A method and system for generating a mid-long term timing scenario

Abstract

The application discloses a kind of medium and long term timing scene generation method and system, comprising: by historical meteorological data, source load is respectively divided into multiple meteorological mode source load, the approximation degree of each source load meteorological mode is measured with meteorological distance, and source load of meteorological mode approximation is aggregated into typical source load coupling mode.Corresponding four seasons typical wind light scene is generated, and matching is carried out according to the distance calculated between typical wind light scene and historical wind light cluster, and finally typical source load coupling scene is obtained.The advantages of the present application are: the generated scene is closer to the actual situation, and the scene in meteorological aspect is enhanced The explainability.Reduces the influence of insufficient data sample quantity on scene quality, so that the generated scene is more accurate and reliable.

Description

Technical Field

[0001] This invention relates to the field of power system source-load power scenario generation technology, and in particular to a method and system for generating medium- and long-term time-series scenarios. Background Technology

[0002] Due to the influence of meteorological changes, the output of renewable energy exhibits strong intermittency and volatility. Meanwhile, with the development of the national economy, the heating and cooling loads, which are significantly affected by temperature, are continuously increasing. Therefore, the impact of meteorology on source-load power in high-proportion renewable energy power systems will gradually strengthen in the future, leading to a more complex coupling relationship among meteorology, source, and load. Furthermore, meteorological changes are often persistent, potentially causing continuous impacts on source-load power over multiple days, making it difficult to analyze the time-series correlations of high-dimensional source-load variables in the medium to long term. Therefore, how to better consider meteorological factors in constructing medium- to long-term source-load time-series scenarios is of significant research importance for addressing the uncertainty of medium- to long-term source-load in high-proportion renewable energy power systems.

[0003] Generative network-based scene generation methods are relatively mature tools among existing methods for analyzing source-load power uncertainty. However, when the time scale corresponding to scene generation increases to the medium to long term, the amount of data samples required for analyzing source-load time-series scenes increases significantly. In practical applications, however, most source-load data come from diverse sources, often resulting in a small number of matching source-load data samples, leading to poor quality of directly generated source-load coupled scenes in existing studies. Existing research rarely analyzes source-load time-series scenes at medium to long-term scales, and most methods lack consideration of the coupling relationship between meteorology, sources, and loads when constructing source-load time-series scenes.

[0004] In summary, there is a lack of a method for generating source-load coupling scenarios that considers the coupling relationship between meteorology, source and load in the medium to long term to address the uncertainty of source-load time-series power.

[0005] Prior art related to this invention

[0006] To address the uncertainty in source load power, a generative network approach is primarily used for scene generation. The specific implementation is as follows:

[0007] First, a suitable generative network is selected, such as a Generative Adversarial Network (GAN) or a Variational Auto-Encoder (VAE). Then, based on the uncertainty of source-load power, a corresponding network structure is designed. Second, historical source-load power curve data is input into the network. The network learns to generate source-load coupling scenes with statistical characteristics similar to historical data. The network parameters are adjusted using a loss function to improve the final scene generation effect. Finally, using the trained generative network model, a large number of source-load coupling scenes are generated, and typical source-load coupling scenes are obtained through scene reduction methods.

[0008] Disadvantages of existing technology 1

[0009] Existing technology one does not analyze source-load coupling time-series scenarios from a medium- to long-term scale. When the time scale increases to the medium to long term, the amount of data samples required to analyze source-load time-series scenarios increases significantly. However, in practical applications, most source-load data come from diverse sources, often resulting in a small number of matching source-load data samples, leading to poor quality of source-load coupling scenarios directly generated in existing technology one.

[0010] Prior art related to this invention

[0011] Considering the impact of weather on source-load power, source-load power curves are classified based on meteorological information, and source-load scenarios under different weather conditions are generated to improve the interpretability of the generated scenarios. The specific format is as follows:

[0012] First, source-load power curves are clustered based on meteorological data. Then, a suitable generative network is selected, such as a Generative Adversarial Network (GAN) or a Variational Auto-Encoder (VAE), and the corresponding network structure is designed. Second, historical source-load power curve data and corresponding meteorological classification labels are input into the network. The network learns to generate source-load scenes with statistical characteristics similar to the historical data of the corresponding labels. The network parameters are adjusted through a loss function to improve the final scene generation effect. Finally, using the trained generative network model and the corresponding meteorological labels, source-load scenes corresponding to the meteorological conditions are generated.

[0013] Disadvantages of existing technology 2

[0014] The existing technology 2 lacks consideration of the coupling between meteorological conditions and source loads on a medium- to long-term scale, and the generated scenario is a short-term scenario of a source or load under a certain meteorological condition. Summary of the Invention

[0015] This invention addresses the shortcomings of existing technologies by providing a method and system for generating medium- to long-term time-series scenarios.

[0016] To achieve the above-mentioned objectives, the technical solution adopted by the present invention is as follows:

[0017] A method for generating medium- to long-term time-series scenes includes the following steps:

[0018] Meteorological data were obtained based on historical source load datasets;

[0019] The historical power data in the historical source-load dataset were clustered using the K-means method to form wind-solar clusters and load clusters. The wind-solar clusters were clustered according to wind speed, surface radiance, and temperature in that order; the load clusters were clustered only according to temperature.

[0020] Based on the obtained wind and solar clusters and load clusters, meteorological models with large sample sizes are selected as typical meteorological models and historical wind and solar clusters according to custom criteria.

[0021] The typical meteorological patterns of the source load are measured by meteorological distance and matched accordingly, and aggregated into multiple source load coupling patterns; the average value of various meteorological data samples in the historical landscape cluster is taken at each time point; the dimensionality of the average value of meteorological data in the historical landscape cluster is reduced by principal component analysis, and the distance matrix between the meteorological patterns of the landscape cluster and the load cluster after dimensionality reduction is calculated; the landscape scenes of the four seasons are generated by generative adversarial network, and the generated landscape scenes are reduced to typical landscape scenes by K-means method;

[0022] The distance matrix between the generated typical landscape scene and the corresponding historical landscape cluster of the season is calculated using the Dynamic Time Warping (DTW) algorithm, and then matched.

[0023] In the generated typical wind and solar scenes, wind power scenes and photovoltaic scenes are horizontally spliced ​​together, and wind power power sequence samples and photovoltaic power sequence samples at the same moment in the historical wind and solar clusters are horizontally spliced ​​together.

[0024] The distance between a typical scenic scene and a sequence of historical scenic clusters is calculated using DTW, and the average of all distance values ​​is taken as the distance between the typical scenic scene and the historical scenic cluster.

[0025] Based on the distance values ​​between the generated typical landscape scenes and historical landscape clusters, the corresponding typical source-load coupling modes are matched, with the one with the smallest distance being matched.

[0026] Based on the correspondence between historical source loads in the model, the typical historical load scenarios are matched with the generated landscape scenarios to obtain the source load coupling scenarios under this model.

[0027] Furthermore, the steps for averaging the samples of various meteorological data within the historical landscape cluster at each time point are as follows: Assume the data is sampled at one point every 1 hour, then the weekly sample is a sequence of 168 sampling points; let the value of a certain type of meteorological data W in the i-th sample within a cluster at time t be... Then, the meteorological data W within this cluster is represented as:

[0028]

[0029] Where n is the number of samples in the cluster;

[0030] The average of all samples of various meteorological data within each cluster is calculated at each time point, i.e.:

[0031]

[0032] In the formula: W mean Let be the mean of all samples within the cluster at each time step, and n be the number of samples within the cluster.

[0033] Furthermore, the specific formula for calculating the distance between the weather models after dimensionality reduction of the wind-solar cluster and the load cluster is as follows:

[0034]

[0035] In the formula: y S Meteorological patterns characterizing wind clusters, y L The meteorological models characterizing the load clusters, with the two clusters reduced to two dimensions by principal component analysis, have x-coordinates a and b respectively. S and a L The ordinates are b S and b L ;

[0036] This yields the meteorological distance matrix between all wind-solar clusters and load clusters;

[0037] The meteorological distances between all meteorological models are normalized, where the approximate distance value between meteorological models is considered to be less than 0.1; wind-solar clusters and load clusters are aggregated according to the matching principle to obtain multiple source-load coupling models.

[0038] This invention also discloses a medium- to long-term time-series scene generation system based on a typical source-load coupling mode, comprising:

[0039] Data preparation module: Obtain meteorological data based on historical source load datasets;

[0040] Meteorological clustering module: The historical power data in the historical source-load dataset is clustered using the K-means method to form wind-solar clusters and load clusters. The wind-solar clusters are clustered according to wind speed, surface radiance, and temperature in that order; the load clusters are clustered only according to temperature.

[0041] Based on the obtained wind and solar clusters and load clusters, meteorological models with large sample sizes are selected as typical meteorological models and historical wind and solar clusters according to custom criteria.

[0042] Source-load coupling mode aggregation module: Measure and match typical meteorological patterns of source loads based on meteorological distance, aggregating them into multiple source-load coupling modes; take the average of various meteorological data samples within the historical wind-and-sun cluster at each time point; reduce the dimensionality of the average meteorological data in the historical wind-and-sun cluster using principal component analysis, and calculate the distance matrix between the dimensionality-reduced meteorological patterns of the wind-and-sun cluster and the load cluster;

[0043] Scene generation module: Generative adversarial network is used to generate landscape scenes of the four seasons;

[0044] Scene reduction module: Uses the K-means method to reduce the generated landscape scenes to typical landscape scenes;

[0045] Typical Scene Matching Module: The distance matrix between the generated typical landscape scenes and the corresponding historical landscape clusters of the season is calculated using the Dynamic Time Warping (DTW) algorithm, and then matched.

[0046] In the generated typical wind and solar scenes, wind power scenes and photovoltaic scenes are horizontally spliced ​​together, and wind power power sequence samples and photovoltaic power sequence samples at the same moment in the historical wind and solar clusters are horizontally spliced ​​together.

[0047] The distance between a typical scenic scene and a sequence of historical scenic clusters is calculated using DTW, and the average of all distance values ​​is taken as the distance between the typical scenic scene and the historical scenic cluster.

[0048] Based on the distance values ​​between the generated typical landscape scenes and historical landscape clusters, the corresponding typical source-load coupling modes are matched, with the one with the smallest distance being matched.

[0049] Based on the correspondence between historical source loads in the model, the typical historical load scenarios are matched with the generated landscape scenarios to obtain the source load coupling scenarios under this model.

[0050] Furthermore, in the source-load coupling mode aggregation module, the average value of various meteorological data samples within the historical landscape cluster is calculated at each time point as follows: assuming the data is sampled at one point every 1 hour, then the weekly sample is a sequence of 168 sample points; assuming the value of a certain type of meteorological data W in the i-th sample within a cluster at time t is... Then, the meteorological data W within this cluster is represented as:

[0051]

[0052] Where n is the number of samples in the cluster;

[0053] The average of all samples of various meteorological data within each cluster is calculated at each time point, i.e.:

[0054]

[0055] In the formula: W mean Let be the mean of all samples within the cluster at each time step, and n be the number of samples within the cluster.

[0056] Furthermore, the distance matrix between the reduced-dimensional meteorological models of the wind-solar cluster and the load cluster is calculated in the source-load coupling mode aggregation module, using the following specific formula:

[0057]

[0058] In the formula: y S Meteorological patterns characterizing wind clusters, y L The meteorological models characterizing the load clusters, with the two clusters reduced to two dimensions by principal component analysis, have x-coordinates a and b respectively. S and a L The ordinates are b S and b L ;

[0059] This yields the meteorological distance matrix between all wind-solar clusters and load clusters;

[0060] The meteorological distances between all meteorological models are normalized, where the approximate distance value between meteorological models is considered to be less than 0.1; wind-solar clusters and load clusters are aggregated according to the matching principle to obtain multiple source-load coupling models.

[0061] The present invention also discloses a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the above-described method for generating medium- and long-term time-series scenarios.

[0062] The present invention also discloses a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the above-described method for generating medium- and long-term time-series scenarios.

[0063] Compared with the prior art, the advantages of the present invention are as follows:

[0064] The coupling characteristics of source-load scenarios are enhanced: By aggregating source-loads under the same meteorological pattern into typical source-load coupling patterns based on meteorological distance, this method and system enhance the coupling characteristics of the final generated typical source-load scenarios. This means that the correlation between sources and loads is more explicit and accurate, making the generated scenarios closer to reality and enhancing the meteorological interpretability of the scenarios.

[0065] Improved Scene Generation Quality: Traditional generative modeling methods often face challenges when generating medium- to long-term source-load coupled scenes, such as diverse source-load data sources and insufficient temporal matching data samples, resulting in low-quality generated scenes. However, the source-load meteorological model matching and aggregation method based on meteorological distance improves upon this. This method reduces the impact of insufficient data samples by aggregating source loads from similar meteorological models, thereby improving the quality of generated scenes and making them more accurate and reliable. Attached Figure Description

[0066] Figure 1 This is a flowchart of scene generation based on a typical source-load coupling mode in an embodiment of the present invention.

[0067] Figure 2 This is a result diagram of the meteorological distance matrix in an embodiment of the present invention.

[0068] Figure 3 This is a typical source-load coupling scenario diagram of an embodiment of the present invention; wherein (a) is a typical source-load scenario in summer corresponding to wind-solar cluster 9 and load cluster 1, and (b) is a typical source-load scenario in summer corresponding to wind-solar cluster 10 and load cluster 2. Detailed Implementation

[0069] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and examples.

[0070] like Figure 1 As shown, this invention provides a method for generating medium- to long-term time-series scenes, comprising the following steps:

[0071] First, the historical source-load dataset was preprocessed. Considering the impact of different seasons and weather changes on the source-load power curve, the data was divided into four seasons and samples were defined according to the length of each week. Since the load exhibits unique fluctuation characteristics from weekdays to weekends and is significantly affected by human factors, the load samples for each week were arranged in order from Monday to Sunday.

[0072] To obtain multiple meteorological models, historical power data of source and load were clustered using the K-means method based on meteorological conditions. Historical source data were clustered sequentially by wind speed, surface radiance, and temperature. Since the amount of load data was less than that of wind and solar data, to ensure sufficient samples for each cluster, load clusters were only clustered by temperature. After clustering, multiple meteorological model clusters for wind / solar and load were formed. To ensure the typicality of the meteorological models, all models were screened based on sample size. Assuming the total source or load sample size is m and the number of meteorological model categories is n, when n ≤ 5, meteorological models with a sample size greater than 20% of the total sample size were retained as typical meteorological models; when n > 5, meteorological models with a sample size greater than m / n were retained as typical meteorological models.

[0073] After obtaining meteorological models from multiple source loads, it is necessary to measure and match these models using meteorological distance to aggregate them into multiple source load coupled models. Assuming the data consists of one sampling point every 1 hour, the weekly sample is a sequence of 168 sampling points. Each cluster contains samples from multiple meteorological data types. Let's assume that the value of a certain type of meteorological data W in the i-th sample within a cluster at time t is... Then the meteorological data W within this cluster can be represented as:

[0074]

[0075] Where n is the number of samples in the cluster. The mean of all samples of various meteorological data within each cluster is calculated at each time point, i.e.:

[0076]

[0077] In the formula: W mean Let be the mean of all samples within the cluster at each time step, and n be the number of samples within the cluster.

[0078] To comprehensively consider various meteorological factors and reduce the complexity of the analysis, principal component analysis (PCA) is used to reduce the dimensionality of the mean values ​​of various meteorological data. The multidimensional meteorological data for each cluster is reduced to a two-dimensional coordinate position (a, b) using PCA. The distance between the reduced meteorological models of each cluster is calculated using Euclidean distance; this distance represents the meteorological distance between the models of each cluster. The corresponding formula is as follows:

[0079]

[0080] In the formula: y S Meteorological patterns characterizing wind clusters, y L The meteorological models characterizing the load clusters, with the two clusters reduced to two dimensions by principal component analysis, have x-coordinates a and b respectively. S and a L The ordinates are b S and bL The final meteorological distance between the two clusters is D(y). L ,y S This yields the meteorological distance matrix between all wind-solar clusters and load clusters. The smaller the distance, the greater the similarity between the meteorological models of the two clusters, and the smaller the meteorological discrepancy. The meteorological distances between all meteorological models are normalized, with a distance value considered to be less than 0.1 for approximate meteorological models. The wind-solar clusters and load clusters are then aggregated according to a matching principle to obtain multiple source-load coupling models.

[0081] Since the sample size of the load is relatively small while the sample size of the source is relatively large, in order to ensure the typicality and rationality of the source and load scenarios, the typical scenarios of the source will be generated by using a convolutional generative adversarial network to generate the scenery scenes of the four seasons, and then reduced to typical scenery scenes using the K-means method. The typical scenarios of the load will be obtained by K-means clustering of historical data.

[0082] After scene generation, the distances between typical seasonal landscape scenes and corresponding historical landscape clusters are calculated and matched using the Dynamic Time Warping (DTW) algorithm. The wind power and photovoltaic scenes within the generated typical landscape scenes are then horizontally stitched together, as are the wind power and photovoltaic power sequence samples from the historical landscape clusters at the same time. After each stitching, the distance between the typical landscape scene and the sequence samples within the landscape cluster is calculated using DTW, and the average of all distance values ​​is taken as the distance between the typical landscape scene and the historical landscape cluster. Typical landscape scenes for each season are matched to corresponding typical source-load coupling modes based on the distances calculated by DTW, with the smallest distance being matched. Then, based on the historical source-load correspondence in the mode, historical typical load scenes are matched with the generated landscape scenes to obtain the source-load coupling scene under this mode.

[0083] Taking data from Japan as an example, historical source-load data and load data are first categorized according to the four seasons, with a sample length of one week, meaning each power sample is 168 hours. Load data is divided according to Monday to Sunday. Then, the historical source-load data for each season are clustered according to meteorological conditions using the K-means method. The maximum silhouette coefficient is used as the number of clusters for all clustering processes. After completing the clustering of each source-load, meteorological distances are calculated according to the procedure, and the meteorological distance matrix results are shown below. Figure 2 .

[0084] like Figure 2As shown in the meteorological distance matrix, the meteorological distance between meteorological models in the same season is relatively small, while the meteorological distance between meteorological models in different seasons is relatively large. The daily average meteorological data corresponding to some matched typical meteorological models are displayed and compared, taking summer weather clusters 9, 1, 10, and 2 as examples. Using 75% and 25% of the daily average maximum value of Japanese summer meteorological data as boundaries, above 75% is defined as high, and below 25% as low. For example, a daily average temperature above 75% of the daily average maximum value represents high temperatures in summer, and below 25% represents low temperatures. The proportion of the daily average values ​​of each meteorological model in the four clusters falling within these two defined ranges is statistically analyzed, and the results are shown in Table 1.

[0085] Table 1. Statistics on the percentage of daily meteorological averages for each cluster.

[0086]

[0087] Overall, the weather conditions in Fengguang Cluster 9 and Load Cluster 1 are moderate in temperature, mostly cloudy, and mostly with light winds and little rain. The weather conditions in Fengguang Cluster 9 and Load Cluster 1 are mostly hot in temperature, mostly cloudy, mostly with light winds and little rain, and less sunny than the previous two clusters.

[0088] After obtaining typical source-load coupling patterns from historical data, a generative adversarial network (GAN) is used to model wind and solar power output. K-means clustering is then used for reduction to generate typical wind and solar scenes for each of the four seasons. These typical scenes are then matched with wind and solar clusters within the typical source-load coupling patterns using distance calculations via DTW (Dynamic Data Wave) to obtain the corresponding source-load coupling scenes. The final typical source-load coupling scenes are shown below. Figure 3 As shown.

[0089] Figure 3 Compared to the scenario load in a Figure 3 Scenario b is generally higher, which corresponds to the temperature conditions in both scenarios; the wind power output in both scenarios is generally lower, corresponding to the mostly light wind conditions in both scenarios; due to Figure 3 The proportion of sunny conditions in scene a is compared to Figure 3 There are more scenarios in scenario b, therefore Figure 3 The peak output of some photovoltaic power in scenario a is compared to Figure 3 The peak value in scenario b is larger. Overall, both source-load coupling scenarios are consistent with their corresponding meteorological conditions, indicating that the present invention can match source loads under the same meteorological conditions and generate source-load coupling scenarios that are interpretable in terms of meteorology.

[0090] In one embodiment of the present invention, a medium- to long-term time-series scene generation system based on a typical source-load coupling mode is provided. This system can be used to implement the aforementioned medium- to long-term time-series scene generation method based on a typical source-load coupling mode, specifically including:

[0091] Data preparation module: Obtain meteorological data based on historical source load datasets;

[0092] Meteorological clustering module: The historical power data in the historical source-load dataset is clustered using the K-means method to form wind-solar clusters and load clusters. The wind-solar clusters are clustered according to wind speed, surface radiance, and temperature in that order; the load clusters are clustered only according to temperature.

[0093] Based on the obtained wind and solar clusters and load clusters, meteorological models with large sample sizes are selected as typical meteorological models and historical wind and solar clusters according to custom criteria.

[0094] Source-load coupling mode aggregation module: Measure and match typical meteorological patterns of source loads based on meteorological distance, aggregating them into multiple source-load coupling modes; take the average of various meteorological data samples within the historical wind-and-sun cluster at each time point; reduce the dimensionality of the average meteorological data in the historical wind-and-sun cluster using principal component analysis, and calculate the distance matrix between the dimensionality-reduced meteorological patterns of the wind-and-sun cluster and the load cluster;

[0095] Scene generation module: Generative adversarial network is used to generate landscape scenes of the four seasons;

[0096] Scene reduction module: Uses the K-means method to reduce the generated landscape scenes to typical landscape scenes;

[0097] Typical Scene Matching Module: The distance matrix between the generated typical landscape scenes and the corresponding historical landscape clusters of the season is calculated using the Dynamic Time Warping (DTW) algorithm, and then matched.

[0098] In the generated typical wind and solar scenes, wind power scenes and photovoltaic scenes are horizontally spliced ​​together, and wind power power sequence samples and photovoltaic power sequence samples at the same moment in the historical wind and solar clusters are horizontally spliced ​​together.

[0099] The distance between a typical scenic scene and a sequence of historical scenic clusters is calculated using DTW, and the average of all distance values ​​is taken as the distance between the typical scenic scene and the historical scenic cluster.

[0100] Based on the distance values ​​between the generated typical landscape scenes and historical landscape clusters, the corresponding typical source-load coupling modes are matched, with the one with the smallest distance being matched.

[0101] Based on the correspondence between historical source loads in the model, the typical historical load scenarios are matched with the generated landscape scenarios to obtain the source load coupling scenarios under this model.

[0102] In one embodiment of the present invention, a terminal device is provided, comprising a processor and a memory. The memory stores a computer program, which includes program instructions. The processor executes the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions to achieve a corresponding method flow or corresponding function. The processor described in this embodiment can be used in the operation of a medium-to-long-term time-series scene generation method based on a typical source-load coupling mode, including the following steps:

[0103] Meteorological data were obtained based on historical source load datasets;

[0104] The historical power data in the historical source-load dataset were clustered using the K-means method to form wind-solar clusters and load clusters. The wind-solar clusters were clustered according to wind speed, surface radiance, and temperature in that order; the load clusters were clustered only according to temperature.

[0105] Based on the obtained wind and solar clusters and load clusters, meteorological models with large sample sizes are selected as typical meteorological models and historical wind and solar clusters according to custom criteria.

[0106] The typical meteorological patterns of the source load are measured by meteorological distance and matched accordingly, and aggregated into multiple source load coupling patterns; the average value of various meteorological data samples in the historical landscape cluster is taken at each time point; the dimensionality of the average value of meteorological data in the historical landscape cluster is reduced by principal component analysis, and the distance matrix between the meteorological patterns of the landscape cluster and the load cluster after dimensionality reduction is calculated; the landscape scenes of the four seasons are generated by generative adversarial network, and the generated landscape scenes are reduced to typical landscape scenes by K-means method;

[0107] The distance matrix between the generated typical landscape scene and the corresponding historical landscape cluster of the season is calculated using the Dynamic Time Warping (DTW) algorithm, and then matched.

[0108] In the generated typical wind and solar scenes, wind power scenes and photovoltaic scenes are horizontally spliced ​​together, and wind power power sequence samples and photovoltaic power sequence samples at the same moment in the historical wind and solar clusters are horizontally spliced ​​together.

[0109] The distance between a typical scenic scene and a sequence of historical scenic clusters is calculated using DTW, and the average of all distance values ​​is taken as the distance between the typical scenic scene and the historical scenic cluster.

[0110] Based on the distance values ​​between the generated typical landscape scenes and historical landscape clusters, the corresponding typical source-load coupling modes are matched, with the one with the smallest distance being matched.

[0111] Based on the historical source-load correspondence in the model, typical historical load scenarios are matched with generated wind and solar scenarios to obtain the source-load coupled scenario under this model. Meteorological data is obtained from the historical source-load dataset.

[0112] The historical power data in the historical source-load dataset were clustered using the K-means method to form wind-solar clusters and load clusters. The wind-solar clusters were clustered according to wind speed, surface radiance, and temperature in that order; the load clusters were clustered only according to temperature.

[0113] Based on the obtained wind and solar clusters and load clusters, meteorological models with large sample sizes are selected as typical meteorological models and historical wind and solar clusters according to custom criteria.

[0114] The typical meteorological patterns of the source load are measured by meteorological distance and matched accordingly, and aggregated into multiple source load coupling patterns; the average value of various meteorological data samples in the historical landscape cluster is taken at each time point; the dimensionality of the average value of meteorological data in the historical landscape cluster is reduced by principal component analysis, and the distance matrix between the meteorological patterns of the landscape cluster and the load cluster after dimensionality reduction is calculated; the landscape scenes of the four seasons are generated by generative adversarial network, and the generated landscape scenes are reduced to typical landscape scenes by K-means method;

[0115] The distance matrix between the generated typical landscape scene and the corresponding historical landscape cluster of the season is calculated using the Dynamic Time Warping (DTW) algorithm, and then matched.

[0116] In the generated typical wind and solar scenes, wind power scenes and photovoltaic scenes are horizontally spliced ​​together, and wind power power sequence samples and photovoltaic power sequence samples at the same moment in the historical wind and solar clusters are horizontally spliced ​​together.

[0117] The distance between a typical scenic scene and a sequence of historical scenic clusters is calculated using DTW, and the average of all distance values ​​is taken as the distance between the typical scenic scene and the historical scenic cluster.

[0118] Based on the distance values ​​between the generated typical landscape scenes and historical landscape clusters, the corresponding typical source-load coupling modes are matched, with the one with the smallest distance being matched.

[0119] Based on the correspondence between historical source loads in the model, the typical historical load scenarios are matched with the generated landscape scenarios to obtain the source load coupling scenarios under this model.

[0120] In one embodiment of the present invention, a storage medium is also provided, specifically a computer-readable storage medium (memory), which is a memory device in a terminal device used to store programs and data. It is understood that the computer-readable storage medium here can include both the built-in storage medium in the terminal device and extended storage media supported by the terminal device. The computer-readable storage medium provides storage space that stores the terminal's operating system. Furthermore, the storage space also stores one or more instructions suitable for loading and execution by a processor, which can be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here can be high-speed RAM or non-volatile memory, such as at least one disk storage device.

[0121] One or more instructions stored in a computer-readable storage medium can be loaded and executed by a processor to implement the corresponding steps of the method for generating medium- to long-term time-series scenarios based on typical source-load coupling modes in the above embodiments; one or more instructions in the computer-readable storage medium are loaded and executed by the processor to perform the following steps:

[0122] Meteorological data were obtained based on historical source load datasets;

[0123] The historical power data in the historical source-load dataset were clustered using the K-means method to form wind-solar clusters and load clusters. The wind-solar clusters were clustered according to wind speed, surface radiance, and temperature in that order; the load clusters were clustered only according to temperature.

[0124] Based on the obtained wind and solar clusters and load clusters, meteorological models with large sample sizes are selected as typical meteorological models and historical wind and solar clusters according to custom criteria.

[0125] The typical meteorological patterns of the source load are measured by meteorological distance and matched accordingly, and aggregated into multiple source load coupling patterns; the average value of various meteorological data samples in the historical landscape cluster is taken at each time point; the dimensionality of the average value of meteorological data in the historical landscape cluster is reduced by principal component analysis, and the distance matrix between the meteorological patterns of the landscape cluster and the load cluster after dimensionality reduction is calculated; the landscape scenes of the four seasons are generated by generative adversarial network, and the generated landscape scenes are reduced to typical landscape scenes by K-means method;

[0126] The distance matrix between the generated typical landscape scene and the corresponding historical landscape cluster of the season is calculated using the Dynamic Time Warping (DTW) algorithm, and then matched.

[0127] In the generated typical wind and solar scenes, wind power scenes and photovoltaic scenes are horizontally spliced ​​together, and wind power power sequence samples and photovoltaic power sequence samples at the same moment in the historical wind and solar clusters are horizontally spliced ​​together.

[0128] The distance between a typical scenic scene and a sequence of historical scenic clusters is calculated using DTW, and the average of all distance values ​​is taken as the distance between the typical scenic scene and the historical scenic cluster.

[0129] Based on the distance values ​​between the generated typical landscape scenes and historical landscape clusters, the corresponding typical source-load coupling modes are matched, with the one with the smallest distance being matched.

[0130] Based on the historical source-load correspondence in the model, typical historical load scenarios are matched with generated wind and solar scenarios to obtain the source-load coupled scenario under this model. Meteorological data is obtained from the historical source-load dataset.

[0131] The historical power data in the historical source-load dataset were clustered using the K-means method to form wind-solar clusters and load clusters. The wind-solar clusters were clustered according to wind speed, surface radiance, and temperature in that order; the load clusters were clustered only according to temperature.

[0132] Based on the obtained wind and solar clusters and load clusters, meteorological models with large sample sizes are selected as typical meteorological models and historical wind and solar clusters according to custom criteria.

[0133] The typical meteorological patterns of the source load are measured by meteorological distance and matched accordingly, and aggregated into multiple source load coupling patterns; the average value of various meteorological data samples in the historical landscape cluster is taken at each time point; the dimensionality of the average value of meteorological data in the historical landscape cluster is reduced by principal component analysis, and the distance matrix between the meteorological patterns of the landscape cluster and the load cluster after dimensionality reduction is calculated; the landscape scenes of the four seasons are generated by generative adversarial network, and the generated landscape scenes are reduced to typical landscape scenes by K-means method;

[0134] The distance matrix between the generated typical landscape scene and the corresponding historical landscape cluster of the season is calculated using the Dynamic Time Warping (DTW) algorithm, and then matched.

[0135] In the generated typical wind and solar scenes, wind power scenes and photovoltaic scenes are horizontally spliced ​​together, and wind power power sequence samples and photovoltaic power sequence samples at the same moment in the historical wind and solar clusters are horizontally spliced ​​together.

[0136] The distance between a typical scenic scene and a sequence of historical scenic clusters is calculated using DTW, and the average of all distance values ​​is taken as the distance between the typical scenic scene and the historical scenic cluster.

[0137] Based on the distance values ​​between the generated typical landscape scenes and historical landscape clusters, the corresponding typical source-load coupling modes are matched, with the one with the smallest distance being matched.

[0138] Based on the correspondence between historical source loads in the model, the typical historical load scenarios are matched with the generated landscape scenarios to obtain the source load coupling scenarios under this model.

[0139] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0140] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0141] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1The function specified in one or more boxes.

[0142] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0143] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the implementation methods of the present invention, and should be understood that the scope of protection of the present invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations based on the technical teachings disclosed in this invention without departing from the spirit of the invention, and these modifications and combinations are still within the scope of protection of the present invention.

Claims

1. A method for generating medium- to long-term time-series scenes, characterized in that, Includes the following steps: Meteorological data were obtained based on historical source load datasets; The historical power data in the historical source-load dataset were clustered using the K-means method to form wind-solar coarse and load clusters. The wind-solar clusters were clustered according to wind speed, surface radiance and temperature in that order. Load clusters are clustered only by temperature; Based on the obtained wind and solar clusters and load clusters, the corresponding meteorological models are selected as typical meteorological models and historical wind and solar clusters according to the custom criteria. The typical meteorological patterns of the source load are measured by meteorological distance and matched accordingly, and aggregated into multiple source load coupling patterns; the average value of various meteorological data samples in the historical landscape cluster is taken at each time point; the dimensionality of the average value of meteorological data in the historical landscape cluster is reduced by principal component analysis, and the distance matrix between the meteorological patterns of the landscape cluster and the load cluster after dimensionality reduction is calculated; the landscape scenes of the four seasons are generated by generative adversarial network, and the generated landscape scenes are reduced to typical landscape scenes by K-means method; The distance matrix between the generated typical landscape scene and the corresponding historical landscape cluster of the season is calculated using the Dynamic Time Warping (DTW) algorithm, and then matched. In the generated typical wind and solar scenes, wind power scenes and photovoltaic scenes are horizontally spliced ​​together, and wind power power sequence samples and photovoltaic power sequence samples at the same moment in the historical wind and solar clusters are horizontally spliced ​​together. The distance between a typical scenic scene and a sequence of historical scenic clusters is calculated using DTW, and the average of all distance values ​​is taken as the distance between the typical scenic scene and the historical scenic cluster. Based on the distance values ​​between the generated typical landscape scenes and historical landscape clusters, the corresponding typical source-load coupling modes are matched, with the one with the smallest distance being matched. Based on the correspondence between historical source loads in the model, the typical historical load scenarios are matched with the generated landscape scenarios to obtain the source load coupling scenarios under this model.

2. The method for generating medium- to long-term time-series scenes according to claim 1, characterized in that... The steps for averaging the samples of various meteorological data within the historical landscape cluster at each time point are as follows: Assume the data is sampled at one point every 1 hour, then the weekly sample is a sequence of 168 sampling points; let the value of a certain type of meteorological data W in the i-th sample within a cluster at time t be... Then, the meteorological data W within this cluster is represented as: Where n is the number of samples in the cluster; The average of all samples of various meteorological data within each cluster is calculated at each time point, i.e.: In the formula: W mean Let be the mean of all samples within the cluster at each time step, and n be the number of samples within the cluster.

3. The method for generating medium- to long-term time-series scenes according to claim 1, characterized in that: The specific formula for calculating the distance between the weather models after dimensionality reduction of the wind-solar cluster and the load cluster is as follows: In the formula: y S Meteorological patterns characterizing wind clusters, y L The meteorological models characterizing the load clusters, with the two clusters reduced to two dimensions by principal component analysis, have x-coordinates a and b respectively. S and a L The ordinates are b S and b L ; This yields the meteorological distance matrix between all wind-solar clusters and load clusters; The meteorological distances between all meteorological models are normalized, where the approximate distance value between meteorological models is considered to be less than 0.1; wind-solar clusters and load clusters are aggregated according to the matching principle to obtain multiple source-load coupling models.

4. A medium- to long-term time-series scene generation system, characterized in that: include: Data preparation module: Obtain meteorological data based on historical source load datasets; Meteorological clustering module: The historical power data in the historical source-load dataset is clustered using the K-means method to form wind-solar coarse and load clusters. The wind-solar clusters are clustered according to wind speed, surface radiance and temperature in that order. Load clusters are clustered only by temperature; Based on the obtained landscape clusters and load clusters, corresponding meteorological models are selected as typical meteorological models and historical landscape clusters according to custom criteria. Source-load coupling mode aggregation module: Measure and match typical meteorological patterns of source loads based on meteorological distance, aggregating them into multiple source-load coupling modes; take the average of various meteorological data samples within the historical wind-and-sun cluster at each time point; reduce the dimensionality of the average meteorological data in the historical wind-and-sun cluster using principal component analysis, and calculate the distance matrix between the dimensionality-reduced meteorological patterns of the wind-and-sun cluster and the load cluster; Scene generation module: Generative adversarial network is used to generate landscape scenes of the four seasons; Scene reduction module: Uses the K-means method to reduce the generated landscape scenes to typical landscape scenes; Typical Scene Matching Module: The distance matrix between the generated typical landscape scenes and the corresponding historical landscape clusters of the season is calculated using the Dynamic Time Warping (DTW) algorithm, and then matched. In the generated typical wind and solar scenes, wind power scenes and photovoltaic scenes are horizontally spliced ​​together, and wind power power sequence samples and photovoltaic power sequence samples at the same moment in the historical wind and solar clusters are horizontally spliced ​​together. The distance between a typical scenic scene and a sequence of historical scenic clusters is calculated using DTW, and the average of all distance values ​​is taken as the distance between the typical scenic scene and the historical scenic cluster. Based on the distance values ​​between the generated typical landscape scenes and historical landscape clusters, the corresponding typical source-load coupling modes are matched, with the one with the smallest distance being matched. Based on the correspondence between historical source loads in the model, the typical historical load scenarios are matched with the generated landscape scenarios to obtain the source load coupling scenarios under this model.

5. The medium-to-long-term time-series scene generation system according to claim 4, characterized in that: In the source-load coupling mode aggregation module, the average value of various meteorological data samples within the historical landscape cluster is calculated at each time point as follows: Assuming the data is sampled at one point every 1 hour, the weekly sample is a sequence of 168 sample points; let the value of a certain type of meteorological data W in the i-th sample within a cluster at time t be... Then, the meteorological data W within this cluster is represented as: Where n is the number of samples in the cluster; The average of all samples of various meteorological data within each cluster is calculated at each time point, i.e.: In the formula: W mean Let be the mean of all samples within the cluster at each time step, and n be the number of samples within the cluster.

6. The medium-to-long-term time-series scene generation system according to claim 4, characterized in that: The source-load coupling mode aggregation module calculates the distance matrix between the reduced-dimensional meteorological models of the wind-solar cluster and the load cluster, using the following formula: In the formula: y S Meteorological patterns characterizing wind clusters, y L The meteorological models characterizing the load clusters, with the two clusters reduced to two dimensions by principal component analysis, have x-coordinates a and b respectively. S and a L The ordinates are b S and b L ; This yields the meteorological distance matrix between all wind-solar clusters and load clusters; The meteorological distances between all meteorological models are normalized, where the approximate distance value between meteorological models is considered to be less than 0.1; wind-solar clusters and load clusters are aggregated according to the matching principle to obtain multiple source-load coupling models.

7. A computer device, characterized in that: The method includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the medium-to-long-term time-series scene generation method according to any one of claims 1 to 3.

8. A computer-readable storage medium, characterized in that: It stores a computer program that, when executed by a processor, implements the medium-to-long-term time-series scene generation method as described in any one of claims 1 to 3.

Images