A wind-solar load combined typical dispatching operation scenario extraction method and system

Abstract

The application discloses a wind-solar-load combined typical scheduling operation scene extraction method and system, and the method comprises the following steps: obtaining original wind-solar-load output data of historical scheduling days and preprocessing; calculating a feature evaluation matrix of each historical scheduling day based on the preprocessed original wind-solar-load output data; screening abnormal historical scheduling days based on the feature evaluation matrix; calculating wind-solar-load matching degree and wind-solar-load matching fluctuation index based on wind-solar-load data of all historical scheduling days, determining an initial clustering number, and clustering, and if the condition is met, merging to generate a final wind-solar-load data clustering set; and extracting a predetermined number of historical scheduling days as typical operation scenes of the corresponding set by calculating the sum of trend perception Euclidean distances between each historical scheduling day in the same set and other historical scheduling days and sorting. Through multi-dimensional feature analysis and trend perception clustering, the application realizes high-precision extraction of wind-solar-load combined typical scheduling operation scenes and significantly improves scene representativeness.

Description

Technical Field

[0001] This invention relates to the field of power system technology, specifically to a method and system for extracting typical dispatching and operation scenarios of wind, solar and load combined operation. Background Technology

[0002] my country's installed capacity of new energy sources continues to grow. By the end of 2023, my country's grid-connected wind power and solar power generation reached 440 million kilowatts and 610 million kilowatts respectively, both ranking first in the world. With the increasing penetration rate of new energy sources, analyzing the fluctuation characteristics of renewable energy output and load power, and generating typical operating sequence scenarios with high coverage, is of great value for power system planning and design, economic operation and dispatch, as well as medium- and long-term electricity market transactions and spot transactions.

[0003] Currently, scholars at home and abroad have conducted extensive research on typical scheduling and operation scenarios such as wind power output, photovoltaic power output, and load. Existing technology 1 (Huang Yixuan, Luan Kaining, Huang Qifeng, Zhang Yanan, Zhou Gan, Yang Shihai, Duan Meimei, Cheng Hanmiao. Research on photovoltaic scenario reduction method based on dynamic time warping and hierarchical aggregation [J]. Power Supply and Utilization, 2023, 40(6): 91-100.) introduces dynamic time warping to measure the similarity of photovoltaic sample time series and uses hierarchical clustering to reduce the photovoltaic scenario dataset. Existing technology 2 (Chen Jianghong, Hu Jiahui, Shi Kanghao, Zheng Xinchao, Ao Zhiqiang. Clustering method of typical wind power scenarios based on improved density peak optimization global K-means [J]. Foreign Electronic Measurement Technology, 2024, 43(12): 71-82.) proposes a clustering method of typical wind power scenarios based on improved density peak optimization global K-means algorithm, which improves the accuracy of extracting typical wind power scenarios. Existing technology 3 (Tian Yunfeng, Xu Man, Shi Yufei, Qiao Yingkuo, Wu Linlin. A method for generating medium- and long-term load scenarios considering the influence of temperature factors [J]. Global Energy Internet, 2024, 7(06): 715-725.) generates daily load scenarios by establishing a joint probability distribution model of daily load characteristic indicators and an hourly load sequence optimization solution model.

[0004] Furthermore, patent CN119940086A proposes a wind-solar scene generation method based on LHS-GRU, and a weighted dimensional processing method for power load influencing factors based on feature correlation, which helps to solve the technical problem that the power points at different times in the scene set generated by statistical methods do not match the actual power output. Patent CN120256939A discloses an invention scheme for extracting extreme scenes involving wind, solar, and load co-location, including a method, device, electronic equipment, and storage medium, which realizes extreme scenes from historical scenes.

[0005] However, the above-mentioned solutions have the following shortcomings: First, the existing technology uses the wind-solar-load vector matrix feature value extraction method in the feature dimensionality reduction stage of wind power, solar power and load data processing. The extracted extreme wind power scenarios, extreme solar power scenarios and typical load scenarios cannot be guaranteed to be simultaneous on the time scale, and the correlation between new energy and load is not fully considered. Second, the existing technology uses Euclidean distance as a criterion when distinguishing scenarios, which cannot measure the trajectory similarity of time series. Summary of the Invention

[0006] To address the technical problems in existing technologies, such as insufficient consideration of the correlation between new energy sources and loads, and the inability to measure the trajectory similarity of time series using Euclidean indicators for scenario differentiation, this invention provides a method and system for extracting typical wind-solar-load joint scheduling operation scenarios. The method includes: acquiring and preprocessing the raw wind, solar, and load output data for historical scheduling days; calculating the feature evaluation matrix for each historical scheduling day based on the preprocessed raw wind, solar, and load output data; performing anomaly screening on historical scheduling days based on the feature evaluation matrix; calculating the wind-solar-load matching degree and wind-solar-load matching fluctuation index based on the wind, solar, and load data of all historical scheduling days, determining the initial number of clusters, and performing clustering; merging clusters if a certain condition is met to generate the final wind, solar, and load data cluster set; and extracting a predetermined number of historical scheduling days as typical operation scenarios for the corresponding set by calculating the sum of the trend-aware Euclidean distances between each historical scheduling day and other historical scheduling days within the same set, and sorting them. This invention achieves high-precision extraction of typical wind, solar, and load joint scheduling operation scenarios through multi-dimensional feature analysis and trend-aware clustering, significantly improving scenario representativeness.

[0007] The present invention adopts the following technical solution.

[0008] The first aspect of the present invention provides a method for extracting typical scheduling and operation scenarios of wind-solar-load joint operation, comprising: S1. Obtain and preprocess the raw wind and solar power output data for historical scheduling days; S2. Based on the preprocessed raw wind and solar load data, calculate the daily wind power generation ratio, wind power maximum power load ratio, solar power generation ratio, solar power maximum power load ratio, and daily average load rate for each historical dispatch day. Then, horizontally concatenate these five indicators according to the date dimension and standardize them to obtain the feature evaluation matrix. S3. Based on the feature evaluation matrix, perform anomaly screening on historical scheduling days; take the feature evaluation vector of each historical scheduling day after removing anomalies as a sample; based on the wind, solar and load data of the historical scheduling days corresponding to all samples, calculate the wind, solar and load matching degree and the wind, solar and load matching fluctuation index, determine the initial number of clusters through the matching parameter-cluster number mapping table, and perform clustering; based on the clustering results, determine whether any two clusters meet the merging conditions, and if so, merge them to generate the final wind, solar and load data cluster set; S4. For each wind, solar and load data cluster set, calculate the trend-aware Euclidean distance between any two historical scheduling days based on the wind, solar and load output data of each historical scheduling day in the set; by calculating the sum of the trend-aware Euclidean distances between each historical scheduling day and other historical scheduling days in the same set, and sorting them, extract a predetermined number of historical scheduling days as typical operating scenarios for the corresponding set.

[0009] Preferably, in S2, the daily power generation ratio of wind power is calculated based on the proportion of wind power output in the load value during historical dispatch days; the maximum value of the ratio of wind power output to load value during historical dispatch days is calculated as the maximum power load ratio of wind power; the daily power generation ratio of photovoltaic power is calculated based on the proportion of photovoltaic power output in the load value during historical dispatch days; the maximum value of the ratio of photovoltaic power output to load value during historical dispatch days is calculated as the maximum power load ratio of photovoltaic power; and the daily average load factor is calculated based on the ratio of average load to maximum load during historical dispatch days.

[0010] Preferably, the process of anomaly screening for historical scheduling days in S3 is as follows: Extract each row of the feature evaluation matrix of the historical scheduling day as the feature evaluation vector of the corresponding historical scheduling day. Treat the feature evaluation vector as a scheduling day feature point in a five-dimensional space. In this space, take any scheduling day feature point as the center and calculate the distance between the point and other scheduling day feature points. Based on the historical scheduling day feature evaluation matrix, a predetermined domain distance is set. When the distance between the center and any other scheduling day feature point is less than the predetermined domain distance, it means that the scheduling day feature point is within the domain of the corresponding center. Each scheduling day feature point is taken as the center, and the total number of scheduling day feature points in its domain is counted. If it is less than the predetermined number threshold, the historical scheduling day corresponding to the center is regarded as an abnormal operation scenario, and the feature evaluation vector of the corresponding historical scheduling day is removed.

[0011] Preferably, the process of determining the initial number of clusters in S3 is as follows: Based on the wind and solar load data of the historical scheduling days corresponding to all samples, the ratio of all wind and solar power output to all load power is calculated as the wind and solar load matching degree; the ratio of the sum of wind and solar power output to load power at each moment of each historical scheduling day is calculated, and the difference between the maximum ratio and the minimum ratio is used as the wind and solar load matching fluctuation index. Set up a matching parameter-cluster number mapping table, and determine the corresponding initial cluster number based on the wind-solar-load matching degree and the wind-solar-load matching fluctuation index.

[0012] Preferably, the clustering process in S3 is as follows: Randomly select a few samples from all samples as the initial cluster centers; For each sample other than the cluster center, calculate the Euclidean distance between each sample and the cluster center; assign each sample to the cluster corresponding to the cluster center with the closest Euclidean distance to generate the initial cluster; for each cluster, based on all samples in the cluster, use the mean of the feature evaluation vectors of all samples as the updated cluster center. Recalculate the Euclidean distance between each sample and the updated cluster center, determine new clusters and update cluster centers, until the Euclidean distance between the cluster center of each cluster and the previous cluster center is less than a predetermined threshold or the predetermined maximum number of iterations is reached, and output all clusters; Based on the clustering results of the feature evaluation matrix, the samples in each cluster are replaced with the original wind and solar load output data of the corresponding historical scheduling day to obtain the initial wind and solar load data cluster set corresponding to each cluster.

[0013] Preferably, the process of determining whether the merging conditions are met in S3 is as follows: For each initial wind-solar load data cluster set, the ratio of the sum of wind and solar power output to the load power of all historical scheduling days in the set is used as the wind-solar load matching degree within the cluster; the sum of wind and solar power output is subtracted from the load power of each historical scheduling day in each time period to obtain the net load power; the mean and standard deviation of the net load power of all time periods of each historical scheduling day are calculated, and the average ratio of the standard deviation to the mean of all historical scheduling days in the set is taken as the net load fluctuation index within the cluster; the load peak-valley difference degree of each historical scheduling day is calculated by subtracting the minimum load power of the day from the maximum load power of the day as the numerator and the average load power of the day as the denominator; the load peak-valley difference degree of the cluster is obtained by averaging the load peak-valley difference degree of all historical scheduling days in the set; Based on the wind-solar-load matching degree within the cluster, the net load fluctuation index within the cluster, and the load peak-valley difference degree within the cluster, the cluster similarity between the two sets is calculated. For any two initial wind-solar-load data cluster sets, calculate the absolute value of the difference in intra-cluster wind-solar-load matching degree between the two sets. When this value is less than a predetermined matching degree threshold, calculate the Euclidean distance between the corresponding cluster centers of the two sets. When this distance is less than a predetermined distance threshold and the cluster similarity is greater than a predetermined similarity threshold, merge the two sets into one set. Continue until no two wind-solar-load data cluster sets meet the merging conditions, and generate the final wind-solar-load data cluster set.

[0014] Preferably, the process of calculating cluster similarity is as follows: For any one of the following indicators—intra-cluster wind-solar-load matching degree, intra-cluster net load fluctuation index, and intra-cluster load peak-valley difference degree—the difference between the indicators of the two initial wind-solar-load data cluster sets is calculated by subtracting the indicators from each set as the numerator, and the difference between the maximum and minimum values ​​of the corresponding indicators in all sets as the denominator. The differences between the three indicators are then squared and weighted, and the square root of the weighted value is used to obtain the operational difference between the two sets. The cluster similarity between the two sets is obtained by subtracting the operational difference from 1.

[0015] Preferably, the process of trend-aware Euclidean distance between any two historical scheduling days in S4 is as follows: Calculate the Euclidean distance between two historical scheduling days as the distance of the wind-solar-load curve; Calculate the differences in wind power, photovoltaic power, and load power between adjacent time periods for each historical scheduling day, and divide by the time difference between adjacent time periods to obtain the differences in wind power, photovoltaic power, and load power for corresponding adjacent time periods. For any two historical scheduling days, calculate the ratio of the differences in wind power, photovoltaic power, and load power between the two historical scheduling days. Set photovoltaic proportion coefficients and wind power proportion coefficients, and weight the ratios in all adjacent time periods to obtain the trend of the wind-solar-load curve. Take the average of the weighted distances between the wind-solar-load curves between historical scheduling days to obtain the average distance between the wind-solar-load curves. By using a set trend similarity coefficient, weight the average distance between the wind-solar-load curves and the trend of the wind-solar-load curves to obtain the trend-aware Euclidean distance between two historical scheduling days.

[0016] Preferably, the process of setting the photovoltaic proportion coefficient and the wind power proportion coefficient is as follows: The photovoltaic power output over all time periods on the two historical dispatch days is summed up as the numerator, and the photovoltaic and wind power output over all time periods on the two historical dispatch days is summed up as the denominator to calculate the photovoltaic proportion coefficient; the wind power output over all time periods on the two historical dispatch days is summed up as the numerator, and the photovoltaic and wind power output over all time periods on the two historical dispatch days is summed up as the denominator to calculate the wind power proportion coefficient.

[0017] A second aspect of the present invention provides a system for extracting typical scheduling operation scenarios of wind-solar-load joint operation, using a method for extracting typical scheduling operation scenarios of wind-solar-load joint operation, including: The data acquisition module acquires and preprocesses the raw wind and solar load data for historical scheduling days; The data evaluation module, based on the preprocessed raw wind and solar load data, calculates the proportion of wind power daily generation, the proportion of wind power maximum power load, the proportion of solar power daily generation, the proportion of solar power maximum power load, and the daily average load rate for each historical dispatch day. These five indicators are then horizontally concatenated according to the date dimension and standardized to obtain a feature evaluation matrix. The data clustering module performs anomaly screening on historical scheduling days based on the feature evaluation matrix. It then uses the feature evaluation vector of each historical scheduling day after anomaly removal as a sample. Based on the wind, solar, and load data for all samples corresponding to the historical scheduling days, it calculates the wind, solar, and load matching degree and the wind, solar, and load matching fluctuation index. Using a matching parameter-cluster number mapping table, it determines the initial number of clusters and performs clustering. Based on the clustering results, it determines whether any two clusters meet the merging conditions; if so, they are merged to generate the final wind, solar, and load data cluster set. The typical scenario extraction module calculates the trend-aware Euclidean distance between any two historical scheduling days based on the wind, solar and load output data of each historical scheduling day in each set for each wind, solar and load data cluster set. By calculating the sum of the trend-aware Euclidean distances between each historical scheduling day and other historical scheduling days in the same set, and sorting them, a predetermined number of historical scheduling days are extracted as typical operating scenarios for the corresponding set.

[0018] A third aspect of the present invention provides a terminal, including a processor and a storage medium; The storage medium is used to store instructions; The processor is configured to operate according to the instructions to execute the steps of a method for extracting typical scheduling operation scenarios of wind, solar and load co-operation.

[0019] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of a method for extracting typical scheduling operation scenarios of wind-solar-load joint operation.

[0020] Compared with the prior art, the beneficial effects of the present invention include at least the following: 1. This invention introduces wind-solar-load matching degree and matching fluctuation index in the clustering stage. The initial number of clusters is determined through a matching parameter-cluster number mapping table, and iterative clustering and adaptive merging are performed based on multidimensional Euclidean distance. This mechanism not only avoids the bias caused by subjective setting of the cluster number, but also dynamically adjusts the clustering results according to the wind and solar power output and load characteristics, improving the stability of typical scenario classification and the accuracy of clustering results, thus enhancing the algorithm's versatility and practicality.

[0021] 2. This invention introduces a trend-aware Euclidean distance model, comprehensively considering the amplitude differences and trends of wind and solar power output curves, and weights them according to the proportion coefficients of photovoltaic and wind power, thus achieving accurate measurement of the trend similarity of different historical scheduling days. Compared with similarity analysis based solely on time-series Euclidean distance, this method can more accurately identify typical scenarios with consistent operating trends, thereby effectively supporting the optimized scheduling of wind, solar, and load joint operations. Attached Figure Description

[0022] Figure 1A flowchart of a method for extracting typical scheduling and operation scenarios of wind-solar-load joint operation provided by the present invention; Figure 2 This is a wind power curve diagram in an example of the present invention; Figure 3 This is a photovoltaic power curve diagram in an example of the present invention; Figure 4 This is a load power curve diagram in an example of the present invention; Figure 5 This is the extraction result of a typical scenario in the wind power curve in the example of this invention; Figure 6 This is the extraction result of a typical scenario in the photovoltaic power curve in the example of this invention; Figure 7 This is the extraction result of a typical scenario in the load power curve in the example of this invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of this invention. The embodiments described in this application are merely some embodiments of this invention, and not all embodiments. Based on the spirit of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of this invention.

[0024] Example 1 Embodiment 1 of the present invention provides a method for extracting typical scheduling and operation scenarios of wind, solar and load co-operation, such as... Figure 1 As shown, the specific implementation steps are as follows: S1. Obtain and preprocess the raw wind and solar load data for historical scheduling days.

[0025] The acquired raw wind and solar load output data includes wind power, photovoltaic and load output data for each time period in the historical scheduling day. These collected data are preprocessed. The preprocessing operations include missing value handling, outlier handling, data alignment, etc. In this embodiment, the number of time periods collected in the historical scheduling day is 96, that is, data is collected once every 15 minutes.

[0026] S2. Based on the preprocessed raw wind and solar load data, calculate the daily wind power generation ratio, wind power maximum power load ratio, solar power generation ratio, solar power maximum power load ratio, and daily average load rate for each historical dispatch day. Then, horizontally concatenate these five indicators according to the date dimension and standardize them to obtain the feature evaluation matrix.

[0027] The daily wind power generation ratio is calculated based on the historical daily wind power output as a percentage of the load. The specific formula for the daily wind power generation ratio is as follows: ; In the formula, This represents the percentage of daily wind power generation, indicating the ratio of wind power generation to daily electricity consumption. Wind power generation is greatly affected by natural factors and fluctuates significantly throughout the day, thus exhibiting distinctive characteristics. The higher the value of this indicator, the stronger the support of wind power for the load. T The number of time periods represents the total number of time periods in this embodiment, which is 96 time periods throughout the day. This represents the wind power output during time period t on that day; Representative on that day t Load values ​​for a given time period.

[0028] The maximum value of the wind power output to load ratio in the historical dispatch days is used as the wind power maximum power load ratio. The specific calculation formula is as follows: ; In the formula, This represents the maximum power-to-load ratio of wind power, which is the maximum value of the ratio of wind power output to load within a day, and is used to express the wind power's ability to support the load.

[0029] The daily photovoltaic (PV) power generation ratio is calculated based on the historical daily dispatch rate of PV output within the load value. The specific formula for the daily PV power generation ratio is as follows: ; In the formula, r s This represents the proportion of daily photovoltaic (PV) power generation. Similar to wind power, PV power generation is also greatly affected by natural factors and fluctuates significantly during the day. The higher the value of this indicator, the stronger the support of PV power generation for the load. Representing the photovoltaic industry on that day t Efforts during specific time periods.

[0030] The maximum value of the ratio of photovoltaic output to load value in historical dispatch days is used as the photovoltaic maximum power load ratio. The specific calculation formula is as follows:

[0031] In the formula, This represents the maximum power-to-load ratio of photovoltaic (PV) power generation, which is the maximum value of the ratio of PV power generation to load within a day, used to express the capacity of PV power generation to support the load. PV power generation typically peaks at midday, while electricity load also typically experiences midday and evening peaks.

[0032] The daily average load factor is calculated based on the ratio of the historical average load to the maximum load. The specific formula for the daily average load factor is as follows: ; In the formula, It represents the daily average load factor and can reflect the load fluctuation throughout the day.

[0033] The calculated percentage of wind power generation, maximum wind power load ratio, percentage of photovoltaic power generation, maximum photovoltaic load ratio, and average daily load factor for each historical dispatch day are concatenated along the date dimension to form a corresponding historical dispatch day feature evaluation matrix, which is then standardized. The specific formula is as follows: ; In the formula, each row represents a different historical scheduling day, and each column represents a different indicator calculated from the historical scheduling working days; n is the total number of historical scheduling days; In this embodiment, the wind and solar load power characteristics of the corresponding historical scheduling days are evaluated using the above five characteristic indicators, thereby expanding the dimensionality of the original wind and solar load data for a scheduling day from... Reduced as a characteristic The problem involves a dimension, forming a feature evaluation vector. The original wind, solar, and load data from n scheduling days then form... The feature evaluation matrix is ​​used, and the maximum value in each column is used as the benchmark value for standardization.

[0034] S3. Based on the feature evaluation matrix, perform anomaly screening on historical scheduling days; take the feature evaluation vector of each historical scheduling day after removing anomalies as a sample; based on the wind, solar and load data of the historical scheduling days corresponding to all samples, calculate the wind, solar and load matching degree and the wind, solar and load matching fluctuation index, determine the initial number of clusters through the matching parameter-cluster number mapping table, and perform clustering; based on the clustering results, determine whether any two clusters meet the merging conditions, and if so, merge them to generate the final wind, solar and load data cluster set.

[0035] As a preferred implementation method, the process of anomaly screening for historical scheduling days is as follows: Each row of the historical scheduling day feature evaluation matrix is ​​extracted as the feature evaluation vector for the corresponding historical scheduling day. This feature evaluation vector is considered as a scheduling day feature point in a five-dimensional space. Within this space, with any scheduling day feature point as the center, the distance between that point and other scheduling day feature points is calculated. The specific formula is as follows: ; In the formula, This represents the distance between the feature point on the i-th scheduling day and the feature point on the j-th scheduling day; Based on the historical scheduling day feature evaluation matrix, the predetermined neighborhood distance is calculated; the specific formula is as follows: ; In the formula, The adjustment coefficient is obtained by fitting experimental data; for example, based on normal and abnormal scheduling days labeled by experts, the corresponding adjustment coefficient is obtained by fitting data. When the distance between the center and any other scheduling day feature point is less than the predetermined domain distance, it means that the scheduling day feature point is within the domain of the corresponding center. Each scheduling day feature point is taken as the center, and the total number of scheduling day feature points in its domain is counted. If it is less than the predetermined threshold, the historical scheduling day corresponding to the center is regarded as an abnormal operation scenario, and the feature evaluation vector of the corresponding historical scheduling day is removed.

[0036] As a preferred implementation method, the process of determining the initial number of clusters is as follows: Based on the wind and solar load data for all samples corresponding to historical scheduling days, the ratio of all wind and solar power output to all load power is calculated as the wind-solar-load matching degree; the ratio of the sum of wind and solar power output to load power at each moment of each historical scheduling day is calculated, and the difference between the maximum and minimum ratios is used as the wind-solar-load matching fluctuation index; the specific formula is as follows: ; In the formula, The wind-solar-load matching degree represents the average ratio of the sum of wind and solar power generation to load in all scheduling periods across N samples. The larger this value, the higher the matching degree between wind and solar power output and load demand. This represents the maximum value of the ratio of the sum of wind and solar power generation to the load in all scheduling periods across N samples. The larger this value, the higher the degree of matching between wind and solar power output and load demand. This represents the minimum ratio of the sum of wind and solar power generation to load in all scheduling periods across N samples. The larger this value, the higher the degree of matching between wind and solar power output and load demand. The wind-solar load matching fluctuation index represents the difference between the maximum and minimum values ​​of the ratio of the sum of wind and solar power generation to load in all scheduling periods of N samples. The larger the value, the greater the fluctuation of the wind-solar load demand matching degree. The output of wind power on historical dispatch day i during time period t; The output of photovoltaic power on historical dispatch day i during time period t; This represents the load power during time period t in historical dispatch day i; Referring to Table 1, a matching parameter-cluster number mapping table is set up. Based on the wind-solar-load matching degree and the wind-solar-load matching fluctuation index, the corresponding initial cluster number is determined. Table 1 uses... and Based on the parameter distribution range, a table of values ​​for the number of clusters K is established. Generally speaking, the number of clusters is distributed between 2 and 6. Table 1. Matching Parameter-Cluster Number Mapping Table

[0037] As a preferred implementation method, the clustering process is as follows: Randomly select a few samples from all samples as the initial cluster centers; For each sample other than the cluster center, calculate the Euclidean distance between each sample and the cluster center; assign each sample to the cluster corresponding to the cluster center with the closest Euclidean distance to generate the initial cluster; for each cluster, based on all samples in the cluster, use the mean of the feature evaluation vectors of all samples as the updated cluster center. Recalculate the Euclidean distance between each sample and the updated cluster center, determine new clusters and update cluster centers, until the Euclidean distance between the cluster center of each cluster and the previous cluster center is less than a predetermined threshold or the predetermined maximum number of iterations is reached, and output all clusters; Based on the clustering results of the feature evaluation matrix, the samples in each cluster are replaced with the original wind and solar load output data of the corresponding historical scheduling day to obtain the initial wind and solar load data cluster set corresponding to each cluster.

[0038] As a preferred implementation method, the process of determining whether the merging conditions are met is as follows: For each initial wind-solar-load data cluster set, the ratio of the sum of wind and solar power output of all historical scheduling days in the set to the load power is used as the wind-solar-load matching degree within the cluster. Subtract the sum of wind and solar power output from the load power for each time period on each historical scheduling day to obtain the net load power for the corresponding time period; calculate the mean and standard deviation of the net load power for all time periods on each historical scheduling day; and take the average of the ratios of the standard deviations to the mean for all historical scheduling days in the cluster as the intra-cluster net load fluctuation index; the specific formula is as follows: ; In the formula, is the intra-cluster net load fluctuation index for cluster u; This represents the total number of historical scheduling days in cluster u; Let be the standard deviation of the net load power on the i-th historical scheduling day in cluster u; Let be the mean net load power of the i-th historical scheduling day in cluster u; For each historical scheduling day, the load peak-valley difference is calculated by subtracting the daily minimum load power from the daily maximum load power as the numerator and the daily average load power as the denominator. The load peak-valley difference is then averaged across all historical scheduling days within the initial wind and solar load data cluster set to obtain the intra-cluster load peak-valley difference for the corresponding set. The specific formula is as follows: ; In the formula, The intra-cluster load peak-valley difference degree of cluster cluster u; and Let these represent the maximum daily load power and the minimum daily load power of the i-th historical scheduling day, respectively. Let be the average daily load power of the i-th historical scheduling day; For any one of the following indicators—intra-cluster wind-solar-load matching degree, intra-cluster net load fluctuation index, and intra-cluster load peak-valley difference degree—the difference between the indicators of the two initial wind-solar-load data cluster sets is calculated by subtracting the indicators from each set as the numerator, and subtracting the minimum value from the maximum value of the corresponding indicator in all sets as the denominator. The squared differences of the three indicators are then weighted, and the square root of the weighted value is used to obtain the operational difference between the two sets. The cluster similarity between the two sets is calculated by subtracting the operational difference from 1. The specific formula is as follows: ; In the formula, The cluster similarity of the initial wind-solar-load data cluster sets u and v; Represents a set The k-th indicator in The degree of matching between wind, solar and load within the cluster; and Let u and v represent the kth indexes of the initial wind-solar-load data cluster sets, respectively. and These represent the maximum and minimum values ​​of the k-th index in all initial wind-solar-load data cluster sets, respectively; The weight of the k-th indicator and ; For any two initial wind-solar-load data cluster sets, calculate the absolute value of the difference in intra-cluster wind-solar-load matching degree between the two sets. When this value is less than a predetermined matching degree threshold, calculate the Euclidean distance between the corresponding cluster centers of the two sets. When this distance is less than a predetermined distance threshold and the cluster similarity is greater than a predetermined similarity threshold, it indicates that the two initial wind-solar-load data cluster sets have similar scenes, and the two sets are merged into one cluster set. This process continues until no two wind-solar-load data cluster sets meet the merging conditions, generating the final wind-solar-load data cluster set.

[0039] S4. For each wind, solar and load data cluster set, calculate the trend-aware Euclidean distance between any two historical scheduling days based on the wind, solar and load output data of each historical scheduling day in the set; by calculating the sum of the trend-aware Euclidean distances between each historical scheduling day and other historical scheduling days in the same set, and sorting them, extract a predetermined number of historical scheduling days as typical operating scenarios for the corresponding set.

[0040] Select wind and solar power output data for a predetermined number of historical scheduling days from each cluster set; in this embodiment, two historical scheduling days are selected, and the specific formula is as follows: ; ; In the formula, and These represent the wind power output during time period T on the i-th historical dispatch day and the j-th dispatch day, respectively; and These represent the photovoltaic output during time period T on the i-th and j-th historical scheduling days, respectively; and These represent the load values ​​during time period T for the i-th and j-th historical scheduling days, respectively. As a preferred implementation method, the process of trend-aware Euclidean distance is as follows: The Euclidean distance between two historical scheduling days is calculated as the wind-solar-load curve distance, and the specific formula is as follows: ; ; ; In the formula, Indicates the distance of the wind power output curve; Indicates the distance between photovoltaic power output curves; Indicates the distance between load curves; Calculate the differences in wind power, solar power, and load power between adjacent time periods for each historical scheduling day, and divide by the time difference between adjacent time periods to obtain the differences in wind power, solar power, and load power for corresponding adjacent time periods. For any two historical scheduling days, calculate the ratio of the differences in wind power, solar power, and load power between the two historical scheduling days. Set solar power proportion coefficients and wind power proportion coefficients, and weight the ratios over all adjacent time periods to obtain the trend of the wind-solar-load curve. Take the average of the weighted average distances between the wind-solar-load curves between historical scheduling days to obtain the average distance between the wind-solar-load curves. Using a set trend similarity coefficient, weight the average distance between the wind-solar-load curves and the trend of the wind-solar-load curves to obtain the trend-aware Euclidean distance between two historical scheduling days. ; In the formula, The trend-aware Euclidean distance representing historical scheduling days i and j; The trend similarity coefficient, representing the trend of the wind-solar-load curve, takes a value in the range [0.5, 1]. , These represent the differences in wind power output on dispatch days i and j during time period t, respectively. , ; , These represent the differences in photovoltaic power on scheduling days i and j during time period t, respectively. , These represent the differences in daily load power of historical scheduling days i and j during time period t, respectively; if If the value is positive, it means that the two have the same trend of change at time t. This method can be used to characterize the similarity of time series on the time scale to make up for the insufficiency of Euclidean distance in measuring the similarity of time series trajectories. , These represent the photovoltaic (PV) and wind power (FH) ratios, respectively, used to measure the proportion of PV and FH in the total power generation of wind and solar power on historical dispatch days i and j. Through these ratios, the differential results for the corresponding time period can be adaptively adjusted based on the reliability of wind and solar power output in different seasons.

[0041] As a preferred implementation method, the process for calculating the photovoltaic proportion coefficient and the wind power proportion coefficient is as follows: The photovoltaic (PV) power output over all time periods on two historical dispatch days is summed as the numerator, and the combined PV and wind power output over all time periods on two historical dispatch days is summed as the denominator to calculate the PV proportion coefficient; similarly, the wind power output over all time periods on two historical dispatch days is summed as the numerator, and the combined PV and wind power output over all time periods on two historical dispatch days is summed as the denominator to calculate the wind power proportion coefficient; the specific formula is as follows: ; ; For each wind-solar-load data cluster, the sum of the trend-aware Euclidean distances between each historical scheduling day and all other scheduling days is calculated and sorted in ascending order. The smaller the value, the stronger the typicality of that day. This allows for the extraction of a predetermined number of typical scheduling operation scenarios from each wind-solar-load data cluster.

[0042] To verify the practical effects of this invention, examples will be provided below. For example... Figure 2 , Figure 3 , Figure 4 The image shows the wind and solar load power curves for 60 scheduling days. This is used as an example to extract typical scheduling operation scenarios for wind, solar, and load coordination. The clustering number is set to 2, therefore two typical scheduling operation scenarios will be extracted, as shown in the image. Figure 5 As shown in the figure, the extracted typical scheduling and operation scenarios are representative.

[0043] Example 2 Embodiment 2 of the present invention provides a typical scheduling and operation scenario extraction system for wind-solar-load joint operation, comprising: The data acquisition module acquires and preprocesses the raw wind and solar load data for historical scheduling days; The data evaluation module, based on the preprocessed raw wind and solar load data, calculates the proportion of wind power daily generation, the proportion of wind power maximum power load, the proportion of solar power daily generation, the proportion of solar power maximum power load, and the daily average load rate for each historical dispatch day. These five indicators are then horizontally concatenated according to the date dimension and standardized to obtain a feature evaluation matrix. The data clustering module performs anomaly screening on historical scheduling days based on the feature evaluation matrix. It then uses the feature evaluation vector of each historical scheduling day after anomaly removal as a sample. Based on the wind, solar, and load data for all samples corresponding to the historical scheduling days, it calculates the wind, solar, and load matching degree and the wind, solar, and load matching fluctuation index. Using a matching parameter-cluster number mapping table, it determines the initial number of clusters and performs clustering. Based on the clustering results, it determines whether any two clusters meet the merging conditions; if so, they are merged to generate the final wind, solar, and load data cluster set. The typical scenario extraction module calculates the trend-aware Euclidean distance between any two historical scheduling days based on the wind, solar and load output data of each historical scheduling day in each set for each wind, solar and load data cluster set. By calculating the sum of the trend-aware Euclidean distances between each historical scheduling day and other historical scheduling days in the same set, and sorting them, a predetermined number of historical scheduling days are extracted as typical operating scenarios for the corresponding set.

[0044] This disclosure can be a system, method, and / or computer program product. A computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for causing a processor to implement various aspects of this disclosure.

[0045] Computer-readable storage media can be tangible devices capable of holding and storing instructions for use by an instruction execution device. Computer-readable storage media can be, for example—but not limited to—electrical storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of computer-readable storage media include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital multifunction disc (DVD), memory sticks, floppy disks, mechanical encoding devices, such as punch cards or recessed protrusions storing instructions thereon, and any suitable combination of the foregoing. The computer-readable storage media used herein are not to be construed as transient signals themselves, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (e.g., light pulses through fiber optic cables), or electrical signals transmitted through wires.

[0046] The computer-readable program instructions described herein can be downloaded from computer-readable storage media to various computing / processing devices, or downloaded via a network, such as the Internet, local area network, wide area network, and / or wireless network, to an external computer or external storage device. The network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards them to the computer-readable storage media in the respective computing / processing device.

[0047] Computer program instructions used to perform the operations of this disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, status setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, etc., and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The computer-readable program instructions may execute entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, electronic circuitry, such as programmable logic circuitry, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), is personalized by utilizing the status information of the computer-readable program instructions to implement various aspects of this disclosure.

[0048] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A method for extracting typical scheduling and operation scenarios of wind-solar-load joint operation, characterized in that, include: S1. Obtain and preprocess the raw wind and solar power output data for historical scheduling days; S2. Based on the preprocessed raw wind and solar load data, calculate the daily wind power generation ratio, wind power maximum power load ratio, solar power generation ratio, solar power maximum power load ratio, and daily average load rate for each historical dispatch day. Then, horizontally concatenate these five indicators according to the date dimension and standardize them to obtain the feature evaluation matrix. S3. Based on the feature evaluation matrix, perform anomaly screening on historical scheduling days; take the feature evaluation vector of each historical scheduling day after removing anomalies as a sample; based on the wind, solar and load data of the historical scheduling days corresponding to all samples, calculate the wind, solar and load matching degree and wind, solar and load matching fluctuation index, determine the initial number of clusters through the matching parameter-cluster number mapping table, and perform clustering. Based on the clustering results, determine whether any two clusters meet the merging conditions. If they do, merge them to generate the final wind, solar and load data cluster set. S4. For each wind, solar and load data cluster set, based on the wind and solar load output data of each historical scheduling day in the set, calculate the trend-aware Euclidean distance between any two historical scheduling days; by calculating the sum of the trend-aware Euclidean distances between each historical scheduling day and other historical scheduling days in the same set, and sorting them, extract a predetermined number of historical scheduling days as typical operating scenarios for the corresponding set. The process of trend-aware Euclidean distance between any two historical scheduling days in S4 is as follows: Calculate the Euclidean distance between two historical scheduling days as the distance of the wind-solar-load curve; Calculate the differences in wind power, photovoltaic power, and load power between adjacent time periods for each historical scheduling day, and divide by the time difference between adjacent time periods to obtain the differences in wind power, photovoltaic power, and load power for corresponding adjacent time periods. For any two historical scheduling days, calculate the ratio of the differences in wind power, photovoltaic power, and load power between the two historical scheduling days. Set photovoltaic proportion coefficients and wind power proportion coefficients, and weight the ratios in all adjacent time periods to obtain the trend of the wind-solar-load curve. Take the average of the weighted distances between the wind-solar-load curves between historical scheduling days to obtain the average distance of the wind-solar-load curve. By setting a trend similarity coefficient, the average distance between the wind-solar-load curves and the trend of the wind-solar-load curves are weighted to obtain the trend-aware Euclidean distance between two historical scheduling days.

2. The method for extracting typical scheduling and operation scenarios of wind-solar-load joint operation according to claim 1, characterized in that, include: In S2, the proportion of wind power output in the load value during historical dispatch days is calculated to determine the proportion of wind power generation per day. The maximum value of the ratio of wind power output to load value in historical dispatch days is used as the wind power maximum power load ratio; the proportion of photovoltaic power output to load value in historical dispatch days is used to calculate the photovoltaic daily power generation ratio; the maximum value of the ratio of photovoltaic power output to load value in historical dispatch days is used as the photovoltaic maximum power load ratio; the daily average load rate is calculated based on the ratio of average load to maximum load in historical dispatch days.

3. The method for extracting typical scheduling and operation scenarios of wind-solar-load joint operation according to claim 1, characterized in that, include: The process of anomaly screening for historical scheduling days in S3 is as follows: Extract each row of the feature evaluation matrix of the historical scheduling day as the feature evaluation vector of the corresponding historical scheduling day. Treat the feature evaluation vector as a scheduling day feature point in a five-dimensional space. In this space, take any scheduling day feature point as the center and calculate the distance between the point and other scheduling day feature points. Based on the historical scheduling day feature evaluation matrix, a predetermined domain distance is set. When the distance between the center and any other scheduling day feature point is less than the predetermined domain distance, it means that the scheduling day feature point is within the domain of the corresponding center. Each scheduling day feature point is taken as the center, and the total number of scheduling day feature points in its domain is counted. If it is less than the predetermined number threshold, the historical scheduling day corresponding to the center is regarded as an abnormal operation scenario, and the feature evaluation vector of the corresponding historical scheduling day is removed.

4. The method for extracting typical scheduling and operation scenarios of wind-solar-load joint operation according to claim 1, characterized in that, include: The process of determining the initial number of clusters in S3 is as follows: Based on the wind and solar load data of the historical scheduling days corresponding to all samples, the ratio of all wind and solar power output to all load power is calculated as the wind and solar load matching degree; the ratio of the sum of wind and solar power output to load power at each moment of each historical scheduling day is calculated, and the difference between the maximum ratio and the minimum ratio is used as the wind and solar load matching fluctuation index. Set up a matching parameter-cluster number mapping table, and determine the corresponding initial cluster number based on the wind-solar-load matching degree and the wind-solar-load matching fluctuation index.

5. The method for extracting typical scheduling and operation scenarios of wind-solar-load joint operation according to claim 1, characterized in that, include: The clustering process in S3 is as follows: Randomly select a few samples from all samples as the initial cluster centers; For each sample other than the cluster center, calculate the Euclidean distance between each sample and the cluster center; assign each sample to the cluster corresponding to the cluster center with the closest Euclidean distance to generate the initial cluster; for each cluster, based on all samples in the cluster, use the mean of the feature evaluation vectors of all samples as the updated cluster center. Recalculate the Euclidean distance between each sample and the updated cluster center, determine new clusters and update cluster centers, until the Euclidean distance between the cluster center of each cluster and the previous cluster center is less than a predetermined threshold or the predetermined maximum number of iterations is reached, and output all clusters; Based on the clustering results of the feature evaluation matrix, the samples in each cluster are replaced with the original wind and solar load output data of the corresponding historical scheduling day to obtain the initial wind and solar load data cluster set corresponding to each cluster.

6. The method for extracting typical scheduling and operation scenarios of wind-solar-load joint operation according to claim 1, characterized in that, include: The process for determining whether the merging conditions are met in S3 is as follows: For each initial wind-solar load data cluster set, the ratio of the sum of wind and solar power output to the load power of all historical scheduling days in the set is used as the wind-solar load matching degree within the cluster; the sum of wind and solar power output is subtracted from the load power of each historical scheduling day in each time period to obtain the net load power; the mean and standard deviation of the net load power of all time periods of each historical scheduling day are calculated, and the average ratio of the standard deviation to the mean of all historical scheduling days in the set is taken as the net load fluctuation index within the cluster; the load peak-valley difference degree of each historical scheduling day is calculated by subtracting the minimum load power of the day from the maximum load power of the day as the numerator and the average load power of the day as the denominator; the load peak-valley difference degree of the cluster is obtained by averaging the load peak-valley difference degree of all historical scheduling days in the set; Based on the wind-solar-load matching degree within the cluster, the net load fluctuation index within the cluster, and the load peak-valley difference degree within the cluster, the cluster similarity between the two sets is calculated. For any two initial wind-solar-load data cluster sets, calculate the absolute value of the difference in intra-cluster wind-solar-load matching degree between the two sets. When this value is less than a predetermined matching degree threshold, calculate the Euclidean distance between the corresponding cluster centers of the two sets. When this distance is less than a predetermined distance threshold and the cluster similarity is greater than a predetermined similarity threshold, merge the two sets into one set. Continue until no two wind-solar-load data cluster sets meet the merging conditions, and generate the final wind-solar-load data cluster set.

7. The method for extracting typical scheduling and operation scenarios of wind-solar-load joint operation according to claim 6, characterized in that, include: The process of calculating cluster similarity is as follows: For any one of the following indicators—intra-cluster wind-solar-load matching degree, intra-cluster net load fluctuation index, and intra-cluster load peak-valley difference degree—the difference between the indicators of the two initial wind-solar-load data cluster sets is calculated by subtracting the indicators from each set as the numerator, and the difference between the maximum and minimum values ​​of the corresponding indicators in all sets as the denominator. The differences between the three indicators are then squared and weighted, and the square root of the weighted value is used to obtain the operational difference between the two sets. The cluster similarity between the two sets is obtained by subtracting the operational difference from 1.

8. The method for extracting typical scheduling and operation scenarios of wind-solar-load joint operation according to claim 1, characterized in that, include: The process of setting the photovoltaic proportion coefficient and the wind power proportion coefficient is as follows: The photovoltaic power output over all time periods on the two historical dispatch days is summed up as the numerator, and the photovoltaic and wind power output over all time periods on the two historical dispatch days is summed up as the denominator to calculate the photovoltaic proportion coefficient; the wind power output over all time periods on the two historical dispatch days is summed up as the numerator, and the photovoltaic and wind power output over all time periods on the two historical dispatch days is summed up as the denominator to calculate the wind power proportion coefficient.

9. A system for extracting typical scheduling and operation scenarios of wind-solar-load joint operation, using the method described in any one of claims 1-8, characterized in that, include: The data acquisition module acquires and preprocesses the raw wind and solar load data for historical scheduling days; The data evaluation module, based on the preprocessed raw wind and solar load data, calculates the proportion of wind power daily generation, the proportion of wind power maximum power load, the proportion of solar power daily generation, the proportion of solar power maximum power load, and the daily average load rate for each historical dispatch day. These five indicators are then horizontally concatenated according to the date dimension and standardized to obtain a feature evaluation matrix. The data clustering module performs anomaly screening on historical scheduling days based on the feature evaluation matrix; it takes the feature evaluation vector of each historical scheduling day after removing anomalies as a sample; based on the wind, solar and load data of the historical scheduling days corresponding to all samples, it calculates the wind, solar and load matching degree and the wind, solar and load matching fluctuation index, determines the initial number of clusters through the matching parameter-cluster number mapping table, and performs clustering. Based on the clustering results, determine whether any two clusters meet the merging conditions. If they do, merge them to generate the final wind, solar and load data cluster set. The typical scenario extraction module calculates the trend-aware Euclidean distance between any two historical scheduling days based on the wind, solar and load output data of each historical scheduling day in each set for each wind, solar and load data cluster set. By calculating the sum of the trend-aware Euclidean distances between each historical scheduling day and other historical scheduling days in the same set, and sorting them, a predetermined number of historical scheduling days are extracted as typical operating scenarios for the corresponding set.

10. A terminal, comprising a processor and a storage medium; characterized in that: The storage medium is used to store instructions; The processor is configured to operate according to the instructions to perform the steps of the method according to any one of claims 1-8.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the steps of the method according to any one of claims 1-8.

Images