Uncertain scene generation and planning method based on source load matching and time sequence characteristics

Abstract

The invention relates to the technical field of energy planning, and discloses an uncertain scene generation and planning method based on source load matching and time sequence characteristics, which effectively reduces the operation performance deviation and improves the engineering implementability of a stochastic planning scheme. According to the method, a representative meteorological year data set, a reference capacity optimization variable set and a whole-year wind-light load initial scene time set are determined through historical data of a target area in previous years; dividing the sample points in each hour into a residual power state region set and a power shortage state region set based on positive and negative source charge difference values; respectively clustering the two types of sets hour by hour by adopting a distance clustering algorithm to obtain an annual typical scene set and a corresponding first probability set; and further generating a year-round disordered boundary sequence, and performing optimization by taking the representative meteorological year data set time sequence as a benchmark to form a year-round scene boundary with time sequence characteristics. According to the method, the source load difference and the time sequence characteristics are effectively reserved, the system reliability evaluation accuracy and the planning implementability are improved, and the operation deviation is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of energy planning, in particular to a source-load matching and time sequence characteristic based uncertainty scenario generation and planning method. BACKGROUND

[0002] With the rapid development of new energy power such as wind energy and light energy with strong randomness and volatility, how to realize large-scale and efficient consumption of new energy power has become a key issue to promote the transformation of China's energy structure. However, with the increasing penetration rate of random and intermittent power sources such as wind and light on the source side, and the increasingly complex and uncertain electricity consumption behavior on the load side, the energy system (such as wind-thermal-storage system, distributed energy system, and integrated energy system) faces significant uncertainty challenges in the planning stage, and therefore the random planning method has attracted widespread attention.

[0003] In the current uncertainty random planning method, the scenario-based analysis method has been widely used by introducing random variables to characterize uncertainty. This method usually includes four steps: probability distribution fitting, scenario generation, scenario reduction, and construction of uncertainty boundary. The existing random planning method uses distance-based clustering algorithm (such as K-means clustering algorithm) in the scenario reduction link, however, the uncertainty boundary constructed by such method has a large unreasonable, resulting in a large relative deviation of the net present value, wind and light curtailment rate and power following deviation from the actual mean value, and the engineering implementability of the random planning scheme is low. In highly nonlinear energy systems, the above limitations are particularly prominent - time-varying operation strategy may further amplify the performance deviation between planning and actual operation, resulting in significant deviation of the planning scheme based on traditional uncertainty optimization from the actual engineering effect. SUMMARY

[0004] The present application provides a source-load matching and time sequence characteristic based uncertainty scenario generation and planning method, which aims to at least solve one of the technical problems existing in the prior art. To this end, the present application proposes a source-load matching and time sequence characteristic based uncertainty scenario generation and planning method, which can distinguish the characteristics of positive / negative differences in source-load matching, and combine the time sequence correlation and operation coupling characteristics between the source and load and themselves, to effectively optimize the uncertainty boundary, improve the rationality, and thus obtain better uncertainty scenarios and planning results, effectively reduce the operation performance deviation, and effectively improve the engineering implementability of the random planning scheme.

[0005] In the first aspect, the technical scheme of the present application relates to a source-load matching and time sequence characteristic based uncertainty scenario generation method, comprising the following steps: According to the historical data of the target area in previous years, a set of representative meteorological year data, a set of reference capacity optimization variables, and a set of initial annual wind-light-load scenarios are determined respectively. Based on the reference capacity optimization variable set, the source-load difference values of the wind and light load sample points of each hour in the initial annual wind and light load scenario time set are calculated, and the wind and light load sample points in each hour are divided into a set of surplus power state regions where source is greater than load and a set of power shortage state regions where source is less than load according to the positive and negative of the corresponding source-load difference values; Based on the distance clustering algorithm, the wind and light load sample points of the corresponding set of surplus power state regions and the set of power shortage state regions are respectively clustered hour by hour, and after iterative convergence, the annual typical scenario time set and the corresponding first probability set are determined; According to the annual typical scenario time set and the first probability set, the annual unordered boundary sequence is determined, and based on the time sequence of the representative meteorological year data set, the annual unordered boundary sequence is optimized according to the genetic algorithm to generate an annual scenario boundary with time sequence characteristics.

[0006] According to some embodiments of the present application, the representative meteorological year data set is determined by the reference year generation method or the typical year construction method based on the historical data of previous years.

[0007] According to some embodiments of the present application, the reference year generation method comprises the following steps: According to the historical data of at least 5 consecutive years in the target region, the average wind speed, annual total radiation and total load of each year are determined; Based on the parameters of the average wind speed, the annual total radiation and the total load of at least 5 consecutive years in the target region, the corresponding parameters of each year are sorted respectively to determine the ranking of the corresponding parameters of each year; Based on the ranking of the corresponding parameters of each year, the comprehensive median deviation D of each year is determined; Based on the comprehensive median deviation of each year, the historical data of the year with the smallest deviation is selected as the representative meteorological year data set; when there are multiple years with the same and smallest deviation, the year with a more moderate annual ranking is selected as the representative meteorological year data set according to a predetermined parameter priority order, and the parameter priority order is the average wind speed, the annual total radiation and the total load.

[0008] According to some embodiments of the present application, the typical year construction method comprises the following steps: The historical data of at least 10 consecutive years in the target region is preprocessed to determine the daily average value of each year and each month of the first target parameter, and the first target parameter includes solar radiation, wind speed, dry bulb temperature and humidity; According to the daily average value of each of the first target parameters, the long-term average cumulative distribution function of each year and each month and the single-year cumulative distribution function of each month are determined respectively; According to the mean square error of the single-year cumulative distribution function and the long-term average cumulative distribution function corresponding to each month of each year, and based on months 1 to 12, the month corresponding to the minimum mean square error of the month is selected as the target month; The hour data of the target month corresponding to months 1 to 12 are spliced and smoothed in order to serve as the representative meteorological year data set.

[0009] According to some embodiments of the present application, the determination of the reference capacity optimization variable set comprises the following steps: The data of the representative meteorological year data set are taken as the system operation boundary, and the system reference capacity configuration is optimized and solved based on a multi-objective optimization algorithm to determine the reference capacity optimization variable set.

[0010] According to some embodiments of the present application, the determination of the initial scene time set of the annual wind and light load comprises the following steps: Based on at least 5 years of historical data of the target region, the Gaussian kernel density estimation method is used to fit the second target parameter in each hour to determine the probability density function corresponding to the second target parameter corresponding to each hour, and the cumulative distribution function corresponding to the second target parameter corresponding to each hour is obtained by integration, the second target parameter includes wind power, solar radiation and power load; Based on the cumulative distribution function corresponding to each hour, the Frank-Copula function is combined to construct the wind and light load three-dimensional joint cumulative distribution function corresponding to each hour in layers; Based on the wind and light load three-dimensional joint cumulative distribution function of each hour, the Monte Carlo or Latin hypercube sampling method is used to generate joint sampling points of each hour, and the corresponding joint sampling points of each hour are inversely transformed to determine the sample values corresponding to each joint sampling point in each hour, the sample values include the first sample value corresponding to the wind power, the second sample value corresponding to the solar radiation and the third sample value corresponding to the power load, and the first sample value, the second sample value and the third sample value corresponding to each hour are taken as the wind and light load sample points in each hour to constitute the initial scene time of wind and light load of each hour; The initial scene time of wind and light load of each hour is combined to constitute the initial all-day scene of wind and light load, and is divided into summer, transition season and winter to generate the initial scene time set of wind and light load corresponding to 24 hours of summer, 24 hours of transition season and 24 hours of winter, respectively, and all the initial scene time set of wind and light load is combined to serve as the initial scene time set of wind and light load of the whole year.

[0011] According to some embodiments of the present application, the determination of the reference capacity optimization variable set comprises the following steps: Based on the distance clustering algorithm, all the wind and light load initial full-day scenes are clustered day by day, and after iterative convergence, a set of annual typical scene days and a corresponding second probability set are determined; According to the index performance of each typical scene day in the set of annual typical scene days, and weighted sum according to the corresponding second probability set, the annual index performance CPWAB is determined; The index performance is input as a system operation boundary, and the reference capacity configuration of the system is optimized and solved based on a multi-objective optimization algorithm to determine the reference capacity optimization variable set.

[0012] According to some embodiments of the application, the set of annual wind and light load initial scenes and the set of annual typical scene times are divided into summer, transition season and winter, wherein summer is the 152th to 243th day of the year, transition season is the 60th to 151th day and the 244th to 334th day of the year, and winter is the 1st to 59th day and the 335th to 365th day of the year.

[0013] According to some embodiments of the application, the step of determining a set of annual unordered boundary sequences according to the set of annual typical scene times and the first probability set, and taking the time sequence of the set of representative meteorological year data as a reference, and optimizing the set of annual unordered boundary sequences according to a genetic algorithm to generate a set of annual scene boundaries with time sequence characteristics includes: According to the set of annual typical scene times and the first probability set, the first probability corresponding to each typical scene time is determined respectively; Each of the first probabilities is multiplied by the corresponding number of seasonal days to obtain the frequency of occurrence of the corresponding typical scene time in the corresponding hours; According to the frequency, each of the typical scene times is repeatedly arranged and combined to form unordered boundary sequences corresponding to summer days x 24 hours, transition season days x 24 hours and winter days x 24 hours respectively, and all the unordered boundary sequences are combined into the set of annual unordered boundary sequences; Taking the time sequence of the set of representative meteorological year data as a reference, and dividing the set of representative meteorological year data according to the same number of seasonal days, the set of annual unordered boundary sequences corresponding to each season are sorted and optimized according to a genetic algorithm to generate the set of annual scene boundaries with time sequence characteristics.

[0014] In a second aspect, the technical scheme of the application also relates to a planning method based on source-load matching and time sequence characteristics, which further includes the following steps: Input the annual scene boundary in the uncertainty scenario generation method based on source-load matching and timing characteristics as the system operation boundary, and optimize and solve the system capacity configuration based on a multi-objective optimization algorithm to determine the capacity random planning variable set of the system.

[0015] The beneficial effects of the present application are as follows: By dividing the sample points in each hour by the source-load difference, the characteristics of positive / negative differences in source-load matching can be distinguished, and by reducing the scene time in each hour, the characteristics of source-load matching can be effectively retained to generate more accurate typical scenarios, improve the accuracy of system reliability evaluation, and use the timing characteristics of historical data for optimization to ultimately generate annual scene boundaries with timing characteristics. The present application combines the timing correlation and operation coupling characteristics of the source, load and the relationship between them, which can effectively improve the rationality of the annual scene boundary, significantly reduce the subsequent operation performance deviation, effectively optimize the uncertainty boundary, improve the rationality, and thus obtain more optimal uncertainty scenarios and capacity random planning variable sets. The results can effectively reduce the operation performance deviation and effectively improve the engineering implementability of the random planning scheme. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 The step flowchart of the uncertainty scenario generation method based on source-load matching and timing characteristics of the present application; Figure 2 The step flowchart of the reference year generation method of the present application; Figure 3 The step flowchart of the typical year construction method of the present application; Figure 4 The step flowchart of the generation of conventional uncertainty boundary; Figure 5 The step flowchart of determining the initial scene time set of annual wind-light-load of the present application; Figure 6 The step flowchart of determining the initial scene time set of annual wind-light-load and the reference capacity optimization variable set of the present application; Figure 7 The step flowchart of reducing the full-day scene corresponding to solar radiation in the initial full-day scene set of wind-light-load of the present application; Figure 8 The step flowchart of reducing the initial scene time set of annual wind-light-load of the present application; Figure 9 The effect influence display diagram of the K-means clustering algorithm of the present application on source-load matching characteristics and timing characteristics; Figure 10 The step flowchart of determining the annual scene boundary of the present application; Figure 11 The source load difference value of this invention DE Curve comparison chart; Figure 12 This is a flowchart illustrating the steps of the planning method based on source-load matching and temporal characteristics of the present invention. Detailed Implementation

[0017] The following will provide a clear and complete description of the concept, specific structure, and technical effects of the present invention in conjunction with the embodiments and accompanying drawings, so as to fully understand the purpose, solution, and effects of the present invention.

[0018] It should be noted that, unless otherwise specified, when a feature is referred to as "fixed" or "connected" to another feature, it can be directly fixed or connected to the other feature, or indirectly fixed or connected to the other feature. It should also be noted that, unless otherwise specified, when a feature is referred to as having an "electrical connection" or "electrical link" with another feature, the two features can be connected directly via pins, via cables, or via wireless transmission. Specific electrical connection methods are common to those skilled in the art, and can be implemented as needed. The singular forms "a," "described," and "the" used herein are also intended to include the plural forms unless the context clearly indicates otherwise. Furthermore, unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in this specification is for the purpose of describing particular embodiments only and not for limiting the invention. The term "and / or" as used herein includes any combination of one or more of the associated listed items.

[0019] It should be understood that although the terms first, second, third, etc., may be used to describe various elements in this disclosure, these elements should not be limited to these terms. These terms are used only to distinguish elements of the same type from one another. For example, a first element may also be referred to as a second element without departing from the scope of this disclosure, and similarly, a second element may also be referred to as a first element. Any and all instances or exemplary language (“e.g.,” “such as,” etc.) provided herein are intended only to better illustrate embodiments of the invention and, unless otherwise required, do not impose a limitation on the scope of the invention.

[0020] Firstly, referring to Figure 1 The technical solution of this invention relates to a method for generating uncertain scenarios based on source-load matching and temporal features, comprising the following steps: Step S100: Based on the historical data of the target area, determine the representative meteorological year data set, the reference capacity optimization variable set, and the annual wind, solar and load initial scenario time set respectively; Step S200, based on the reference capacity optimization variable set, the annual wind light load initial scene time set is calculated DE , according to the positive and negative of the corresponding source load difference DE , the wind light load sample points in each hour are divided into source greater than load surplus state region set and source less than load power shortage state region set according to the positive and negative of the corresponding source load difference Step S300, based on the distance clustering algorithm, the wind light load sample points of the corresponding surplus state region set and the power shortage state region set are clustered hour by hour, and after iterative convergence, the annual typical scene time set and the corresponding first probability set are determined; Step S400, according to the annual typical scene time set and the first probability set, the annual unordered boundary sequence is determined, and according to the genetic algorithm, the annual unordered boundary sequence is optimized based on the time sequence of the representative meteorological year data set to generate the annual scene boundary with time sequence characteristics.

[0021] In step S100, the historical data of the target area in the past years at least includes wind speed, solar radiation, power load and other conventional climate parameters in the corresponding year, and in this embodiment, the annual wind light load initial scene time set is the set of wind light load initial scene time sets corresponding to 24 hours in summer, 24 hours in transition season and 24 hours in winter. In this embodiment, the summer is the 152th to 243rd day of the year, the transition season is the 60th to 151st day and the 244th to 334th day of the year, and the winter is the 1st to 59th day and the 335th to 365th day of the year. It should be noted that the representative meteorological year data set, the reference capacity optimization variable set and the annual wind light load initial scene time set can be determined by analyzing the historical data of the target area in the past years by means commonly used by those skilled in the art.

[0022] In step S200 of the present application, the sample points in each hour are divided by the source load difference, which can distinguish the characteristics of positive / negative difference of source load matching. In step S300, the scene time in each hour is reduced, which can effectively retain the characteristics of source load matching, generate more accurate typical scenes and improve the accuracy of system reliability evaluation. The execution of step S400 can retain the time sequence characteristics of historical data and generate the annual scene boundary with time sequence characteristics. The present application combines the time sequence correlation and operation coupling characteristics of source, load and their relationship, which can effectively improve the rationality of the annual scene boundary, significantly reduce the subsequent operation performance deviation, effectively optimize the uncertain boundary, improve the rationality, and further obtain better uncertainty scene and planning result, effectively reduce the operation performance deviation, and effectively improve the engineering implementability of the random planning scheme.

[0023] It can be known that in step S100, the representative meteorological year data set can be determined by the previous year historical data based on the reference year generation method or the typical year construction method.

[0024] Referring to Figure 2 In some embodiments of the present application, the reference year generation method comprises the following steps: Step S111, determining the average wind speed, annual total radiation and total load of each year from the historical data of at least 5 consecutive years in the target area; Step S112, sorting the parameters corresponding to each year based on the parameters of the average wind speed, annual total radiation and total load of at least 5 consecutive years in the target area, and determining the ranking of the corresponding parameters of each year; Step S113, determining the comprehensive median deviation of each year based on the ranking of the corresponding parameters of each year D ; Step S114, selecting the year with the smallest deviation D， of the comprehensive median deviation D of each year as the representative meteorological year data set; when there are multiple years with the same and smallest deviation D , select the year with the more central annual ranking as the representative meteorological year data set according to the preset parameter priority order, and the parameter priority order is the average wind speed, the annual total radiation and the total load.

[0025] In combination with steps S111 to S114, at least 5 years of historical data of the target area are selected, and the average wind speed, annual total radiation and total load corresponding to each year are calculated, and the selected years are independently sorted according to the corresponding variables of the average wind speed, annual total radiation and total load, and the comprehensive median deviation of the average wind speed, annual total radiation and total load corresponding to each year is calculated D , the formula for calculating the deviation D of the comprehensive median deviation is: (1); In the formula, the average wind speed, the annual total radiation and the total load of the corresponding year are respectively R WS , R DNI and R Load . MThe median value is selected, if the same deviation occurs, the year with the more central wind speed ranking (the time sequence volatility of wind speed is greater) is selected preferentially, if the wind speed ranking is still the same, the year with the more central annual total irradiance is selected, if the annual total irradiance ranking is also the same, the year with the more central total load is selected, finally, if the total load is still the same, the data of a year is selected randomly as the representative meteorological year data set.

[0026] It should be noted that in steps S111 to S114, a person skilled in the art can select the variable to be referred to from the historical data according to the actual situation, and it is not necessarily limited to the average wind speed, the annual total irradiance and the total load, and corresponding adjustment can be made in combination with the demand of energy planning.

[0027] Reference Figure 3 In some embodiments of the present application, the typical year construction method comprises the following steps: Step S121, pre-processing the historical data of at least 10 consecutive years of the target region to determine the daily average value of each month of each year of the first target parameter, the first target parameter including solar irradiance, wind speed, dry bulb temperature and humidity; Step S122, determining the long-term average cumulative distribution function of each year and each month and the single-year cumulative distribution function of each month according to the daily average value of each first target parameter; Step S123, selecting the month with the minimum mean square error of the corresponding month as the target month based on months 1 to 12 according to the mean square error of the single-year cumulative distribution function and the long-term average cumulative distribution function corresponding to each month; Step S124, splicing and smoothing the hour data of the target month corresponding to months 1 to 12 in order to serve as a representative meteorological year data set.

[0028] In the execution of steps S121 and S122, the long-term average cumulative distribution function is: (2); For each month m All years are arranged in ascending order according to the corresponding daily average value, and the long-term average cumulative distribution function is calculated according to formula (2) C long ( p, m, i ), in formula (2): R ( i ) represents the ranking of the i-th daily average value in all years of the month, N is the total number of days in the month in all years, m is the value corresponding to months 1 to 12; The single-year cumulative distribution function of each month is: (3); By the first year m Arrange the daily averages for each month in ascending order, and calculate the annual cumulative distribution function for each month according to formula (3). C single ( p, y, m, i In the formula: r ( i ) indicates the year's number m Mid-month i Sorting of daily averages n This represents the number of days in the current month.

[0029] Then, steps S123 and S124 are executed to select the month with the smallest mean square error as the target month, and time smoothing is performed on the hourly data at the splicing points between the target months to ensure the climate continuity when spliced ​​into a complete year. The resulting typical meteorological year dataset retains key climate characteristics, including the average value, frequency distribution, correlation between parameters, and temporal consistency of the entire annual cycle. Finally, the typical meteorological year dataset is combined into a representative meteorological year dataset.

[0030] Reference Figure 4 The conventional steps for generating uncertain boundaries mainly include four steps: probability distribution fitting, sampling, clustering, and boundary construction. Therefore, when determining the set of reference capacity optimization variables, this invention can use conventional uncertain boundaries as the operating boundaries of the system and optimize them through a multi-objective optimization algorithm to determine the corresponding parameters.

[0031] Reference Figure 5 In some embodiments of the present invention, determining the initial scene set of wind, solar and monsoon throughout the year includes the following steps: Step S131: Based on historical data of at least 5 consecutive years for the target area, the second target parameter within each hour is fitted using the Gaussian kernel density estimation method to determine the probability density function corresponding to the second target parameter for each hour, and the cumulative distribution function corresponding to the second target parameter for each hour is obtained by integration. The second target parameters include wind power, solar irradiance and power load. Step S132: Based on the cumulative distribution function corresponding to each hour, construct the three-dimensional joint cumulative distribution function of wind, solar and load corresponding to the second objective parameter of each hour in a hierarchical manner using the Frank-Copula function; Step S133: Based on the three-dimensional joint cumulative distribution function of wind and solar load per hour, use Monte Carlo or Latin hypercube sampling methods to generate joint sampling points for each hour, and perform inverse transformation on the corresponding joint sampling points for each hour to determine the sample values ​​corresponding to each joint sampling point within each hour. The sample values ​​include the first sample value corresponding to wind power, the second sample value corresponding to solar irradiance, and the third sample value corresponding to power load. The first sample value, the second sample value, and the third sample value corresponding to each hour are used as the wind and solar load sample points for each hour to form the initial scenario of wind and solar load for each hour. Step S134: Combine the initial scene time of each hour of the wind and light lotus by day to form the initial full-day scene of the wind and light lotus, and divide it into summer, transition season and winter, and generate summer scene respectively. 24 Hours, transitional season 24 Hours and winter 24 The initial scene time set corresponding to each hour is used to combine all the initial scene time sets of the wind, light and lotus to form the initial scene time set of the wind, light and lotus for the whole year.

[0032] In step S131, the formula for the probability density function is: (4); Formula (4) represents the point x The Gaussian kernel probability density function at point is given by the formula. n For the sample size, h For bandwidth, x i For the first one drawn from the set i There are 10 independent samples; the bandwidth can be determined using the thumb method, and the optimal bandwidth is: (5); In formula (5) σ The standard deviation of the data set corresponding to the observed samples. n The number of samples; In step S132, the formula for the three-dimensional joint cumulative distribution function of wind, solar, and load is: (6); Among them, in formula (6) α 1 , α 2 They are respectively F 1 , F 2 as well as C 12 and F 3Frank-Copula parameter between them, the formula is to use Copula function to establish joint distribution function, through two levels of Frank-Copula function to describe the dependence structure between wind and light load.

[0033] In the execution step 131, based on the target area at least 5 years of historical data, each year is also divided into summer, transition season and winter three seasons corresponding to the number of days, in each season inside the Gaussian kernel density estimation method is used to each hour corresponding to the wind power, solar radiation and power load parameters are respectively modeled, that is, according to formula (4) and formula (5), the probability density function corresponding to each parameter is determined f 1 、 f 2 and f 3 , the probability density function f 1 corresponding to the wind power parameter, the probability density function f 2 corresponding to the solar radiation parameter, the probability density function f 3 corresponding to the power load parameter; and the integral of the probability density function f 1 、 f 2 and f 3 is carried out, and the corresponding cumulative distribution function F 1 、 F 2 and F 3 is obtained, and F 1 、 F 2 and F 3 corresponding to formula (6) respectively, then step S132 is executed, that is, the calculation according to formula (6) is carried out, then the hourly joint cumulative distribution function can be obtained.

[0034] In the execution step S133, first, Monte Carlo or Latin hypercube combined with Copula joint sampling method is used to sample the hourly joint distribution function obtained in step S132, to generate a large number of hourly joint sampling points, and the sample value corresponding to the cumulative probability is calculated by means of cubic spline interpolation method, that is, inverse transformation, that is, the first sample value of wind power x 1 , the second sample value of solar radiation x 2 and the third sample value of power loadx 3 Specifically, the cumulative probability of a sample can be approximated using cubic spline interpolation. The specific formula is as follows: (7); Among them, in formula (7) u 1 , u 2 and u 3 , respectively, correspond to the cumulative probability values ​​of wind power, solar irradiance and power load, and are generated by combining Monte Carlo or Latin hypercube sampling techniques using formula (6); a 1 , a 2 , a 3 , b 1 , b 2 , b 3 , c 1 , c 2 , c 3 , d 1 , d 2 as well as d 3 These are the coefficients of the cubic spline interpolation, which can be pre-calculated using the cubic spline interpolation method; F 1 -1 , F 2 -1 and F 3 -1 These are the cumulative distribution functions. F 1 , F 2 and F 3 The corresponding inverse function; Then the first sample value within each hour x 1 Second sample value x 2 and the third sample value x 3 As the sample points of wind and solar load corresponding to each hour, to form the initial scene of wind and solar load for each hour; Finally, according to step S134, combination and division are performed, and then the initial all-day scene of wind and light load and the initial scene of wind and light load of the whole year can be correspondingly constituted.

[0035] With reference to Figure 6 In some embodiments of the present application, the determination of the reference capacity optimization variable set comprises the following steps: Step S135, based on the distance clustering algorithm, all the initial all-day scenes of wind and light load are clustered day by day, and after iterative convergence, the set of typical scene days of the whole year and the corresponding second probability set are determined; Step S136, according to the index performance of each typical scene day in the set of typical scene days of the whole year, and the weighted sum is performed according to the corresponding second probability set, the index performance of the whole year is determined C PWAB ; Step S137, the index performance C PWAB is input as the system operation boundary, and the reference capacity configuration is optimized and solved based on the multi-objective optimization algorithm to determine the reference capacity optimization variable set.

[0036] Wherein, after step S134 is executed, step S135 is continued to be executed, in step S135, the distance clustering algorithm can be a conventional distance clustering algorithm such as K-means clustering method or hierarchical clustering method, and the set of initial all-day scenes of wind and light load in step S134 can be reduced to obtain the typical scene days and the corresponding second probability, and all the typical scene days and the corresponding second probability are combined to constitute the set of typical scene days of the whole year and the corresponding second probability set.

[0037] The present application adopts the K-means clustering method in the reduction mode, and reduces the set of initial all-day scenes of wind and light load in units of "day", and the reference Figure 7 , Figure 7 is the solar radiation determined after steps 133 to 135 DNI corresponding to the step flow effect diagram of all-day scene reduction.

[0038] Wherein, the index performance C PWAB The formula is: (8); The C ( t, j ) in formula (8) is the index performance corresponding to the t hour in the j typical scene day; P j is the index performance corresponding to the jThe probability of occurrence of a typical scenario.

[0039] In determining the corresponding index performance C PWAB Then, step S137 is performed, and when the system operation performance is evaluated, in this embodiment, the net present value, the wind and light curtailment rate, and the power following deviation are taken as evaluation indexes to confirm the reference capacity optimization variable set, and the specific steps are as follows: The index performance is taken as the system operation boundary input, a multi-objective optimization algorithm is used to optimize and solve the system reference capacity configuration, and the ideal point is formed by the optimal values of each single objective in the Pareto solution set, and the decision variable combination corresponding to the solution with the minimum Euclidean distance from the point is taken as the reference capacity optimization variable set; the multi-objective optimization algorithm can be an NSGA-II algorithm or a particle swarm algorithm based on the Pareto frontier to perform capacity optimization, in this embodiment, the NSGA-II algorithm based on the Pareto frontier is used, the maximum net present value, the minimum wind and light curtailment rate, and the minimum power following deviation are taken as the targets, the ideal point is finally formed by the optimal values of each single objective in the Pareto solution set, and the decision variable combination corresponding to the solution with the minimum Euclidean distance from the point is taken as the reference capacity optimization variable set, in this embodiment, the reference capacity optimization variable set includes: Reference wind power installed capacity N w1 , unit: [unit]; Reference photovoltaic installed capacity C pv1 , unit: [MW]; Reference thermal power installed capacity C csp1 , unit: [MW]; Reference heat storage time TES 1 , unit: [h]; Reference mirror field magnification SM 1 .

[0040] In some embodiments of the present application, the determination of the reference capacity optimization variable set includes the following steps: Step S141, taking the data set of representative meteorological year data as the system operation boundary, and based on a multi-objective optimization algorithm, the system capacity configuration is optimized and solved to determine the reference capacity optimization variable set.

[0041] It can be known that in step S141, the multi-objective optimization algorithm used can also be the NSGA-II algorithm or the particle swarm algorithm based on the Pareto frontier for capacity optimization, and the specific calculation method belongs to the routine technical means of those skilled in the art, so reference can be made to step 137 above to determine the parameters of the reference capacity optimization variable set, which will not be described in detail here.

[0042] It can be known that the reference capacity optimization variable set can be calculated according to step S141 or step S137, and those skilled in the art can select appropriate calculation according to the situation to determine.

[0043] In some embodiments of the present application, the source-load difference value calculated in step S200 DE is calculated according to the following formula: (9); wherein the source-load difference value DE is in [MW]; N w1 and C pv1 correspond to the reference wind power installed capacity N w1 and the reference photovoltaic installed capacity C pv1 respectively calculated in step S141 or step S137. W p is the sample point corresponding to the wind power in each hour, in [MW]; Load is the sample point corresponding to the power load in each hour, in [MW]; The numerical value 5.56 is the component area required for each kilowatt of photovoltaic installed capacity, in [m 2 ]; When the source-load difference value DE > 0, the corresponding wind-light-load sample point is collected in the surplus power state region set; When the source-load difference value DE < 0, the corresponding wind-light-load sample point is collected in the power shortage state region set.

[0044] Step S200 calculates each sample point in each hour according to formula (9), and then the sample points in each hour can be divided into two different areas according to the source-load matching state in each hour, so that the subsequent K-means clustering can avoid dividing the sample points with different states into the same category, neutralize the source-load matching characteristics of the sample points in the hour, retain the source-load matching characteristics, and then more accurately capture the volatility of renewable energy and load, so as to improve the accuracy of system reliability evaluation.

[0045] In some embodiments of the application, the distance clustering algorithm in step S300 can also be a conventional distance clustering algorithm such as K-means clustering method or hierarchical clustering method, and a suitable algorithm can be selected according to the needs in the art to reduce the initial scene time set of wind-light-load throughout the year to obtain the typical scene time and the corresponding first probability, and then the annual typical scene time set and the corresponding first probability set can be formed. In this embodiment, the K-means clustering method is used in step S300 to reduce the initial scene time set of wind-light-load throughout the year by taking hours as the target.

[0046] The core idea of the K-means algorithm is to divide the samples into several clusters and use the center of each cluster to represent the characteristic distribution of the cluster, and in the process of assigning samples to clusters, each sample is assigned to the nearest cluster center.

[0047] Taking a certain uncertain parameter sampling data set X in a certain hour as an example (the same processing method is used for the other two dimensions of the uncertain parameter), assuming that the data X=[ x 1 , x 2 ,..., x n ], randomly select k samples from the data as the initial clustering centers v [ v 1 , v 2 ,..., v k ]. Calculate the distance from the data x i to each clustering center: (10); According to the distance from each sample to the clustering center, the data points are divided into the cluster corresponding to the nearest clustering center. Then, the mean value of the data in each cluster is calculated as the new clustering center: (11); wherein:N j indicates the number of samples contained in the j

[0048] The above category division and cluster center updating steps are repeatedly performed until the clustering center no longer changes or the maximum number of iterations preset in advance is reached, and the algorithm is terminated. The k (12).

[0049] Based on step S300, the corresponding wind and light load sample points of the surplus power state region set and the power shortage state region set are clustered hour by hour, and after iterative convergence, the typical scene time of each hour and the corresponding probability are obtained. This step does not involve the time sequence characteristics of the data, that is, the fluctuation trend combined between hours, and can independently process the influence of clustering on the source load matching characteristics.

[0050] Step S300 is based on the core idea of the above-mentioned K-means algorithm, and the sample data of the surplus power state region set and the power shortage state region set in each hour are clustered respectively. After iterative convergence, the typical scene time and the corresponding first probability are obtained. By combining all the typical scene times and the corresponding first probabilities, the annual typical scene time set and the corresponding first probability set can be determined. It should be noted that step S300 is a clustering algorithm, which is clustered in units of "hours". For reference, Figure 8 is an effect diagram of step S300 scene reduction.

[0051] It can also be known that step S135 is also the same algorithm, but step S135 is clustered in units of "days", which is different from step S300. In order to facilitate subsequent description, it can be understood that step S300 implements a "point clustering" mode, and the clustering of step S135 is a "line clustering" mode.

[0052] Specifically, after steps S200 and S300, by comparing Figure 7 , Figure 8 and Figure 9 , it can be known that the present application can effectively retain the source load matching difference characteristics, and at the same time, it does not involve the time sequence fluctuation, and can provide a clear and typical data basis for the time sequence analysis and optimization of step S400.

[0053] The annual wind and light load initial scene time set in step S100 includes wind and light load data of 24 hours in 3 seasons, that is, summer 24 hours, transition season 24 hours and winter 24 ​​The initial scene time set of the wind and light lotus for each hour, after executing steps S200 and S300 respectively on the initial scene time set of the wind and light lotus for the whole year, can yield a size of k×24×3 The typical scenario time set throughout the year and the corresponding first probability set.

[0054] Reference Figure 10 In some embodiments of the present invention, step S400 includes: Step S410: Based on the set of typical scenarios throughout the year and the set of first probabilities, determine the first probability corresponding to each typical scenario. Step S420: Multiply each first probability by the corresponding number of days in the season to obtain the frequency of occurrence of the corresponding typical scenario within the corresponding hour; Step S430: Based on frequency, repeatedly arrange and combine each typical scenario to form the number of summer days. ×24 Hours, number of days in the transition season ×24 Hours and winter days ×24 The disordered boundary sequence corresponding to each hour is combined into a whole year disordered boundary sequence. Step S440: Based on the time series of the representative meteorological year data set, the representative meteorological year data set is divided according to the same number of days in each season. The unordered boundary sequence of the whole year corresponding to each season is sorted and optimized according to the genetic algorithm to generate the whole year scene boundary with time series characteristics.

[0055] The formula for frequency calculation in step S420 is as follows: (13); In formula (13): N f,j For the first j The frequency of occurrence of a typical scenario within a certain hour; P j For the first j The probability corresponding to each typical scenario; Day season In this embodiment, the year is divided into summer, transition season and winter, corresponding to the number of days in each season. Summer is the 152nd to 243rd day of the year, transition season is the 60th to 151st day and the 244th to 334th day of the year, and winter is the 1st to 59th day and the 335th to 365th day of the year. After determining the frequency of each hour, step S430 is executed, which involves repeatedly arranging the corresponding typical scene times according to their frequency to form a complete sequence for a specific hour within that season. This process is repeated for each hour, resulting in the corresponding sequence of sizes. Day season ×24The original disordered boundary sequence of each season is combined into a full-year disordered boundary sequence.

[0056] The genetic algorithm model formula of step S440 is: (14); In formula (14), F ED represents the fitness function, D(X,Y) is a standard Euclidean distance function, D WP (X,Y), D DNI (X,Y), D Load (X,Y), D DE (X,Y) and respectively correspond to the wind and light loads, and D DE Source load difference DE The weighted Euclidean distance between the reordered sequence and the reference sequence is composed, and the reordered sequence is based on the original disordered boundary sequence, and the reference sequence is the data vector sequence corresponding to the representative meteorological year data set in step S100. The representative meteorological year data is also divided according to the same number of seasonal days, and the corresponding data vectors are taken as reference sequences; X, Y and respectively are the data vectors corresponding to the corresponding reordered sequence and reference sequence; the corresponding X The reordered sequence can be determined from the corresponding data vector of the full-year disordered boundary sequence. It should be noted that the parameters corresponding to the reference sequence and the matched reordered sequence should be of the same type of parameter type; 、 、 、 and respectively are weight coefficients. Here, since the wind and light are more volatile, they are given higher decision weights. In actual application, it can be adjusted according to actual needs.

[0057] Therefore, combined with formula (14), the optimization of the full-year disordered boundary sequence can be completed to obtain the ordered sequence of the corresponding season Day season ×24 ] opt ; and then combined into 365×24 the full-year scene boundary according to the seasonal day order. The full-year scene boundary has better source-load matching characteristics and time sequence characteristics compared with the traditional uncertain boundary. The full-year scene boundary formula is: (15).

[0058] Reference Figure 11 , the reference year DE curve in the figure is one of the representative meteorological year data sets, it can be seen that the corresponding load source difference DE curve in the summer of the application has the same time sequence characteristics as the reference sequence after time sequence optimization by the genetic algorithm, so the curves of other parameters after optimization will also have the same characteristics, which will not be specifically shown this time, and the final annual scene boundary has the corresponding time sequence characteristics.

[0059] Therefore, compared with the conventional generated uncertain boundary, the annual scene boundary obtained by the steps S100 to S400 of the application can retain the source-load matching characteristics and time sequence characteristics, which is more reasonable and can effectively reduce the error of system operation.

[0060] In a second aspect, with reference to Figure 10 , the technical scheme of the application also relates to a planning method based on source-load matching and time sequence characteristics, which further comprises the following steps: Step S500, input the annual scene boundary in the above-mentioned source-load matching and time sequence characteristic-based uncertain scene generation method as the system operation boundary, and optimize and solve the system capacity configuration based on a multi-objective optimization algorithm to determine the capacity random planning variable set of the system.

[0061] Specifically, during optimization, the maximum net present value of the system, the minimum wind and light curtailment rate and the minimum power following deviation are taken as the optimization objectives, and a multi-objective optimization algorithm is used for optimization and solution to obtain the capacity random planning variable set of the system. Similarly, the annual scene boundary in the above-mentioned step S400 can be input as the system operation boundary, a multi-objective optimization algorithm is used to optimize and solve the system capacity configuration, and the ideal point is formed by the optimal values of each single objective in the Pareto solution set, and the decision variable combination corresponding to the solution with the minimum Euclidean distance from the point is taken as the capacity random planning variable set. In this embodiment, the NSGA-II algorithm based on the Pareto front is also used, the maximum net present value, the minimum wind and light curtailment rate and the minimum power following deviation are taken as the objectives, and finally the ideal point is formed by the optimal values of each single objective in the Pareto solution set, and the decision variable combination corresponding to the solution with the minimum Euclidean distance from the point is taken as the capacity random planning variable set. In this embodiment, the capacity random planning variable set includes: Number of wind power installed units N w2 , unit: [unit]; Photovoltaic installed capacity C pv2 , unit: [MW]; Thermal power installed capacity C csp2 , unit: [MW]; duration of heat storage TES 2 in [h]; mirror field magnification SM 2 .

[0062] Meanwhile, it is worth noting that the reference capacity optimization variable set determined by step S100 can be determined by analyzing step S141 or step S137. The detailed steps for determining the reference capacity optimization variable set are the technical means for determining the capacity stochastic programming variable set in the industry, but in the present application, the decision variable set obtained by the conventional technical means is taken as the reference variable. The present application takes the reference capacity optimization variable set as the reference value in the source-load matching of step S200, so as to confirm a more reasonable capacity stochastic programming variable set, effectively reduce the operation performance deviation, and effectively improve the engineering implementability of the stochastic programming scheme.

[0063] Based on the above, the following is based on the source-load matching and the uncertainty scenario generation and planning method based on the timing characteristics, and combined with actual data for simulation analysis, as follows: In this embodiment, the meteorological data (including wind speed, radiation, temperature and other parameters) of Xilingol, China (43°93'N, 116°05'E) from 2009 to 2018 for 8760h x 10 years and the typical year 8760h wind-light-load data and conventional uncertain boundary are adopted. First, the deterministic capacity optimization is carried out under the boundary of the typical year, and the optimal decision planning obtained is the deterministic capacity optimization result as follows: 75 wind turbines (112.5 MW), 92 MW of photovoltaic installation, 92 MW of photo-thermal installation, 16 hours of heat storage duration, and 1.823 of mirror field magnification; The deterministic capacity optimization result is taken as the reference capacity optimization variable set, and then the annual scenario boundary generated based on steps S100 to S400 of the present application is taken as the improved uncertain boundary.

[0064] On this basis, combined with the scheme of the present application: 1. The first uncertain boundary is obtained by combining only steps S200 and S300, i.e. only combining source-load matching and "point clustering", and then not considering timing optimization in step S400, so as to obtain the corresponding first uncertain boundary; 2. Without considering the source-load matching problem, the corresponding typical scenario set and probability set are obtained by directly performing "line clustering" for scenario reduction in the conventional manner, i.e. taking "day" as the unit, and then only combining the timing optimization of step S400 to obtain the corresponding uncertain boundary as the second uncertain boundary.

[0065] In order to highlight the rationality and effectiveness of the annual scene boundary, i.e. the uncertain boundary generation scheme of the present application, the above-mentioned corresponding system operation index performance is compared with the average performance of the index under the actual boundary from 2019 to 2023 under the same decision planning scheme [75, 92, 92, 16, 1.823], and the results are shown in Table 1: From the comparison results in Table 1, it can be seen that the relative error of the uncertain boundary generated by the conventional random planning scheme in the performance of the three system operation indicators is more than 10% compared with the actual average, and the relative error of the wind and light rejection rate is even as high as 99.03%, which shows that the conventional uncertain boundary is not reasonable, resulting in serious operation performance deviation of the system. Meanwhile, referring to the first uncertain boundary and the second uncertain boundary, the corresponding error is also high, which shows that considering the source-load matching feature, "point clustering" or time sequence feature alone cannot completely solve the problems in the conventional scheme, and the three need to be combined simultaneously, i.e. the present application simultaneously combines the steps S200 of distinguishing the source-load matching, the step S300 of reducing the scene in the manner of "point clustering", and the step S400 of "optimizing in combination with the time sequence feature" to achieve the optimal effect. By using the scheme of the present application, the relative deviation of the net present value, the wind and light rejection rate and the power following deviation from the actual average is 1.65%, 6.15% and 0.86%, which is reduced by 13.76%, 99.28% and 13.22% compared with the conventional scheme, fully showing that the present application is more reasonable and can effectively reduce the operation performance deviation.

[0066] And the uncertainty capacity planning scheme determined based on the steps S100 to S500 of the present application, i.e. the capacity random planning variable set is: There are 73 wind power units (109.5 MW), 101 MW of photovoltaic power, 100 MW of photo-thermal power, 16 hours of heat storage time, and 1.499 of mirror field magnification; The system index comparison under different boundaries under the capacity random planning variable set is shown in Table 2: From Table 2, it can be seen that under the planning method based on source-load matching and time sequence feature of the present application, the index performance deviation of the improved uncertain boundary from the actual average is still within 10%, which fully shows the rationality of the method of the present application again.

[0067] According to the embodiments of the present application, some effects can be achieved as follows: by dividing the sample points in each hour by the source-load difference value, the characteristics of positive / negative differences of source-load matching can be distinguished, and by reducing the scene in each hour respectively, the characteristics of source-load matching can be effectively retained to generate more accurate typical scenes, the accuracy of system reliability evaluation is improved, and the historical data is optimized by using the time sequence characteristics, and finally the annual scene boundary with time sequence characteristics can be generated, the source, load and the time sequence correlation and operation coupling characteristics between the source and load are combined, the rationality of the annual scene boundary is effectively improved, the subsequent operation performance deviation is significantly reduced, the uncertain boundary is effectively optimized, the rationality is improved, and then the results of the optimal uncertainty scene and capacity random planning variable set can be obtained, the operation performance deviation is effectively reduced, and the engineering implementability of the random planning scheme is effectively improved.

[0068] It should be appreciated that the method steps in the embodiments of the present application can be realized or implemented by computer hardware, a combination of hardware and software, or through computer instructions stored in a non-transitory computer readable memory. The uncertainty scene generation method based on source-load matching and time sequence characteristics can use standard programming techniques. Each program can be implemented in a high-level procedural or object-oriented programming language to communicate with a computer system. However, if necessary, the program can be implemented in assembly or machine language. In any case, the language can be a compiled or interpreted language. In addition, the program can be run on a programmed special-purpose integrated circuit for this purpose.

[0069] In addition, the operations of the processes described herein can be performed in any suitable order unless otherwise indicated herein or otherwise clearly contradicted by context. The processes described herein (or variations and / or combinations thereof) can be performed under the control of one or more computer systems configured with executable instructions (e.g., executable instructions, one or more computer programs, or one or more applications) to perform operations, and can be implemented as code (e.g., executable instructions, one or more computer programs, or one or more applications) executing collectively on one or more processors, by hardware, or combinations thereof. The computer programs include a plurality of instructions executable by one or more processors.

[0070] Further, the methods can be implemented in any type of computing platform operably connected to a suitable computing platform, including but not limited to a personal computer, mini-computer, mainframe, workstation, network or distributed computing environment, separate or integrated computer platforms, or in communication with charged particle tools or other imaging devices, and the like. Aspects of the present application can be implemented in machine-readable code stored on a non-transitory storage medium or device, whether removable or integrated to the computing platform, such as a hard disk, optical read and / or write storage media, RAM, ROM, and the like, such that it can be read by a programmable computer to configure and operate the computer to perform the processes described herein when the storage medium or device is read by the computer. In addition, the machine-readable code, or portions thereof, can be transmitted over wired or wireless networks. The present application described herein includes these and other different types of non-transitory computer readable storage media when such media include instructions or programs that implement the steps described above in conjunction with a microprocessor or other data processor. The present application can also include the computer itself when programmed in accordance with the methods and techniques described herein.

[0071] The computer program can be applied to input data to perform the functions described herein to transform the input data to generate output data that is stored to non-volatile memory. The output information can also be applied to one or more output devices such as a display. In a preferred embodiment of the present application, the transformed data represents a physical and tangible object, including a particular visual depiction of the physical and tangible object produced on a display.

[0072] The above description is only preferred embodiments of the present application, and the present application is not limited to the above-described embodiments, but any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the present application. The technical solutions and / or embodiments of the present application can have various modifications and changes within the scope of the present application.

Claims

1. A method for generating uncertainty scenarios based on source load matching and timing characteristics, characterized in that, The method comprises the following steps: According to the historical data of the target area in previous years, a representative meteorological year data set, a reference capacity optimization variable set, and a full-year wind and light load initial scenario time set are determined respectively; Based on the reference capacity optimization variable set, the source-load difference value of the wind and light load sample point of each hour of the full-year wind and light load initial scenario time set is calculated, and according to the positive and negative of the corresponding source-load difference value, the wind and light load sample points in each hour are divided into a surplus power state region set where source is greater than load and a power shortage state region set where source is less than load; Based on the distance clustering algorithm, the wind and light load sample points of the corresponding surplus power state region set and the power shortage state region set are clustered hour by hour respectively, and after iterative convergence, a full-year typical scenario time set and a corresponding first probability set are determined; According to the full-year typical scenario time set and the first probability set, a full-year unordered boundary sequence is determined, and based on the time sequence of the representative meteorological year data set, the full-year unordered boundary sequence is optimized according to a genetic algorithm to generate a full-year scenario boundary with time sequence characteristics. 2.The source-load matching and timing feature based uncertainty scenario generation method of claim 1, wherein, The representative meteorological year data set is determined by the reference year generation method or the typical year construction method based on historical data in previous years. 3.The source-load matching and timing feature based uncertainty scenario generation method of claim 2, wherein, The reference year generation method comprises the following steps: According to the historical data of the target area in at least 5 consecutive years, the average wind speed, annual total radiation, and total load of each year are determined; Based on the parameters of the average wind speed, the annual total radiation, and the total load of the target area in at least 5 consecutive years, the corresponding parameters of each year are sorted respectively to determine the ranking of the corresponding parameters of each year; Based on the ranking of the corresponding parameters of each year, the comprehensive median deviation of each year is determined. Based on the comprehensive median deviation degree of each year ， The historical data of the year with the minimum deviation degree is selected as the representative meteorological year data set; when there are multiple years with the same and minimum deviation degree, the year with a more moderate annual ranking is selected as the representative meteorological year data set according to a preset parameter priority order, and the parameter priority order is the average wind speed, the total annual radiation and the total load. 4.The source-load matching and timing feature uncertainty scenario generation method according to claim 2, wherein, The typical year construction method comprises the following steps: The historical data of the target area in at least 10 consecutive years is preprocessed to determine the daily average value of the first target parameter of each month of each year, and the first target parameter includes solar radiation, wind speed, dry bulb temperature, and humidity; According to the daily average value of each of the first target parameters, the long-term average cumulative distribution function of each month of each year and the single-year cumulative distribution function of each month are determined respectively; According to the mean square error of the single-year cumulative distribution function and the long-term average cumulative distribution function corresponding to each month, and based on months 1 to 12, the month with the smallest mean square error corresponding to the month is selected as the target month; The hour data of the target month corresponding to months 1 to 12 are spliced and smoothed in order to serve as the representative meteorological year data set. 5.The source-load matching and timing feature based uncertainty scenario generation method of claim 1, wherein, The determination of the reference capacity optimization variable set comprises the following steps: The data of the representative meteorological year data set is taken as the system operation boundary, and the reference capacity configuration of the system is optimized and solved based on a multi-objective optimization algorithm to determine the reference capacity optimization variable set. 6.The source-load matching and timing feature based uncertainty scenario generation method of claim 1, wherein, The determination of the full-year wind and light load initial scenario time set comprises the following steps: Based on the historical data of the target area for at least 5 consecutive years, the Gaussian kernel density estimation method is used to fit the second target parameter in each hour to determine the probability density function corresponding to the second target parameter in each hour, and the cumulative distribution function corresponding to the second target parameter in each hour is obtained by integration, the second target parameter including wind power, solar radiation and power load; Based on the cumulative distribution function corresponding to each hour, the Frank-Copula function is combined to construct the wind-solar-load three-dimensional joint cumulative distribution function corresponding to the second target parameter in each hour; Based on the wind-solar-load three-dimensional joint cumulative distribution function in each hour, Monte Carlo or Latin hypercube sampling method is used to generate joint sampling points in each hour, and inverse transformation is performed on the corresponding joint sampling points in each hour to determine the sample value corresponding to each joint sampling point in each hour, the sample value including the first sample value corresponding to wind power, the second sample value corresponding to solar radiation and the third sample value corresponding to power load, and the first sample value, the second sample value and the third sample value corresponding to each hour are taken as the wind-solar-load sample point in each hour to constitute the wind-solar-load initial scene in each hour; The wind-solar-load initial scene in each hour is combined by day to constitute the wind-solar-load initial full-day scene, and is divided into summer, transition season and winter to generate the wind-solar-load initial scene set corresponding to 24 hours in summer, 24 hours in transition season and 24 hours in winter, and all the wind-solar-load initial scene sets are combined to serve as the annual wind-solar-load initial scene set. 7.The source-load matching and timing feature based uncertainty scenario generation method of claim 6, wherein, The determination of the reference capacity optimization variable set includes the following steps: Based on the distance clustering algorithm, all the wind-solar-load initial full-day scenes are clustered day by day, and after iterative convergence, a set of annual typical scene days and a corresponding second probability set are determined; According to the index performance of each typical scene day in the set of annual typical scene days, and according to the weighted sum of the corresponding second probability set, the annual index performance is determined; The index performance is taken as the system operation boundary input, and the reference capacity configuration is optimized and solved based on the multi-objective optimization algorithm to determine the reference capacity optimization variable set. 8.The source-load matching and timing feature uncertainty scenario generation method according to claim 1, wherein, The annual wind-solar-load initial scene set and the annual typical scene set are divided into summer, transition season and winter, wherein summer is the 152th to 243rd day of the year, transition season is the 60th to 151st day and the 244th to 334th day of the year, and winter is the 1st to 59th day and the 335th to 365th day of the year. 9.The source and load matching and timing feature uncertainty scenario generation method according to any one of claims 1 to 8, characterized in that, The step of determining the annual unordered boundary sequence according to the annual typical scene set and the first probability set, and optimizing the annual unordered boundary sequence according to the genetic algorithm based on the time sequence of the representative meteorological year data set to generate the annual scene boundary with time sequence characteristics includes: According to the annual typical scene set and the first probability set, the first probability corresponding to each typical scene is determined. multiply each of the first probability by corresponding seasonal day number, to obtain corresponding frequency of occurrence in corresponding hour in corresponding typical scene; According to the frequency, each of the typical scene time repeating permutation combination is formed into summer days ×24 hours, transition season days ×24 hours and winter days ×24 hours, the disordered boundary sequence corresponding to the disordered boundary sequence is combined into the whole year disordered boundary sequence; According to genetic algorithm, each seasonal corresponding annual unordered boundary sequence is sorted and optimized respectively to generate the annual scene boundary with time sequence characteristics.

10. A method for planning based on source load matching and timing characteristics, the method comprising: determining a plurality of source load matching and timing characteristics; and generating a plan based on the plurality of source load matching and timing characteristics. The method comprises the following steps: The annual scene boundary in the source load matching and time sequence characteristic based uncertainty scene generation method according to any one of claims 1 to 9 is input as a system operation boundary, and the system capacity configuration is optimized and solved based on a multi-objective optimization algorithm to determine the capacity random planning variable set of the system.

Images