation efficiency of the hydropower station is ignored, assuming a linear relationship between the power generation output and water flow of the same hydropower station; and reservoir capacity loss caused by natural evaporation and seepage is ignored. Considering that the error caused by the above simplifications is far less than the source-load prediction deviation, it is acceptable. Taking Southwest my country as an example, the average precipitation in the past five years differs by nearly double between relatively dry and rainy years, causing a huge prediction deviation. Hydropower stations are mainly run-of-river stations and short-term regulating stations, and the seasonal periodic fluctuations in the water storage of large reservoirs are obvious. Water storage plans are also constrained by flood control, navigation, irrigation, and ecology, resulting in a smaller deviation of the head height from the multi-year operating curve, thus leading to a relatively smaller error.

[0141] S25. Introduce upward adaptive capacity gap and downward adaptive capacity gap to relax the power supply and demand balance constraint. Multiply the upward adaptive capacity gap and downward adaptive capacity gap by their respective weight coefficients and add them to the objective function. Dynamically adjust the values ​​of the upward adaptive capacity gap and downward adaptive capacity gap through the optimization process of the objective function to automatically adapt to the fluctuation range of uncertain parameters and obtain an adaptive range robust optimization model.

[0142] Considering the possibility of extreme weather, the uncertain parameters on both the load and power supply sides fluctuate greatly, which, reflected in the power supply and demand balance constraints, easily leads to P... lack or P wasteThe value is very large. However, in reality, the probability of the extreme case where both load and power supply simultaneously reach the interval boundary is extremely small. It is clearly inappropriate to achieve robustness within the fluctuation range without considering the grid's carrying capacity. Therefore, to relax the model constraints and automatically adapt to the supply and demand uncertainty fluctuation range, two non-negative physical quantities are introduced: upward adaptive capacity gap ΔP. + and the extreme downward adaptive capacity gap ΔP - This reflects the difference between the system's reserved backup power capacity and the capacity demand caused by source load fluctuations. This difference is added to the power supply and demand balance constraints to relax the constraints.

[0143] ;

[0144] In the formula, To provide an upward adaptive capacity gap; This is a downward adaptive capacity gap;

[0145] The relaxed electricity supply and demand balance constraint is decomposed as follows:

[0146] ;

[0147] As can be seen, the first inequality is the power balance formula based on the expected value of source load prediction, corresponding to the traditional power balance formula without considering probabilistics. The following two equations are the calculation formulas for upward and downward adaptive capacity gaps.

[0148] The upward adaptive capacity gap and the downward adaptive capacity gap are weighted and added to the objective function. ΔP is then dynamically adjusted through the optimization process of the objective function. + and ΔP - The value of is used to influence the system's decision variables in turn, so as to adapt to the larger confidence interval under extreme weather conditions. The adaptive interval robust optimization model is as follows:

[0149] ;

[0150] In the formula, F back To adapt to capacity gap risk; λ lack The load power shortage loss coefficient, i.e., the economic loss per unit of electricity, can be analyzed and estimated based on local economic data and historical power outage data; λ waste The cost per kilowatt-hour of clean energy curtailment; λ + and λ - These are the upward and downward adaptive capacity gap risk weight coefficients for the system, with values ​​ranging from 0 to 1. Essentially, they balance between economic efficiency and conservatism, with values ​​closer to 1 indicating a greater bias towards conservatism.

[0151] It can be seen that, due to the introduction of F into the objective function backDuring the solution process, this model will try its best to meet the system's backup capacity requirements under extreme weather conditions, so as to maximize ΔP + and ΔP - As small as possible. When F back When λ = 0, this model is equivalent to the model described in the single-layer structure optimization problem, and is reduced to a traditional robust optimization model. Meanwhile, due to the risk weight coefficient λ... + and λ - A lower-than-actual power outage and curtailment loss coefficient can increase ΔP when the grid's reserve capacity is insufficient. + and ΔP - By narrowing the robust interval range and prioritizing the demand for deterministic loads and clean energy consumption, automatic adaptation to the confidence interval is achieved.

[0152] S3. Based on the aforementioned adaptive interval robust optimization model, a multi-timescale power balance optimization model for the power system under extreme weather conditions is established. This multi-timescale power balance optimization model is then solved to obtain the output plan for each generating unit and the power supply and demand balance dispatch scheme. Extreme weather profoundly affects the source-load characteristics of the new power system. Its most significant feature is breaking traditional temporal patterns, exacerbating randomness, and triggering cascading failures. Therefore, when establishing a power balance model, it is necessary to shift from deterministic programming to probabilistic programming, taking extreme weather conditions into account. Since the impact of extreme weather has significant temporal heterogeneity, a single-timescale model cannot meet the requirements, necessitating the establishment of a multi-timescale model. This embodiment uses the annual and weekly timescales as examples to establish an annual power balance model and a weekly power balance model.

[0153] The annual power balance model schedules the annual production plan using a one-week cycle.

[0154] Under extreme weather conditions, the power grid faces multiple possibilities, requiring a trade-off between security and economy, and between clean energy curtailment and load shortages, such as allowing moderate shortages to reduce excessive clean energy curtailment. To achieve these requirements, a multi-objective optimization method is adopted, with the model's objective functions being to minimize load shortages, minimize renewable energy curtailment, and minimize generation costs. By assigning appropriate coefficients to the weights of each objective function, it is transformed into a cost that can be comprehensively optimized, namely:

[0155] ;

[0156] In the formula, F1 is the load power shortage loss; F2 is the cost of clean energy curtailment; F3 is the unit power generation cost; F backTo address adaptive capacity gap risk, this model considers the overall socio-economic aspects, incorporating both the socio-economic losses caused by power outages and the expected losses potentially caused by adaptive capacity gap risk into the overall cost, thus providing a higher level of decision-making. Unless otherwise specified, all costs referred to below are broadly defined costs encompassing the concepts mentioned above.

[0157] ;

[0158] In the formula, nt is the number of scheduling cycles, which is 52 cycles for the annual power balance model; T is the cycle duration, which is one week for the annual model, i.e., 604,800 s;

[0159] ;

[0160] In the formula, F w F s F h These represent the total costs of wind curtailment, solar curtailment, and hydropower curtailment, respectively. The maximum output capacity of the i-th wind or solar power unit during time period t; Q represents the actual output of the i-th wind or solar power unit during time period t; waste (i,t) represents the water discharge of the i-th hydropower station during time period t; Ω w Ω s Ω h Let represent the collection of wind turbines, photovoltaic units, and hydropower stations, respectively; h(i) and η(i) are the average head and power generation efficiency of the hydropower station, respectively; ρ is the density of water; and g is the gravitational acceleration constant.

[0161] ;

[0162] In the formula, F th and These are the generation costs of thermal power units and energy storage power stations, respectively. The generation cost of thermal power is approximately considered to be proportional to the amount of electricity generated. , They are collections of thermal power units and energy storage power stations; λ th The unit cost of electricity generation for thermal power units is λ. The cost of electricity generation for energy storage power stations stems from the energy conversion efficiency in the energy storage and generation processes, and the cost is directly proportional to the amount of electricity lost. sto The cost per unit of electricity consumed by the energy storage unit; P sto (i,t) represents the power consumption of the energy storage power station when it is in energy storage mode; η sto (i) and η g (i) represents the energy conversion efficiency of the energy storage power station in the two processes of energy storage and power generation.

[0163] The pre-week power balance model schedules a week's power generation plan with an hourly scheduling cycle, while the weekly time-scale power balance model, based on more accurate source-load forecast data, schedules a week's power generation plan and potential load shedding plans with an hourly (or shorter time scale) scheduling cycle. The optimization objective function of the pre-week power balance model is basically the same as that of the annual power balance model, only requiring corresponding changes to the number of scheduling cycles and cycle duration.

[0164] The key feature is the uncertainty on both the source and load sides; therefore, the method of this invention focuses on the overall balance of power supply and demand, without considering grid structure and transmission flow. The constraints of the annual power balance model and the pre-week power balance model include:

[0165] With system power balance constraints, ignoring line losses, we have:

[0166] ;

[0167] In the formula, The predicted load for time period t;

[0168] The upper and lower limits of generator unit output are constrained by economic considerations or maintenance needs, typically involving the shutdown of some generator units during certain periods of the year. These constraints are determined through optimization within an annual time-scale power balance model. Considering the unit shutdown status, the constraints on the upper and lower limits of generator unit output are as follows:

[0169] ;

[0170] In the formula, S g (i,t) is a 0-1 variable, representing the i-th... The operating status of each unit during time period t, with 1 indicating that it is operating; This is the upper limit of the unit's output. For thermal and hydropower units, it is the rated power, and for wind and photovoltaic units, it is the maximum output power given by the predicted data. The lower limit of unit output is generally 40% to 60% of the rated power for thermal power units for economic reasons, while it is 0 for wind power, photovoltaic and hydropower units. This refers to the total number of weeks the unit will be out of service within a year, which is usually determined in advance by the unit to which the equipment belongs, based on factors such as maintenance and upgrade needs.

[0171] Water and electricity constraints:

[0172] ;

[0173] Taking into account the needs of flood control, navigation, irrigation, and ecology under extreme weather conditions, the maximum and minimum water storage capacity and discharge capacity of the reservoir are strictly limited, and the upper and lower limits change dynamically throughout the year:

[0174] ;

[0175] In the formula, V(i,t) is the first... The water storage capacity of each reservoir at the end of time period t; Q in (i,t) and Q out (i,t) represent the average inflow and outflow of the reservoir during this period; Q(i,t) represents the water flow for power generation at the hydropower station, i.e., the average amount of water used for power generation per second. and These are the upper and lower limits of the reservoir's water storage capacity, respectively. and These are the upper and lower limits of the reservoir's discharge capacity, respectively.

[0176] Spatiotemporal coupling constraints on flow at cascade hydropower stations: In power systems with a high proportion of hydropower, hydropower stations typically occur in upstream and downstream cascade hydropower groups, with runoff between upstream and downstream stations influencing each other spatiotemporally. Therefore, runoff correlation constraints for cascade hydropower stations are introduced to reflect the cascading effects of upstream and downstream reservoirs under extreme weather conditions. A typical watershed relationship for a cascade hydropower station is as follows: Figure 2 As shown, the inflow to the downstream hydropower station consists of the discharge from the nearest upstream hydropower station and the natural runoff of the basin between the two stations. If tributaries exist, the nearest hydropower station on all tributaries must be considered. Due to spatial differences, the discharge from the upstream hydropower station takes a considerable amount of time to reach the downstream station. Therefore, the spatiotemporal coupling relationship between the flows of the cascade hydropower stations is as follows:

[0177] ;

[0178] In the formula, The average natural runoff of the i-th reservoir during time period t is the portion of the flow that is not controlled by the upstream reservoir, consisting of regional precipitation, domestic drainage, and tributaries within the territory. Let t be the set of upstream reservoirs of the i-th reservoir; lag (j,i) represents the time delay required for water to flow from the j-th hydropower station to the i-th hydropower station;

[0179] Constraints of energy storage units: Different types of energy storage power stations have different energy storage media and working mechanisms, but from the perspective of system power balance, they can share a common energy storage-discharge model, with the following constraints:

[0180] ;

[0181] In the formula, E(i,t) is the first... The energy stored in a single energy storage power station at the end of time period t; and These represent the upper and lower limits of the energy storage capacity of the power station during that time period; and These represent the upper and lower limits of the power storage capacity of the power station during that time period; S sto (i,t) is a 0-1 variable, representing the state of the energy storage power station during that time period, with 1 indicating that it is storing energy.

[0182] In addition to satisfying all the constraints in the annual power balance model, the pre-week power balance model also requires the following constraints:

[0183] Meet the boundary conditions specified in the annual production plan:

[0184] ;

[0185] In the formula, Ω S=0 For units that need to be shut down this week according to the annual production plan; V sta (i) and V end (i) are respectively the first The water storage levels of the reservoirs at the beginning of this week (i.e., last weekend) and the end of this week; E sta (i) and E end (i) are respectively the first The energy storage capacity of an energy storage station at the beginning and end of the week; , , , The first The water storage capacity of the reservoir at the initial and final moments, and the water storage capacity of the first reservoir at the last moment. The energy stored in a single energy storage power station at the initial and final moments;

[0186] Furthermore, due to the significantly higher rates of wind and solar power output and load fluctuations in the new power system under extreme weather conditions compared to normal conditions, it is necessary to consider the limitations on the ramp-up rate of thermal power unit output, as well as the impact of the inability of energy storage power stations to quickly switch operating states. The constraints on the ramp-up rate of thermal power unit output are as follows:

[0187] ;

[0188] In the formula, r(i) is the ramp rate of the i-th unit; T int (i) is a non-negative integer variable, representing the i-th... The time required for an energy storage power station to switch between energy storage and power generation states; t gap As a temporary variable, it represents any variable not greater than T in the formula. int (i) are natural numbers, and N represents the set of natural numbers.

[0189] The solution method for the multi-timescale power balance optimization model is as follows:

[0190] S310. Collect the data required for the annual power balance model and preprocess it, and preset the core indicators (such as load shedding rate, clean energy curtailment rate, etc.) and the upper limit of the number of iterations for optimization calculation;

[0191] S320. Solve the annual time-scale power balance model, formulate the annual production plan, preliminarily formulate the output curves of each generator unit based on the current uncertainty parameter prediction range, and output the annual decision variables. The annual decision variables include the start-up and shutdown arrangements of each generator unit and the water storage plans of each large reservoir, forming the boundary conditions of the weekly power balance model.

[0192] S330. Collect this week's power supply and load forecast data, and combine them with the boundary conditions determined by the annual production plan to solve the pre-week power balance model.

[0193] S340. The deviation between the reservoir water storage and energy storage capacity in the previous week's power balance model and the annual production plan is multiplied by a penalty coefficient and added to the objective function. The deviation result is then fed back into the annual production plan for adjustment. Considering the complexity of extreme weather conditions, strictly adhering to the boundary conditions output by the annual plan may lead to an unsolvable problem. To increase the flexibility of the new power system in dispatching, the boundary conditions defined in the annual production plan are relaxed, allowing the reservoir water storage and energy storage capacity to deviate from the annual production plan under extreme weather conditions. The deviation is multiplied by a larger penalty coefficient and added to the objective function, and the deviation result is fed back into the annual production plan for adjustment.

[0194] S350. Calculate the core indicators of load power shortage loss, clean energy curtailment cost and unit power generation cost based on the solution results, and determine whether they meet the preset requirements.

[0195] S360. If the requirements are not met and the number of iterations has not reached the upper limit, then the boundary conditions of the power balance model from the previous week are corrected according to the core indicators that are not met, and S330 to S350 are repeated for iterative calculation until the requirements are met or the maximum number of iterations is reached.

[0196] S370: Output this week's production plan and arrange the actual output curves of each power supply.

[0197] Examples and exemplifications

[0198] The total installed capacity expected to be operational by 2030 is 173.94 million kW, including 23.56 million kW of thermal power, 107.16 million kW of hydropower, 17.27 million kW of wind power, 14.6 million kW of photovoltaic power, and 11.35 million kW of planned installed capacity of various types of centralized energy storage units, mainly pumped storage.

[0199] This provincial power grid is a typical new power system with a high proportion of clean energy, with hydropower, wind power, and solar power accounting for 80% of the total installed capacity. Hydropower accounts for the absolute majority, and various types of hydropower stations of different sizes are distributed in a multi-tiered manner. According to the reservoir capacity regulation time scale, run-of-river hydropower stations, short-term regulating hydropower stations, and medium- and long-term regulating hydropower stations each account for about one-third of the installed capacity, of which the installed capacity of hydropower with medium- and long-term regulation capabilities is 37.95 million kW.

[0200] Taking into account the electricity exchanged with the external grid, the maximum annual load is estimated at 122 million kW, and the annual electricity consumption is estimated at 657 billion kWh.

[0201] The annual load and clean energy resource normalization curves are generated from the forecast data, such as... Figure 3 As shown.

[0202] The load curve exhibits distinct peaks in winter and summer, and a trough around the Spring Festival. The highest load is expected to occur in week 34, with the average load at 89.49% of the maximum load, primarily driven by cooling loads. The maximum usable power generation of wind and solar power is highly volatile, with overall annual usable hours being relatively low. The natural runoff curve, normalized to maximum flow, shows that the minimum weekly average runoff during the dry season is only 22% of the maximum weekly average runoff during the wet season, indicating significant seasonal fluctuations in hydropower generation.

[0203] From an annual timescale perspective, the peak load period and the high-water season largely overlap, thus playing a positive role in peak shaving. However, looking at different years, there is a certain negative correlation between load and hydropower changes: in high-water years, due to abundant rainfall, there are usually fewer high-temperature days, resulting in relatively lower loads; and vice versa. This cross-year source-load reverse fluctuation characteristic exacerbates the uncertainty of power balance, bringing severe pressure on ensuring power supply and facilitating the absorption of clean energy.

[0204] Scene setup and simulation methods

[0205] The medium- and long-term source-load time series is decomposed into four parts: trend term, seasonal cycle term, extreme weather disturbance term, and random uncertainty term. Figure 3 The predicted curve shown already includes trend and seasonal cycle terms. As an "ideal scenario", the method described in this invention is used to perform annual time-scale power balance calculations and generate an annual production plan.

[0206] exist Figure 3 Extreme weather disturbances and random uncertainties were added to the predicted curves shown to simulate real-world scenarios, and three representative scenarios were selected from these:

[0207] 1. Typical scenario: Only add a normal random uncertainty term with a standard deviation of a certain percentage of the predicted value. In this scenario, the uncertainty parameters deviate little from the prediction curve.

[0208] 2. Extreme high temperature and drought scenario (hereinafter referred to as "high temperature scenario"): Based on the ordinary scenario, extreme weather disturbances are added to significantly reduce summer runoff and increase summer load to simulate a continuous extreme high temperature and drought scenario in summer;

[0209] 3. Extreme low temperature and high water level scenario (hereinafter referred to as "high water level scenario"): Based on the ordinary scenario, extreme weather disturbances are added to significantly reduce the summer load and increase the inflow of water to simulate the extreme scenario of continuous heavy rain and generally low temperature in summer.

[0210] Load and flow curves for different scenarios are as follows Figure 4 and Figure 5 As shown;

[0211] Next, the second phase of power balance calculations will be carried out weekly, simulating operation under different scenarios according to the pre-established annual production plan.

[0212] Finally, the power balance results of the second phase of the 52 cycles throughout the year are summarized to obtain the annual simulation operation results.

[0213] In comparison, we conducted power balance optimization calculations and simulations using both the existing two-stage robust optimization method based on columns and constraints (hereinafter referred to as the "existing robust method") and the traditional method that does not consider uncertainties (hereinafter referred to as the "traditional method"), to evaluate the calculation results of different methods under different scenarios.

[0214] Calculation results

[0215] The annual simulation curves for different scenarios and algorithms are as follows: Figures 6 to 17 As shown.

[0216] As can be seen, the results of the ideal scenario and the normal scenario are basically consistent, indicating that the normal fluctuation of the source load parameters in a normal year has little impact on the annual power balance; however, the results of the two extreme scenarios of high temperature and drought and low temperature and abundant water deviate significantly from the ideal scenario, and there are also significant differences between the three different methods.

[0217] The annual simulation operation statistics for different scenarios and algorithms are summarized and presented in Table 1.

[0218] Table 1. Statistical data of simulation operation throughout the year for different scenarios and algorithms.

[0219]

[0220] The total cost consists of load shortage losses, renewable energy curtailment costs, thermal power generation costs, and energy storage costs. Based on electricity economic data and analysis methods, these costs are estimated at 9 yuan / kWh, 0.52 yuan / kWh, 0.34 yuan / kWh, and 0.52 yuan / kWh, respectively.

[0221] As can be seen, in an ideal scenario without random fluctuations in parameters, the total costs of the three methods are very similar, and all operate economically at near-minimum costs.

[0222] In normal scenarios, the operating costs of this method and existing robust methods are slightly higher, mainly due to the curtailment of clean energy and the increase in thermal power generation caused by random fluctuations in parameters. The operating costs of traditional methods increase sharply, with a load shortage of 1.36 billion kWh, resulting in losses accounting for more than 30% of the total annual cost, exposing the limitations of traditional methods in dealing with uncertain parameter fluctuations.

[0223] In high-temperature scenarios, due to insufficient overall power supply, all three methods exhibit significant load gaps, constituting a major portion of the total annual cost. Among them, the existing robust method, due to its most conservative planning, has the lowest load gap and total cost. This method is similar to the existing robust method, while the traditional method has the highest cost. The load gap curves for different methods in high-temperature scenarios are shown below. Figure 18 As shown.

[0224] As can be seen, the load gap is mainly concentrated in the high-temperature period of summer from week 20 to week 38, which basically coincides with the peak load period.

[0225] In scenarios with abundant water resources, all three methods result in significant power waste due to surplus clean energy. This method, however, has the lowest overall power waste and the lowest cost.

[0226] Overall, this method, as an adaptive two-stage robust optimization method between traditional and existing robust methods, can balance the economy and conservatism of the computational results. It has the lowest total cost in ordinary and water-rich scenarios, and the total cost is slightly higher than existing robust methods in high-temperature scenarios, but much lower than traditional methods.

[0227] When faced with source-load uncertainty, ensuring power supply to the load and ensuring the consumption of clean energy are essentially contradictory requirements. Considering that the loss from shedding loads under the same power volume is far greater than the loss from curtailing clean energy, existing robust methods tend to favor ensuring power supply, further increasing the risk of curtailing clean energy. This method improves this problem to some extent.

[0228] In terms of computation time, using the same computer configuration (Intel Core i7-13700H 32 GB) and the CPLEX 12.8 solver, this method takes less than one minute in all scenarios. Due to the simplification of existing robust methods, the computation time is significantly reduced; in the three stochastic scenarios of this example, the time is approximately one-quarter of that of existing robust methods. Considering the high complexity of real-world power systems and the need for real-time rolling calculations based on updated data, adopting this method to improve computational efficiency has practical engineering significance.

[0229] The above description is merely a preferred embodiment of the present invention and does not constitute any limitation on the present invention. Any person skilled in the art can make many possible variations and modifications to the technical solution of the present invention, or modify it into equivalent embodiments, without departing from the scope of the present invention. Therefore, any modifications, equivalent changes, and alterations made to the above embodiments based on the technology of the present invention without departing from the scope of the present invention are within the protection scope of the present invention.

Claims

1. An adaptive robust balancing method for power supply and demand uncertainties in extreme weather, characterized by: The method includes the following steps: S1. Obtain the operating parameters of the power system, including load forecast data, power supply side forecast data, and operating constraint parameters of each generator set; S2. Construct an adaptive interval two-stage robust optimization model based on confidence intervals. The method for constructing the model includes: S21. Determine the confidence region for the uncertain parameter at a given confidence level; S22. Decompose the uncertain parameters into load-side uncertain parameters and power-side uncertain parameters, and use two extreme cases to cover the fluctuation range of the uncertain parameters. The two extreme cases are: the load value is at its maximum and the power value is at its minimum, and the load value is at its minimum and the power value is at its maximum. S23. Transform the two-stage robust optimization problem into a single-level deterministic optimization problem. The objective function of this single-level deterministic optimization problem is to minimize the sum of the costs of the decision variables in the first stage and the decision variables in the second stage. S24. Extract the power supply and demand balance constraints from the single-layer deterministic optimization problem, and calculate the maximum supply and demand imbalance based on the power supply and demand balance constraints. S25. Introduce upward adaptive capacity gap and downward adaptive capacity gap to relax the power supply and demand balance constraint. Multiply the upward adaptive capacity gap and downward adaptive capacity gap by their respective weight coefficients and add them to the objective function of the single-layer deterministic optimization problem. Dynamically adjust the values ​​of the upward adaptive capacity gap and downward adaptive capacity gap through the optimization process of the objective function to automatically adapt to the fluctuation range of uncertain parameters and obtain an adaptive interval robust optimization model. S3. Based on the adaptive interval robust optimization model, establish a multi-time-scale power balance optimization model for the power system under extreme weather conditions, and solve the multi-time-scale power balance optimization model to obtain the output plan of each generator unit and the power supply and demand balance scheduling scheme.

2. The adaptive robust balancing method for adapting to the uncertainty of power supply and demand in extreme weather as described in claim 1, characterized in that: The transformation of the two-stage robust optimization problem into a single-layer deterministic optimization problem is as follows: ; wherein, F is the objective function; x is the vector of decision variables in the first stage; y is the vector of adjusted decision variables in the second stage; m and n are the coefficient vectors of the objective function, and the superscript T thereon represents transpose; A, B, C and D are constraint coefficient matrices; e and h are vectors of right-hand side terms of constraints, , , are respectively the expected value and the upper and lower fluctuation intervals of the load-side uncertain parameter at a given confidence level K (0<K≤1); , , are respectively the expected value and the upper and lower fluctuation intervals of the power-side uncertain parameter at the given confidence level K.

3. The adaptive robust balancing method for adapting to the uncertainty of power supply and demand in extreme weather as described in claim 2, characterized in that: The power supply and demand balance constraint is as follows: ; In the formula, P G This represents the total output of conventional deterministic generating units; Reserved positive and backup power capacity for the system; The negative backup power capacity reserved for the system; P lack P represents the maximum possible load gap. waste The maximum possible power surplus; D G For the power supply side uncertain parameter ξ G Factors affecting power output; The positive and negative backup power capacity are calculated as follows: ; In the formula, P g (i) represents the current output of the unit; and These are the unit's rated maximum and minimum output, respectively; This is the positive backup power capacity; This is the negative backup power capacity.

4. The adaptive robust balancing method for adapting to the uncertainty of power supply and demand in extreme weather as described in claim 3, characterized in that: The relaxation of the power supply and demand balance constraint is as follows: ; In the formula, To provide an upward adaptive capacity gap; This is a downward adaptive capacity gap; The relaxed electricity supply and demand balance constraint is decomposed as follows: ; By weighting the upward adaptive capacity gap and the downward adaptive capacity gap and adding them to the objective function, we obtain the adaptive interval robust optimization model: ; In the formula, F back To adapt to capacity gap risk; λ lack λ is the load power shortage loss coefficient; waste The cost per kilowatt-hour of clean energy curtailment; λ + and λ - These are the upward and downward adaptive capacity gap risk weight coefficients for the system, with values ​​ranging from 0 to 1.

5. The adaptive robust balancing method for adapting to the uncertainty of power supply and demand in extreme weather as described in claim 1, characterized in that: The multi-timescale power balance optimization model includes an annual power balance model and a pre-week power balance model. The annual power balance model arranges the annual production plan with a one-week scheduling cycle, while the pre-week power balance model arranges the power generation plan for one week with an hourly scheduling cycle.

6. The adaptive robust balancing method for adapting to the uncertainty of power supply and demand in extreme weather as described in claim 1, characterized in that: The objective functions of the annual power balance model and the pre-week power balance model are: ; In the formula, F1 is the load power shortage loss; F2 is the cost of clean energy curtailment; F3 is the unit power generation cost; F back To mitigate the risk of adaptive capacity gaps; ; In the formula, nt is the number of scheduling cycles; T is the cycle duration; ; In the formula, F w F s F h These represent the total costs of wind curtailment, solar curtailment, and hydropower curtailment, respectively. The maximum output capacity of the i-th wind or solar power unit during time period t; Q represents the actual output of the i-th wind or solar power unit during time period t; waste (i,t) represents the water discharge of the i-th hydropower station during time period t; Ω w Ω s Ω h Let represent the collection of wind turbines, photovoltaic units, and hydropower stations; h(i) and η(i) represent the average head and power generation efficiency of the hydropower station, respectively; ρ is the density of water; and g is the gravitational acceleration constant. ; In the formula, F th and The respective power generation costs of thermal power units and energy storage power stations; λ th The unit cost of electricity generation for thermal power units; λ sto The cost per unit of electricity consumed by the energy storage unit; P sto (i,t) represents the power consumption of the energy storage power station when it is in energy storage mode; η sto (i) and η g (i) represents the energy conversion efficiency of the energy storage power station in the two processes of energy storage and power generation, respectively; , They are a combination of thermal power units and energy storage power stations.

7. The adaptive robust balancing method for adapting to the uncertainty of power supply and demand in extreme weather as described in claim 6, characterized in that: The constraints of the annual power balance model and the pre-week power balance model include: System power balance constraints: ; In the formula, The predicted load for time period t; Generator output upper and lower limit constraints: ; In the formula, S g (i,t) are 0-1 variables, representing the i-th... The operating status of each unit during time period t, with 1 indicating that it is operating; This is the upper limit of the unit's output. This represents the lower limit of the unit's output. This represents the total number of weeks the unit is out of service within a year. Water and electricity constraints: ; ; In the formula, V(i,t) is the first... The water storage capacity of each reservoir at the end of time period t; Q in (i,t) and Q out (i,t) represent the average inflow and outflow of the reservoir during this period, respectively; Q(i,t) represents the water flow for power generation at the hydropower station. and These are the upper and lower limits of the reservoir's water storage capacity, respectively. and These are the upper and lower limits of the reservoir's discharge capacity, respectively. Spatiotemporal coupling constraints of flow rates in cascade hydropower stations: ; In the formula, Let be the average natural runoff of the i-th reservoir during time period t; Let t be the set of upstream reservoirs of the i-th reservoir; lag (j,i) represents the water flow from the j-th hydropower station to the j-th hydropower station. The delay time required for each hydroelectric power station; Constraints of energy storage units: ; In the formula, E(i,t) is the first... The energy stored in a single energy storage power station at the end of time period t; and These represent the upper and lower limits of the energy storage capacity of the power station during that time period; and These represent the upper and lower limits of the power storage capacity of the power station during that time period; S sto (i,t) is a 0-1 variable, representing the state of the energy storage power station during that time period, with 1 indicating that it is storing energy.

8. The adaptive robust balancing method for adapting to the uncertainty of power supply and demand in extreme weather as described in claim 7, characterized in that: The constraints of the pre-week power balance model also include: Meet the boundary conditions specified in the annual production plan: ; In the formula, Ω S=0 For units that need to be shut down this week according to the annual production plan; V sta (i) and V end (i) are respectively the first The water levels in the reservoirs at the beginning and end of this week; E sta (i) and E end (i) are respectively the first The energy storage capacity of an energy storage station at the beginning and end of the week; , , , The first The water storage capacity of the reservoir at the initial and final moments, and the water storage capacity of the first reservoir at the last moment. The energy stored in a single energy storage power station at the initial and final moments; Thermal power unit output ramp-up rate constraint: ; In the formula, r(i) is the ramp rate of the i-th unit; T int (i) is a non-negative integer variable, representing the i-th... The time required for an energy storage power station to switch between energy storage and power generation states; This is a temporary variable; N represents the set of natural numbers.

9. The adaptive robust balancing method for adapting to the uncertainty of power supply and demand in extreme weather as described in claim 8, characterized in that: The solution method for the multi-timescale power balance optimization model is as follows: S310. Collect the data required for the annual power balance model and preprocess it, and preset the core indicators and the upper limit of the number of iterations for optimization calculation; S320. Solve the annual time-scale power balance model, formulate the annual production plan, and output the annual decision variables. The annual decision variables include the start-up and shutdown arrangements of each generator unit and the water storage plans of each large reservoir, forming the boundary conditions of the weekly power balance model. S330. Collect this week's power supply and load forecast data, and combine them with the boundary conditions determined by the annual production plan to solve the pre-week power balance model. S340. Multiply the deviation between the reservoir water storage and energy storage capacity in the previous week's power balance model and the annual production plan by a penalty coefficient and add it to the objective function, and feed the deviation result back to the annual production plan for adjustment. S350. Calculate the core indicators of load power shortage loss, clean energy curtailment cost and unit power generation cost based on the solution results, and determine whether they meet the preset requirements. S360. If the requirements are not met and the number of iterations has not reached the upper limit, then the boundary conditions of the power balance model from the previous week are corrected according to the core indicators that are not met, and S330 to S350 are repeated for iterative calculation until the requirements are met or the maximum number of iterations is reached. S370: Output this week's production plan and arrange the actual output curves of each power supply.