Wind-solar-hydro combined short-term optimal dispatching method based on wind and solar uncertainty prediction scenarios

Abstract

The application discloses a wind-solar-water combined short-term optimization scheduling method based on wind-solar uncertainty prediction scenes, which comprises the following steps: pre-processing and dividing levels of historical wind-solar output data; establishing a wind-solar prediction scene construction model and a reduction model to generate a wind-solar prediction typical scene set; establishing a scene migration probability calculation model to obtain the migration probability between different time periods of actual wind-solar output values; updating the prediction scene according to the wind-solar prediction typical scene set, combining the hourly wind-solar output measured data and the scene migration probability information; and performing wind-solar-water combined short-term optimization scheduling. The application is based on the prediction and measured data of historical wind-solar output, considers the influence of the uncertainty of wind-solar output on the wind-solar-water combined short-term optimization scheduling by constructing the wind-solar output prediction scene, and provides a new method and technology for the optimization scheduling of a multi-energy complementary system connected with new energy.

Description

Technical Field

[0001] This invention relates to a short-term optimization scheduling method for wind, solar and hydropower combined, taking into account the uncertainties of wind and solar power, specifically a short-term optimization scheduling method for wind, solar and hydropower combined, based on a scenario with uncertain wind and solar power forecasts. Background Technology

[0002] Renewable energy is a green and low-carbon energy source and an important component of my country's multi-pronged energy supply system. It is of great significance for improving the energy structure, protecting the ecological environment, addressing climate change, and achieving sustainable economic and social development. In the past decade, renewable energy has become the mainstay of my country's newly installed power generation capacity. The increase in renewable energy power generation will account for more than 50% of the increase in total electricity consumption, and wind and solar power generation will more than double. However, wind, solar, and photovoltaic power generation inherently involve uncertainties, and direct grid connection would significantly impact the power grid. However, by leveraging the flexible output of hydropower and bundling wind, solar, and hydropower for transmission, the volatility of wind and solar power can be mitigated to some extent, reducing the impact of their uncertainties.

[0003] Currently, research on short-term optimal scheduling methods for wind-solar-hydro combined operations, considering wind and solar uncertainties, by scholars both domestically and internationally, mainly focuses on modeling methods based on Bayesian theory and fuzzy mathematics. For example, Liu Qiaobo (2018) fitted the distribution of prediction errors for wind and solar power at different output levels based on historical data, introduced Spearman correlation coefficients to describe the complementarity of wind and solar prediction errors seasonally, and established a wind-solar joint prediction error distribution model; Zhang Juntao (2020) used coupled quantile regression based on stochastic programming theory to transform historical statistical information from deterministic prediction sequences into scenario sets; Xu Yechi (2022) established a stochastic optimal scheduling model considering frequency response by generating wind power output scenarios that describe the stochasticity of wind power based on the analysis of wind power stochasticity in the prediction error distribution. However, the above methods all describe individual scenarios or multiple unrelated independent scenarios based on historical data for formulating optimal scheduling schemes. They cannot consider the correlation between scenarios and the optimization of the scheduling system by the measured values ​​updated over time. The accuracy of characterizing wind and solar uncertainties will affect the formulation of optimal scheduling schemes, ultimately causing deviations from the scheduling objectives. Summary of the Invention

[0004] Objective of the Invention: To address the problems and shortcomings of existing technologies, this invention provides a short-term optimized scheduling method for wind, solar, and hydropower joint operations based on uncertain wind and solar power forecast scenarios. Based on historical wind and solar power output prediction and measured data, a wind and solar scenario forecasting model, a scenario reduction model, and a scenario migration probability calculation model are established. By updating the typical wind and solar forecast scenario set and its migration probability coefficients hourly, the optimized scheduling decision is calculated and updated hourly, enabling short-term optimized scheduling of wind, solar, and hydropower joint operations.

[0005] Technical solution: A short-term optimization scheduling method for wind, solar, and hydropower joint operation based on uncertain wind and solar forecast scenarios, comprising the following steps:

[0006] S1. Select historical landscape output data, process missing and outlier values, normalize the processed data, and divide the data into sample levels.

[0007] S2. Establish a scene-based wind and solar scene forecasting model, use sample data at each level to perform sample feature analysis within the level, and use the analysis results to perform wind and solar scene forecasting based on the day-ahead forecast data of wind and solar power output, so as to obtain a wind and solar forecast scene set.

[0008] S3. Establish a scene reduction model to reduce the initial set of wind and light forecast scenes, and iteratively calculate and select the most representative typical wind and light forecast scenes to form a typical set of wind and light forecast scenes.

[0009] S4. Establish a scene migration probability calculation model based on the historical measured values ​​of wind and solar power output, divide the measured values ​​into intervals, and calculate the migration probability between measured values ​​that have time correlation.

[0010] S5. Based on the results of typical scenarios in the wind and solar forecast, and based on the hourly measured data of wind and solar power output, the scheduling decision is updated and optimized hourly. In accordance with the principle of best source-load matching, short-term optimized scheduling of wind, solar and hydropower is carried out.

[0011] Further, step S1 specifically involves: historical wind and solar power output data mainly including day-ahead forecast wind power output data and measured wind power output data with consistent spatiotemporal correlation, as well as day-ahead forecast photovoltaic power output data and measured photovoltaic power output data with consistent spatiotemporal correlation; the method for handling missing values ​​is: data supplementation is performed using linear interpolation based on two adjacent sets of data; the method for handling outliers is: firstly, outliers are deleted, and missing data is supplemented using the method for missing values; the processed data is then normalized using the following method:

[0012] s′=(s-μ) / σ

[0013] In the formula, s′ is the normalized sample value; s is the original sample data; μ is the mean of the original sample set; and σ is the standard deviation of the original sample set.

[0014] Furthermore, the structure of the scene-based landscape scene forecasting model established in step S2 is as follows:

[0015] Based on the forecast and measured datasets obtained after sample normalization, the datasets are classified into levels according to the magnitude of the forecast values. The feature values ​​within each level are calculated using the following method:

[0016]

[0017] In the formula, ε′ i,j sp′ is the relative error value of the j-th sample within the i-th level. i,j The normalized value for the predicted sample; st′ i,j This is the normalized value of the measured sample.

[0018]

[0019] In the formula, n is the first feature value of the hierarchy; i denoted as the number of relative error values ​​for the samples at level i.

[0020]

[0021] In the formula, It is the second feature value of the level.

[0022] Furthermore, based on two feature values ​​in this sample level, super Latin square sampling is used to generate a wind and solar power output forecast scene set.

[0023] Furthermore, step S3 specifically involves using K-means random centroid clustering to reduce the number of clusters, with the clustering results controlled between 4 and 8.

[0024] Furthermore, step S4 specifically involves: dividing the data into intervals based on the normalized results of the measured wind and solar power output, and establishing a scene migration probability calculation model. The probability calculation method is as follows:

[0025] st″=(st′-μ′) / σ′

[0026] In the formula, st″ represents the standard score of the standard normal distribution; μ′ represents the mean of the normalized value of the measured sample; and σ′ represents the standard deviation of the normalized value of the measured sample.

[0027]

[0028] In the formula, Pst′ (u,d) This represents the probability that the next normalized measured value is st′ when the current measured value is in the interval (u,d) after normalization.

[0029] Furthermore, step S5 specifically involves: based on the results of the typical scenario set in the wind and solar forecast, calculating the probability of each typical scenario occurring hourly, and continuously generating decision scenarios for short-term optimization scheduling calculations. The calculation method is as follows:

[0030]

[0031]

[0032] In the formula, P′ q Let q be the migration probability coefficient of the qth scene in the typical scene set for wind and light forecasting.

[0033]

[0034] In the formula, x′ p This represents the wind and solar forecast scene value for the p-th hour within the optimized scheduling period; x p,q This represents the value of the w-th scene in the p-th hour of the typical scene values ​​for the weather forecast.

[0035] Furthermore, a joint short-term optimization scheduling model for wind, solar, and hydropower is constructed with the goal of achieving optimal source-load matching. Based on continuously generated short-term optimization scheduling decision scenarios, a joint short-term optimization scheduling model for wind, solar, and hydropower considering the uncertainties of wind and solar power is realized. The calculation method is as follows:

[0036]

[0037] Nh p =N w +N s +N h

[0038]

[0039] In the formula, N v Source-load matching rate; Ny p The total system output for hour p; Nh p The system load in hour p; N w N s N h These are the power output values ​​for wind power, solar power, and hydropower, respectively. H represents the power generation flow rate of the hydropower station at time j. j The head of the hydropower station at time j is the head of the generator.

[0040] A short-term joint wind-solar-hydropower scheduling system based on uncertain wind-solar forecast scenarios includes:

[0041] The data processing module is used to hierarchically divide the input historical landscape power output data and calculate the feature values ​​of the historical data at each level for scene generation.

[0042] The scene generation module is used to establish a landscape scene forecast model and a scene reduction model. Based on the day-ahead forecast data, it generates a scene set using the feature values ​​of samples at each level, and uses iterative clustering to reduce the scene set, outputting a typical landscape forecast scene set.

[0043] The scene migration probability calculation module is used to calculate the migration probability between measured values ​​that have time correlation, and outputs the migration probability coefficient of each scene in the typical scene set of wind and light forecasts.

[0044] The optimization scheduling module is used to formulate a joint wind, solar and hydropower optimization scheduling plan. Based on the typical wind and solar forecast scenario set and the migration probability coefficient of each scenario, it performs short-term joint wind, solar and hydropower optimization scheduling calculations according to the principle of best source-load matching.

[0045] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the short-term optimization scheduling method for wind, solar, and water joint based on uncertain wind and solar forecast scenarios as described above.

[0046] A computer-readable storage medium storing a computer program that executes the short-term optimization scheduling method for wind, solar and hydropower based on a wind-solar uncertainty forecast scenario as described above.

[0047] Beneficial effects: Compared with the prior art, the technical effects of the present invention are as follows: (1) Based on the historical wind and solar power output prediction and measured data, the present invention establishes a wind and solar scene forecast model, a scene reduction model and a scene migration probability calculation model based on scene construction. By generating a typical wind and solar forecast scene set, a joint wind, solar and hydropower scheduling plan is formulated, and a new wind, solar and hydropower joint short-term optimization scheduling method considering uncertainty is proposed accordingly; (2) By updating the typical wind and solar forecast scene set and its migration probability coefficient hourly, the optimization scheduling decision can be calculated and updated hourly. The data required for the uncertainty information extraction part of this method are all historical data, and a PYTHON program can be written to handle the joint wind, solar and hydropower short-term optimization scheduling calculation under different input data conditions in any scheduling period; (3) The principle is simple, the operation is simple and flexible, and it is easy to implement. This technical method is based on scene construction and reduction and scene migration probability model for optimization scheduling calculation. The calculation speed is fast and the response time is short; (4) It can provide support for the subsequent access of other renewable energy sources. This technology provides a new method and technology for the optimization scheduling of multi-energy complementary systems with new energy access, and also lays a good foundation for the consumption of new energy in multi-energy complementary systems, reduction of grid impact and optimization of energy structure. Attached Figure Description

[0048] Figure 1 This is a flowchart of the method of the present invention;

[0049] Figure 2 This is a system structure block diagram of the present invention. Detailed Implementation

[0050] The present invention will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. After reading the present invention, any modifications of the present invention in various equivalent forms by those skilled in the art will fall within the scope defined by the appended claims.

[0051] Based on historical wind and solar power output prediction and measured data, this invention establishes a wind and solar scene forecasting model, a scene reduction model, and a scene migration probability calculation model based on scene construction. By updating the typical wind and solar forecast scene set and its migration probability coefficients hourly, the invention calculates, updates, and optimizes scheduling decisions hourly, and performs short-term optimized scheduling of wind, solar, and water resources.

[0052] like Figure 1 As shown, the present invention provides a short-term optimization scheduling method for wind, solar, and water joint operation based on uncertain wind and solar forecast scenarios, comprising the following steps:

[0053] S1. Select historical landscape output data, process missing and outlier values, normalize the processed data, and divide the data into sample strata. The specific steps are as follows:

[0054] S1.1 First, historical wind and solar power output data are screened, and atmospheric environmental measurement data near relevant wind farms and photovoltaic power stations are selected for calculation. The data accuracy is controlled within 1-15 minutes, while ensuring the overall consistency of the data samples to ensure their spatiotemporal correlation.

[0055] S1.2 After obtaining the historical wind and solar power output prediction and measured data, it is necessary to process the missing values ​​and outliers. For wind and solar power output data, there will be no negative values ​​or data exceeding the range of the measuring instrument. In particular, for photovoltaic power output data, the measured value before sunrise should always be zero. Therefore, the selected historical data needs to be complete and relatively reasonable wind and solar power output prediction and measured data.

[0056] S1.3. Normalize the processed samples. The sample normalization method is as follows:

[0057] s′=(s-μ) / σ

[0058] In the formula, s′ is the normalized sample value; s is the original sample data; μ is the mean of the original sample set; and σ is the standard deviation of the original sample set.

[0059] S2. Establish a scene-based wind and solar scene forecasting model, perform sample feature analysis within each level using sample data from each level, and use the analysis results to forecast wind and solar scenes based on the day-ahead forecast data of wind and solar power output, thus obtaining a wind and solar forecast scene set. The specific steps are as follows:

[0060] S2.1 Calculate the forecast error of the forecast values ​​at each time point in the dataset, and divide the data into levels according to the statistical results of the forecast error. The error values ​​can be used to divide the data into levels or the error samples can be divided equally. Determine the upper and lower boundaries of the samples within each level for subsequent sampling to generate the scene.

[0061] S2.2 Calculate the feature values ​​of the data within each level. The feature value calculation method is as follows:

[0062]

[0063] In the formula, ε′ i,j sp′ is the relative error value of the j-th sample within the i-th level. i,j The normalized value for the predicted sample; st′ i,j This is the normalized value of the measured sample.

[0064]

[0065] In the formula, n is the first feature value of the hierarchy; i denoted as the number of relative error values ​​for the samples at level i.

[0066]

[0067] In the formula, It is the second feature value of the level.

[0068] S2.3. Based on the two feature values ​​in this sample level, super Latin square sampling (LHS) is used to generate a wind and solar power output forecast scene set. Super Latin square sampling accurately establishes the input distribution with fewer iterations. Compared to Monte Carlo methods, its key lies in stratifying the input probability and establishing intervals with equal cumulative probability scales. By sampling from each interval or layer of the input distribution, the sampling result is forced to represent the value of each interval, thus forcibly reconstructing the input probability distribution. The generation strategy for the wind, solar, and photovoltaic forecast scene set is as follows: the entire sampling process adopts the "sampling without replacement" principle, and the number of strata in the cumulative distribution is the same as the number of iterations in the entire execution process. It is important to note that when using this method for sampling, it is necessary to maintain the independence between variables. Maintaining independence is achieved by randomly selecting sampling intervals for each variable, which effectively avoids unintentional correlations between variables.

[0069] S3. Establish a scene reduction model to reduce the initial wind and light forecast scene set, and iteratively calculate and select the most representative typical wind and light forecast scenes to form a typical wind and light forecast scene set.

[0070] The K-means algorithm, a fundamental algorithm in clustering, belongs to the category of unsupervised learning. Its basic principle is to randomly determine k initial points as cluster centroids, then calculate the distance between each point in the sample data and the cluster centroid, and classify the samples based on this distance. By iterating the position of the cluster centroids, better clustering results can be obtained.

[0071] The specific steps are as follows:

[0072] S3.1 Based on the initial landscape forecast scene set, we select 4-8 initial centroids, which are the typical cluster centroids.

[0073] S3.2 Calculate the distances of other elements in the entire scene set from the centroid of the typical cluster. Here, the distances are calculated using Euclidean distance.

[0074] S3.3 Select a new cluster centroid and calculate the classification of the entire scene set under the new cluster centroid. By repeatedly performing S3.2-S3.3, guide the classification results of the entire scene set to converge, which indicates that the clustering is complete.

[0075] S4. Establish a scene migration probability calculation model based on historical measured values ​​of wind and solar power output. Divide the measured values ​​into intervals and calculate the migration probability between measured values ​​that are time-dependent. The specific steps are as follows:

[0076] S4.1 Calculate the standard normal distribution score of the corresponding cases for each scenario in the typical scenario set. The calculation method is as follows:

[0077] st″=(st′-μ′) / σ′

[0078] In the formula, st″ represents the standard score of the standard normal distribution; μ′ represents the mean of the normalized value of the measured sample; and σ′ represents the standard deviation of the normalized value of the measured sample.

[0079] S4.2. Based on the normalization results of the measured wind and light data, divide the data into intervals and establish a scene migration probability calculation model. The probability calculation method is as follows:

[0080]

[0081] In the formula, Pst′ (u,d) This represents the probability that the next normalized measured value is st′ when the current measured value is in the interval (u,d) after normalization.

[0082] S5. Based on the results of typical scenarios in the wind and solar power forecast, and using hourly measured data of wind and solar power output, the scheduling decision is updated and optimized hourly. Following the principle of best source-load matching, short-term optimized scheduling of wind, solar, and hydropower is carried out. The specific steps are as follows:

[0083] S5.1 Based on the migration probability values ​​of each scene in the typical scene set for wind and solar prediction, calculate the migration probability coefficient of each scene. The calculation method is as follows:

[0084]

[0085]

[0086] In the formula, P′ q Let q be the migration probability coefficient of the qth scene in the typical scene set for wind and light forecasting.

[0087] S5.2 Based on the migration probability coefficients of each scenario, generate computational scenarios for short-term optimization scheduling. The calculation method is as follows:

[0088]

[0089] In the formula, x′ p This represents the wind and solar forecast scene value for the p-th hour within the optimized scheduling period; x p,q This represents the value of the p-th hour in the q-th scene of the typical weather forecast.

[0090] S5.3. Construct a short-term joint wind-solar-hydropower scheduling model with the goal of achieving optimal source-load matching. Based on continuously generated short-term optimization scheduling decision scenarios, realize short-term joint wind-solar-hydropower scheduling considering the uncertainties of wind and solar power. The calculation method is as follows:

[0091]

[0092] Nh p =N w +N s +N h

[0093]

[0094] In the formula, N v Source-load matching rate; Ny p The total system output for hour p; Nh p The system load in hour p; N w N s N h These are the power output values ​​for wind power, solar power, and hydropower, respectively. H represents the power generation flow rate of the hydropower station at time j. j The head of the hydropower station at time j is the head of the generator.

[0095] This invention, based on historical wind and solar power output prediction and measured data, establishes a scene-based wind and solar scenario forecasting model, a scenario reduction model, and a scenario migration probability calculation model. By updating the typical wind and solar forecast scenario set and its migration probability coefficients hourly, it calculates and updates the optimized scheduling decisions hourly, enabling short-term optimized scheduling of wind, solar, and hydropower combined. This technology provides a new method and technique for the optimized scheduling of multi-energy complementary systems that integrate new energy sources, and lays a solid foundation for the absorption of new energy, reduction of grid impact, and optimization of energy structure in multi-energy complementary systems.

[0096] like Figure 2 As shown, a short-term joint wind-solar-hydropower optimization scheduling system based on uncertain wind-solar forecast scenarios includes:

[0097] The data processing module is used to hierarchically divide the input data and calculate the feature values ​​of the historical data at each level for scene generation.

[0098] The scene generation module is used to establish a landscape scene forecast model and a scene reduction model. Based on the day-ahead forecast data, it generates a scene set using the feature values ​​of samples at each level, and uses iterative clustering to reduce the scene set, outputting a typical landscape forecast scene set.

[0099] The scene migration probability calculation module is used to calculate the migration probability between measured values ​​that have time correlation, and outputs the migration probability coefficient of each scene in the typical scene set of wind and light forecasts.

[0100] The optimization scheduling module is used to formulate a joint wind, solar and hydropower optimization scheduling plan. Based on the typical wind and solar forecast scenario set and the migration probability coefficient of each scenario, it performs short-term joint wind, solar and hydropower optimization scheduling calculations according to the principle of best source-load matching.

[0101] The implementation process and methods of the system are the same, and will not be repeated here.

[0102] Obviously, those skilled in the art should understand that the steps of the wind-solar-hydro joint short-term optimization scheduling method based on wind-solar uncertainty forecast scenarios and the modules of the wind-solar-hydro joint short-term optimization scheduling system based on wind-solar uncertainty forecast scenarios described in the above embodiments of the present invention can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. Optionally, they can be implemented using computer-executable program code, thereby storing them in a storage device for execution by the computing device. In some cases, the steps shown or described can be executed in a different order than those described herein, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the embodiments of the present invention are not limited to any specific hardware and software combination.

Claims

1. A short-term optimization scheduling method for wind-solar-hydro joint operation based on uncertain wind-solar forecast scenarios, characterized in that, Includes the following steps: S1. Select historical landscape output data, process missing and outlier values, normalize the processed data, and divide the data into sample levels. S2. Establish a scene-based wind and solar scene forecasting model, use sample data at each level to perform sample feature analysis within the level, and use the analysis results to perform wind and solar scene forecasting based on the day-ahead forecast data of wind and solar power output, so as to obtain a wind and solar forecast scene set. S3. Establish a scene reduction model to reduce the initial set of wind and light forecast scenes, and iteratively calculate and select the most representative typical wind and light forecast scenes to form a typical set of wind and light forecast scenes. S4. Establish a scene migration probability calculation model based on the historical measured values ​​of the scenery, divide the measured values ​​into intervals, and calculate the migration probability between measured values ​​that have time correlation. S5. Based on the results of typical wind and solar forecast scenarios, and based on the hourly measured data of wind and solar power output, the scheduling decision is updated and optimized hourly. In accordance with the principle of best source-load matching, short-term optimized scheduling of wind, solar and water is carried out. In step S4, intervals are divided based on the normalization results of the measured wind and solar data, and a scene migration probability calculation model is established. The probability calculation method is as follows: In the formula, The standard score represents the standard normal distribution. This represents the mean of the normalized values ​​of the measured sample. This represents the standard deviation of the normalized values ​​of the measured sample. In the formula, This indicates that the current measured value, after normalization, is located at... When the interval is reached, the next measured value after normalization is: The probability of; In step S5, based on the results of the typical scenario set of wind and solar forecasts, the probability of each typical scenario occurring is calculated hourly, and decision scenarios for short-term optimization scheduling calculations are continuously generated. The calculation method is as follows: In the formula, For typical scenarios of wind and light forecasting, the first The migration probability coefficient for each scenario; In the formula, Indicates the first [number]th ... Scenery forecast values ​​for each hour; This represents the first value in the typical scene values ​​for wind and light forecasts. In the first scenario The value of an hour; In step S5, a joint short-term optimization scheduling model for wind, solar, and hydropower is constructed with the goal of achieving optimal source-load matching. Based on the continuously generated decision scenarios for short-term optimization scheduling, joint scheduling of wind, solar, and hydropower is performed. The calculation method is as follows: In the formula, Source-load matching rate; For the first Total system output per hour; For the first System load per hour; , , These represent the output values ​​of wind power, solar power, and hydropower, respectively. For the first hydroelectric power station The power generation flow rate at that time; For the first hydroelectric power station The head of the generator at that time.

2. The short-term joint optimization scheduling method for wind, solar, and hydropower based on uncertain wind and solar forecast scenarios as described in claim 1, characterized in that, In S1, the historical wind and solar power output data includes: day-ahead forecast wind power output and measured wind power output data with consistent spatiotemporal correlation, and day-ahead forecast photovoltaic power output data and measured photovoltaic power output data with consistent spatiotemporal correlation; the method for handling missing values ​​is: data supplementation is performed by linear interpolation based on two adjacent sets of data; the method for handling outliers is: first, outliers are deleted, and their missing data is supplemented using the method for missing values; the sample normalization method is: In the formula, These are the normalized sample values; This refers to the original data of the sample. This is the mean of the original sample set; denoted as the standard deviation of the original sample set.

3. The short-term joint optimization scheduling method for wind, solar, and water resources based on uncertain wind and solar forecast scenarios as described in claim 1, characterized in that, In S2, the structure of the landscape scene forecasting model based on scene construction is as follows: Based on the forecast and measured datasets obtained after sample normalization, the datasets are classified into levels according to the magnitude of the forecast values. The feature values ​​within each level are calculated using the following method: In the formula, For the first The first in the hierarchy The relative error value of each sample; For the normalized value of the predicted sample; This represents the normalized value of the measured sample; In the formula, The first feature value of the hierarchy; For the first The number of relative error values ​​for samples at each level; In the formula, It is the second feature value of the level; Based on two feature values ​​in the sample hierarchy, a super Latin square sampling method is used to generate a wind and solar power output forecast scene set.

4. The short-term joint optimization scheduling method for wind, solar, and hydropower based on uncertain wind and solar forecast scenarios as described in claim 1, characterized in that, The scene reduction method used in S3 is K-means clustering, which uses random centroids for clustering, and the clustering results are controlled between 4 and 8.

5. A system implementing the short-term joint wind-solar-hydropower optimization scheduling method based on wind-solar uncertainty forecast scenarios according to claim 1, characterized in that, include: The data processing module is used to hierarchically divide the input historical landscape power output data and calculate the feature values ​​of the historical data at each level for scene generation. The scene generation module is used to establish a landscape scene forecast model and a scene reduction model. Based on the day-ahead forecast data, it generates a scene set using the feature values ​​of samples at each level, and uses iterative clustering to reduce the scene set, outputting a typical landscape forecast scene set. The scene migration probability calculation module is used to calculate the migration probability between measured values ​​that have time correlation, and outputs the migration probability coefficient of each scene in the typical scene set of wind and light forecasts. The optimization scheduling module is used to formulate a joint wind, solar and hydropower optimization scheduling plan. Based on the typical wind and solar forecast scenario set and the migration probability coefficient of each scenario, it performs short-term joint wind, solar and hydropower optimization scheduling calculations according to the principle of best source-load matching.

6. A computer device, characterized in that, The computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the short-term optimization scheduling method for wind, solar and water joint based on the wind and solar uncertainty forecast scenario as described in any one of claims 1-4.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that executes the short-term optimization scheduling method for wind, solar and hydropower based on wind and solar uncertainty forecast scenarios as described in any one of claims 1-4.

Images