A distributed resource timing scenario generation method

Abstract

This invention discloses a method for generating time-series scenarios of distributed resources, belonging to the field of grid analysis technology for distributed energy access. The method includes: acquiring historical operating data, historical meteorological data, and meteorological forecast data of distributed photovoltaic (PV) and loads within the area to be analyzed; generating historical sample daily feature vectors and scenario-to-be-generated daily feature vectors based on the acquired data, and determining a set of similar days; acquiring historical operating data of distributed PV and loads on similar days within the area to be analyzed based on the set of similar days, analyzing their power fluctuation range distribution characteristics and power probability distribution characteristics, and obtaining the upper and lower bounds of fluctuation and the power probability density function of distributed PV and loads; and obtaining a set of time-series power scenarios for distributed PV and loads considering fluctuation range correction based on the upper and lower bounds of fluctuation and the power probability density function of distributed PV and loads. This invention can fully utilize historical power information and fluctuation information to generate high-precision uncertainty scenarios.

Description

Technical Field

[0001] This invention relates to scenario uncertainty analysis technology for distributed resources, belonging to the field of robust scheduling optimization technology for microgrids with high proportion of distributed energy access, and particularly to a method for generating distributed resource time-series scenarios. Background Technology

[0002] With the integration of numerous distributed power sources into the distribution network and microgrids, their inherent stochastic and fluctuating characteristics will amplify the uncertainty in microgrid operation and control. Simultaneously, due to the introduction of load control measures such as demand response, traditional source-load structures and deterministic optimization methods are no longer adequate to reflect the shift in the role of distribution networks and microgrids from traditional energy consumers to energy progenitors in day-ahead dispatch.

[0003] Therefore, using time-series scenarios to quantify typical operating conditions of loads and photovoltaics in microgrids can provide valuable boundary constraints for system operation. Simultaneously, scenario-driven uncertainties should be effectively utilized to reduce the overly conservative tendency of traditional robust optimization while ensuring system stability, and to fully leverage the flexibility of the system's distributed resources. Summary of the Invention

[0004] The purpose of this invention is to provide a method for generating distributed resource time-series scenarios, which can fully utilize historical power information and fluctuation information to generate high-precision uncertain scenarios. The technical solution adopted by this invention is as follows.

[0005] On one hand, the present invention provides a method for generating distributed resource time-series scenarios, including:

[0006] Acquire historical operational data of distributed photovoltaic and load, historical meteorological data, and meteorological forecast data for the area to be analyzed;

[0007] Based on the acquired data, generate historical sample day feature vectors and scene-specific day feature vectors to be generated, and determine the set of similar days;

[0008] Based on the set of similar days, historical operating data of distributed photovoltaic and load on similar days in the region to be analyzed are obtained, and the distribution characteristics of their power fluctuation range and power probability distribution characteristics are analyzed to obtain the upper and lower bounds of power fluctuation and power probability density function of distributed photovoltaic and load.

[0009] Based on the upper and lower bounds of power fluctuations and the power probability density function of the distributed photovoltaic and load, a time-series power scenario set for distributed photovoltaic and load considering fluctuation range correction is obtained.

[0010] Optionally, the meteorological characteristic data types in the historical meteorological data and meteorological forecast data include: maximum temperature, minimum temperature, average temperature, precipitation, relative humidity, average irradiance, average wind speed, and average air pressure.

[0011] The process of generating historical sample daily feature vectors and scene-specific daily feature vectors based on the acquired data includes:

[0012] Calculate the daily average photovoltaic power and load power based on historical operating data of distributed photovoltaic and load.

[0013] Calculate the Spearman correlation coefficients between the daily average photovoltaic power and load power and various types of historical meteorological characteristic data sequences;

[0014] For each meteorological feature data type, the corresponding Spearman correlation coefficient calculation results are compared with the correlation coefficient threshold, and the feature vector type is selected based on the comparison results;

[0015] Based on the selected feature vector type, the feature vectors for each historical sample day and the day the scene is to be generated are obtained, as follows:

[0016]

[0017] In the formula, Indicates the number of historical sample days. The feature vector representing the day to be generated for the scene. Indicates the first A historical day sample feature vector, Indicates the first The m-th feature value in the feature vector of historical day samples.

[0018] In the above technical solution, the correlation coefficient threshold can be set or adjusted according to experience or needs, and meteorological feature data types with Spearman correlation coefficient calculation results not less than the correlation coefficient threshold are selected for generating feature vectors.

[0019] Optionally, determining the set of similar days includes:

[0020] Based on the feature vectors of each historical sample day and the day to be generated in the scenario, calculate the comprehensive correlation coefficient between the meteorological features of the day to be generated in the scenario and the meteorological features of each historical sample day.

[0021] Historical samples with a comprehensive correlation coefficient higher than the comprehensive correlation coefficient threshold are combined with the date the scene is to be generated to form a set of similar dates.

[0022] Optionally, the formula for calculating the comprehensive correlation coefficient is:

[0023]

[0024] in, For the first The comprehensive correlation coefficient between historical samples and the meteorological characteristics of the day the scene is to be generated. For the first The geometric correlation between historical samples and the date the scene is to be generated. For the first The distance correlation between each historical sample and the date the scene is to be generated is expressed as follows:

[0025] ,

[0026] ;

[0027] Among them, intermediate variables , .

[0028] Optionally, historical operating data of distributed photovoltaic and load on similar days within the region to be analyzed can be used to analyze the distribution characteristics of their power fluctuation ranges, including:

[0029] For the distributed photovoltaic (PV) and load time-series data of each historical sample day within the similar day set, the data fluctuation value for each time period is calculated to obtain the fluctuation range of the distributed PV and load data for each time period, expressed as:

[0030]

[0031] in, ,

[0032] In the formula, This represents the fluctuation value of distributed photovoltaic or load data within time period t. Let be the power fluctuation value from the start time to the end time of time period t within day d, T be the sampling time interval between adjacent time series data, and S be the set of similar days.

[0033] Optionally, analyze the power probability distribution characteristics of historical operating data for similar daily distributed photovoltaic and loads within the area to be analyzed, including:

[0034] Choose the kernel function and its bandwidth;

[0035] For each data point on multiple similar days, the local density of each data point is estimated using the selected kernel function;

[0036] The global density estimate is obtained by weighted averaging the local densities of all data points.

[0037] Among them, for distributed photovoltaic or load datasets at any data point t on multiple similar days. The probability density function of this dataset is estimated using the Gaussian kernel function as follows:

[0038]

[0039] In the formula, It is at point The estimated density value at that location; It represents the number of historical sample days; Is at each data point The Gaussian kernel function applied at the given location; h is the kernel function bandwidth.

[0040] Optionally, based on the upper and lower bounds of the power fluctuations and the power probability density function of the distributed photovoltaic and load, a time-series power scenario set considering fluctuation interval correction is obtained, including:

[0041] Based on the fluctuation range and power probability density function of distributed photovoltaic / load data for each time period, the probability density sampling correction space for each power scenario is calculated for each time period.

[0042] Simple random sampling is performed based on the probability density sampling correction space to obtain photovoltaic / load power sampling values ​​for each time period, further resulting in the time-series power scenario set. This time-series power scenario set includes multiple power time-series data subsets corresponding to different scenarios, and each power time-series data subset includes power data corresponding to each time period.

[0043] Optionally, the step of calculating the probability density sampling correction space for each power scenario in each time period includes:

[0044] The power fluctuation range for time period t+1 is defined as follows:

[0045]

[0046] in, and These represent the power sample values ​​for time periods t and t+1 under the s-th power scenario, respectively;

[0047] Will Substituting the endpoints of the power fluctuation range into the probability density function for time period t+1, we obtain the probability density sampling correction space for time period t+1 under the s-th power scenario, expressed as:

[0048]

[0049] .

[0050] In the above technical solutions, the sampling interval of each time series scenario is corrected according to the fluctuation range of adjacent time moments, and then random sampling is performed within the corrected interval to effectively simulate the fluctuation characteristics of photovoltaic load power within adjacent time intervals.

[0051] Optionally, the method further includes: taking the time-series power scenario set as the time-series original scenario set, forming a scenario-driven original uncertainty set based on the time-series original scenario set, and using finite coverage theory to solve for several simplified uncertainty sets covering the original uncertain scenarios.

[0052] Optionally, the original uncertainty set driven by the scenario is represented as:

[0053]

[0054] In the formula, and These represent the number of distributed photovoltaic (PV) and load generation scenarios, respectively. , and , Let be the sampled value and the average value of the distributed photovoltaic and load power scenarios in time period t, respectively, for the i-th time series scenario. Let Γ be the initial set of uncertainties driven by the scenario, and Γ be the budget parameter, indicating that there are at most Γ time periods when the maximum uncertainty range can be reached. , These represent the sampled values ​​of distributed photovoltaic power and load at time t, respectively.

[0055] The method of using finite coverage theory to solve for several simplified uncertainty sets covering the original uncertain scenario includes:

[0056] The transformed optimization problem, based on the original uncertain set driven by the scenario, is expressed as:

[0057]

[0058] In the formula,

[0059] and Let be the upper and lower bounds of the power value of the scene in time period t within the nth finite coverage set;

[0060] and Let be the upper and lower bounds of the scene at time t within the original scene set formed by the s-th sample, that is: , ; M is the balance factor; M is a very large number (Big M method). It is a 0-1 variable that guides the mapping of the original temporal scene set to the finite coverage scene set N; Indicates the balance factor; Represents a finite covering set;

[0061] Solving the optimization problem yields an uncertain set that covers the scenario-driven conditions. n indeterminate sets of boxes ,Right now: .

[0062] The objective function of the above optimization problem consists of two terms: the first term minimizes the total size of all uncertain sets, the second term minimizes the size of the largest uncertain set, and the balance factor. This term is used to balance the weights of the first and second terms in the MILP objective function. The first term controls the complexity of the uncertainty set, while the second term balances the size of a single uncertainty set, reducing the running cost of considering the uncertainty set in subsequent robust optimizations.

[0063] In a second aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the distributed resource time-series scenario generation method as described in the first aspect.

[0064] Beneficial effects

[0065] The distributed resource time-series scene generation method of the present invention takes into account the reduction of the generated scene width by utilizing the fluctuation information of historical data. It proposes a scene generation method based on fluctuation interval correction. By constructing a daily feature vector, historical data samples with high correlation with the day to be generated are selected based on the comprehensive correlation coefficient to form a similar day set. Based on the similar day set and the historical data of distributed photovoltaic-load, the source-load time-series power probability density function is fitted and the historical fluctuation interval is calculated. The upper and lower bounds of the fluctuation interval are used to correct the sampling interval based on the probability density function of source-load power, thereby forming a set of scenes before the current day, which improves the quality of scene generation.

[0066] On the other hand, based on the obtained day-ahead scenario set, this invention defines a scenario-driven original uncertainty set. By drawing on the finite coverage theory, a complex source-load scenario uncertainty set is covered by a finite number of simpler uncertainty sets, transforming the scenario-driven uncertainty set into multiple simple uncertainty sets that jointly cover it. This can reduce the conservatism of robust optimization modeling for microgrid configuration and scheduling while ensuring system robustness. Attached Figure Description

[0067] Figure 1 The diagram shown is a schematic flowchart of a sampling method based on fluctuation range correction in one embodiment of the method of the present invention.

[0068] Figure 2 The diagram shown is a flowchart of an embodiment of the method of the present invention. Detailed Implementation

[0069] The following description, in conjunction with the accompanying drawings and specific embodiments, provides further details.

[0070] Example 1

[0071] This embodiment introduces a distributed resource time-series scenario generation method suitable for day-ahead scheduling. It can generate a high-precision set of day-ahead scenarios based on the characteristics of historically similar days, providing relatively accurate reference data for day-ahead scheduling. The method of this embodiment includes:

[0072] Acquire historical operational data of distributed photovoltaic and load, historical meteorological data, and meteorological forecast data for the area to be analyzed;

[0073] Based on the acquired data, generate historical sample day feature vectors and scene-specific day feature vectors to be generated, and determine the set of similar days;

[0074] Based on the set of similar days, historical operating data of distributed photovoltaic and load on similar days in the region to be analyzed are obtained, and the distribution characteristics of their power fluctuation range and power probability distribution characteristics are analyzed to obtain the upper and lower bounds of power fluctuation and power probability density function of distributed photovoltaic and load.

[0075] Based on the upper and lower bounds of power fluctuations and the power probability density function of the distributed photovoltaic and load, a time-series power scenario set for distributed photovoltaic and load considering fluctuation range correction is obtained.

[0076] The specific implementation of this embodiment includes the following parts.

[0077] I. Formation of Similar Day Sets Based on Comprehensive Correlation Coefficient

[0078] This embodiment first obtains historical operating data of distributed photovoltaic and load in the region, as well as historical meteorological data and weather forecasts. Then, based on the numerical weather forecast and historical sample meteorological data of the day to be generated in the scenario, it forms the feature vector of the historical sample day and the feature vector of the day to be generated in the scenario. By calculating the meteorological comprehensive correlation coefficient, it selects samples with high comprehensive similarity as the constituent elements of the similar day set, thus obtaining the historical similar sample set.

[0079] The meteorological characteristic data types in the historical meteorological data and meteorological forecast data include: maximum temperature, minimum temperature, average temperature, precipitation, relative humidity, average irradiance, average wind speed, and average air pressure.

[0080] This section specifically includes the following steps:

[0081] Step 1: Construct the historical sample daily feature vector and the scene's daily feature vector to be generated, including:

[0082] Calculate the daily average photovoltaic power / load power based on historical operating data of distributed photovoltaic and load;

[0083] Calculate the Spearman correlation coefficient between the daily average photovoltaic power / load power and various types of historical meteorological characteristic data sequences;

[0084] For each meteorological feature data type, the corresponding Spearman correlation coefficient calculation result is compared with the correlation coefficient threshold. Based on the comparison result, the feature vector type is selected. The correlation coefficient threshold can be set or adjusted according to experience or needs. Meteorological feature data types with Spearman correlation coefficient calculation results not less than the correlation coefficient threshold are selected for generating feature vectors.

[0085] Based on the selected feature vector type, the feature vectors for each historical sample day and the day the scene is to be generated are obtained, as follows:

[0086]

[0087] In the formula, Indicates the number of historical sample days. The feature vector representing the day to be generated for the scene. Indicates the first A historical day sample feature vector, Indicates the first The m-th feature value in the feature vector of historical day samples.

[0088] Step two involves determining the set of similar days, including: calculating the comprehensive correlation coefficient between the meteorological characteristics of the day to be generated and the meteorological characteristics of each historical sample day, based on the feature vectors of each historical sample day and the day to be generated; and forming a set of similar days with historical samples whose comprehensive correlation coefficient is higher than the threshold. The calculation of the comprehensive correlation coefficient includes the following steps:

[0089] 1.2.1) Calculation of geometric correlation degree:

[0090] intermediate variables

[0091]

[0092] In the formula, That is, the first The geometric correlation between historical samples and the date the scene is to be generated.

[0093] 1.2.2) Distance correlation calculation:

[0094] intermediate variables

[0095]

[0096] In the formula, That is, the first The distance correlation between historical samples and the date the scene is to be generated.

[0097] 1.2.3) Combine geometric correlation and distance correlation to calculate the comprehensive correlation. This is used to describe the overall correlation between the date the scene is to be generated and historical samples, and is expressed as:

[0098] .

[0099] If the comprehensive correlation coefficient between the feature vector of the i-th historical sample and the feature vector of the scene to be generated day is higher than a certain threshold, it is considered that there is a similarity effect; otherwise, they are not similar. In this case, historical samples similar to the scene to be generated day can be combined into a similar day set.

[0100] II. Volatility Analysis and Power Probability Density Modeling of Similar Day Sets

[0101] Based on the similarity set of historical samples, we analyze the distribution characteristics of power fluctuation intervals and power probability distribution characteristics to obtain the upper and lower bounds of power fluctuations for distributed photovoltaic and load. We then use kernel density estimation to model the probability density fitting function of distributed photovoltaic and load power.

[0102] The analysis process for the upper and lower bounds of power fluctuations in distributed photovoltaic systems and loads is as follows:

[0103] Based on the day-ahead scheduling, a day is divided into 96 time periods. In the set of similar days, the source load power fluctuation between adjacent time periods can be expressed as:

[0104]

[0105]

[0106] In the formula, This represents the fluctuation value of distributed photovoltaic or load data within time period t. Let S be the power fluctuation value from the start time to the end time of time period t within day d, where T is the sampling time interval between adjacent time series data, and S is the set of similar days. This yields the upper and lower bounds of the power fluctuation of distributed photovoltaic / load systems.

[0107] .

[0108] In a set of similar days, the probability density function modeling process for the distributed photovoltaic power / load power of any time period is as follows:

[0109] 2.2.1) Choose a kernel function; the Gaussian kernel function is typically used.

[0110]

[0111] The probability density function of the dataset can be estimated using the Gaussian kernel function as follows:

[0112]

[0113] In the formula, This is the estimated density value at data point 𝑥, where 𝑛 is the sample size. At data point 𝑥 i The Gaussian kernel function applied is h, which is the bandwidth, controlling the width of each kernel function and the smoothness of the density estimate.

[0114] 2.2.2) Smoothing Bandwidth. The width (bandwidth) of the kernel function controls the degree of smoothing. The larger the bandwidth, the smoother the estimated density curve; the smaller the bandwidth, the closer the curve will be to the local features of the original data.

[0115] 2.2.3) Overlay all data points. For each data point, a kernel function is used to estimate the local density of that point, and then a weighted average of the local densities of all data points is performed to form a global density estimate.

[0116] In this embodiment, a distributed photovoltaic or load dataset with multiple similar days at any data point t is used. The probability density function of this dataset is estimated using the Gaussian kernel function as follows:

[0117]

[0118] In the formula, It is at point The estimated density value at that location; It represents the number of historical sample days; Is at each data point The Gaussian kernel function applied at the given location; h is the kernel function bandwidth.

[0119] III. Generation of the original scenario of distributed photovoltaic and load considering fluctuation range correction: Based on the historical probability distribution and historical fluctuation range of the power of the distributed photovoltaic and load, the upper and lower bounds of the fluctuation range at time t are used to correct the sampling interval of the time series power of the distributed photovoltaic and load at time t+1, so as to obtain the original set of the time series power scenario of distributed photovoltaic and load considering fluctuation range correction.

[0120] This embodiment, based on the probabilistic modeling of photovoltaic / load power in each time period within a similar day set, uses a simple random sampling method for each of the 96 time periods to generate day-ahead scenarios. To narrow the scenario width, the sampling interval of each sampling time point is corrected based on the fluctuation amplitude of adjacent times, and then random sampling is performed within the corrected power probability interval, effectively simulating the fluctuation characteristics of photovoltaic load power in adjacent time periods. Specifically, when sampling the s-th scenario time period by time, adjacent power values ​​must satisfy:

[0121]

[0122] Will Substitution In this process, the power probability density sampling interval correction expression for time period t+1 is obtained:

[0123]

[0124]

[0125] At this point, based on the sampling results at time t+1, the sampling interval of the power probability density function of photovoltaic and load in time period t+1 is adjusted from [0,1] to [c1,c2]. Then, the corrected photovoltaic / load power sampling value sequence can be obtained further through a simple random sampling method. The specific sampling process is as follows: Figure 1 As shown. (The sentence is incomplete and requires more context.) Figure 1 Following the random sampling process shown, multiple power time-series data subsets corresponding to different scenarios can be obtained. Each power time-series data subset includes power data for each corresponding time period. The number of scenarios corresponding to distributed photovoltaic and load power sampling is [number missing]. , The number of sampling periods can be set as needed. The value is 96, which corresponds to the number of data points in the power time-series data subset.

[0126] Example 2

[0127] To reduce the conservatism of robust optimization modeling for microgrid configuration and scheduling while ensuring system robustness, this embodiment, based on Embodiment 1, defines a scenario-driven primordial uncertainty set based on the day-ahead scenario set obtained therein. Drawing on finite coverage theory, a complex source-load scenario uncertainty set is covered by a finite number of simpler uncertainty sets, transforming the scenario-driven uncertainty set into a set covered by multiple simpler uncertainty sets. Specifically, the time-series power scenario set is used as the time-series primordial scenario set. Based on this set, a scenario-driven primordial uncertainty set is formed. Finite coverage theory is used to solve the corresponding mixed-integer linear programming problem, obtaining the number of box-shaped uncertainty sets and corresponding parameters of the finitely covered primordial uncertainty scenario, thus completing the transformation from a complex primordial uncertainty set to a finite number of simplified box-shaped uncertainty sets.

[0128] Specifically, based on the sampling results, the original uncertainties of scenario-driven distributed photovoltaic power and loads under constraints can be expressed as:

[0129]

[0130] In the formula, and These represent the number of distributed photovoltaic (PV) and load generation scenarios, respectively. , and , Let be the sampled value and the average value of the distributed photovoltaic and load power scenarios in time period t, respectively, for the i-th time series scenario. Let Γ be the initial set of uncertainties driven by the scenario, and Γ be the budget parameter, indicating that there are at most Γ time periods when the maximum uncertainty range can be reached. , These represent the sampled values ​​of distributed photovoltaic power and load at time t, respectively.

[0131] Based on existing sampling scenarios And its corresponding upper and lower bounds of fluctuation. , This leads to a method for generating finite-coverage uncertainty sets based on historical databases and mixed-integer linear programming (MILP) problems. The optimization problem for uncertainty set transformation can be formulated as:

[0132]

[0133] In the formula,

[0134] and Let these be the upper and lower bounds of the scene at time t within the nth finite coverage set;

[0135] and Let be the upper and lower bounds of the scene at time t within the original scene set formed by the s-th sample, that is: , ; M is the balance factor; M is a very large number (Big M method). It is a 0-1 variable that guides the mapping of the original scene set S to the finitely covered scene set N.

[0136] The objective function consists of two terms. The first term minimizes the total size of all uncertain sets, and the second term minimizes the size of the largest uncertain set, with a parameter adjusting the weights of the two terms. It's important to note that the second term balances the size of individual uncertain sets. The reason for introducing the second term is that the required flexibility will be limited by the maximum volatility. This means that optimizing the largest uncertain set is important for reducing the operating cost of the considered uncertain sets. The following three lines of constraints ensure that all possible scenarios and their volatility can be contained within a finite coverage set. M is a sufficiently large number (Big M method). By solving the above mixed-integer linear programming problem, the original uncertain set U(S) can be transformed into a finite number (specifically, the number of optimization solutions) of simple box-shaped uncertain sets, which reduces the conservatism of robust optimization while ensuring coverage of the original uncertain sets.

[0137] Thus, the uncertain set based on scenario-driven principles... Can be an indeterminate set of n boxes Coverage, that is: .

[0138] In the implementation process described above, this invention not only considers the influencing factors of distributed photovoltaic and load to construct a set of similar days, but also takes into account the use of historical data fluctuation information to compress the sampling interval, thereby reducing the scene width of photovoltaic and load. It can effectively compress the uncertainty of scene generation by using two methods, namely the sampling interval and limited coverage, while ensuring that historical data information is fully considered. It fully explores and compresses the uncertainty of distributed resource operation scenarios. Its analysis method can be effectively applied to the characterization of uncertainty of distributed resources at the microgrid and distribution network levels, providing an accurate description of uncertainty for subsequent uncertain optimization models such as robust scheduling models, robust configuration models, and robust planning models.

[0139] Example 3

[0140] Based on the same inventive concept as Embodiments 1 and 2, this embodiment introduces a computer-readable storage medium storing a computer program that, when executed by a processor, implements the distributed resource time-series scenario generation and its uncertain set transformation method as described in Embodiment 1.

[0141] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0142] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0143] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0144] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0145] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims. All of these forms are within the protection scope of the present invention.

Claims

1. A method for generating distributed resource time-series scenarios, characterized in that, include: Acquire historical operational data of distributed photovoltaic and load, historical meteorological data, and meteorological forecast data for the area to be analyzed; Based on the acquired data, generate historical sample day feature vectors and scene-specific day feature vectors to be generated, and determine the set of similar days; Based on the set of similar days, historical operating data of distributed photovoltaic and load on similar days in the region to be analyzed are obtained, and the distribution characteristics of their power fluctuation range and power probability distribution characteristics are analyzed to obtain the upper and lower bounds of power fluctuation and power probability density function of distributed photovoltaic and load. Based on the upper and lower bounds of power fluctuations and the power probability density function of the distributed photovoltaic and load, a time-series power scenario set considering fluctuation range correction is obtained. Specifically, this includes: calculating the probability density sampling correction space for each power scenario for each time period based on the fluctuation range and power probability density function of the distributed photovoltaic / load data for each time period; performing simple random sampling based on the probability density sampling correction space to obtain the photovoltaic / load power sampling values ​​for each time period, and further obtaining the time-series power scenario set. Among them, the historical operating data of similar daily distributed photovoltaic and loads in the area to be analyzed are used to analyze the distribution characteristics of their power fluctuation range, including: For the distributed photovoltaic (PV) and load time-series data of each historical sample day within the similar day set, the data fluctuation value for each time period is calculated to obtain the fluctuation range of the distributed PV and load data for each time period, expressed as: ， in, , In the formula, This represents the fluctuation value of distributed photovoltaic or load data within time period t. Let T represent the power fluctuation value from the start time to the end time of time period t within day d, where T is the sampling time interval between adjacent time series data. For a set of similar days; Historical operating data of similar daily distributed photovoltaic and load data within the region to be analyzed are used to analyze the power probability distribution characteristics, including: Choose the kernel function and its bandwidth; For each data point on multiple similar days, the local density of each data point is estimated using the selected kernel function; The global density estimate is obtained by weighted averaging the local densities of all data points. Among them, for distributed photovoltaic or load datasets at any data point t on multiple similar days. The probability density function of this dataset is estimated using the Gaussian kernel function as follows: ， In the formula, It is at point The estimated density value at that location; It represents the number of historical sample days; Is at each data point The Gaussian kernel function applied at the location; h is the kernel function bandwidth; For each time period, the probability density sampling correction space for each power scenario is calculated separately, including: The power fluctuation range for time period t+1 is defined as follows: ， in, and These represent the power sample values ​​for time periods t and t+1 under the s-th power scenario, respectively; Will Substituting the endpoints of the power fluctuation range into the probability density function for time period t+1, we obtain the probability density sampling correction space for time period t+1 under the s-th power scenario, expressed as: ， 。 2. The method according to claim 1, characterized in that, The meteorological characteristic data types in the historical meteorological data and meteorological forecast data include: maximum temperature, minimum temperature, average temperature, precipitation, relative humidity, average irradiance, average wind speed, and average air pressure. The process of generating historical sample daily feature vectors and scene-specific daily feature vectors based on the acquired data includes: Calculate the daily average photovoltaic power and load power based on historical operating data of distributed photovoltaic and load. Calculate the Spearman correlation coefficients between the daily average photovoltaic power and load power and various types of historical meteorological characteristic data sequences; For each meteorological feature data type, the corresponding Spearman correlation coefficient calculation results are compared with the correlation coefficient threshold, and the feature vector type is selected based on the comparison results; Based on the selected feature vector type, the feature vectors for each historical sample day and the day the scene is to be generated are obtained, as follows: ， In the formula, Indicates the number of historical sample days. The feature vector representing the day to be generated for the scene. Indicates the first A historical day sample feature vector, Indicates the first The m-th feature value in the feature vector of historical day samples.

3. The method according to claim 2, characterized in that, The determination of the similar day set includes: Based on the feature vectors of each historical sample day and the day to be generated in the scenario, calculate the comprehensive correlation coefficient between the meteorological features of the day to be generated in the scenario and the meteorological features of each historical sample day. Historical samples with a comprehensive correlation coefficient higher than the comprehensive correlation coefficient threshold are combined with the date the scene is to be generated to form a set of similar dates; The formula for calculating the comprehensive correlation coefficient is as follows: ， in, For the first The comprehensive correlation coefficient between historical samples and the meteorological characteristics of the day the scene is to be generated. For the first The geometric correlation between historical samples and the date the scene is to be generated. For the first The distance correlation between each historical sample and the date the scene is to be generated is expressed as follows: ， ； Among them, intermediate variables , .

4. The method according to claim 3, characterized in that, it further... include: The time-series power scenario set is used as the time-series original scenario set. Based on the time-series original scenario set, a scenario-driven original uncertainty set is formed. The finite coverage theory is used to solve for several simplified uncertainty sets covering the original uncertain scenarios.

5. The method according to claim 4, characterized in that, The original uncertainty set driven by the scenario is represented as: ， In the formula, and These represent the number of distributed photovoltaic (PV) and load generation scenarios, respectively. , and , Let be the sampled value and the average value of the distributed photovoltaic and load power scenarios in time period t, respectively, for the i-th time series scenario. Let Γ be the initial set of uncertainties driven by the scenario, and Γ be the budget parameter, indicating that there are at most Γ time periods when the maximum uncertainty range can be reached. , These represent the sampled values ​​of distributed photovoltaic power and load at time t, respectively. The method of using finite coverage theory to solve for several simplified uncertainty sets covering the original uncertain scenario includes: The transformed optimization problem, based on the original uncertain set driven by the scenario, is expressed as: ， In the formula, and Let be the upper and lower bounds of the power value of the scene in time period t within the nth finite coverage set; and The original scene set formed by the s-th sample The upper and lower bounds of the scene's power value at time t, i.e.: , ; As a balance factor; M is a large number determined according to the Big M method; It is the original scene set for guiding the time sequence. 0-1 variables mapped to a finite coverage scene set N; Indicates the balance factor; Indicates a limited set of coverage scenarios; Solving the optimization problem yields an uncertain set that covers the scenario-driven conditions. n indeterminate sets of boxes ,Right now: .

6. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the distributed resource time-series scenario generation method as described in any one of claims 1-5.

Images