Wind power group power prediction method, system, equipment and medium

By constructing a meteorological feature classification model and a graph convolution model, the problems of insufficient spatiotemporal continuity and the applicability of meteorological feature classification in wind power group power prediction are solved, and higher accuracy wind power group power prediction is achieved.

CN119029831APending Publication Date: 2024-11-26CHINA ELECTRIC POWER RESEARCH INSTITUTE CO LTD +2
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202310588263.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2023-05-23
Publication Date
2024-11-26

AI Technical Summary

Technical Problem

Existing wind power group power prediction methods fail to effectively consider the spatiotemporal continuity of wind resources, and the applicability of meteorological feature classification is not strong, resulting in insufficient prediction accuracy.

Method used

By constructing meteorological feature classification models and graph convolution models, and utilizing historical predicted power error data of wind power clusters and gridded numerical weather forecast data, meteorological feature matching and model selection are performed to improve the spatial continuity of wind power cluster forecasts and the adaptability of meteorological features.

Benefits of technology

It improves the accuracy and efficiency of wind power cluster power prediction, reduces data requirements, and enhances the analysis of the spatial continuity of wind power clusters.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119029831A_ABST
    Figure CN119029831A_ABST
Patent Text Reader

Abstract

The invention provides a wind power group power prediction method, system and device and a medium, and the method comprises the steps: carrying out the matching of the grid numerical weather forecast data of a prediction time point of a wind power group with different meteorological feature libraries through a pre-constructed meteorological feature classification model, obtaining meteorological characteristics corresponding to the gridding numerical weather forecast data at each prediction time point; selecting a wind power group prediction model from a plurality of pre-constructed wind power group prediction models according to the meteorological characteristics, and calculating a wind power prediction value through the wind power group prediction model based on the gridding numerical weather forecast data of the prediction time point; the grid numerical weather forecast data of the forecast time point of the wind power group is input, compared with traditional single-point numerical weather forecast, the continuity of the space where the wind power group is located is improved, and the meteorological feature classification model improves the division adaptability of meteorological features.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the field of renewable energy power generation and dispatching operation, and specifically relates to a method, system, equipment and medium for predicting wind power group power. Background Art

[0002] Wind power output has the characteristics of randomness, intermittency and volatility. In order to ensure the safe and stable operation of the power grid after large-scale wind power is connected to the grid, the power grid dispatching department needs to formulate a dispatching plan based on wind power forecast data. At the same time, with the continuous increase in wind power penetration, under the uncertainty constraints that the power system can bear, the accuracy of wind power prediction is required to be continuously improved.

[0003] From the perspective of actual application needs, power grid dispatching agencies need the predicted wind power of the entire network to develop dispatching plans. Therefore, the predicted power of wind power groups in the entire network or in different regions is the data they are more concerned about. The existing wind power group power prediction methods are mainly based on the accumulation of single-station predictions. With the gradual maturity of deep learning related technologies, wind power group predictions based on deep learning have also had preliminary applications. In summary, the existing methods have the following problems: First, whether it is single-station accumulation or deep learning methods, the current methods all use the single-point numerical weather forecast corresponding to the wind farm as input. Since wind resources have temporal and spatial continuity, the single-point numerical weather forecast does not consider the continuity of the space where the group is located; second, under different meteorological characteristics, wind power output has certain distribution characteristics and the differences are large, and the division of meteorological characteristics is not very applicable. Summary of the invention

[0004] In order to overcome the above-mentioned deficiencies of the prior art, the present invention proposes a method for wind power group power prediction, comprising:

[0005] The gridded numerical weather forecast data of the wind power group at the forecast time point is matched with different meteorological feature libraries through the pre-built meteorological feature classification model to obtain the meteorological features corresponding to the gridded numerical weather forecast data at each forecast time point;

[0006] According to the meteorological characteristics, a corresponding wind power group prediction model is selected from a plurality of pre-constructed wind power group prediction models, and based on the gridded numerical weather forecast data at the prediction time point, the wind power prediction value is calculated through the corresponding wind power group prediction model;

[0007] The meteorological characteristic classification model is constructed based on the historical forecast power error data of the wind power group and the historical gridded numerical weather forecast data through the meteorological characteristic-error characteristic matching mapping method;

[0008] The wind power group prediction model is based on meteorological characteristics and is constructed using a graph convolution model.

[0009] Preferably, the construction of the meteorological feature classification model includes:

[0010] Based on the acquired historical predicted power error data of the wind power group, obtaining the historical predicted power error waveform characteristic data after data dimension reduction through the first autoencoder model, and determining the error type of the historical predicted power error waveform characteristic data;

[0011] Based on the historical gridded numerical weather forecast data of the wind power group, the historical gridded numerical weather forecast data after data dimension reduction is obtained through the second autoencoder model;

[0012] The historical gridded numerical weather forecast data after data dimensionality reduction are matched with the meteorological characteristics corresponding to the error type by a meteorological characteristic-error characteristic matching mapping method, and the corresponding meteorological characteristics are marked on the historical gridded numerical weather forecast data after data dimensionality reduction;

[0013] The historical gridded numerical weather forecast data corresponding to each type of meteorological feature after dimension reduction is stored in the corresponding meteorological feature library to complete the construction of the meteorological feature classification model;

[0014] Among them, the first autoencoder model is constructed based on the historical predicted power error data of the wind power group through the first autoencoder and the first decoder; the second autoencoder model is constructed based on the historical gridded numerical weather forecast data of the wind power group through the second autoencoder and the second decoder.

[0015] Preferably, the historical predicted power error data of the wind power group obtained is used to obtain the historical predicted power error waveform feature data after data dimension reduction through the first autoencoder model, including:

[0016] Based on the historical predicted power error data of the wind power group, the historical predicted power error waveform feature data is extracted through the first autoencoder;

[0017] Based on the historical predicted power error waveform characteristic data, obtaining the historical predicted power error waveform characteristic data after data dimension reduction through a first decoder;

[0018] The historical predicted power error waveform characteristic data after the data dimension reduction is divided into multiple categories by a K-means clustering method, and the error type of each category is determined;

[0019] Among them, the first autoencoder is constructed by a feature extraction function; the first decoder is constructed by a time dimension reduction function.

[0020] Preferably, the historical gridded numerical weather forecast data based on the wind power group is obtained through a second autoencoder model to obtain the historical gridded numerical weather forecast data after data dimensionality reduction, including:

[0021] Based on the historical gridded numerical weather forecast data of the wind power group, a data dimension reduction is performed according to the time section through the second autoencoder to obtain the historical gridded numerical weather forecast data after the data dimension reduction;

[0022] Based on the historical gridded numerical weather forecast data after the primary data dimensionality reduction, a secondary data dimensionality reduction is performed through a second decoder to obtain the historical gridded numerical weather forecast data after the secondary data dimensionality reduction;

[0023] Among them, the second autoencoder is constructed based on the time section dimensionality reduction function; the second decoder is constructed based on the longitude and latitude dimensionality reduction function.

[0024] Preferably, the construction of the wind power group prediction model includes:

[0025] Extracting historical gridded numerical weather forecast sample data corresponding to the meteorological feature library of various meteorological features, and obtaining wind power cluster power data at the time corresponding to the historical gridded numerical weather forecast sample data;

[0026] The correspondence between the historical gridded numerical weather forecast sample data and the wind power cluster power data at the corresponding time is constructed by the graph convolution method, and a wind power group prediction model corresponding to various meteorological characteristics is constructed according to the correspondence.

[0027] Preferably, the historical predicted power error data is obtained by making a difference between the historical actual total power data of the wind power group and the corresponding historical predicted power data.

[0028] Based on the same inventive concept, the present invention also provides a system for wind power group power prediction, including: a meteorological feature acquisition module and a power prediction module;

[0029] The meteorological feature acquisition module is used to match the gridded numerical weather forecast data at the forecast time point of the wind power group with different meteorological feature libraries through a pre-built meteorological feature classification model to obtain the meteorological features corresponding to the gridded numerical weather forecast data at each forecast time point;

[0030] The power prediction module is used to select a corresponding wind power group prediction model from a plurality of pre-constructed wind power group prediction models according to the meteorological characteristics, and calculate the wind power prediction value through the corresponding wind power group prediction model based on the gridded numerical weather forecast data at the prediction time point;

[0031] The meteorological characteristic classification model is constructed based on the historical forecast power error data of the wind power group and the historical gridded numerical weather forecast data through the meteorological characteristic-error characteristic matching mapping method;

[0032] The wind power group prediction model is based on meteorological characteristics and is constructed using a graph convolution model.

[0033] Preferably, the construction of the meteorological feature classification model in the meteorological feature acquisition module includes:

[0034] Based on the acquired historical predicted power error data of the wind power group, obtaining the historical predicted power error waveform characteristic data after data dimension reduction through the first autoencoder model, and determining the error type of the historical predicted power error waveform characteristic data;

[0035] Based on the historical gridded numerical weather forecast data of the wind power group, the historical gridded numerical weather forecast data after data dimension reduction is obtained through the second autoencoder model;

[0036] The historical gridded numerical weather forecast data after data dimensionality reduction are matched with the meteorological characteristics corresponding to the error type by a meteorological characteristic-error characteristic matching mapping method, and the corresponding meteorological characteristics are marked on the historical gridded numerical weather forecast data after data dimensionality reduction;

[0037] The historical gridded numerical weather forecast data corresponding to each type of meteorological feature after dimension reduction is stored in the corresponding meteorological feature library to complete the construction of the meteorological feature classification model;

[0038] Among them, the first autoencoder model is constructed based on the historical predicted power error data of the wind power group through the first autoencoder and the first decoder; the second autoencoder model is constructed based on the historical gridded numerical weather forecast data of the wind power group through the second autoencoder and the second decoder.

[0039] Preferably, the meteorological feature acquisition module obtains the historical predicted power error waveform feature data after data dimension reduction based on the acquired historical predicted power error data of the wind power group through the first autoencoder model, and determines the error type of the historical predicted power error waveform feature data, including:

[0040] Based on the historical predicted power error data of the wind power group, the historical predicted power error waveform feature data is extracted through the first autoencoder;

[0041] Based on the historical predicted power error waveform characteristic data, obtaining the historical predicted power error waveform characteristic data after data dimension reduction through a first decoder;

[0042] The historical predicted power error waveform characteristic data after the data dimension reduction is divided into multiple categories by a K-means clustering method, and the error type of each category is determined;

[0043] Among them, the first autoencoder is constructed by a feature extraction function; the first decoder is constructed by a time dimension reduction function.

[0044] Preferably, the meteorological feature acquisition module obtains the historical gridded numerical weather forecast data after data dimension reduction based on the historical gridded numerical weather forecast data of the wind power group through the second autoencoder model, including:

[0045] Based on the historical gridded numerical weather forecast data of the wind power group, a data dimension reduction is performed according to the time section through the second autoencoder to obtain the historical gridded numerical weather forecast data after the data dimension reduction;

[0046] Based on the historical gridded numerical weather forecast data after the primary data dimensionality reduction, a secondary data dimensionality reduction is performed through a second decoder to obtain the historical gridded numerical weather forecast data after the secondary data dimensionality reduction;

[0047] Among them, the second autoencoder is constructed based on the time section dimensionality reduction function; the second decoder is constructed based on the longitude and latitude dimensionality reduction function.

[0048] Preferably, the construction of the wind power group prediction model of the power prediction module includes:

[0049] Extracting historical gridded numerical weather forecast sample data corresponding to the meteorological feature library of various meteorological features, and obtaining wind power cluster power data at the time corresponding to the historical gridded numerical weather forecast sample data;

[0050] The correspondence between the historical gridded numerical weather forecast sample data and the wind power cluster power data at the corresponding time is constructed by the graph convolution method, and a wind power group prediction model corresponding to various meteorological characteristics is constructed according to the correspondence.

[0051] Preferably, the historical predicted power error data of the meteorological feature acquisition module is obtained by making a difference between the historical actual total power data of the wind power group and the corresponding historical predicted power data.

[0052] Based on the same inventive concept, the present invention also provides a computer device, including: one or more processors;

[0053] A memory for storing one or more programs;

[0054] When the one or more programs are executed by the one or more processors, a method for wind power group power prediction as described above is implemented.

[0055] Based on the same inventive concept, the present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed, the method for predicting wind power group power as described above is implemented.

[0056] Compared with the closest prior art, the present invention has the following beneficial effects:

[0057] The present invention provides a method, system, device and medium for predicting wind power group power, comprising: matching gridded numerical weather forecast data at a prediction time point of a wind power group with different meteorological feature libraries through a pre-constructed meteorological feature classification model to obtain meteorological features corresponding to the gridded numerical weather forecast data at each prediction time point; selecting a corresponding wind power group prediction model from a plurality of pre-constructed wind power group prediction models according to the meteorological features, and calculating a wind power prediction value through the corresponding wind power group prediction model based on the gridded numerical weather forecast data at the prediction time point; wherein the meteorological feature classification model is constructed based on historical prediction power error data of the wind power group and historical gridded numerical weather forecast data through a meteorological characteristic-error characteristic matching mapping method; the wind power group prediction model is constructed based on meteorological features and adopts a graph convolution model; the present invention inputs the gridded numerical weather forecast data at the prediction time point of the wind power group, which improves the continuity of the space where the wind power group is located compared to the traditional single-point numerical weather forecast, and the meteorological feature classification model improves the adaptability of the division of meteorological features. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 A schematic flow chart of a method for wind power group power prediction provided by the present invention;

[0059] Figure 2 A meteorological feature classification method based on wind power prediction power error characteristics and a wind power prediction flow chart provided by the present invention;

[0060] Figure 3 A comparison chart of prediction curves of the two prediction methods provided by the present invention;

[0061] Figure 4 A schematic diagram of a system for wind power group power prediction provided by the present invention. DETAILED DESCRIPTION

[0062] The specific implementation modes of the present invention are further described in detail below with reference to the accompanying drawings.

[0063] Embodiment 1:

[0064] The present invention provides a method for predicting wind power group power. Figure 1As shown, including:

[0065] Step 1: Match the gridded numerical weather forecast data of the wind power group at the forecast time point with different meteorological feature libraries through the pre-built meteorological feature classification model to obtain the meteorological features corresponding to the gridded numerical weather forecast data at each forecast time point;

[0066] Step 2: According to the meteorological characteristics, a corresponding wind power group prediction model is selected from a plurality of pre-constructed wind power group prediction models, and based on the gridded numerical weather forecast data at the prediction time point, the wind power prediction value is calculated through the corresponding wind power group prediction model;

[0067] The meteorological characteristic classification model is constructed based on the historical forecast power error data of the wind power group and the historical gridded numerical weather forecast data through the meteorological characteristic-error characteristic matching mapping method;

[0068] The wind power group prediction model is based on meteorological characteristics and is constructed using a graph convolution model.

[0069] Before step 1, it is necessary to build a meteorological feature classification model in advance, including:

[0070] Taking a wind power group consisting of N wind farms as the object, collect historical actual total power data of the wind power group and the corresponding historical predicted power data for at least one year, and obtain the historical predicted power error data of the wind power group by subtracting the historical actual total power data from the corresponding historical predicted power data;

[0071] The first autoencoder model includes at least one or more of the following: a first autoencoder and a first decoder; based on the historical predicted power error data, the feature extraction function of the first autoencoder is used to extract the historical predicted power error waveform feature data, and the historical predicted power error waveform feature data is divided at one day as the time interval. At this time, the data dimension is 24×A, where A is the number of days. The first autoencoder is constructed by the feature extraction function, and based on the historical predicted power error waveform feature data, the time dimension reduction function of the first decoder is used to obtain the historical predicted power error waveform feature data after data dimension reduction. At this time, the data dimension is 4×A, and the first decoder is constructed by the time dimension reduction function;

[0072] The K-means clustering method is used to divide the historical predicted power error waveform feature data after data dimension reduction into multiple categories, and the error type of each category is determined;

[0073] The historical gridded numerical weather forecast data covering the spatial geographical distribution of all wind farms in the wind power group are collected. The historical gridded numerical weather forecast data are stored in the format of four-dimensional data, and the data dimension is M×N×K×T, where M is the number of grids in the latitude direction, N is the number of grids in the longitude direction, K is the number of meteorological elements, and T is the number of sections at a time. The historical gridded numerical weather forecast data at the same time include the longitude and latitude of the upper left corner of the grid and all meteorological parameters of different grids. The historical gridded numerical weather forecast data of sections at different times are superimposed to form the final historical gridded numerical weather forecast data. The second autoencoder model includes a second autoencoder and a second decoder. Based on The historical gridded numerical weather forecast data is obtained through the time section dimension reduction function of the second autoencoder to obtain the historical gridded numerical weather forecast data after the first data dimension reduction. The data dimension at this time is M×N×K. Based on the historical gridded numerical weather forecast data after the first data dimension reduction, the latitude and longitude dimension reduction function of the second decoder is used to perform a second data dimension reduction to obtain the historical gridded numerical weather forecast data after the second data dimension reduction. The data dimension at this time is 25×K, that is, the data dimension of each meteorological element in the historical gridded numerical weather forecast data is reduced to 25. The second autoencoder is constructed based on the time section dimension reduction function, and the second decoder is constructed based on the latitude and longitude dimension reduction function.

[0074] Through the meteorological characteristics-error characteristics matching mapping method and supervised machine learning algorithm, a pre-trained matching mapping model is constructed, and the historical gridded numerical weather forecast data after secondary data dimensionality reduction and multiple error types are input into the matching mapping model, and the meteorological characteristics corresponding to each gridded numerical weather forecast data label are output;

[0075] The present invention takes the whole-grid wind power of a certain place as an example, and divides the historical predicted power error waveform feature data after data dimension reduction into three categories through the K-means clustering method, and determines the error type of each category, for example: the three error types are positive error type, no error type and negative error type, and according to the three error types, divides the historical gridded numerical weather forecast data after secondary data dimension reduction into historical gridded numerical weather forecast data with meteorological characteristics of generating positive errors, historical gridded numerical weather forecast data with meteorological characteristics of generating no errors and historical gridded numerical weather forecast data with meteorological characteristics of generating negative errors, and annotates the corresponding meteorological characteristics for each historical gridded numerical weather forecast data, and stores the gridded numerical weather forecast data corresponding to each type of meteorological characteristics in the corresponding meteorological characteristic library;

[0076] The number of meteorological elements of the historical gridded numerical weather forecast data of the present invention is 7. The historical gridded numerical weather forecast data at the same time includes the longitude and latitude of the upper left corner of the grid, and all meteorological parameters of different grids include one or more of the following: relative humidity, 10m wind speed, 10m wind direction, temperature, air pressure, 100m wind speed, 100m wind direction;

[0077] Specifically, step 1 includes:

[0078] The gridded numerical weather forecast data of the wind power group's forecast time point are matched with different meteorological feature libraries to obtain the meteorological features corresponding to the gridded numerical weather forecast data at the forecast time point, wherein the forecast time point includes at least one or more of the following: every day and every hour; the present invention inputs the gridded numerical weather forecast data of the wind power group's forecast time point, which improves the continuity of the space where the wind power group is located compared to the traditional single-point numerical weather forecast, and when constructing the meteorological feature classification model, the operating data of all sites in the group is not required, and the data demand is small.

[0079] Before step 1, it is necessary to build a wind power group prediction model in advance, including:

[0080] Extract the historical gridded numerical weather forecast sample data corresponding to the meteorological feature library of multiple types of meteorological features, and obtain the wind power cluster power data at the time corresponding to the historical gridded numerical weather forecast sample data;

[0081] The correspondence between the historical gridded numerical weather forecast sample data and the wind power cluster power data at the corresponding time is constructed by the graph convolution method, and a wind power cluster prediction model corresponding to multiple types of meteorological characteristics is constructed according to the correspondence.

[0082] The present invention takes the whole-grid wind power of a certain place as an example, extracts the historical grid numerical weather forecast sample data of the meteorological feature library corresponding to the three types of meteorological characteristics, and obtains the wind power cluster power data at the corresponding time of the historical grid numerical weather forecast sample data;

[0083] The correspondence between the historical gridded numerical weather forecast sample data and the wind power cluster power data at the corresponding time is constructed by the graph convolution method, and the wind power group prediction models corresponding to three types of meteorological characteristics are constructed according to the correspondence, namely the first wind power group prediction model CM1, the second wind power group prediction model CM2 and the third wind power group prediction model CM3;

[0084] Specifically, step 2 includes: selecting a corresponding wind power group prediction model from a plurality of pre-constructed wind power group prediction models according to the meteorological characteristics corresponding to the gridded numerical weather forecast data at the prediction time point, and calculating the wind power prediction value for the next several days through the corresponding wind power group prediction model based on the gridded numerical weather forecast data at the prediction time point; the present invention classifies each meteorological characteristic to construct a wind power group prediction model and performs wind power prediction, and Figure 2 The flowchart shown is used to construct each model and predict the power of the wind power group, which improves the efficiency of constructing each model and the accuracy of power prediction.

[0085] Embodiment 2:

[0086] Taking the wind power of a certain place as an example, the actual operation data from May 5 to August 22, 2022 are extracted, and the current mature neural network modeling method and the method of the present invention are compared. The prediction performance indicators are shown in Table 1, and the prediction curve is shown in Figure 3 As shown, the effect of the method of the present invention is the best compared with other methods.

[0087] Table 1 Performance indicators of the two prediction methods

[0088]

[0089] Embodiment 3:

[0090] Based on the same inventive concept, the present invention also provides a system for predicting wind power group power, such as Figure 4 As shown, it includes: a meteorological feature acquisition module and a power prediction module;

[0091] The meteorological feature acquisition module is used to match the gridded numerical weather forecast data at the forecast time point of the wind power group with different meteorological feature libraries through a pre-built meteorological feature classification model to obtain the meteorological features corresponding to the gridded numerical weather forecast data at each forecast time point;

[0092] The power prediction module is used to select a corresponding wind power group prediction model from a plurality of pre-constructed wind power group prediction models according to the meteorological characteristics, and calculate the wind power prediction value through the corresponding wind power group prediction model based on the gridded numerical weather forecast data at the prediction time point;

[0093] The meteorological characteristic classification model is constructed based on the historical forecast power error data of the wind power group and the historical gridded numerical weather forecast data through the meteorological characteristic-error characteristic matching mapping method;

[0094] The wind power group prediction model is based on meteorological characteristics and is constructed using a graph convolution model.

[0095] Preferably, the construction of the meteorological feature classification model in the meteorological feature acquisition module includes:

[0096] Based on the acquired historical predicted power error data of the wind power group, obtaining the historical predicted power error waveform characteristic data after data dimension reduction through the first autoencoder model, and determining the error type of the historical predicted power error waveform characteristic data;

[0097] Based on the historical gridded numerical weather forecast data of the wind power group, the historical gridded numerical weather forecast data after data dimension reduction is obtained through the second autoencoder model;

[0098] The historical gridded numerical weather forecast data after data dimensionality reduction are matched with the meteorological characteristics corresponding to the error type by a meteorological characteristic-error characteristic matching mapping method, and the corresponding meteorological characteristics are marked on the historical gridded numerical weather forecast data after data dimensionality reduction;

[0099] The historical gridded numerical weather forecast data corresponding to each type of meteorological feature after dimension reduction is stored in the corresponding meteorological feature library to complete the construction of the meteorological feature classification model;

[0100] Among them, the first autoencoder model is constructed based on the historical predicted power error data of the wind power group through the first autoencoder and the first decoder; the second autoencoder model is constructed based on the historical gridded numerical weather forecast data of the wind power group through the second autoencoder and the second decoder.

[0101] Preferably, the meteorological feature acquisition module obtains the historical predicted power error waveform feature data after data dimension reduction based on the acquired historical predicted power error data of the wind power group through the first autoencoder model, and determines the error type of the historical predicted power error waveform feature data, including:

[0102] Based on the historical predicted power error data of the wind power group, the historical predicted power error waveform feature data is extracted through the first autoencoder;

[0103] Based on the historical predicted power error waveform characteristic data, obtaining the historical predicted power error waveform characteristic data after data dimension reduction through a first decoder;

[0104] The historical predicted power error waveform characteristic data after the data dimension reduction is divided into multiple categories by a K-means clustering method, and the error type of each category is determined;

[0105] Among them, the first autoencoder is constructed by a feature extraction function; the first decoder is constructed by a time dimension reduction function.

[0106] Preferably, the meteorological feature acquisition module obtains the historical gridded numerical weather forecast data after data dimension reduction based on the historical gridded numerical weather forecast data of the wind power group through the second autoencoder model, including:

[0107] Based on the historical gridded numerical weather forecast data of the wind power group, a data dimension reduction is performed according to the time section through the second autoencoder to obtain the historical gridded numerical weather forecast data after the data dimension reduction;

[0108] Based on the historical gridded numerical weather forecast data after the primary data dimensionality reduction, a secondary data dimensionality reduction is performed through a second decoder to obtain the historical gridded numerical weather forecast data after the secondary data dimensionality reduction;

[0109] Among them, the second autoencoder is constructed based on the time section dimensionality reduction function; the second decoder is constructed based on the longitude and latitude dimensionality reduction function.

[0110] Preferably, the construction of the wind power group prediction model of the power prediction module includes:

[0111] Extracting historical gridded numerical weather forecast sample data corresponding to the meteorological feature library of various meteorological features, and obtaining wind power cluster power data at the time corresponding to the historical gridded numerical weather forecast sample data;

[0112] The correspondence between the historical gridded numerical weather forecast sample data and the wind power cluster power data at the corresponding time is constructed by the graph convolution method, and a wind power group prediction model corresponding to various meteorological characteristics is constructed according to the correspondence.

[0113] Preferably, the historical predicted power error data of the meteorological feature acquisition module is obtained by making a difference between the historical actual total power data of the wind power group and the corresponding historical predicted power data.

[0114] Embodiment 4:

[0115] Based on the same inventive concept, the present invention also provides a computer device, which includes a processor and a memory, wherein the memory is used to store a computer program, the computer program includes program instructions, and the processor is used to execute 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 (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, which is suitable for implementing one or more instructions, and is specifically suitable for loading and executing one or more instructions in the computer storage medium to implement the corresponding method flow or corresponding function, so as to implement the steps of a method for wind power group power prediction in the above embodiment.

[0116] Embodiment 5:

[0117] Based on the same inventive concept, the present invention also provides a storage medium, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device for storing programs and data. It can be understood that the computer-readable storage medium here can include both built-in storage media in a computer device and, of course, extended storage media supported by the computer device. The computer-readable storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by a processor are also stored in the storage space, and these instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk memory. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the steps of a method for wind power group power prediction in the above embodiment.

[0118] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented 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.

[0119] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0120] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0121] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0122] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit its protection scope. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that after reading the present invention, those skilled in the art can still make various changes, modifications or equivalent substitutions to the specific implementation methods of the application, but these changes, modifications or equivalent substitutions are all within the protection scope of the claims to be approved.

Claims

1. A method for predicting wind power group power, characterized in that: include: The gridded numerical weather forecast data of the wind power group at the forecast time point is matched with different meteorological feature libraries through the pre-built meteorological feature classification model to obtain the meteorological features corresponding to the gridded numerical weather forecast data at each forecast time point; According to the meteorological characteristics, a corresponding wind power group prediction model is selected from a plurality of pre-constructed wind power group prediction models, and based on the gridded numerical weather forecast data at the prediction time point, the wind power prediction value is calculated through the corresponding wind power group prediction model; The meteorological characteristic classification model is constructed based on the historical forecast power error data of the wind power group and the historical gridded numerical weather forecast data through the meteorological characteristic-error characteristic matching mapping method; The wind power group prediction model is based on meteorological characteristics and is constructed using a graph convolution model.

2. The method according to claim 1, characterized in that The construction of the meteorological feature classification model includes: Based on the acquired historical predicted power error data of the wind power group, obtaining the historical predicted power error waveform characteristic data after data dimension reduction through the first autoencoder model, and determining the error type of the historical predicted power error waveform characteristic data; Based on the historical gridded numerical weather forecast data of the wind power group, the historical gridded numerical weather forecast data after data dimension reduction is obtained through the second autoencoder model; The historical gridded numerical weather forecast data after data dimensionality reduction are matched with the meteorological characteristics corresponding to the error type by a meteorological characteristic-error characteristic matching mapping method, and the corresponding meteorological characteristics are marked on the historical gridded numerical weather forecast data after data dimensionality reduction; The historical gridded numerical weather forecast data corresponding to each type of meteorological feature after dimension reduction is stored in the corresponding meteorological feature library to complete the construction of the meteorological feature classification model; Among them, the first autoencoder model is constructed based on the historical predicted power error data of the wind power group through the first autoencoder and the first decoder; the second autoencoder model is constructed based on the historical gridded numerical weather forecast data of the wind power group through the second autoencoder and the second decoder.

3. The method according to claim 2, characterized in that The method of obtaining the historical predicted power error data of the wind power group through the first autoencoder model to obtain the historical predicted power error waveform feature data after data dimension reduction includes: Based on the historical predicted power error data of the wind power group, the historical predicted power error waveform feature data is extracted through the first autoencoder; Based on the historical predicted power error waveform characteristic data, obtaining the historical predicted power error waveform characteristic data after data dimension reduction through a first decoder; The historical predicted power error waveform characteristic data after the data dimension reduction is divided into multiple categories by a K-means clustering method, and the error type of each category is determined; Among them, the first autoencoder is constructed by a feature extraction function; the first decoder is constructed by a time dimension reduction function.

4. The method according to claim 2, characterized in that The historical gridded numerical weather forecast data based on the wind power group is obtained through the second autoencoder model to obtain the historical gridded numerical weather forecast data after data dimension reduction, including: Based on the historical gridded numerical weather forecast data of the wind power group, a data dimension reduction is performed according to the time section through the second autoencoder to obtain the historical gridded numerical weather forecast data after the data dimension reduction; Based on the historical gridded numerical weather forecast data after the primary data dimensionality reduction, a secondary data dimensionality reduction is performed through a second decoder to obtain the historical gridded numerical weather forecast data after the secondary data dimensionality reduction; Among them, the second autoencoder is constructed based on the time section dimensionality reduction function; the second decoder is constructed based on the longitude and latitude dimensionality reduction function.

5. The method according to claim 2, characterized in that The construction of the wind power group prediction model includes: Extracting historical gridded numerical weather forecast sample data corresponding to the meteorological feature library of various meteorological features, and obtaining wind power cluster power data at the time corresponding to the historical gridded numerical weather forecast sample data; The correspondence between the historical gridded numerical weather forecast sample data and the wind power cluster power data at the corresponding time is constructed by the graph convolution method, and a wind power group prediction model corresponding to various meteorological characteristics is constructed according to the correspondence.

6. The method according to claim 2, characterized in that The historical predicted power error data is obtained by subtracting the historical actual total power data of the wind power group from the corresponding historical predicted power data.

7. A system for predicting wind power group power, characterized in that: include: Meteorological characteristics acquisition module and power prediction module; The meteorological feature acquisition module is used to match the gridded numerical weather forecast data at the forecast time point of the wind power group with different meteorological feature libraries through a pre-built meteorological feature classification model to obtain the meteorological features corresponding to the gridded numerical weather forecast data at each forecast time point; The power prediction module is used to select a corresponding wind power group prediction model from a plurality of pre-constructed wind power group prediction models according to the meteorological characteristics, and calculate the wind power prediction value through the corresponding wind power group prediction model based on the gridded numerical weather forecast data at the prediction time point; The meteorological characteristic classification model is constructed based on the historical forecast power error data of the wind power group and the historical gridded numerical weather forecast data through the meteorological characteristic-error characteristic matching mapping method; The wind power group prediction model is based on meteorological characteristics and is constructed using a graph convolution model.

8. The system according to claim 7, characterized in that The construction of the meteorological feature classification model in the meteorological feature acquisition module includes: Based on the acquired historical predicted power error data of the wind power group, obtaining the historical predicted power error waveform characteristic data after data dimension reduction through the first autoencoder model, and determining the error type of the historical predicted power error waveform characteristic data; Based on the historical gridded numerical weather forecast data of the wind power group, the historical gridded numerical weather forecast data after data dimension reduction is obtained through the second autoencoder model; The historical gridded numerical weather forecast data after data dimensionality reduction are matched with the meteorological characteristics corresponding to the error type by a meteorological characteristic-error characteristic matching mapping method, and the corresponding meteorological characteristics are marked on the historical gridded numerical weather forecast data after data dimensionality reduction; The historical gridded numerical weather forecast data corresponding to each type of meteorological feature after dimension reduction is stored in the corresponding meteorological feature library to complete the construction of the meteorological feature classification model; Among them, the first autoencoder model is constructed based on the historical predicted power error data of the wind power group through the first autoencoder and the first decoder; the second autoencoder model is constructed based on the historical gridded numerical weather forecast data of the wind power group through the second autoencoder and the second decoder.

9. The system according to claim 8, characterized in that The meteorological feature acquisition module obtains the historical predicted power error waveform feature data after data dimension reduction based on the acquired historical predicted power error data of the wind power group through the first autoencoder model, and determines the error type of the historical predicted power error waveform feature data, including: Based on the historical predicted power error data of the wind power group, the historical predicted power error waveform feature data is extracted through the first autoencoder; Based on the historical predicted power error waveform characteristic data, obtaining the historical predicted power error waveform characteristic data after data dimension reduction through a first decoder; The historical predicted power error waveform characteristic data after the data dimension reduction is divided into multiple categories by a K-means clustering method, and the error type of each category is determined; Among them, the first autoencoder is constructed by a feature extraction function; the first decoder is constructed by a time dimension reduction function.

10. The system according to claim 8, characterized in that The meteorological feature acquisition module obtains the historical gridded numerical weather forecast data after data dimension reduction based on the historical gridded numerical weather forecast data of the wind power group through the second autoencoder model, including: Based on the historical gridded numerical weather forecast data of the wind power group, a data dimension reduction is performed according to the time section through the second autoencoder to obtain the historical gridded numerical weather forecast data after the data dimension reduction; Based on the historical gridded numerical weather forecast data after the primary data dimensionality reduction, a secondary data dimensionality reduction is performed through a second decoder to obtain the historical gridded numerical weather forecast data after the secondary data dimensionality reduction; Among them, the second autoencoder is constructed based on the time section dimensionality reduction function; the second decoder is constructed based on the longitude and latitude dimensionality reduction function.

11. A computer device, characterized in that: include: one or more processors; A memory for storing one or more programs; When the one or more programs are executed by the one or more processors, a method for wind power group power prediction according to any one of claims 1 to 6 is implemented.

12. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed, a method for wind power group power prediction according to any one of claims 1 to 6 is implemented.